{"task_id": "format-code-task-000001", "source_id": "format-code-task-000001", "domain": "code", "task_path": "tasks/format-code-task-000001", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f0701aeb7d0813fe06c2b72bb6b558df48cefb21be6a283fbb0250b317b45e12", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Stuck transactions caused by excessive gas limits\n\nWe're hitting a nasty bug: a transaction submitted with an absurdly large gas\nlimit (well above the block gas limit) gets permanently stuck, and worse, it\nblocks every later transaction from the same account.\n\nHere's what happens. When the block builder pops a transaction and writes it\ninto the block being assembled, the write fails because the requested gas can\nnever fit in a block. But the builder treats *every* write failure identically:\nit puts the transaction back and tries again on the next block. So this\ntransaction is retried forever, and since the pool already advanced the\naccount's expected nonce when it accepted the transaction, no later transaction\nfrom that account can ever be processed.\n\nThe underlying problem is two-fold and I'd like both addressed:\n\n1. **The state executor's transaction-write path doesn't distinguish \"retry\"\n from \"discard\" failures.** When writing a transaction to a block fails, the\n caller needs to know whether the failure is *recoverable* — the transaction\n might succeed in a later block, so it should be retried — or *non-recoverable*\n — the transaction can never be included and must be dropped.\n\n A write failure is **non-recoverable** precisely when the transaction's gas\n requirement exceeds the block's gas limit (no block could ever hold it).\n Every other failure encountered while writing is **recoverable**, including:\n an incorrect nonce, the sender not being able to afford the gas cost, and a\n transaction that would fit a full block but not the gas remaining in the\n current (already partly filled) one.\n\n Expose this through the value returned by the write call: it must still\n behave as an `error` (and be `nil` on success), but when non-nil it must\n carry an exported `IsRecoverable` boolean field reporting the above.\n\n2. **The pool can't roll back an account's expected nonce.** The pool tracks the\n next expected nonce per account. When a transaction it tracked is discarded,\n that counter has to be rolled back, otherwise the account stays blocked. Add\n a method `DecreaseAccountNonce(tx *types.Transaction)` to the transaction pool\n that decrements by one the next expected nonce it tracks for the sender of\n `tx`. If the pool isn't tracking that account, it must be a safe no-op.\n\nWith these in place, the block builder can drop a permanently-invalid\ntransaction (and roll back the account nonce) while still retrying transactions\nthat merely didn't fit the current block.\n"} {"task_id": "format-code-task-000002", "source_id": "format-code-task-000002", "domain": "code", "task_path": "tasks/format-code-task-000002", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:22a911b303065341fa7cf655d07e201b08d0714420cc176ece7e06c46650614c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\non internal,\n\nrequest:\n\n```json\n{\n\t\"jsonrpc\":\"2.0\",\n\t\"method\":\"zkevm_batchNumberByBlockNumber\",\n\t\"params\":[\n\t\t\"87377\"\n\t],\n\t\"id\":1\n}\n```\n\nresponse:\n\n```json\n{\n \"jsonrpc\": \"2.0\",\n \"id\": 1,\n \"error\": {\n \"code\": -32000,\n \"message\": \"failed to get batch number from block number\"\n }\n}\n```"} {"task_id": "format-code-task-000003", "source_id": "format-code-task-000003", "domain": "code", "task_path": "tasks/format-code-task-000003", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:89d2302961adfe5b768b28b72e5c7a227e24e32d923f2a996b077c85fa3bc428", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## SHL opcode not implemented\n\nI'm running some Solidity-compiled bytecode through smol-evm and it bails out partway with an `UnknownOpcode` error. Disassembling the bytecode I can see it's hitting a `SHL` (shift left) — and looking through `opcodes.py` I notice `SHR` is implemented but `SHL` isn't there at all.\n\nCould we add it? The Solidity compiler emits `SHL` pretty routinely (bit packing, struct layout, masking, etc.), so right now smol-evm can't really get through most real contract bytecode. Behavior should match what the yellow paper / EVM spec defines for `SHL`."} {"task_id": "format-code-task-000004", "source_id": "format-code-task-000004", "domain": "code", "task_path": "tasks/format-code-task-000004", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2777136eb50329705f24b2400b691aade6b6942117011fbb9399aa3ad3cc1bd0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problème\n\nDans Pix Admin, sur l'écran de recherche d'organisations, quand je saisis du texte (par exemple `foo`) dans le champ ID puis que je lance la recherche, l'interface affiche le message générique **« Erreur dans Ember »** et aucun résultat ne s'affiche.\n\n## Reproduction\n\n1. Se connecter à Pix Admin avec un compte Pix Master\n2. Aller sur la liste des organisations\n3. Dans le filtre par ID, saisir une valeur non numérique (ex. `foo`)\n4. Lancer la recherche\n\n→ L'appel à `GET /api/organizations?filter[id]=foo` part vers l'API et la réponse fait planter l'UI (\"Erreur dans Ember\").\n\n## Comportement attendu\n\nSaisir un ID non numérique est une simple erreur de frappe côté utilisateur ; ça ne devrait pas faire planter quoi que ce soit. La requête devrait être rejetée proprement par l'API au lieu d'aller jusqu'au bout du traitement.\n\nLes autres cas de recherche doivent continuer à fonctionner normalement :\n- aucun critère de recherche\n- ID numérique existant\n- recherche par nom"} {"task_id": "format-code-task-000007", "source_id": "format-code-task-000007", "domain": "code", "task_path": "tasks/format-code-task-000007", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:39abf5a6f56accf125a99c5047a0c738af4ff1b920c29267a64b2a0901854f73", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Filling an order against a non-contract takerAsset silently \"succeeds\"\n\nI was stress-testing edge cases against `fillOrderArgs` and noticed that the protocol does not behave like SafeERC20 when the taker asset address has no contract code at it.\n\n### Repro\n\n1. Build a regular order, but set `takerAsset` to an address that has no code deployed at it on the current network. Realistic ways this happens:\n - typo in the token address,\n - signing an order on mainnet for a token that only exists on another chain,\n - a token that has been self-destructed.\n2. Maker signs and the order goes on the book.\n3. A taker calls `fillOrderArgs` for that order.\n\n### What I see\n\nThe fill goes through. `OrderFilled` is emitted, the maker's `makerAsset` is transferred out to the taker, and on the maker side I see no `takerAsset` arrived (which makes sense — there's no token contract there to actually move balances). From the maker's point of view they just gave funds away for nothing.\n\n### What I expected\n\nThe taker→maker transfer leg should be treated as failed and the whole `fillOrderArgs` call should revert (with the existing `TransferFromTakerToMakerFailed`), so the maker's asset is never moved. That's what OpenZeppelin's `SafeERC20.safeTransferFrom` does — if the call to `transferFrom` returns no data, it still requires the callee to actually be a contract; otherwise the \"success\" is meaningless.\n\n### Why I think this is in scope here\n\nThe protocol uses an internal helper (`_callTransferFromWithSuffix`) for the taker→maker transfer instead of the standard SafeERC20 path, presumably so it can append the taker asset suffix. That helper's success check is more permissive than SafeERC20's — calling into an EOA returns no return data and is being accepted as a successful ERC20 transfer. The helper should have the same \"no code, no deal\" guarantee that SafeERC20 has, otherwise orders against bogus / non-existent token addresses will keep draining makers."} {"task_id": "format-code-task-000008", "source_id": "format-code-task-000008", "domain": "code", "task_path": "tasks/format-code-task-000008", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f83df347542976209efc13596eabae5e46f35b4330f7fae6fa9e487c5493f0dd", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `manimgl` and `manim-render` commands to support opening or revealing rendered output files directly from the CLI. When I run `manimgl path/to/scene.py SceneName --open` or `manimgl path/to/scene.py SceneName -o`, the command should render the scene to a media file just like `--write_file`, and after the movie or final-frame image has been saved it should launch that saved file with the platform's default opener. Passing `--open` by itself should imply file writing; users should not also have to pass `--write_file`.\n\nI also want `manimgl path/to/scene.py SceneName --finder` to imply file writing and then reveal the saved output in the operating system's file browser when rendering completes. If the run saves only the final frame with `--skip_animations`, the image file should be opened or revealed; otherwise the movie file should be opened or revealed. These flags should appear in `manimgl --help` with `-o, --open` described as automatically opening the saved file and `--finder` described as showing the output file in Finder/file browser. A successful render that opens or reveals the output should still exit with code 0."} {"task_id": "format-code-task-000011", "source_id": "format-code-task-000011", "domain": "code", "task_path": "tasks/format-code-task-000011", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c8522aaf8f02f486d9dfadc34812d082efbca8b337cf987d2940045cd93f00a3", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## 1277 and 338 directories are placeholders — please fill them in\n\nI was browsing this repo looking for the Go solutions to a couple of LeetCode problems and noticed that **1277. Count Square Submatrices with All Ones** and **338. Counting Bits** both appear to be unfinished placeholders.\n\nIn `leetcode/1201-1300/1277.Count-Square-Submatrices-with-All-Ones/` and `leetcode/301-400/0338.Counting-Bits/`, the contents look like this:\n\n- `README.md` has the `> [!WARNING|style:flat]` banner saying *\"This question is temporarily unanswered if you have good ideas. Welcome to Create Pull Request PR\"*.\n- The `## Description` section is mostly empty, and the only example given is:\n ```\n Input: a = \"11\", b = \"1\"\n Output: \"100\"\n ```\n which clearly belongs to some other problem (looks like a string-addition / add-binary kind of question), not to 1277 or 338.\n- `Solution.go` only contains the dummy stub:\n ```go\n package Solution\n\n func Solution(x bool) bool {\n return x\n }\n ```\n The `bool` signature obviously doesn't match either problem — 1277 takes a 2D matrix and returns a count, 338 takes an integer `n` and returns a slice.\n\nCould someone fill these two in properly? Specifically, for each of the two problems:\n\n- update `README.md` so the description and example(s) actually match what LeetCode is asking for,\n- and replace the stub `Solution.go` with a real implementation using a function signature that fits the problem's input/output.\n\nHappy to review if someone picks them up. Thanks!"} {"task_id": "format-code-task-000013", "source_id": "format-code-task-000013", "domain": "code", "task_path": "tasks/format-code-task-000013", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7f52223c1e597cfb277e8814377ecc2166d70708273938e0b8d60631d84a9470", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nThe small `lib/lru` package is now being used by callers that need to inspect and manage a cache without reaching into its implementation. Extend its public API while fixing the byte-accounting bug when an existing key is updated.\n\nAdd these methods to `*Cache`: `Bytes() int64`, `Peek(key string) (Value, bool)`, `Keys() []string`, `Remove(key string) (Value, bool)`, `Clear()`, `Resize(maxBytes int64)`, `Stats()` returning a struct value with exported `Hits` and `Misses` fields of type `uint64`, and `ResetStats()`.\n\nThe observable contract is:\n\n- An entry consumes `len(key) + value.Len()` bytes. `Bytes` reports the sum for entries currently retained. Replacing an existing key must account for both larger and smaller values, make that key most recently used, and must not itself invoke `OnEvicted`.\n- A positive byte limit is enforced after every add or replacement by evicting least-recently-used entries until the cache fits. If one entry is larger than the limit, it is added and then evicted through the normal callback path. A limit of zero continues to mean unlimited capacity.\n- `Get` keeps its current lookup and recency behavior and increments exactly one counter: `Hits` for a found key or `Misses` for an absent key. `Peek` returns the same value/found pair without changing recency or either counter. `Keys` returns the current keys from most to least recently used and does not change recency or counters. `ResetStats` zeros both counters without changing entries.\n- `Remove` returns the removed value and `true`, updates byte usage, and invokes `OnEvicted` once when the key exists. For a missing key it returns the zero `Value` and `false` without a callback.\n- `Clear` removes every entry, leaves both `Len()` and `Bytes()` at zero, and invokes `OnEvicted` once per entry in least-to-most-recently-used order.\n- `Resize` changes the byte limit immediately. A positive limit evicts least-recently-used entries, using `OnEvicted`, until the cache fits; resizing to zero disables the limit without discarding retained entries.\n\nKeep the existing `New`, `Add`, `Get`, `RemoveOldest`, `Len`, `Value`, and `OnEvicted` surfaces compatible. In particular, `Get` and replacement still promote an entry, `RemoveOldest` removes the least-recently-used entry and invokes the callback, and existing zero-limit callers remain unlimited. The cache may remain documented as unsafe for concurrent access."} {"task_id": "format-code-task-000014", "source_id": "format-code-task-000014", "domain": "code", "task_path": "tasks/format-code-task-000014", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2c7c021648118dd49b4be28cd89fd85d6e510e31a6a95588847e3959b64fe03f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `github.com/8treenet/freedom/infra/kafka` package to provide a stateful Kafka producer API for Freedom applications. The entry point should be `kafka.GetProducer() kafka.Producer`, where `Producer` exposes `Start(addrs []string, config *sarama.Config)`, `NewMsg(topic string, content []byte) *kafka.Msg`, and `Restart() error`; the concrete `*kafka.ProducerImpl` should also support `Close() error` for lifecycle cleanup.\n\nCalling `Start([]string{\"127.0.0.1:9092\"}, cfg)` should remember the broker addresses and Sarama config for the producer lifecycle, and it should force `cfg.Producer.Return.Errors` and `cfg.Producer.Return.Successes` to `true` so synchronous sends can report their result. After the Freedom application boots with configured broker addresses, publishing `kafka.GetProducer().NewMsg(\"orders\", []byte(\"paid\")).SetMessageKey(\"order-1\").SetHeader(map[string]interface{}{\"x-request-id\":\"r1\", \"attempt\": 2}).Publish()` should send one Sarama producer message to topic `orders`, with value `paid`, key `order-1`, and headers `x-request-id: r1` and `attempt: 2`. If no message key is set, `Publish` should generate a non-empty uppercase UUID-style key with dashes removed before sending.\n\n`Msg.SetHeader` should be chainable and should merge additional header maps into any existing headers. `Msg.GetHeader`, `Msg.SetMessageKey`, and `Msg.GetMessageKey` should expose the current message headers and key consistently before publish. If `Publish` is called before the producer has an initialized Sarama sync producer, it should return an error equivalent to `producer is not initialized` instead of panicking.\n\nI also need producer middleware support through `kafka.InstallMiddleware(...kafka.ProducerHandler)`, where `ProducerHandler` is `func(*kafka.Msg)`. Installed handlers should run in order when `Msg.Publish` calls `Msg.Next`; a handler can call `Next` to continue, call `Stop` to prevent the send, and inspect `IsStopped` or `GetExecution` for message state. `Restart()` should close the current producer connection and then dial again using the stored addresses and config, and `Close()` should be a no-op when no producer connection is active."} {"task_id": "format-code-task-000016", "source_id": "format-code-task-000016", "domain": "code", "task_path": "tasks/format-code-task-000016", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5f46605969f46946830b492ccca83d68404d50cce97b3796b8bd2257715d0ba1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Sync IDA Pro type definitions across collaborating analysts\n\nWe're using Polichombr to collaborate on reversing a sample with two\nother analysts. The current server-side IDA actions cover most of what\nwe need — global names, comments, and structs all sync between our IDA\ninstances through the API — but there's one IDA action that doesn't get\nshared: **type definitions** (the ones you set by hitting `Y` on a\nvariable, function, or address in IDA Pro).\n\n### What we'd like to do\n\nIn IDA, I'll often refine a function prototype on a `sub_xxxxxx`, e.g.\nturn the default `int __cdecl sub_401000(int, int)` into something\nmeaningful like `BOOL __stdcall ParseConfigBlob(char *blob, size_t\nlen)`. Then I want my teammates to pick that up automatically the next\ntime their plugin pulls from the server, the same way they pick up the\nnames and comments I made.\n\nRight now, there's no way to do this through Polichombr. Names get\npushed and pulled fine, structs get pushed and pulled fine, but the\ntype info I set with `Y` just stays local in my `.idb` and the rest of\nthe team has to re-discover it independently. For functions with\nnon-trivial prototypes (lots of pointers, callbacks, custom typedefs)\nthis ends up being a real source of duplicated work.\n\n### What we'd expect\n\nIdeally the server should treat applied type definitions as just\nanother kind of IDA action, alongside names / comments / structs:\n\n- the plugin should be able to push a type definition for a given\n sample at a given address,\n- and a teammate's plugin should be able to query, for a given sample,\n the type definitions that have been recorded — with the same kind of\n filtering you already support for names and comments (by address, or\n only the ones added after a given timestamp, so we don't keep\n re-pulling the whole history on every sync).\n\nWould it be possible to add this? Happy to test against any branch.\n\n(For API shape, mirroring the existing `/samples//names/` style would make sense — e.g. a `types`-flavored endpoint that accepts an `address` + `typedef` payload on push and returns the recorded `typedefs` on pull.)"} {"task_id": "format-code-task-000018", "source_id": "format-code-task-000018", "domain": "code", "task_path": "tasks/format-code-task-000018", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:627524c653f2b1669cafec6ac48ceeadbc3cd695383a5954f060fb4ee4d1f4a5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Persist sublattice metadata atomically and track the root dispatch\n\nWhen the dispatcher runs a workflow that contains sublattices, each sublattice is\nitself dispatched and gets its own lattice record in the results database. Two\nthings are currently awkward about how those records are created.\n\nFirst, the database column that links a sublattice's lattice record back to the\nparent electron that spawned it (`electron_id`) is only filled in *after* the\nsublattice has already been written to the database. There is a window during\nwhich a sublattice record exists with no link to its parent, so an interruption\ncan leave it orphaned. We want the parent electron id to be written in the *same*\ninsertion that creates the lattice record, not in a later update.\n\nSecond, we are about to need a way to identify, for any (sub)lattice in a\nhierarchy, the dispatch id of the top-level (\"root\") workflow that ultimately\nkicked everything off. A nested sublattice three levels deep should still be able\nto report the dispatch id of the original top-level dispatch.\n\nPlease make the following behavior available.\n\n## `root_dispatch_id` on the result object\n\nThe result object should expose a read-only `root_dispatch_id`. For a result\nobject that represents a top-level dispatch, this equals its own dispatch id. A\nresult object constructed for a given dispatch id reports that same id as its\nroot dispatch id until it is told otherwise (see the factory below).\n\n## Persisting a parent electron id atomically\n\nPersisting a result object should optionally accept the database id of the parent\nelectron that spawned this workflow. When provided, that id must be stored on the\nlattice record in the very same transaction that first creates the lattice\nrecord — not written by a subsequent update. When it is not provided, the\nlattice record's `electron_id` stays null. Persisting a top-level workflow must\ncontinue to behave exactly as before (its `electron_id` is null).\n\n## A factory for building a result object from a serialized lattice\n\nAdd a convenience factory, importable as\n`covalent._results_manager.result.initialize_result_object`, that constructs and\npersists a result object from a JSON-serialized lattice. It takes the JSON\nlattice and, optionally, a parent result object and a parent electron id:\n\n```\ninitialize_result_object(json_lattice, parent_result_object=None, parent_electron_id=None)\n```\n\nIt must:\n\n- deserialize the lattice and assign the new result object a freshly generated,\n unique dispatch id (two calls never collide);\n- initialize the result object's nodes and persist it before returning, passing\n along the parent electron id so it lands atomically on the lattice record as\n described above;\n- when a parent result object is supplied (i.e. this lattice is a sublattice),\n make the new result object inherit the parent's `root_dispatch_id` rather than\n using its own dispatch id;\n- return the fully constructed result object.\n\nThe returned result object's own dispatch id must still be distinct from its\nparent's.\n"} {"task_id": "format-code-task-000019", "source_id": "format-code-task-000019", "domain": "code", "task_path": "tasks/format-code-task-000019", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:78f404abfd84d841daf74e13995a6fc5c6f7fc25031414ddfa21030eca5aea78", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Bulk-delete dispatches by status filter\n\nRight now the dispatch-management backend can only soft-delete dispatches one at\na time, by passing an explicit list of dispatch ids. The UI needs a \"delete all\"\ncapability so users can clear out every dispatch that matches whatever they are\ncurrently looking at — i.e. the active status filter plus the search box — in a\nsingle request.\n\nAdd a bulk deletion operation to the dispatch summary data-access layer (the\nsame component that already handles per-id deletion and the dispatch listing).\nIt should accept a request object `DeleteAllDispatchesRequest` with two optional\nfields:\n\n- `status_filter`: a dispatch status enum value, defaulting to a new `ALL`\n selector that means \"every status\".\n- `search_string`: a string, defaulting to `\"\"`.\n\nExpose the behavior as a `delete_all_dispatches` method that takes such a request\nand returns the same response shape used by the existing per-id delete\n(`success_items`, `failure_items`, `message`).\n\nBehavior:\n\n- Selection. Only currently-active dispatches are eligible. A dispatch is\n selected when its status matches the filter **and** the search string is a\n case-insensitive substring of either its name or its dispatch id. An empty\n search string matches everything.\n- Status filter semantics:\n - `ALL` selects dispatches in any status.\n - `COMPLETED` is a group selector: it selects dispatches that are `COMPLETED`\n as well as those in the post-processing states `POSTPROCESSING`,\n `POSTPROCESSING_FAILED`, and `PENDING_POSTPROCESSING`.\n - Any other status value selects only dispatches in exactly that status.\n- Deletion is a soft delete, consistent with the existing per-id delete: each\n selected dispatch and its electrons are marked inactive. Dispatches that are\n already inactive are never re-selected or reported.\n- Response. `success_items` contains the dispatch ids (as UUIDs) that were\n deleted; order is not significant. When at least one dispatch was deleted the\n `message` is `\"Dispatch(es) have been deleted successfully!\"`. When nothing\n matched, `success_items` is empty and the `message` is\n `\"No dispatches were deleted\"`.\n\nThe existing per-id delete operation must keep working unchanged.\n"} {"task_id": "format-code-task-000020", "source_id": "format-code-task-000020", "domain": "code", "task_path": "tasks/format-code-task-000020", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e8f805c67ada7162c7dfea3a3d48857b90ac0c5376b4822500c9053bf1987127", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI'm working with a custom ACSet implementation, and `elements(mySet)` immediately throws a MethodError; I hit the same kind of MethodError when passing one of these ACSet instances as the sample `typ` to `inverse_elements` for both an Elements object and an Elements morphism.\n\n## Expected Outcomes\n\n- `elements` should work for ACSet instances beyond the built-in struct-backed cases, producing an `Elements` value that reflects the input instance instead of failing during method lookup.\n- `inverse_elements` should accept an ACSet instance as the sample `typ` when converting an `AbstractElements` value back, and should return the corresponding ACSet result rather than a method-dispatch failure.\n- `inverse_elements` should also accept an ACSet instance as the sample `typ` when converting an Elements morphism back, and should return the corresponding ACSet transformation rather than a method-dispatch failure.\n- Existing behavior for struct-backed ACSets should remain compatible.\n\n## Implementation Notes\n\nThe concrete data structures, validation locations, and code organization are up to the implementer. The fix should be driven by the public behavior of ACSet values participating in the category-of-elements round trip, not by special-casing only one test fixture."} {"task_id": "format-code-task-000022", "source_id": "format-code-task-000022", "domain": "code", "task_path": "tasks/format-code-task-000022", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e5915160e52cff553af177130ed7a1e58011a84b9f7d1efe47800d77bba975c1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Test accounts can't sign EIP712 messages\n\nI'm writing tests for a contract that uses EIP712 signatures (it's an ERC-2612 `permit`-style flow, so I need to construct a typed `Permit` message and have an account sign it). For the live network the signing works through my real account, but in the ape test suite I want to use one of the built-in test accounts so the test is hermetic.\n\nRoughly what I'm doing:\n\n```python\nfrom eip712.messages import EIP712Message\n\nclass Permit(EIP712Message):\n _name_ = \"MyToken\"\n _version_ = \"1\"\n _chainId_ = 1\n _verifyingContract_ = token.address\n\n owner: \"address\"\n spender: \"address\"\n value: \"uint256\"\n nonce: \"uint256\"\n deadline: \"uint256\"\n\ndef test_permit(accounts, token):\n owner = accounts[0] # an ape_test TestAccount\n spender = accounts[1]\n\n msg = Permit(\n owner=owner.address,\n spender=spender.address,\n value=1000,\n nonce=0,\n deadline=2**32,\n )\n\n sig = owner.sign_message(msg)\n assert sig is not None # <- fails\n token.permit(owner, spender, 1000, 2**32, sig.v, sig.r, sig.s, sender=spender)\n```\n\n`owner.sign_message(msg)` just gives me back `None`, so the assertion blows up and I can never get to the `permit` call. Plain string messages signed against the same test account work fine, it's specifically the EIP712 typed message that comes back empty.\n\nIt'd be great if the `ape_test` accounts could sign EIP712 messages the same way they handle the other message types, so tests that exercise `permit` / EIP712-based auth flows can actually run end-to-end without a real key."} {"task_id": "format-code-task-000023", "source_id": "format-code-task-000023", "domain": "code", "task_path": "tasks/format-code-task-000023", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:df1ec1afb6406e966b4e45334caf0d1ca873927e0fdb1203f91e0097ace9da10", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nThe project compiler cache is used on every project load, but its current behavior is fragile: an empty contracts directory is treated as present, diagnostics crash when the directory is absent, filesystem traversal makes compiler input order vary, malformed cache files prevent a rebuild, and recompiling one source can leave contract types produced by an older version of that source in the manifest. These problems make a clean checkout noisy and can expose stale artifacts after normal edits or deletions.\n\nMake the public `ProjectManager` behavior deterministic and self-healing while preserving the existing compiler/plugin interfaces. `sources_missing` must report true for a missing or empty `contracts/` directory. `sources` must recursively discover files whose suffix has a registered compiler and return them in stable lexicographic order by their path relative to `contracts/`. `extensions_with_missing_compilers` must recursively collect unique unsupported file suffixes in the same stable order and return an empty list, rather than raising, when `contracts/` is missing or empty.\n\n`load_contracts(use_cache=True)` must treat a malformed local manifest cache (for example, invalid JSON) as a cache miss and rebuild from current sources. For a valid cache, unchanged sources must be reused without invoking their compiler again, while new or content-changed sources are compiled. Results belonging to a changed source must replace that source's prior contract types, and results belonging to deleted sources must disappear from the returned mapping and the persisted manifest. The persisted cache must describe the current source set so a subsequent load sees the same result. `use_cache=False` remains the explicit force mode and must recompile every current source, in deterministic order."} {"task_id": "format-code-task-000024", "source_id": "format-code-task-000024", "domain": "code", "task_path": "tasks/format-code-task-000024", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:97ec31a98b47222fb0b4208d1a14059d422c308e0ed347624f312f64b7ffd7cb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\n我在用 create-market 表单建分类市场（categorical），发现校验有点漏。比如我两个 outcome 填了一模一样的答案，它居然能过；还有我把某个 outcome 留空，它也不一定报错，就这么放过去了。感觉像是分类答案那块的校验没盖全。最好能让重复的、空的答案都老老实实报出来，别让我糊里糊涂就提交了。\n\n## Expected outcomes\n\n- Blank categorical market outcomes are reported with the user-facing error message `Answer cannot be blank`.\n- Categorical market outcomes with the same case-sensitive value are reported with the user-facing error message `Category must be unique` for each submitted answer involved in the duplication.\n- The categorical-outcome error output lets callers associate each reported error with the submitted answer that caused it, without reordering or losing the submitted outcome positions.\n- Missing or empty categorical-outcome input is treated as having no categorical outcome errors.\n- The create-market form’s categorical step runs categorical-outcome validation whenever categorical outcomes are present in form state.\n\n## Implementation notes\n\nThe exact module organization, helper functions, and placement of validation logic are up to the implementer. Preserve the existing create-market form validation style and public user-facing messages, but avoid relying on any particular internal file layout or refactor path."} {"task_id": "format-code-task-000025", "source_id": "format-code-task-000025", "domain": "code", "task_path": "tasks/format-code-task-000025", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4fe75c4a4c4e6210d4a87aa10998fcb93c66a6ec1f851bdcbd96415bf286c4bd", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nts def of updateOne&findOneAndUpdate returns FlattenMap instead of UpdateWriteOpResult, when lean() is used\n### Prerequisites\n\n- [X] I have written a descriptive issue title\n- [X] I have searched existing issues to ensure the bug has not already been reported\n\n\n### Mongoose version\n\n7.1.0\n\n### Node.js version\n\n16.20\n\n### MongoDB server version\n\n5.x\n\n### Typescript version (if applicable)\n\n4.9.5\n\n### Description\n\nI tried upgrading to mongoose 7.x, but now I'm stuck with some typescript issues.\n![image](https://user-images.githubusercontent.com/5757263/236601137-fd8e2776-2b12-427b-bc02-e81ee801791d.png)\n\nsomehow updateOne returns a FlattenMap and not the ResultType defined in updateOne:\n![image](https://user-images.githubusercontent.com/5757263/236601177-2fcb360b-5dab-4cef-91ae-066cc6c3f34b.png)\n\nI would expect a UpdateWriteOpResult here.\n\n\n### Steps to Reproduce\n\nmy UserModel is defined as \n\n```\nconst userSchema = new Schema({...});\n\nexport interface IDBUser { .. };\n\nconst DBUserModel: Model = model('User', userSchema);\n\n\n```\n\njust run updateOne on it and check the return type.\n\n### Expected Behavior\n\ntypescript should correctly return the UpdateWriteOpResult"} {"task_id": "format-code-task-000026", "source_id": "format-code-task-000026", "domain": "code", "task_path": "tasks/format-code-task-000026", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:77d20cca70ee937251907e49fbfead1a4d3e80f34905562725c99d78f9d1bd4f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nTypescript: Wrong return type when using lean in different way\n### Prerequisites\n\n- [X] I have written a descriptive issue title\n- [X] I have searched existing issues to ensure the bug has not already been reported\n\n\n### Mongoose version\n\n6.X.X - 7.X.X\n\n### Node.js version\n\nat least 16\n\n### MongoDB server version\n\nat least 4\n\n### Typescript version (if applicable)\n\n4.7 - latest\n\n### Description\n\nThe return type of \n\n```ts\nconst a = await Model.findById('my-id', undefined, { lean: true })\n```\n\nis the same as\n\n```ts\nconst a = await Model.findById('my-id')\n```\n\nit's:\n```ts\n// ^? const a: (mongoose.Document & { name: string; } & { _id: mongoose.Types.ObjectId; }) | null\n```\n\n\n### Steps to Reproduce\n\nSee [ts playground](https://www.typescriptlang.org/play?ts=5.1.6#code/JYWwDg9gTgLgBCCA7A5hCBnApgGjgbwAKBlAYwAssQBDOAXzgDMoIQ4ByRVdbdgKD6lkGeBgpVaAXjhIsAdzhlKNABT4+cGdRBYAXARgBPMHsUwowVHihYAjgFdgNgCb7z9rHRx86ASgFCSCJwALIQzlgANnDSXGiYWAB0iBGRKuwAElGREOx4YsrU-oLC8FJw1HLUwPBhqYmMls4AQoYAks7p1Oz+1HwA9P1wAHoA-AGlcABGMRVVNaHhUQ1NrR1deXD2SBGNss54+HCRWNRIblAe9P5TA0NjE0HwpLOV1bVLkSs7a53s3b5Eiczip-KQ7iNRkA)\n\n### Expected Behavior\n\nThe return type should be the same as\n\n```ts\nconst a = await Model.findById('a').lean()\n```\n\nit's:\n\n```ts\n// ^? const a: (mongoose.FlattenMaps<{ name: string; }> & { _id: mongoose.Types.ObjectId; }) | null\n```"} {"task_id": "format-code-task-000027", "source_id": "format-code-task-000027", "domain": "code", "task_path": "tasks/format-code-task-000027", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b342c819994822f62255717bd8ed7917e91965ca9bc7fab36f339edd596bdb2f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nSupport string initialization for map in schema\n### Prerequisites\n\n- [X] I have written a descriptive issue title\n- [X] I have searched existing issues to ensure the feature has not already been requested\n\n\n### 🚀 Feature Proposal\n\nMy project uses String for setting the type in schema of each keys like this:\n```js\nusername: { type: 'string', required: true }\n```\n\nIn case of map, I just tried this style of code:\n```js\ninstance: {\n type: Schema.Types.Map,\n of: Schema.Types.Mixed,\n default: new Map(),\n}\n```\nThis is working code. But if I use 'Map' instead of `Schema.Types.Map`, compiler aborts to compile.\n```js\ninstance: {\n type: 'Map',\n of: 'Mixed',\n default: new Map(),\n}\n```\n\nSo, is it able to allow with setting the map type with String?\n\n### Motivation\n\n_No response_\n\n### Example\n\n_No response_"} {"task_id": "format-code-task-000028", "source_id": "format-code-task-000028", "domain": "code", "task_path": "tasks/format-code-task-000028", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:233c258f8339f4f14aa2ba6fc7b2ed89c16f92a03f87c35f254bfbc2eb35f18f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a `validateAllPaths` validation option for documents\n\nMongoose documents already let you narrow down what gets validated. For example, a schema\nconfigured with `validateModifiedOnly: true` only runs validators on paths that were actually\nmodified, and callers can pass `pathsToValidate` / `pathsToSkip` to further restrict the work.\n\nWe need the opposite escape hatch: a way to force validation of **every** path defined on the\nschema, regardless of what has (or hasn't) been modified.\n\nAdd a `validateAllPaths` boolean option that is accepted by both the asynchronous `validate()`\nmethod and the synchronous `validateSync()` method (via their options object). When\n`validateAllPaths` is `true`:\n\n- Validation runs against every path declared in the schema, including paths whose values were\n never modified. This takes precedence over a schema-level `validateModifiedOnly` setting, so an\n unmodified-but-invalid path is reported as an error even when the schema would normally skip it.\n- Each element of an array path is validated too, so element-level validators (e.g. `enum` on the\n items of a string array) run against every element. A failing element is reported under its\n indexed path (for example `tags.0`).\n- A document whose paths are all valid still passes: `validateSync()` returns `undefined` and\n `validate()` resolves.\n\nThe option is mutually exclusive with the existing path-narrowing options. Combining\n`validateAllPaths` with any of the following must raise a `TypeError` (and `validate()` must reject\nwith it) before any validation runs, using these exact messages:\n\n- with `pathsToSkip` → `Cannot set both \\`validateAllPaths\\` and \\`pathsToSkip\\``\n- with `pathsToValidate` → `Cannot set both \\`validateAllPaths\\` and \\`pathsToValidate\\``\n- with a `validateModifiedOnly` option → `Cannot set both \\`validateAllPaths\\` and \\`validateModifiedOnly\\``\n\nThese rules apply identically to `validate()` and `validateSync()`.\n"} {"task_id": "format-code-task-000029", "source_id": "format-code-task-000029", "domain": "code", "task_path": "tasks/format-code-task-000029", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:83c68bed4b1c6ec51847df71e135dd5635b377dfabfcc7f856f03e6203e1621e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Statics that override built-in model/document methods are getting wrapped with hooks unexpectedly\n\nI'm using a plugin similar to `mongoose-delete` that overrides some built-in methods (`save`, `remove`, etc.) by attaching them through `schema.statics`. My schema also defines normal hooks like `pre('save')` / `post('save')` for unrelated business logic on documents.\n\nRoughly:\n\n```js\nconst schema = new Schema({ name: String });\n\n// plugin-style: override built-in via statics\nschema.statics.save = function() { /* custom impl */ };\n\n// my own business hook for document save\nschema.pre('save', function() { /* ... */ });\nschema.post('save', function() { /* ... */ });\n\nconst Model = mongoose.model('Test', schema);\n```\n\nWhen I call the overridden static, the document-level `save` hooks fire on it, which is clearly wrong — the static is its own method and shouldn't be running document middleware. The same kind of weirdness shows up if I do this with names like `insertMany` or `bulkWrite`.\n\nMongoose already handles this correctly when the colliding name is a **query** or **aggregate** middleware (e.g. defining `schema.statics.findOne = ...` doesn't auto-wrap it with the `findOne` hook). I'd expect the same behavior for model-level (`bulkWrite`, `insertMany`, `createCollection`) and document-level (`save`, `validate`, `remove`, `updateOne`, `deleteOne`, `init`) middleware: if a user-defined static happens to share a name with one of those, don't auto-apply the corresponding middleware to it.\n\nThis currently makes plugins that override built-in methods through statics very awkward to combine with normal hooks."} {"task_id": "format-code-task-000030", "source_id": "format-code-task-000030", "domain": "code", "task_path": "tasks/format-code-task-000030", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e8a9ca146420b25fb170db312469d9b11d8eb9783ac4586ec2e33a33c0bf4311", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## CORS AllowedOrigins doesn't support wildcard subdomains (e.g. `http://*.contoso.com`)\n\nI'm using Azurite as a local emulator for Azure Storage during development. On our real storage account in Azure we have CORS rules configured for the Blob service that allow a wildcard subdomain pattern, something like:\n\n```xml\n\n \n http://*.contoso.com, http://www.fabrikam.com\n PUT,GET\n x-ms-meta-data*,x-ms-meta-target*,x-ms-meta-abc\n x-ms-meta-*\n 200\n \n\n```\n\nThis is documented as supported by Azure Storage — see [Enabling CORS for Azure Storage](https://learn.microsoft.com/en-us/rest/api/storageservices/cross-origin-resource-sharing--cors--support-for-the-azure-storage-services#enabling-cors-for-azure-storage), which explicitly mentions that `*` can be used to allow any subdomain of a given domain.\n\nWhen I push the same service properties (with the same CORS rules) into Azurite and then hit the blob endpoint from a page served at e.g. `http://app.contoso.com`, the browser's preflight request to Azurite fails and the request is rejected as a CORS violation. If I change AllowedOrigins to the exact origin `http://app.contoso.com` it works, and if I switch the endpoint back to the real Azure Storage account (same wildcard config), it also works.\n\nSo it looks like Azurite is treating the configured allowed origin as a literal string rather than honoring the `*` wildcard pattern that the Azure Storage service supports.\n\nIt would be great if Azurite's Blob CORS handling matched the real service here, so the same CORS configuration works against both Azurite and Azure Storage."} {"task_id": "format-code-task-000031", "source_id": "format-code-task-000031", "domain": "code", "task_path": "tasks/format-code-task-000031", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a4ea0e9a25cca6314783da9a232e7cec11d1d7d106aeedfa684ee15383ad804d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## waagent crashes / mis-behaves when publishing hostname on a VM whose network interface isn't ready yet\n\nI'm running WALinuxAgent on an Azure Linux VM. On some of my VMs, when the agent tries to publish the hostname (e.g. after a hostname change, or early in the boot cycle), it blows up or silently fails to actually restart the right interface.\n\nWhat I observe:\n\n- On a \"good\" VM, publishing the hostname works — the DHCP hostname is sent and the interface comes back up with the new name.\n- On a \"bad\" VM (same image, just different luck with timing on boot), the same code path errors out of `publish_hostname` instead of finishing. After that, the new hostname never makes it out to DHCP for that boot.\n- I can also reproduce something similar on a host where the primary interface recorded by waagent doesn't match anything in the current `ifconf` listing — the agent logs a warning about the primary interface not being found and then bails out of the whole flow with an exception instead of falling back to whatever non-loopback interface is actually present.\n\nThe `get_mac_addr()` side of the world seems to cope with the \"interface isn't ready yet\" situation fine — I never see it fail this way. It's the hostname publishing path (which ends up calling into the same lower-level \"give me an interface name\" helper) that doesn't tolerate the interface being momentarily unavailable.\n\nExpected: querying the interface name during `publish_hostname` should be just as tolerant as the MAC-address query is. If no usable interface is available right now, the agent should wait it out / skip gracefully, not raise an exception that aborts the whole publish step. And the underlying helper shouldn't be throwing in a way that takes down callers that just want a best-effort answer.\n\nRepro is annoyingly timing-dependent on a real VM, but you can see the shape of it by forcing the situation where the recorded primary interface isn't in the current ifconf list — the agent ends the call with an exception instead of returning whatever non-loopback interface it found (or nothing at all)."} {"task_id": "format-code-task-000033", "source_id": "format-code-task-000033", "domain": "code", "task_path": "tasks/format-code-task-000033", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e35fa3a5c59289dc1bfffb59d95ece7d93d1f2ac169851c1b24b297beb4ef2da", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a configurable Azure cloud abstraction\n\nRight now azd assumes it is always talking to the Azure public cloud — portal links,\nstorage and container-registry hostnames, and the SDK client configuration are all\nhard-wired to the public cloud's values. We want to support the sovereign clouds\n(Azure China and Azure US Government) too, so the rest of the codebase can ask for the\ncorrect endpoints instead of hard-coding `portal.azure.com`, `core.windows.net`, etc.\n\nIntroduce a small self-contained package, importable as\n`github.com/azure/azure-dev/cli/azd/pkg/cloud`, that models a target cloud and lets\ncallers resolve one from user/project configuration.\n\n## What a \"cloud\" carries\n\nA cloud is represented by a `Cloud` value exposing, as readable fields:\n\n- `Configuration` — the underlying `azcore` cloud configuration (the\n `cloud.Configuration` type from\n `github.com/Azure/azure-sdk-for-go/sdk/azcore/cloud`), so SDK clients can be pointed at\n the right authority/management endpoints,\n- `PortalUrlBase` — the base URL of the cloud's web portal,\n- `StorageEndpointSuffix` — the DNS suffix used for the cloud's storage endpoints,\n- `ContainerRegistryEndpointSuffix` — the DNS suffix used for the cloud's container\n registry endpoints.\n\nProvide a no-argument constructor for each of the three well-known clouds —\n`AzurePublic()`, `AzureChina()`, and `AzureGovernment()` — each returning a `*Cloud`\npopulated with the correct values:\n\n| constructor | azcore configuration | portal base URL | storage suffix | container registry suffix |\n|----------------------|-------------------------|----------------------------|--------------------------|---------------------------|\n| `AzurePublic()` | `cloud.AzurePublic` | `https://portal.azure.com` | `core.windows.net` | `azurecr.io` |\n| `AzureChina()` | `cloud.AzureChina` | `https://portal.azure.cn` | `core.chinacloudapi.cn` | `azurecr.cn` |\n| `AzureGovernment()` | `cloud.AzureGovernment` | `https://portal.azure.us` | `core.usgovcloudapi.net` | `azurecr.us` |\n\n## Resolving a cloud from configuration\n\nA cloud is selected by a stable, configuration-friendly name. Expose these as exported\nstring constants:\n\n- `AzurePublicName` = `AzureCloud`\n- `AzureChinaCloudName` = `AzureChinaCloud`\n- `AzureUSGovernmentName` = `AzureUSGovernment`\n\nDefine a `Config` struct that holds a single name field, serializable to/from both JSON\nand YAML under the key `name`.\n\n`NewCloud(config *Config) (*Cloud, error)` builds a cloud from such a config:\n\n- a recognized name resolves to the matching cloud,\n- an empty name defaults to the Azure public cloud,\n- any other value is an error whose message includes the offending name.\n\n`ParseCloudConfig(partialConfig any) (*Config, error)` takes a loosely-typed\nconfiguration value — the kind of thing you get back when reading a node out of azd's\nJSON/YAML config (e.g. a `map[string]any{\"name\": \"AzureChinaCloud\"}`) — and turns it\ninto a `*Config`.\n\nFinally, expose the configuration key under which the cloud node lives as an exported\nconstant `ConfigPath` whose value is `cloud`.\n"} {"task_id": "format-code-task-000034", "source_id": "format-code-task-000034", "domain": "code", "task_path": "tasks/format-code-task-000034", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:97cb9a9081323508cf4eaa282765cfe9253530e431d20cd5e74758bc96193edb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI’m seeing our vsrpc endpoint fall over when one of my RPC methods hits a panic — the client just gets a dropped/hung connection instead of a normal JSON-RPC failure for that call. It’s also awkward to verify because the debug service doesn’t seem to have a simple RPC method I can call that deliberately panics with a message.\n\n# Expected outcomes\n\n- Panic handling for vsrpc calls\n - If a vsrpc RPC method panics while handling a request, that panic should be reported to the caller as a JSON-RPC error for that request instead of causing the endpoint, process, or connection to fail.\n - After one RPC call panics and returns an error, the same server/connection should remain usable for subsequent RPC calls.\n - This behavior should apply consistently to RPC methods adapted through the existing vsrpc handler helper APIs, whether the method returns only an error or returns a value plus an error.\n\n- Error details\n - The JSON-RPC error for a recovered panic should use `jsonrpc2.InternalError`.\n - The error message should start with `panic:`, include the panic value, and include stack trace information.\n\n- Debug service verification\n - The debug service should expose a callable vsrpc method named `TestPanicAsync`.\n - Calling `TestPanicAsync` with a message should deliberately trigger the panic path and return the same JSON-RPC internal error behavior, including the supplied message in the error details.\n\n# Implementation notes\n\nThe specific recovery mechanism, helper structure, and placement of validation or error conversion are left to the implementer. Preserve the existing request unmarshalling, cancellation, and normal success/error behavior for non-panicking calls."} {"task_id": "format-code-task-000035", "source_id": "format-code-task-000035", "domain": "code", "task_path": "tasks/format-code-task-000035", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f0d2850673fd9e046920cdde1d05976e8e941d2d4aa08931551ba0a76b7b8a40", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nBBC SVG logo doesnt show when printing\n**Describe the bug**\nWhen you want to print out an article, onto paper, like its 1976, the icon doesn't show\n\n**To Reproduce**\nSteps to reproduce the behaviour:\n\n1. Go to an article with the Brand in\n2. Try to print\n3. See the svg isn't visible\n\n**Expected behaviour**\nThe SVG is visible\n\n**Screenshots**\n![image](https://user-images.githubusercontent.com/11341355/50640807-e2071180-0f5d-11e9-8162-80e75551c99b.png)\n\n**Desktop (please complete the following information):**\n\n- OS: Mac\n- Browser: Chrome\n- Version: 71.0.3578.98\n\n- [x] Initially labelled with [\"bug\"](https://github.com/BBC-News/psammead/labels/bug)"} {"task_id": "format-code-task-000036", "source_id": "format-code-task-000036", "domain": "code", "task_path": "tasks/format-code-task-000036", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b9ed31bc889585e5ab98623fda9cfce7988a3309a16e09d5e79be6d6e955e110", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nUse agreed LeadingStory layouts\n**Is your feature request related to a problem? Please describe.**\nFollowing discussion with UX, we have updated the layouts for `StoryPromo`s of type `'leading'`.\n\n**Describe the solution you'd like**\nOn all breakpoints, the DOM ordering will be `Info` component (containing Headline, Summary, Timestamp) followed by the `Image` component. Internally this means the `TextGridItem` followed by `ImageGridItem`.\n\n| Breakpoint | `TextGridItem` columns | `ImageGridItem` columns | Total columns/line |\n|------------|-------------------------|---------------------------|-------------------|\n| > 1007px | 2 | 4 | 6 |\n| 600px - 1007px | 3 | 3 | 6 |\n| < 600px | 6 | 6 | 6 |\n\nThis will also need the relevant fallbacks added.\n\n**Describe alternatives you've considered**\nA clear and concise description of any alternative solutions or features you've considered.\n\n**Testing notes**\n[Tester to complete]\n\nDev insight: Will there be any potential regression? etc\n\n- [x] This feature is expected to need manual testing.\n\n**Additional context**\nTablet: \n![image](https://user-images.githubusercontent.com/43134742/73276695-ff4df400-41e0-11ea-876e-31091df336d5.png)\n\nMobile (note the padding under the timestamp is not correct, this was a quick prototype):\n![image](https://user-images.githubusercontent.com/43134742/73276441-99fa0300-41e0-11ea-921a-5ed36d04bcd2.png)"} {"task_id": "format-code-task-000037", "source_id": "format-code-task-000037", "domain": "code", "task_path": "tasks/format-code-task-000037", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3920c30b14a52e951a66b08c4fd83ea4591a427f0376ad54cd7813f86ae3eae2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nUpdate fallback text colour in Social Embed\n### Is your feature request related to a problem? Please describe.\nFollowing UX review, we should update the text colour in the Social Embed fallback/Notice.\n\n### Describe the solution you'd like\nAll standard text within the social embed notice component (not the link) should be `#3F3F42` (SHADOW).\nNote: The current link colour is correct.\n\n### Describe alternatives you've considered\nn/a\n\n### Testing notes\n[Tester to complete]\n\nDev insight: Will there be any potential regression? etc\n\n- ~~[ ] This feature is expected to need manual testing.~~\n- [x] This feature is expected to have a UX review.\n\n### Additional context\nhttps://bbc.github.io/psammead/?path=/story/components-socialembed-canonical--unsupported-provider - See storybook for what this currently looks like."} {"task_id": "format-code-task-000039", "source_id": "format-code-task-000039", "domain": "code", "task_path": "tasks/format-code-task-000039", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:36ca2c575bd37e62dc399f5da74ebc9aceb950e4f418103b3cfa3e96e993557b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nWhen I open an AOPP link while my BitBox is already connected and unlocked, the wallet jumps straight into picking an account, which feels a bit too automatic. I’d like it to pause first and show me who is asking, with a chance to continue or cancel before I get taken further into the flow.\n\n## Expected outcomes\n\n- AOPP requests opened while a keystore is already connected must first enter a visible approval step instead of immediately continuing to account selection.\n- The approval step must expose the request host and provide both Cancel and Continue actions.\n- The AOPP state returned to clients must include `state: 'user-approval'` while the request is waiting for this explicit decision.\n- Calling `POST /aopp/approve` must approve an AOPP request that is currently waiting for user approval and then let the normal AOPP flow continue to the next appropriate state.\n- Calling `POST /aopp/approve` when the AOPP flow is not waiting for user approval must not advance the flow.\n- The web AOPP API should expose `approve(): Promise` for the Continue action, while the existing cancel behavior remains available for Cancel.\n\n## Implementation notes\n\n- The exact internal state-machine structure, handler wiring, and UI component organization are up to the implementer.\n- Preserve existing AOPP behavior for requests that still need a keystore to be connected or unlocked.\n- Validate the behavior through externally observable state transitions, API responses, and UI actions rather than relying on internal naming."} {"task_id": "format-code-task-000041", "source_id": "format-code-task-000041", "domain": "code", "task_path": "tasks/format-code-task-000041", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4fb3e32d0e57e2c2f5b8819202ed4c5bf622e8b40adb5c787aead7f9885efa61", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nEODataAccessGateway constructor fails on AWS Lambda\n**Describe the bug**\n\n`makedirs` [invocation](https://github.com/CS-SI/eodag/blob/develop/eodag/api/core.py#L98) within the EODataAccessGateway constructor causes `OSError: [Errno 30] Read-only file system` when run on AWS Lambda\n\n**Code To Reproduce**\nCLI commands or Python code snippet to reproduce the bug. Please use maximum verbosity using:\n\n```py\nfrom aws_lambda_powertools.logging import Logger\nfrom aws_lambda_powertools.tracing import Tracer\nfrom eodag import EODataAccessGateway\n\nlogger = Logger()\ntracer = Tracer()\n\n\n@tracer.capture_lambda_handler\n@logger.inject_lambda_context\ndef lambda_handler(event, context):\n EODataAccessGateway()\n```\n\n**Output**\nCompete output obtained with maximal verbosity.\n\n```\n[ERROR] OSError: [Errno 30] Read-only file system: '/home/sbx_user1051'\nTraceback (most recent call last):\n File \"/opt/python/aws_lambda_powertools/tracing/tracer.py\", line 305, in decorate\n response = lambda_handler(event, context, **kwargs)\n File \"/opt/python/aws_lambda_powertools/logging/logger.py\", line 438, in decorate\n return lambda_handler(event, context, *args, **kwargs)\n File \"/var/task/, in lambda_handler\n EODataAccessGateway()\n File \"/opt/python/eodag/api/core.py\", line 91, in __init__\n makedirs(self.conf_dir)\n File \"/opt/python/eodag/utils/__init__.py\", line 496, in makedirs\n os.makedirs(dirpath)\n File \"/var/lang/lib/python3.9/os.py\", line 215, in makedirs\n makedirs(head, exist_ok=exist_ok)\n File \"/var/lang/lib/python3.9/os.py\", line 215, in makedirs\n makedirs(head, exist_ok=exist_ok)\n File \"/var/lang/lib/python3.9/os.py\", line 225, in makedirs\n mkdir(name, mode)\n```\n\n**Environment:**\n\n - Python version: 3.9\n - EODAG version: 2.5.2\n - Runtime: AWS Lambda\n\n**Additional context**\n\nFile system on AWS Lambda is read-only. It has \"ephemeral storage\" that could be accessed in `/tmp` ([1](https://aws.amazon.com/blogs/aws/aws-lambda-now-supports-up-to-10-gb-ephemeral-storage/), [2](https://aws.amazon.com/blogs/aws/aws-lambda-now-supports-up-to-10-gb-ephemeral-storage/)), however this class does not accept any configuration to where to point to `makedirs`. \n\nThe workaround that heavily relies on the [current implementation](https://github.com/CS-SI/eodag/blob/develop/eodag/api/core.py#L97) is to change `$HOME` env var before invoking the constructor. But it is barely acceptable and extremely inconvenient, since this variable is used by many tools locally and in automated tools along the way."} {"task_id": "format-code-task-000042", "source_id": "format-code-task-000042", "domain": "code", "task_path": "tasks/format-code-task-000042", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:df6d414cfb63da9b2b2766392a7de94f58a57b4f6d1656766599ceace007f75c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Uploading a DataFrame with non-ASCII text fails\n\nI have a pandas DataFrame with some string columns containing accented characters (Spanish place names like `\"Málaga\"`, `\"A Coruña\"`, plus a few rows with Asian characters). I'm trying to push it to CARTO with something along the lines of:\n\n```python\nfrom cartoframes.data import Dataset\n\nds = Dataset(df)\nds.upload(table_name='my_places', credentials=creds, if_exists='replace')\n```\n\nThe DataFrame itself looks fine in pandas (the strings display correctly, no weird mojibake), but the upload blows up partway through instead of finishing. If I strip the rows that contain accented / non-ASCII characters and re-run with the same code path, everything uploads without complaint, so it seems specific to non-ASCII content in the data.\n\nI'd expect `Dataset.upload` to handle arbitrary unicode text in string columns — having to pre-sanitize / drop rows with accents before every upload is pretty painful when the whole point of the dataset is Spanish-speaking locations."} {"task_id": "format-code-task-000043", "source_id": "format-code-task-000043", "domain": "code", "task_path": "tasks/format-code-task-000043", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:22dbc5e3bb2b375a9032d568dcf57e02f1fcc0391f31cdc9ae5106cb4ef46404", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Fix DataFrame uploads with reserved/geometry columns\n\nUploading a `pandas`/`geopandas` DataFrame to CARTO via `Dataset(df).upload(...)` is\ngenerating inconsistent and sometimes broken SQL. For example, a DataFrame whose only\ncolumns are `cartodb_id` and `the_geom` ends up producing a `CREATE TABLE` statement with\na dangling leading comma like `CREATE TABLE t (, the_geom geometry(...))`, which the\ndatabase rejects. The root cause is that the column list used to create the table, the\ncolumn list used in the `COPY` statement, and the values written to each row are all\ncomputed independently and can disagree about how many columns there are and what they are\ncalled.\n\nMake the upload path derive a single, consistent description of the destination columns\nfrom the DataFrame and use it everywhere. The observable behavior of an upload (the\n`CREATE TABLE` statement and the `COPY ... FROM stdin` statement together with the rows of\nCSV data streamed to the backend) must satisfy the following contract.\n\n## Column selection and ordering\n\n- Every column of the DataFrame is uploaded, in the DataFrame's original column order. In\n particular `cartodb_id` is treated as a normal column and is kept.\n- A column named `the_geom_webmercator` (matched case-insensitively) is always dropped and\n never appears in the table, the `COPY` column list, or the data.\n\n## Geometry handling\n\n- Exactly one column is treated as the geometry. If the DataFrame is a GeoDataFrame with an\n active geometry column, that column is used; otherwise the first column whose lower-cased\n name is one of `the_geom`, `geom`, `geometry` is used, considered in that order of\n priority (so if several are present, `the_geom` wins, then `geom`, then `geometry`).\n- The chosen geometry column is named `the_geom` in the destination, declared with type\n `geometry(, 4326)` where `` is detected from the data (e.g.\n `Point`), and its values are emitted as `SRID=4326;`. A null/missing geometry value\n produces an empty field.\n- Any other column that merely happens to be named like a geometry but was not selected as\n *the* geometry column is uploaded as an ordinary column: it keeps its (normalized) name\n and its value is written verbatim, not re-encoded.\n\n## Ordinary columns\n\n- Non-geometry column names are normalized to SQL-safe names (lower-cased, spaces and other\n unsupported characters replaced, etc.) in both the `CREATE TABLE` and the `COPY` column\n list. Values are written in their string form; a null value produces an empty field.\n- Column SQL types in `CREATE TABLE` are derived from the DataFrame dtypes:\n `float64`/`float32` → `numeric`, `int64` → `bigint`, `int32` → `integer`,\n `object` → `text`, `bool` → `boolean`, datetime dtypes → `timestamp`; anything else\n defaults to `text`.\n\n## `with_lnglat`\n\n- When `with_lnglat=(lng_col, lat_col)` is passed, an extra geometry column named\n `the_geom` of type `geometry(Point, 4326)` is appended after the regular columns, built\n per row as `SRID=4326;POINT ( )`. The `lng_col` and `lat_col` columns are still\n uploaded as ordinary columns.\n- When `with_lnglat` is used, any geometry column that would otherwise have been detected in\n the DataFrame is dropped, so the synthesized point is the only geometry uploaded.\n\n## Statement format\n\nThe `COPY` statement keeps its existing shape:\n\n```\nCOPY (,,...) FROM stdin WITH (FORMAT csv, DELIMITER '|');\n```\n\nwith the destination column names joined by commas (no spaces). Each streamed row is the\ncolumn values joined by `|` and terminated by a newline, encoded to bytes.\n"} {"task_id": "format-code-task-000044", "source_id": "format-code-task-000044", "domain": "code", "task_path": "tasks/format-code-task-000044", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:119cc3e86df0e2ee6fe6fc7862ed26dda72bbbae0585f9a9088a41c877a9337f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: heading-pitch-roll transforms using aircraft (NED) convention\n\nI'm building a flight / UAV visualization on top of Cesium. Most of my pose data comes from autopilot logs and flight simulators, where heading / pitch / roll are defined in the standard aviation way — i.e. relative to a **North-East-Down** local frame at the aircraft's position (heading rotates about the local Down axis, pitch about local East, roll about local North).\n\nI tried to drive a model's `modelMatrix` with the existing helpers:\n\n```js\nvar origin = Cesium.Cartesian3.fromDegrees(lon, lat, alt);\nvar heading = aircraft.heading; // from sim, defined in NED\nvar pitch = aircraft.pitch;\nvar roll = aircraft.roll;\n\nvar m = Cesium.Transforms.headingPitchRollToFixedFrame(origin, heading, pitch, roll);\n// also tried Cesium.Transforms.headingPitchRollQuaternion(...)\n```\n\nThe orientation that comes out doesn't match what the aircraft is actually doing. After staring at it for a while I realized the existing `headingPitchRollToFixedFrame` / `headingPitchRollQuaternion` interpret the angles relative to a local **East-North-Up** frame, which is a different convention from what aviation / autopilot data uses. So the same numeric (heading, pitch, roll) triple means different physical rotations in the two conventions, and feeding aircraft-convention values into the ENU-based helper gives the wrong attitude.\n\nRight now, to get correct attitude I have to roll my own helper on top of `Transforms.northEastDownToFixedFrame` and manually build the rotation from heading/pitch/roll, which feels like something the library should expose directly — especially since the ENU-based versions are already there.\n\n### Ask\n\nCould `Transforms` provide companions to the existing heading/pitch/roll helpers that interpret the angles using the aviation (NED) convention? Concretely, I'd like to be able to write something like:\n\n```js\nvar origin = Cesium.Cartesian3.fromDegrees(lon, lat, alt);\n\n// 4x4 world transform from aircraft-convention HPR\nvar transform = /* aircraft-HPR -> fixed frame */(origin, heading, pitch, roll);\n\n// and a quaternion form, mirroring the existing ENU pair\nvar q = /* aircraft-HPR -> quaternion */(origin, heading, pitch, roll);\n```\n\nso that anyone consuming flight / drone telemetry can convert directly to a world-space transform without having to re-derive the NED axis swap themselves. Defaults should match the rest of `Transforms` (WGS84 ellipsoid, optional `result` out-param).\n\nThe names I'd expect for these new helpers are something like `aircraftHeadingPitchRollToFixedFrame` and `aircraftHeadingPitchRollQuaternion`, mirroring the existing `headingPitchRoll*` pair."} {"task_id": "format-code-task-000045", "source_id": "format-code-task-000045", "domain": "code", "task_path": "tasks/format-code-task-000045", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fcb5d0910fc4725ca505d3cfb8e05b990140a83fe1c17f0152db8c742982feab", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAdd support for content_type: \"location\" and \"image_url\" on Quick Replies\nWe need to refactor the `_formatQuickReplies` method to support an object with an open format. New types of quick replies format added since we implemented this methods are:\n\nLocation:\n\n```\n {\n \"content_type\":\"location\",\n }\n```\n\nText with Image:\n\n```\n {\n \"content_type\":\"text\",\n \"title\":\"Green\",\n \"payload\":\"DEVELOPER_DEFINED_PAYLOAD_FOR_PICKING_GREEN\",\n \"image_url\":\"http://petersfantastichats.com/img/green.png\"\n }\n```\n\nBut more formats will likely be added in the future so we should support whatever object the user sends (or auto-format like we do now if it's a string).\n\nFacebook Docs: https://developers.facebook.com/docs/messenger-platform/send-api-reference/quick-replies"} {"task_id": "format-code-task-000046", "source_id": "format-code-task-000046", "domain": "code", "task_path": "tasks/format-code-task-000046", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ceeb9bea228a86c93e3a983e9b9e1ec7ff8d761feccc194c57d110cc75b4e519", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Separator DSL methods produce incorrect lookahead / possible tokens\n\nWhen a rule uses the separator variants of the repetition DSL methods (`MANY_SEP` / `AT_LEAST_ONE_SEP`), the computation of the possible next tokens / lookahead paths gives wrong results.\n\nFor example, given a rule along the lines of:\n\n```ts\nthis.RULE(\"list\", () => {\n this.MANY_SEP({\n SEP: Comma,\n DEF: () => { this.CONSUME(Identifier) }\n })\n})\n```\n\nAfter consuming an `Identifier`, the set of possible following tokens should include the separator (`Comma`) as one of the options (since the repetition may continue), in addition to whatever can follow the rule. In practice the separator is missing / not produced correctly by the path computation, which then breaks downstream features that rely on it (syntactic content assist, lookahead, etc.).\n\nThe non-separator variants (`MANY` / `AT_LEAST_ONE`) behave correctly for the same shape of grammar — the problem is specific to the `*_SEP` DSL methods.\n\nCould the path / lookahead computation be fixed so that separator repetition methods produce the same kind of correct results as their non-separator counterparts?"} {"task_id": "format-code-task-000047", "source_id": "format-code-task-000047", "domain": "code", "task_path": "tasks/format-code-task-000047", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1b67b35306e761ceb70521c02b9cb92a908cccf12262767ed336012e4e76ff93", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nWindows runner: set Windows component for Docker Desktop to optional\nHey @kichik,\n\nI`m able to build the image for the windows runner in eu-central-1 now without issues.\nBut what I recognized is that Docker desktop will be installed out of the box.\nhttps://github.com/CloudSnorkel/cdk-github-runners/blob/c64683b62802de1817e173029a6d9a6695fb5fe9/src/providers/image-builders/ami.ts#L254\n\nDocker desktop requires dependent on company size a subscription when it`s installed on windows or mac.\n\nCan we make the installation of docker desktop optional for the windows runner ?"} {"task_id": "format-code-task-000048", "source_id": "format-code-task-000048", "domain": "code", "task_path": "tasks/format-code-task-000048", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fe519e917b08748f24a113fb9890783ce1d1e103ccb2ebd062d382d1e5f626b0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: iterate over a numeric range from a template\n\nI'd like to be able to loop over a numeric range directly inside a jet template. Typical use cases are rendering pagination links, repeating a UI element N times, or emitting a numbered list.\n\nRight now I don't see a way to do this from the template side alone. What I end up doing is building a throwaway slice in Go just so the template has something to `range` over:\n\n```go\n// in the handler\nnums := make([]int, 10)\nfor i := range nums {\n nums[i] = i\n}\nvars.Set(\"nums\", nums)\n```\n\n```\n{{ range i := nums }}\n {{ i }}\n{{ end }}\n```\n\nThis feels backwards — the template is the side that knows it wants to render \"page 1 through page N\", but it can't express that without Go-side glue. Every place that needs a numeric loop ends up repeating the same boilerplate, and it leaks template concerns into handler code.\n\nOther template engines provide something for this out of the box (Python's `range`, Twig's `..` operator, etc.). Could jet have an equivalent builtin so a template can produce a numeric sequence on its own, without the caller having to prepare a slice for it? I'd expect something like `ints(from, to)` usable directly in a `range`."} {"task_id": "format-code-task-000049", "source_id": "format-code-task-000049", "domain": "code", "task_path": "tasks/format-code-task-000049", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:61f60677512e77b14dbd2882794c2534c481ea5ed7a45fc7921aee940804de83", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI'm trying to spin up containers on my Flocker cluster via the HTTP API but I can't find a way to do it — the configuration endpoints let me create datasets but there's nothing for containers as far as I can tell. Can we get a POST under `/v1/configuration/containers` so I can declare a container (host + name + image) the same way I declare datasets? Also it'd be nice if it actually rejected duplicate names instead of silently letting two containers share one.\n\nOne small thing while you're in there: the duplicate-dataset_id 409 returns its error under a `message` key, but every other error I've hit uses `description` — kinda annoying to special-case in my client, would be great to make that consistent.\n\n# Expected outcomes\n\n- Container configuration creation:\n - A `POST` to `/v1/configuration/containers` with a JSON body containing `host`, `name`, and `image` creates a container entry in the cluster configuration.\n - A successful container creation returns HTTP `201 Created`.\n - The successful response body includes the same `host`, `name`, and `image` values supplied by the client.\n - The created container is persisted in the configuration for the specified host.\n\n- Duplicate container names:\n - Creating a container with a `name` that is already used by any configured container is rejected.\n - Duplicate container-name requests return HTTP `409 Conflict`.\n - Duplicate container-name error responses use a `description` field that explains the duplicate-name conflict.\n\n- Dataset conflict error format:\n - Duplicate `dataset_id` configuration requests continue to return HTTP `409 Conflict`.\n - The duplicate-`dataset_id` conflict response uses a `description` field rather than a `message` field.\n\n# Implementation notes\n\n- Match the existing HTTP API conventions in this repository for routing, JSON request/response validation, persistence, and error formatting.\n- The internal data structures, helper functions, and exact placement of validation or collision checks are implementation details; preserve the externally observable API behavior described above."} {"task_id": "format-code-task-000050", "source_id": "format-code-task-000050", "domain": "code", "task_path": "tasks/format-code-task-000050", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5c76dad6079c8df3a423a02bbde6d2127bec286baff71f3de5411b1ebade3ffc", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a scoped configuration system for cogs\n\nRed's cogs currently have no consistent way to declare the settings they use or\nto read and write those settings at different scopes (the bot as a whole, a\nparticular guild, a channel, a role, a user, or a specific member of a guild).\nWe want a small configuration subsystem that gives every cog a single object to\nwork with.\n\nAdd a `Config` class, importable as `from core.config import Config`, with the\nbehaviour described below.\n\n## Obtaining a config object\n\n`Config.get_conf(cog_name, unique_identifier=0, force_registration=False)`\nreturns a fresh config object for a cog. `cog_name` is a string; the other two\narguments are keyword options. `force_registration` controls the strict mode\ndescribed further down and defaults to `False`.\n\n## Registering defaults\n\nA cog declares the keys it uses, per scope, by registering defaults:\n\n- `register_global(**defaults)`\n- `register_guild(**defaults)`\n- `register_channel(**defaults)`\n- `register_role(**defaults)`\n- `register_user(**defaults)`\n- `register_member(**defaults)`\n\nEach call records the given keys together with their default values for that\nscope. Calling a register method more than once merges the new keys with the\nones already registered for that scope (later calls override earlier defaults\nfor the same key). Defaults for one scope are independent from every other\nscope.\n\n## Reading and writing values\n\nThe config object itself represents the **global** scope. To work with another\nscope you ask for it by id:\n\n- `config.guild(guild_id)`\n- `config.channel(channel_id)`\n- `config.role(role_id)`\n- `config.user(user_id)`\n- `config.member(guild_id, member_id)`\n\nAn id may be given either as an integer or as an object that exposes an `id`\nattribute (the object's `id` is used). Each of these returns a scoped view of\nthe configuration.\n\nReading a value is done by calling the key as a method on the relevant scope.\nOn the global scope `config.foo()` returns the value of `foo`; on another scope\n`config.guild(123).foo()` returns the value of `foo` for guild `123`. If no\nvalue has been stored for that key in that scope, the registered default for the\nscope is returned. If the key was never registered (and strict mode is off), the\nreader returns `None`.\n\nWriting a value is asynchronous: `await scope.set(key, value)` stores `value`\nfor `key` in that scope, e.g. `await config.set(\"foo\", 1)` for the global scope\nor `await config.guild(123).set(\"foo\", 1)` for a guild. After a value has been\nset, reading the same key in the same scope returns the stored value.\n\n`await scope.clear()` removes every stored value for that exact scope, so\nsubsequent reads fall back to the registered defaults again.\n\n## Scope isolation\n\nScopes and individual ids are fully isolated. Storing a value for one guild must\nnot change what another guild, the global scope, or any other scope reports.\nMembers are identified by the pair `(guild_id, member_id)`, so the same member\nid under two different guilds is two distinct scopes.\n\n## Strict mode\n\nWhen a config object is created with `force_registration=True`, reading or\nwriting a key that has not been registered for the scope being used raises\n`AttributeError`. When `force_registration` is `False` (the default), reading an\nunregistered key returns `None` and writing an unregistered key is allowed.\n"} {"task_id": "format-code-task-000051", "source_id": "format-code-task-000051", "domain": "code", "task_path": "tasks/format-code-task-000051", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f190d89c473987498ee878e62c22a35f63f1a081bc5045e031430d2cdb13d0dc", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nWhen I write a `file_owner` rule, I'm stuck specifying the expected owner as a literal numeric UID via `fileuid`. That's brittle for accounts whose UID isn't guaranteed to be the same everywhere — I really just want to say the owner should be `root` (or some other username) and have the check figure out the actual UID at scan time. Could the template accept a username there too, not just a hard-coded number?\n\n## Expected outcomes\n\n- The `file_owner` template supports a `uid_or_name` variable for the expected owner.\n- A `uid_or_name` value that is a numeric UID continues to generate owner checks equivalent to the previous numeric-UID behavior.\n- A `uid_or_name` value that is a username generates a check that compares against that account’s UID as resolved on the scanned system, so the same rule can work across systems where the account has different numeric UIDs.\n- Bundled OpenShift rules that use the `file_owner` template express their expected owner through `uid_or_name` instead of the old numeric-only `fileuid` variable, without changing which owner they require.\n\n## Implementation notes\n\nThe exact implementation strategy, data flow, and validation location are up to the implementer. Preserve the existing `file_owner` template behavior for path matching, recursive checks, and regular-expression-based file selection while extending only how the expected owner can be specified."} {"task_id": "format-code-task-000052", "source_id": "format-code-task-000052", "domain": "code", "task_path": "tasks/format-code-task-000052", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fdaea6844a8b24ec7fc21bdf61ceb8d299120dfcaab5e7baec72098620ffedfd", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Run custom database-initialization SQL after a cluster comes up\n\nUsers want a way to have the operator run their own SQL against a freshly\ninitialized PostgresCluster — e.g. to `CREATE EXTENSION`, seed reference tables,\nor create application objects — without manually `exec`-ing into a pod. We want\nto support pointing the cluster at a ConfigMap that holds a SQL file and have\nthe operator execute it once the cluster is running.\n\n## API\n\nAdd an optional field to the `PostgresCluster` **spec** called `databaseInitSQL`.\nIt is an object that references a ConfigMap living in the same namespace as the\ncluster, with two required string sub-fields:\n\n- `name` — the name of the ConfigMap.\n- `key` — the data key inside that ConfigMap whose value is the SQL to run.\n\nAdd an optional **status** field, also called `databaseInitSQL` (a string). Its\npresence records that the initialization SQL has been applied successfully.\n\nBoth new types must support the project's deep-copy conventions (copying a\ncluster must produce an independent copy of these values).\n\n## Behavior\n\nHook this into the reconcile loop as an additional step. Expose that step as a\nmethod on the reconciler with the signature\n\n```go\nfunc (r *Reconciler) reconcileDatabaseInitSQL(ctx context.Context,\n cluster *v1beta1.PostgresCluster, instances *observedInstances) error\n```\n\nso it can be driven directly. The `instances` argument is the set of observed\ninstances for the cluster. The step must behave as follows:\n\n- **Spec absent.** If `spec.databaseInitSQL` is not set, the status field must be\n cleared (set back to nil) and the step returns without doing anything else. In\n particular it must never try to execute SQL.\n\n- **Already applied.** If `spec.databaseInitSQL` is set but the status field is\n already set, the step is a no-op: it must not execute SQL and must leave the\n status unchanged.\n\n- **Needs applying.** If `spec.databaseInitSQL` is set and the status is not yet\n set, the step attempts to apply the SQL:\n - Fetch the named ConfigMap from the cluster's namespace. If it cannot be\n fetched, return that error (e.g. a not-found error from the API surfaces\n unchanged) without executing SQL.\n - If the ConfigMap exists but does not contain the requested `key`, return an\n error whose message identifies the missing key, without executing SQL.\n - Find the instance that is running, non-terminating, and writable (the\n primary), and execute the SQL string against its database container, passing\n the SQL on standard input. If no such instance is currently available, the\n step returns without error and without setting the status (so it is retried\n on a later reconcile).\n - If execution succeeds, set the status field so the SQL will not run again.\n If execution fails, return the error and leave the status unset so it is\n retried.\n\nBecause the status can be lost or the field re-added, the SQL may run more than\nonce; that is acceptable and is the user's responsibility to make idempotent.\n"} {"task_id": "format-code-task-000053", "source_id": "format-code-task-000053", "domain": "code", "task_path": "tasks/format-code-task-000053", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:07d3d16e52e278a171f897cf33c907494c3389a4df4f6675896a1661003c823e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Operator-managed Pods rejected under the Restricted Pod Security Standard\n\nWe're rolling postgres-operator out into a cluster where namespaces enforce the\nKubernetes [Restricted Pod Security Standard][pss]. When the operator brings up\nits Postgres pods in such a namespace, the API server rejects them (or, with\n`warn`/`audit` configured, surfaces a warning for every container) because the\ncontainers' SecurityContext does not satisfy the Restricted profile.\n\nLooking at `RestrictedSecurityContext()` in `internal/initialize/security.go`,\nmost of what Restricted requires is already there — `runAsNonRoot`, dropped\ncapabilities, no privilege escalation, read-only root filesystem, etc. — but\nthe resulting pods still trip the Restricted check, so the defaults aren't\nquite \"Restricted-compliant\" out of the box.\n\nFor an operator that's clearly trying to ship a hardened SecurityContext\n(`RestrictedSecurityContext` is even the function name), it would be great\nif pods produced by it could be deployed into a Restricted-enforcing namespace\nwithout the user having to layer extra mutating webhooks or post-process the\ngenerated specs.\n\n[pss]: https://kubernetes.io/docs/concepts/security/pod-security-standards/#restricted"} {"task_id": "format-code-task-000054", "source_id": "format-code-task-000054", "domain": "code", "task_path": "tasks/format-code-task-000054", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a5ce2ed54740af4bafad19d1ccf3d0cb768e5dbd40063e012eb4112bd285100a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Unified node CRUD on the folder-tree service\n\nThe folder-tree service holds the explorer's file/folder hierarchy in its state under\n`folderTree.data`. That value is a list of root-folder nodes; every node may carry a\n`children` list of nested nodes, and each node has a numeric `id`, a `name`, and a\n`fileType` of `File`, `Folder`, or `RootFolder`.\n\nRight now there is no convenient, general-purpose way to look up or mutate an arbitrary node\nin that hierarchy — callers have to reach for several narrow helpers and reimplement tree\ntraversal themselves. Add a small, unified set of node operations to the service so that\nconsumers can manage the tree through one consistent API:\n\n- **get** — given a node id, return the matching node from anywhere in the hierarchy\n (any root, at any depth). Return `null` when no node has that id.\n\n- **add** — insert a new node.\n - When given a reference id that points to a **folder** (a `Folder` or `RootFolder`),\n the new node becomes a child of that folder.\n - When given a reference id that points to a **file**, the new node is placed alongside\n it, inside the same parent folder.\n - When called without a reference id, the new node is added as a new top-level root entry\n (appended after any existing roots).\n\n- **update** — given a node carrying an `id` plus any fields to change, merge those fields\n into the existing node in place, leaving the node's other fields and its position in the\n tree untouched. Do nothing if no node has that id.\n\n- **remove** — given a node id, delete that node (and its entire subtree) from the\n hierarchy. Sibling nodes are unaffected. Do nothing if no node has that id.\n\nAfter any of `add`, `update`, or `remove`, the change must be reflected in the service's\nstate, so a subsequent **get** (and reading `folderTree.data`) observes it.\n\nThese operations work across a hierarchy that may contain multiple root folders and\narbitrarily nested children.\n"} {"task_id": "format-code-task-000055", "source_id": "format-code-task-000055", "domain": "code", "task_path": "tasks/format-code-task-000055", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:8df5e78b8414a0f6363a4ef1f328de3b545b5a7eb781bc4d954961a54312baf0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## RUM events being ingested without a `sessionId`\n\nWhile looking at the RUM events our app sends to the Datadog intake, I noticed a non-trivial number of them are missing a `sessionId` (the field is just absent / undefined on the event). These same events still have a view id and look otherwise valid, they just have no session attached, which makes them useless for session-level analysis.\n\nAfter some digging I managed to reproduce it with the following setup:\n\n1. Open our app in two tabs, both initialized with the browser RUM SDK. The session is initially **not** sampled / not tracked.\n2. Wait for the session to expire.\n3. In tab A, do something that causes the SDK to renew the session, and this time the new session lands in the *tracked* bucket.\n4. Switch to tab B (do **not** interact with it — don't click, don't navigate, just let background things like XHRs, long tasks, errors etc. happen).\n\nTab B keeps sending RUM events to the intake, and those are the ones showing up without a `sessionId`. Tab B never had a chance to react to the renewed session — from its point of view it's still on the last view of the now-expired session — yet events from it keep flowing through.\n\nI'd expect the SDK to simply **not send** RUM events when it can't attach them to a real session. A tab that hasn't seen any actual user interaction since the session was renewed shouldn't be silently producing orphan events in the backend."} {"task_id": "format-code-task-000056", "source_id": "format-code-task-000056", "domain": "code", "task_path": "tasks/format-code-task-000056", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d026ee734aefb4a241fb86d03b1543ebd95f6a7259109c02a4be47500d8b2f74", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\n我们 Lambda 用的内存设得比较小（128MB 那种），跑 Node 的函数 OOM 的时候 Datadog 里基本看不到 `aws.lambda.enhanced.out_of_memory` 这个指标，感觉是因为内存太小直接被 kill 了、function 日志里压根没有 OOM 的堆栈，extension 就漏掉了。但其实 AWS 给的 platform.report 日志里 status 是 error、而且 maxMemoryUsed 已经顶到 memorySize 了，这种情况能不能也算成 OOM 给我报出来？另外要注意同一个 requestID 别重复计数，一次 invocation 最多报一次就行。\n\n# Expected outcomes\n\n- Platform report OOM detection: when a Lambda platform report for a request indicates an errored invocation and reports memory usage that reaches or exceeds the configured memory size, the extension emits the same enhanced OOM signal as it does for an OOM found in function logs.\n- Existing OOM detection remains intact: function logs that already match the existing out-of-memory detection still emit `aws.lambda.enhanced.out_of_memory` and the associated error enhanced metric.\n- Per-request de-duplication: if multiple log records for the same request indicate OOM, including a function log and a platform report for the same request, `aws.lambda.enhanced.out_of_memory` is emitted at most once for that request.\n- Request-scoped de-duplication state persistence: the request identity used to avoid duplicate OOM reporting survives the existing execution-context save/restore flow, so restored processing does not count the same request again.\n- Non-qualifying platform reports must not be counted as OOM solely because they are report logs.\n\n# Implementation notes\n\n- The exact data structures, helper boundaries, API shape, and validation location are up to the implementer.\n- The platform-report path and function-log path should share the same externally visible OOM metric semantics.\n- The de-duplication should be based on the request identity, not on log ordering or on a single log source."} {"task_id": "format-code-task-000057", "source_id": "format-code-task-000057", "domain": "code", "task_path": "tasks/format-code-task-000057", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9e1d3d6484d0d443aca802000639cf5901c142a256071f20732052cfb972204b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Custom auto multi-line samples 通过环境变量配置时不生效\n\n我在 Kubernetes 里跑 datadog-agent，所有配置都走环境变量（不挂 datadog.yaml）。最近想用 auto multi-line detection 的 custom samples 功能聚合一些应用日志。\n\n按照文档把配置写在 datadog.yaml 里测试是没问题的，比如：\n\n```yaml\nlogs_config:\n auto_multi_line_detection_custom_samples:\n - sample: \"2024-01-01 00:00:00\"\n label: start_group\n - regex: \"^\\\\[ERROR\\\\]\"\n label: start_group\n```\n\nagent 重启后能正确识别我的多行日志。\n\n但是搬到容器部署、改成环境变量后就不工作了：\n\n```\nDD_LOGS_CONFIG_AUTO_MULTI_LINE_DETECTION_CUSTOM_SAMPLES='[{\"sample\":\"2024-01-01 00:00:00\",\"label\":\"start_group\"},{\"regex\":\"^\\\\[ERROR\\\\]\",\"label\":\"start_group\"}]'\n```\n\nagent 起来后这些 custom samples 完全没生效，日志还是按默认逻辑拆，并且 agent 日志里能看到一行 unmarshal custom samples 失败的错误。\n\n其它列表型的配置（比如 tags、`DD_CONTAINER_EXCLUDE` 那一类）以 JSON 字符串形式传 env var 都是正常工作的，所以希望这个配置项也能支持同样的用法 —— 容器化部署里没法只靠 YAML。"} {"task_id": "format-code-task-000058", "source_id": "format-code-task-000058", "domain": "code", "task_path": "tasks/format-code-task-000058", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:91b7266714fc6d0a66653e4b15a81820625e6019b8510e625560989946c22c47", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Feature request: expose the agent process start time to Python checks\n\nI'm writing a custom Python check and I need to know when the agent process itself started up (i.e. roughly the time the agent was (re)started on the host). Use cases:\n\n- Skipping / filtering out events or log entries from before the agent came up, so a freshly-restarted agent doesn't re-emit old stuff.\n- Computing how long the agent has been running, for sanity-check metrics in the check itself.\n- Distinguishing \"the agent just started, this is the first run\" from \"the agent has been running for a while\" when deciding what state to publish.\n\nFrom inside a check (Python side), I can get things like the hostname, the cluster name, the agent version, etc. via the `datadog_agent` module, but I can't find anything that tells me when the agent process actually started. As far as I can tell there is no way to get this value from Python today — the agent obviously knows it internally (it has to, for its own status page / flare), it's just not surfaced to checks.\n\nCould the agent expose its process start time through the `datadog_agent` module so checks can read it? A timestamp (seconds since the epoch) would be the most useful shape — that's trivial to compare against `time.time()` from inside a check. Something like `datadog_agent.get_process_start_time()` would be ideal."} {"task_id": "format-code-task-000059", "source_id": "format-code-task-000059", "domain": "code", "task_path": "tasks/format-code-task-000059", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:bd1d7cd57d1a31aa1b5c6a49553932a1c321d040c15b0b962045f5a25a859667", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI'm running DatadogAgent v2alpha1 and need to control how the operator rolls out the node agent DaemonSet and cluster-agent-related Deployments; right now if I tweak the workload update strategy directly, the operator just reverts it. Can we make that configurable from the DatadogAgent overrides so I can choose the rollout style and tune how aggressive it is?\n\nExpected outcomes:\n- DatadogAgent v2alpha1 component overrides should accept an `updateStrategy` setting in the CR, including a strategy type and rolling-update limits.\n- When an override sets that strategy, the managed node-agent DaemonSet or cluster-agent-related Deployment should reflect the requested rollout behavior instead of being forced back to the operator default.\n- Rolling-update limits for unavailable pods and surge pods should preserve Kubernetes `IntOrString` semantics, so both numeric counts and percentage-style values are accepted and round-trip cleanly.\n- Supported strategy values should remain usable in the normal Kubernetes sense for the target workload, including rolling-style and non-rolling-style choices where applicable.\n- Unrelated override fields should continue to behave as they do today.\n\nImplementation notes:\n- The exact reconciliation structure, helper layout, generated-code organization, and internal data flow are implementation choices.\n- Validation, defaulting, and object copying may be organized however is most natural for the codebase, as long as the observable CR/API behavior and resulting workload spec match the expectations above."} {"task_id": "format-code-task-000060", "source_id": "format-code-task-000060", "domain": "code", "task_path": "tasks/format-code-task-000060", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c3b5d988c97689ca6a49626f265b52954d31b1110062a6df7f52ec16e8f8cb6f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Support `scheduling_options` / cumulative evaluation window on `DatadogMonitor`\n\nI'm using the operator to manage monitors via GitOps. I'd like to set up a few \"calendar-aligned\" cumulative monitors — e.g. a monthly cost / usage budget alert that resets on the 1st of each month, and a daily SLO-like accumulator that resets at 00:00 UTC.\n\nIn the Datadog UI (and via the public Datadog API) this is exposed through `scheduling_options.evaluation_window`, where you can pick a cumulative window aligned to the start of the hour, day, or month (`hour_starts` / `day_starts` / `month_starts`). The behaviour is different from a rolling timeframe — values accumulate from the alignment point instead of sliding.\n\nThe problem is I don't see any way to express this on a `DatadogMonitor` CR. `spec.options` covers a lot of the usual stuff (`timeoutH`, `requireFullWindow`, `notificationPresetName`, thresholds, renotify settings, etc.) but there's no field for the evaluation window / scheduling options. If I add it under `spec.options` anyway, the K8s API server rejects it as an unknown field, and even if I bypass that the controller has no logic to forward it to the Datadog API — the resulting monitor in Datadog ends up with the default rolling window instead of the cumulative one I asked for.\n\nFor folks managing budget / quota style monitors as code, this means we currently have to click those monitors together in the UI (or maintain them through a separate Terraform pipeline) instead of keeping them next to the rest of our `DatadogMonitor` manifests.\n\nCould the operator expose the scheduling options / cumulative evaluation window on `DatadogMonitor` so it reaches parity with the Datadog API? All three alignment kinds (hourly / daily / monthly) should be configurable, and the controller should pass them through when creating / updating the monitor."} {"task_id": "format-code-task-000062", "source_id": "format-code-task-000062", "domain": "code", "task_path": "tasks/format-code-task-000062", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:431223d61ff3999242835c72529da923868e2f7beb5a260d2e75136d8e96af77", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nproposal: profiler: add CPUProfileRate option\nCalling [`runtime.SetCPUProfileRate`](https://pkg.go.dev/runtime#SetCPUProfileRate) with a non-zero value starts profiling at the given rate, if it hasn't already been called. This rate is effective until profiling is stopped (by setting the rate to 0). Because the profiler continually starts and stops CPU profiling, every round of CPU profiling after the first one will instead run with the default rate of 100 Hz. The user has no way to change this other than completely stopping our profiling. We should have a profiler option to give a rate that will be set each time before CPU profiling is started."} {"task_id": "format-code-task-000063", "source_id": "format-code-task-000063", "domain": "code", "task_path": "tasks/format-code-task-000063", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:970d881585867c9bba743a1f6e24488c955db4ddb2e187c291a7d1e25b5f7a61", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nioredis package is logging AUTH commands\n**Describe the bug**\nioredis instrumentation is logging all commands to redis including the `AUTH` command which contains the password.\n\nhttps://github.com/DataDog/dd-trace-js/blob/v0.36.2/packages/datadog-plugin-ioredis/src/index.js#L13\n\n**Environment**\n\n* **Operation system:** Linux\n* **Node version:** 14.17.5\n* **Tracer version:** 0.36.2\n* **Agent version:** 7.31.0"} {"task_id": "format-code-task-000064", "source_id": "format-code-task-000064", "domain": "code", "task_path": "tasks/format-code-task-000064", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:082e16aba3c87b2dfb43b032cbcaa1d5b2c35f408b3a0fd69c9f09ffa5159f61", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI noticed our Datadog CI Visibility events are showing credentials from URLs, like the token in our Git remote ending up in `git.repository_url`, and even `GITHUB_SERVER_URL` leaking into the GitHub Actions CI tags/env vars. Can dd-trace-js strip the username/password/token parts before reporting those URLs?\n\n## Expected Outcomes\n\n- Git repository URLs reported in CI Visibility metadata, including `git.repository_url`, should not include URL username, password, or token userinfo when the source URL contains credentials.\n- CI Visibility URL metadata should remove URL userinfo consistently across supported repository and CI URL sources, not only one provider-specific path.\n- GitHub Actions URLs derived from `GITHUB_SERVER_URL`, including `ci.pipeline.url` and `ci.job.url`, should not expose credentials from the server URL.\n- The `_dd.ci.env_vars` payload should report `GITHUB_SERVER_URL` without URL userinfo.\n- Sanitized URLs should retain the non-sensitive address information needed to identify the repository or CI run.\n\n## Implementation Notes\n\n- Apply the sanitization before CI Visibility URL metadata is reported.\n- Keep the implementation free to choose its helper structure, parsing strategy, and validation location."} {"task_id": "format-code-task-000065", "source_id": "format-code-task-000065", "domain": "code", "task_path": "tasks/format-code-task-000065", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6cc1ad55d1fd9ee8a37653b599d7e8fe777eef4745dd1a1ea6efb1b72b273f7e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nHey, I'm using dd-trace in a long-running Node service and I'm seeing memory creep up over time. When I take a heap snapshot, a lot of it is span tag/metric objects sticking around — they seem to stay alive as long as I hold any reference to a span, even well after the trace has already been flushed and sent. I've also noticed something weird with child spans started after the parent's trace was finished: they end up on what feels like a stale/detached trace instead of continuing on the parent's, so the relationship gets a bit funky. Ideally once a trace is shipped off, all that tag/metric data should be free to GC even if some span ref is still lying around, and a child should just keep using the same trace as its parent without any surprises. Can you take a look?\n\n# Expected outcomes\n\n- Post-flush memory release:\n - After a completed trace has been flushed/sent, retaining references to spans from that trace should not retain their old tag or metric payloads.\n - The flushed trace should no longer keep references to the spans that were part of that completed trace.\n - This cleanup should apply consistently across completed traces that are flushed.\n\n- Parent/child trace continuity:\n - A child span created from a parent span should continue on the same underlying trace as the parent rather than being attached to a detached or reset trace.\n - This should remain true even when the child is created after the parent’s earlier trace work has already completed and been flushed.\n\n# Implementation notes\n\n- The exact cleanup location, data structures, and lifecycle hooks are implementation details; prefer the smallest change that preserves normal trace formatting/sending while allowing already-flushed span metadata to be collected.\n- Do not change the user-facing tracing semantics beyond the lifecycle behavior described above."} {"task_id": "format-code-task-000067", "source_id": "format-code-task-000067", "domain": "code", "task_path": "tasks/format-code-task-000067", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5987f5e3ecbdc9568af6bd08225a63fb8c4daba8bb466709a64be3a195ed15e6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `directory` check stops scanning the rest of the tree after one inaccessible subdirectory\n\nWe're using the `directory` integration with `recursive: true` to track file counts and total size on some large filesystem trees. Most of the tree is readable by the agent user, but the trees do contain a handful of subdirectories with restricted permissions (different ACLs, a few mount points the agent user isn't on the allow list for, that kind of thing).\n\nThe metrics we're getting back are way off:\n\n- `system.disk.directory.files`\n- `system.disk.directory.folders`\n- `system.disk.directory.bytes`\n\n…all report numbers much smaller than what's actually in the configured root. After bumping the agent log level we see a single error line about a permission problem while traversing the root path, and then nothing more — it looks like the check stops walking the rest of the tree as soon as it bumps into one entry it can't read, even though the vast majority of subdirectories under that same root are perfectly fine to access.\n\nFor a monitoring check this is pretty surprising. If one subtree can't be read we'd expect just that subtree to be skipped (with the error logged so we know it happened) and the rest of the tree to still be counted toward the metrics. As things stand, a single restricted folder somewhere under the configured root silently makes our totals wrong, and there's no easy way to tell from the reported numbers that anything is off.\n\nOther filesystem-traversal tools (`find`, `du`, etc.) report errors on a per-entry basis and keep going. Could the recursive walk in this check be made resilient to per-entry errors the same way — skip the unreadable parts, log them, and continue traversing the rest so the metrics reflect everything the agent *could* see?"} {"task_id": "format-code-task-000069", "source_id": "format-code-task-000069", "domain": "code", "task_path": "tasks/format-code-task-000069", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:25f55976fd5a55e4280cb4ec3c6e7273e7f58d066037d5282d9428d38cafa679", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Deterministic cluster-state translation for the partition mapper\n\nThe `mapper` package turns Kafka cluster state into the objects our rebalancing/reassignment\nlogic consumes. Two pieces of that translation are currently underspecified and need to be made\nreliable.\n\n## 1. Stable ordering when building a partition map from topic states\n\n`mapper.PartitionMapFromTopicStates` translates a `kafkaadmin.TopicStates` into a\n`*mapper.PartitionMap`. Today the partitions in the returned map come out in whatever order the\nunderlying maps happen to iterate, so two calls with the same input can produce maps whose\npartition lists differ in order. That makes downstream diffs and output noisy and forces every\ncaller to re-sort.\n\nMake the returned partition map deterministic:\n\n- The `Partitions` list must be sorted ascending by topic name, and within a topic ascending by\n partition ID, regardless of the iteration order of the input.\n- Each partition's replica list must preserve the order given in the source partition state\n (the replicas are reported preferred-leader-first; don't reorder them).\n- An empty or `nil` `TopicStates` yields an empty but non-`nil` `*PartitionMap` with no\n partitions and a `nil` error.\n\n## 2. Merging stored metrics into broker metadata\n\nBroker metadata assembled from the cluster needs storage metrics (collected out-of-band) merged\nin before the planner can use it. Add a way to populate a `mapper.BrokerMetaMap` from a\n`mapper.BrokerMetricsMap`:\n\n- For every broker present in the broker-metadata map, look up its entry in the metrics map and\n set the broker's `StorageFree` from it.\n- If a broker in the metadata map has no corresponding metrics entry, mark that broker's metadata\n as having incomplete metrics and record an error identifying the broker by its numeric ID. The\n broker's `StorageFree` is left untouched in that case.\n- Brokers that appear in the metrics map but not in the metadata map are ignored — they neither\n change anything nor produce an error.\n- The returned errors must be ordered ascending by broker ID, with exactly one error per broker\n that is missing metrics. When every broker has metrics, no errors are returned.\n- The broker-metadata map is updated in place.\n\nExpose this as a method on `BrokerMetaMap` with the signature\n`LoadMetrics(metrics BrokerMetricsMap) []error`.\n"} {"task_id": "format-code-task-000070", "source_id": "format-code-task-000070", "domain": "code", "task_path": "tasks/format-code-task-000070", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5b2ac87e6bad019cfdc5301229d4fb7fe775042326f7a93c4779f65c17482d38", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Manage the `orchestrion.tool.go` file incrementally\n\n`orchestrion pin` records which tracer integrations are enabled in a module by\nmaintaining a build-tagged Go source file named `orchestrion.tool.go`. Today\nthat file is rewritten from scratch on every run, which throws away anything a\nuser added by hand. We want to manage it incrementally instead.\n\nAdd a small, reusable piece of functionality (in package\n`github.com/DataDog/orchestrion/internal/pin/toolfile`) that creates or refreshes\nthis file for a given module directory. It should expose:\n\n```go\ntype Options struct {\n NoGenerate bool\n}\n\nfunc Update(dir string, opts Options) error\n```\n\n`Update` operates on `/orchestrion.tool.go` and must behave as follows:\n\n- **Creation.** When the file does not exist, create it as a valid Go source\n file carrying a `//go:build tools` build constraint, declared in\n `package tools`, and containing a blank import (`_ \"...\"`) of\n `github.com/DataDog/orchestrion`.\n\n- **Incremental update.** When the file already exists, parse it and keep every\n import it already declares — including blank imports a user added manually —\n while ensuring the blank import of `github.com/DataDog/orchestrion` is present.\n Each imported package must appear exactly once (no duplicates), and the result\n must remain a valid Go source file in `package tools` with the\n `//go:build tools` constraint.\n\n- **Generate directive.** Unless `Options.NoGenerate` is set, the file must\n contain a `//go:generate go run github.com/DataDog/orchestrion pin` directive.\n When `NoGenerate` is true, the directive must not be present — and if it was\n there before, it must be removed.\n\n- **Idempotency.** Running `Update` repeatedly with the same options must\n converge to a stable result: a second invocation leaves the file byte-for-byte\n identical to the first.\n\n- **Bad input.** If the file exists but cannot be parsed as Go source, `Update`\n must return an error and leave the file untouched.\n"} {"task_id": "format-code-task-000071", "source_id": "format-code-task-000071", "domain": "code", "task_path": "tasks/format-code-task-000071", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d97d9bf30c418a4d3a4732a5cdcbbcf3af21b918750549629861ab7ec24bac84", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## geomap widget should support formulas and functions\n\nMost of the dashboard widget types (`timeseries_definition`,\n`query_value_definition`, etc.) accept a `request` configured in the\nformulas-and-functions style — one or more `query { ... }` blocks plus\noptional `formula { ... }` blocks. `geomap_definition` doesn't: its\n`request` only exposes the older `q`, `log_query`, and `rum_query`\narguments.\n\nThis means I can't move my geomap widgets to formulas/functions even\nthough the Datadog API and UI support it. Here is the kind of `request`\nbody I'd write for a timeseries widget today:\n\n```hcl\nrequest {\n query {\n metric_query {\n name = \"q1\"\n data_source = \"metrics\"\n query = \"avg:system.cpu.user{*} by {country-iso-code}\"\n }\n }\n formula {\n formula_expression = \"q1\"\n }\n}\n```\n\nIf I drop the same body into `geomap_definition.request`, `terraform plan`\nrejects it — the geomap request schema doesn't know about `query` or\n`formula`.\n\nCould the geomap widget's request gain the same formula+query support that\nother widget types already have? Ideally with the same set of query\nsources (metric / events platform / process queries) so the config looks\nconsistent across widget types, and so the existing `formula` options\n(alias, limit, etc.) work the same way here."} {"task_id": "format-code-task-000072", "source_id": "format-code-task-000072", "domain": "code", "task_path": "tasks/format-code-task-000072", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:84c6874aa8a23cdd4015851adbb0a016f40f74bc5489fa540974537a253114a3", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nHi there,\n\nThis is a followup to https://github.com/DataDog/terraform-provider-datadog/issues/196.\n\nPlease add the possibility to enable the automuting feature in Terraform.\nWhen setting up the the integration the automuting feature is currently disabled by default.\n\nThanks!\n\nDocs: https://registry.terraform.io/providers/DataDog/datadog/latest/docs/resources/integration_azure\n\n![grafik](https://user-images.githubusercontent.com/88875030/138053563-80a223ad-6d88-4e75-a3a4-dceaeebe40f2.png)"} {"task_id": "format-code-task-000073", "source_id": "format-code-task-000073", "domain": "code", "task_path": "tasks/format-code-task-000073", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:942c824088d3f8711fae9910bd12d3f7e6c70adc5d086b61a828fb154b33b895", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## MD029 and MD030 fire on ordered lists inside blockquotes\n\nI'm using markdownlint on my project's documentation. I often quote example snippets inside a blockquote, and sometimes those examples are ordered lists. markdownlint flags these as violations even though the markdown looks fine and renders correctly.\n\n### Repro 1 — a single-level ordered list in a blockquote\n\n```md\n> 1. The simplest ordered list in blockquote\n```\n\nThis gets flagged as **MD029 / Ordered list item prefix**. But there's only one item and it starts at `1.`, so I'm not sure what's wrong with it.\n\n### Repro 2 — a nested ordered list in a blockquote\n\n```md\n> 1. blockquote-ol-li\n> 1. blockquote-ol-li-ol-li\n```\n\nThis trips both **MD029** and **MD030 / Spaces after list markers**. The inner list item is indented by 3 spaces under `1. ` just like a plain (non-blockquoted) nested ordered list would be, and the numbering restarts at `1.` as expected for a nested list. Outside of a blockquote the equivalent markdown is happily accepted.\n\n(For what it's worth, the same examples also trigger MD027 when there's the extra space after `>`, but I understand that one — it's the MD029 and MD030 reports that look wrong to me, because the lists themselves are written correctly.)\n\nCould MD029 and MD030 be taught to recognize ordered lists that live inside a blockquote?"} {"task_id": "format-code-task-000075", "source_id": "format-code-task-000075", "domain": "code", "task_path": "tasks/format-code-task-000075", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c7dab750bea597d686b55b6072e32cbc6daec8afd526ba9645c5b26715b8ec70", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI'm running DiceDB with `EvictionPolicy` set to `allkeys-lfu`, but it doesn't actually seem to evict based on access frequency — feels like the LFU policy just isn't doing anything. I'd really like keys I hit a lot to stick around and the rarely-touched ones to get kicked out first when memory fills up. Could you make `allkeys-lfu` actually work as a real frequency-based eviction strategy? It'd also be nice if I could tune how fast the frequency counter ramps up via some config, and honestly LFU feels like a saner default than LRU for my workload.\n\n## Expected outcomes\n\n- LFU eviction behavior\n - When `EvictionPolicy` is configured as `allkeys-lfu`, memory-pressure eviction should prefer removing keys with lower observed access frequency.\n - If candidate keys have equivalent access frequency, eviction should fall back to recency so that the longer-idle key is preferred for eviction.\n - Frequently accessed keys should remain more likely to survive eviction than rarely accessed keys across representative workloads, not only for one hard-coded key pattern.\n\n- Default and configuration behavior\n - When no eviction policy is explicitly configured, DiceDB should default to LFU-style all-keys eviction.\n - DiceDB should accept an `allkeys-lfu` eviction policy value alongside the existing eviction policy values.\n - A server configuration option for the LFU counter growth factor should be available, with a default value of `10`.\n - Increasing the LFU counter growth factor should make frequency counter growth slower; decreasing it should make growth faster.\n\n- Counter and recency behavior\n - LFU frequency tracking should remain bounded under repeated access and must not wrap around in a way that makes very frequently accessed keys look rarely used.\n - LFU frequency bookkeeping must not corrupt recency-based idle-time behavior used for LRU ordering or LFU tie-breaking.\n\n## Implementation notes\n\n- The exact data structures, sampling strategy, counter representation, and update locations are implementation choices.\n- The LFU implementation may use an approximate frequency counter, but externally observable eviction behavior should match the outcomes above.\n- Preserve existing eviction modes other than the requested LFU/default behavior changes."} {"task_id": "format-code-task-000076", "source_id": "format-code-task-000076", "domain": "code", "task_path": "tasks/format-code-task-000076", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4f92177ad8e43e011f3805cf5a3b5919165d7172dcec59ddd6b23e7669d2ec11", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `color_deconvolution` fails when passing a stain matrix with only two stains\n\nThe docstring of `color_deconvolution` says:\n\n> For two stain images the third column is zero and will be complemented using cross-product. At least two of the three columns must be non-zero.\n\nSo I expected that I could just give it the two stain vectors I care about (e.g. hematoxylin and eosin) and let the function fill in the residual third stain itself.\n\nWhat I actually tried:\n\n```python\nimport numpy as np\nfrom histomicstk.preprocessing.color_deconvolution import color_deconvolution\n\n# only the two stains I'm interested in (H and E), as columns\nw = np.array([\n [0.650, 0.072],\n [0.704, 0.990],\n [0.286, 0.105],\n])\n\nStains, StainsFloat, Wc = color_deconvolution(im_rgb, w)\n```\n\nThis blows up inside `color_deconvolution` — it clearly assumes `w` already has three columns and tries to look at the third one directly, so a 3x2 input never gets a chance to be complemented.\n\nThe only way I've found to make it work is to manually pad `w` with a zero column before calling the function:\n\n```python\nw3 = np.zeros((3, 3))\nw3[:, :2] = w\ncolor_deconvolution(im_rgb, w3) # works, third stain gets filled in\n```\n\nBut based on the docstring (and the fact that `complement_stain_matrix` already handles building the residual stain) I'd expect passing a 3x2 matrix directly to just work — i.e. the function should accept a stain matrix that contains only the two real stain columns and complement the missing one on its own."} {"task_id": "format-code-task-000077", "source_id": "format-code-task-000077", "domain": "code", "task_path": "tasks/format-code-task-000077", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:037a645bb68ee6e855c3d56e969cd20567e7f2691d3cbc387778a426c9b8575f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nTraceback when \"Union[Member, User]\" is used as type hint in slash command parameters\n### Summary\n\nWhen \"Union[Member, User]\" is used as a type hint in a slash command parameter, a traceback from disnake is shown when the slash command is invoked targeting a user who should resolve to a User object, not a Member object..\n\n### Reproduction Steps\n\nReference the below Minimal Reproducible Code and invoke the relevant slash command against a user who is not currently in the server. This issue does *not* occur when the target of the slash command is in the server (and resolves to a Member object).\n\n### Minimal Reproducible Code\n\n```python\n@discord_bot.slash_command()\nasync def whois(\n inter: ApplicationCommandInteraction,\n user: Union[Member, User] = commands.Param(\n description=\"The member to display information about\"\n ),\n) -> None:\n pass\n```\n\n\n### Expected Results\n\nThe slash command should be invoked successfully without a traceback when targeting a user who is not in the server. This used to work in v2.4.0 of disnake - not positive if it is broken in v2.5.0 as well.\n\n### Actual Results\n\nThe following traceback is observed when this slash command is invoked:\n```\nTraceback (most recent call last):\n File \"/usr/local/lib/python3.8/site-packages/disnake/ext/commands/interaction_bot_base.py\", line 1264, in process_application_commands\n await app_command.invoke(interaction)\n File \"/usr/local/lib/python3.8/site-packages/disnake/ext/commands/slash_core.py\", line 680, in invoke\n await call_param_func(self.callback, inter, self.cog, **kwargs)\n File \"/usr/local/lib/python3.8/site-packages/disnake/ext/commands/params.py\", line 811, in call_param_func\n kwargs[param.param_name] = await param.convert_argument(\n File \"/usr/local/lib/python3.8/site-packages/disnake/ext/commands/params.py\", line 464, in convert_argument\n return await self.verify_type(inter, argument)\ndisnake.ext.commands.errors.MemberNotFound: Member \"redacted_valid_snowflake_here\" not found.\n```\n\n### Intents\n\ndefault, members, message_content\n\n### System Information\n\n```markdown\n$ python -m disnake -v\n- Python v3.8.10-final\n- disnake v2.5.1-final\n - disnake pkg_resources: v2.5.1\n- aiohttp v3.7.4.post0\n- system info: Linux 5.4.0-120-generic #136-Ubuntu SMP Fri Jun 10 13:40:48 UTC 2022\n```\n```\n\n\n### Checklist\n\n- [X] I have searched the open issues for duplicates.\n- [X] I have shown the entire traceback, if possible.\n- [X] I have removed my token from display, if visible.\n\n### Additional Context\n\nTalked with Mari about this over Discord, she said she'd open up an issue for this. I opened this up to save her some time 😅"} {"task_id": "format-code-task-000078", "source_id": "format-code-task-000078", "domain": "code", "task_path": "tasks/format-code-task-000078", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a5414de8f9504855344787400ecfe484631d88378f5e153229a32c4cc6ac7242", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Allow overriding the config directory, not just the config file\n\nRight now the only way I can override where the CLI keeps its config is the `--configuration` flag, which expects the full path to the config file, e.g.\n\n```\ndoppler --configuration=/some/path/.doppler.yaml me\n```\n\nThis is awkward for a few reasons:\n\n- I don't really care what the file is called — that's an internal detail of the CLI. The CLI already owns a *directory* (`~/.doppler/` by default), and the file inside it is something the CLI manages. Asking me to repeat the filename in every invocation feels backwards.\n- It's easy to get the filename wrong (is it `.doppler.yaml`? `doppler.yaml`? something else?), and there's no real reason I should have to know.\n- In setups where I want to point Doppler at a non-default location (custom mount, shared machine, container with a writable directory that isn't `$HOME`, etc.), I'd much rather say \"use *this* directory\" and let the CLI figure out the rest.\n\nIt would be nice to have a flag that just takes a directory, something I can use like:\n\n```\ndoppler --=/some/path me\n```\n\n…and have the CLI place its config file inside that directory like it normally does under `~/.doppler/`.\n\nExisting behavior (default location of `~/.doppler/`) should keep working unchanged when no override is passed.\n\nI'd expect the new flag to be named something like `--config-dir`."} {"task_id": "format-code-task-000079", "source_id": "format-code-task-000079", "domain": "code", "task_path": "tasks/format-code-task-000079", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9da17888de8857509d0ef862b893f0e099e9c6477bd794b73ff9c8f6487ecd70", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nRight now circuit strings only seem to take single-letter element names like R, C, L — but I want to define my own custom elements with longer names (like \"RR\") and use them in a circuit, and at the moment it just doesn't parse them right. Could the parser be made to handle multi-letter element names? Also it'd be great if it correctly read multi-digit parameter indices too, since right now anything past single digits seems to break. Ideally custom element names would be restricted to plain uppercase letters so there's a clear rule about what counts as a valid name.\n\n## Expected outcomes\n\n- Multi-letter uppercase element names are supported by the public circuit-building APIs: a user-defined element such as `RR` can be made available and then used in a circuit string without being misparsed as separate one-letter elements.\n- Circuit parsing correctly preserves element names and numeric suffixes when extracting elements from circuit strings, including suffixes with more than one digit.\n- Element names are valid only when they consist of uppercase letters; names containing lowercase letters or digits in the name portion are rejected by validation or parsing.\n- The `RR` element is available and behaves like a resistor scaled by a factor of 100, so an `RR` circuit can produce the same impedance as an equivalent resistor circuit when its parameter is adjusted accordingly.\n\n## Implementation notes\n\n- The internal representation, lookup strategy, parsing approach, and validation location are up to the implementer.\n- Preserve existing behavior for existing single-letter elements and circuit syntax while extending support for the new naming and indexing cases.\n- Error types and messages may follow the repository’s existing conventions, as long as invalid names are rejected rather than silently accepted or misparsed."} {"task_id": "format-code-task-000080", "source_id": "format-code-task-000080", "domain": "code", "task_path": "tasks/format-code-task-000080", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a3b7c0a528714bb1188a746b1313d6cfb39977fb528582821d54458706181c5b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Harden the hsm_secret tool's `encrypt` flow and file validation\n\nOur standalone HSM secret command-line utility (`tools/hsmtool`) lets operators\nencrypt and decrypt their node's `hsm_secret` file with a password. Two rough\nedges keep biting people, and I'd like both fixed.\n\n## 1. Confirm the password when encrypting\n\nRight now `encrypt ` asks for the password exactly once and\nimmediately encrypts the seed with it. A single typo means the operator locks\nthemselves out of a seed they can never recover.\n\nMake the `encrypt` subcommand ask for the password a second time to confirm it,\nthe way well-behaved tools do. The behavior must be:\n\n- The two entries must match. If they differ, the tool aborts with a non-zero\n exit status, prints an error that makes clear the confirmation did not match,\n and leaves the `hsm_secret` file **completely untouched** (still the original\n plaintext seed, same bytes, same size).\n- If the two entries match, encryption proceeds exactly as before, producing a\n valid encrypted `hsm_secret` that `decrypt` can later turn back into the\n original seed using the same password (a full encrypt → decrypt round-trip\n must recover the original bytes).\n\n## 2. Be strict about what counts as a valid `hsm_secret`\n\nA plaintext seed is exactly 32 bytes and an encrypted one is exactly 73 bytes.\nToday the tool only checks \"bigger than 32 means encrypted\", so files of any\nother length get misinterpreted (treated as encrypted, or partially read) and\nproduce confusing failures.\n\nWhenever a subcommand needs to read an existing `hsm_secret` (for example\n`encrypt` and `decrypt`), it must first reject any file whose size is neither 32\nnor 73 bytes. Rejection means: print an error that identifies the file as an\n**invalid** `hsm_secret`, exit with a non-zero status, and leave the file\nunchanged. A genuine 32-byte plaintext seed and a genuine 73-byte encrypted seed\nmust keep working exactly as they do today.\n\nPasswords are read from the terminal with echo disabled, as before.\n"} {"task_id": "format-code-task-000081", "source_id": "format-code-task-000081", "domain": "code", "task_path": "tasks/format-code-task-000081", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:18ac41f5a2e66c62b657c71671521f8d8a9dead4a8676bae43908df5bfeb1779", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `doctest_module(..., 'all')` stops at the first failure instead of running the whole batch\n\nI'm using xdoctest to run all the doctests inside one of my modules:\n\n```python\nfrom xdoctest import doctest_module\ndoctest_module('mypkg.mymod', 'all')\n```\n\nThe module has ~10 doctests. As soon as one of the earlier ones fails, the\nrunner raises out of `_run_examples` and the rest of the doctests are never\nexecuted. So I see exactly one traceback, no summary, no idea which of the\nremaining tests pass or fail. To find out, I have to fix that one test and\nre-run the whole batch, hit the next failure, fix, re-run, etc.\n\nWhat I'd expect from an \"all\" run is the usual test-runner behavior: keep\ngoing on failure, run every collected example, and at the end print the\nfailures along with a \"N / M passed\" summary so I can see the whole picture\nin one pass. Failing fast still makes sense when there's only a single\nexample being run (e.g. when I'm debugging one specific doctest by name) —\nin that case I do want the exception to propagate so I get the traceback\nimmediately.\n\n---\n\nWhile poking at this I also noticed `doctest_module`'s `argv` parameter\ndoesn't really behave like one would expect. Two things:\n\n1. Passing `argv=['--verbose']` (or `argv=['all', '--quiet']`, etc.) from\n Python doesn't change the verbosity / style. The flags I put in `argv`\n are ignored and it looks at the real process `sys.argv` instead. That\n makes it hard to drive xdoctest programmatically from another script /\n test harness where I don't want to mutate `sys.argv`.\n\n2. If the first entry of `argv` is a flag (e.g. `argv=['--verbose']` with\n no explicit command), it ends up being treated as the test name to run\n and of course matches nothing. I'd expect flags to be recognized as\n flags regardless of position, and the command to default to \"no command\n given\" in that case.\n\nBoth feel like the `argv` handling should consistently use the argv that\nwas passed in."} {"task_id": "format-code-task-000082", "source_id": "format-code-task-000082", "domain": "code", "task_path": "tasks/format-code-task-000082", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:eadb7b98150fdecfa17f54df7616bb6b2fb5067853f86a29f0f6788a2f52202d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add an SEI message package for AVC and HEVC\n\nThe library can already pull SEI (Supplementary Enhancement Information) NAL units out\nof a stream, but there is no shared, codec-aware way to parse their contents. I'd like a\nnew package, importable as `github.com/edgeware/mp4ff/sei`, that turns the raw bytes of an\nSEI NAL unit into typed, inspectable messages and can re-encode them.\n\n## Splitting a NAL unit into messages\n\nExpose `func ExtractSEIData(rs io.ReadSeeker) ([]SEIData, error)`. The input is the rbsp\nbyte stream of one SEI NAL unit **with the NAL unit header already removed**. An SEI NAL\nunit packs one or more messages back to back, each laid out as:\n\n- a payload **type**: read bytes one at a time and sum them; keep going while a byte equals\n `0xff`, stop after the first byte that is not `0xff`.\n- a payload **size** (in bytes): encoded the same way (sum of bytes, continuing while `0xff`).\n- exactly *size* bytes of payload.\n\nThe stream is emulation-prevention encoded (the `0x000003` escaping used by Annex B byte\nstreams), so a `0x03` that follows two `0x00` bytes is not part of the payload and must not\nbe counted or returned. Keep reading messages until there is no more rbsp data (the\nremaining bits are just the rbsp stop bit and zero padding). Return the messages in stream\norder. The payload stored for each message is the de-emulated rbsp payload, so its length\nequals the declared size.\n\n`SEIData` is the raw, undecoded form of a message. Provide a constructor\n`func NewSEIData(payloadType uint, payload []byte) *SEIData` and make it satisfy the\n`SEIMessage` interface below, with `Type()` returning the payload type, `Size()` the payload\nlength in bytes, and `Payload()` the raw payload bytes.\n\n## Typed messages\n\nDefine the interface every message implements:\n\n```go\ntype SEIMessage interface {\n Type() uint // SEI payload type\n Size() uint // size in bytes of the rbsp payload\n Payload() []byte // the rbsp payload\n String() string // human-readable description\n}\n```\n\nAdd a codec discriminator `type Codec` with exported values `AVC` and `HEVC`, and a decoder\n\n```go\nfunc DecodeSEIMessage(sd *SEIData, codec Codec) (SEIMessage, error)\n```\n\nthat promotes a raw `SEIData` to a typed message. For payload types this package does not\nspecifically understand, it must fall back to returning a message that preserves the original\ntype and payload (never an error). Implement specific decoding for the two HDR static-metadata\nmessages:\n\n- **Mastering display colour volume**, payload type **137**. Its 24-byte payload is, in order\n and big-endian: for each of three colour primaries a 16-bit X then a 16-bit Y value\n (`DisplayPrimariesX[0]`, `DisplayPrimariesY[0]`, … `DisplayPrimariesX[2]`,\n `DisplayPrimariesY[2]`), then a 16-bit white point X and white point Y, then a 32-bit max\n display mastering luminance and a 32-bit min display mastering luminance. Decode it to a\n `*MasteringDisplayColourVolumeSEI` exposing those values as exported fields\n `DisplayPrimariesX [3]uint16`, `DisplayPrimariesY [3]uint16`, `WhitePointX uint16`,\n `WhitePointY uint16`, `MaxDisplayMasteringLuminance uint32`, `MinDisplayMasteringLuminance uint32`.\n Its `Size()` is always 24.\n\n- **Content light level information**, payload type **144**. Its 4-byte payload is two 16-bit\n big-endian values: max content light level then max picture average light level. Decode it\n to a `*ContentLightLevelInformationSEI` with exported fields `MaxContentLightLevel uint16`\n and `MaxPicAverageLightLevel uint16`. Its `Size()` is always 4.\n\nFor both of these, decoding a payload whose length does not match the fixed size must return\nan error rather than a message. `Payload()` on a decoded message must reproduce the exact\npayload bytes it was decoded from (round-trip), and must also work for a message built\ndirectly from its fields.\n\n## Type names\n\nProvide a named type `type SEIType uint` whose `String()` method returns a human-re"} {"task_id": "format-code-task-000083", "source_id": "format-code-task-000083", "domain": "code", "task_path": "tasks/format-code-task-000083", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a8f751a0c49ad6174b15b2bceae819cf3753adfd1603479ab98fa3bdd20480e2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## ASM policies SDK is missing the Vulnerability Assessment endpoint\n\nI'm using the SDK to script ASM policy management on a BIG-IP. For a given policy I can already get to most of the sub-resources through the lazy attributes — `policy_builder`, `signatures_s`, `blocking_settings`, `history_revisions_s`, etc. — but I can't get to the `vulnerability-assessment` sub-resource.\n\nThe endpoint itself is there on the box, e.g.\n\n```\nGET /mgmt/tm/asm/policies//vulnerability-assessment\n```\n\nreturns the VA config just fine when I hit it directly. But via the SDK there's no corresponding attribute on the policy object, so I have no way to `load()` it / interact with it the way I do with the other policy sub-resources.\n\nCould the Vulnerability Assessment resource be added to the ASM policies module so it works consistently with the other policy sub-resources?"} {"task_id": "format-code-task-000084", "source_id": "format-code-task-000084", "domain": "code", "task_path": "tasks/format-code-task-000084", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:797d8aa16af182eaf8d75fdda727bacf73c78368e0d6a35703b4290e62622cfe", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `wrappers.vector.HumanRendering` doesn't actually render\n\nI'm trying to use `gymnasium.wrappers.vector.HumanRendering` to show\nseveral CartPole envs tiled in a single pygame window:\n\n```python\nimport gymnasium as gym\nfrom gymnasium.wrappers.vector import HumanRendering\n\nenvs = gym.make_vec(\"CartPole-v1\", num_envs=3, render_mode=\"rgb_array\")\nenvs = HumanRendering(envs)\n\nobs, info = envs.reset()\nfor _ in range(100):\n envs.step(envs.action_space.sample())\nenvs.close()\n```\n\nThe very first `reset()` call blows up and I never get a window. I\ntried a few other configurations — different `num_envs`, passing a\ncustom `screen_size`, leaving it as `None` — and the failure mode\nshifts around but the wrapper is never actually usable for me. I\ncan't get a single rendered frame out of it for any setup I tried.\n\nCould someone take a look? Based on what I'm seeing it doesn't look\nlike the vector `HumanRendering` wrapper works at all in its current\nstate."} {"task_id": "format-code-task-000085", "source_id": "format-code-task-000085", "domain": "code", "task_path": "tasks/format-code-task-000085", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a11491cccf73e66a7e48338d50bd3787a1d7ebb25b096a0a00b185889bd9034a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[Proposal] Add transitional probabilities to Taxi and Cliff Walking toy text environments\n### Proposal\n\nOnly Frozen Lake in the toy text grid world environments implements transitional probabilities. \n\nTaxi is supposed to have it based on the previous documentation but has never been implemented, always returning 1.0. \n\nAt the same time cliff walking could also be set up to use transitional probabilities using the same approach.\n\n### Motivation\n\nAdding transitional probabilities to taxi will close out a TODO that has been on the list for a long time. It will also bring the environment in line with the source paper, The Fickle Taxi Task - Section 7.1 of Hierarchical Reinforcement Learning with the MAXQ Value Function Decomposition (https://www.jair.org/index.php/jair/article/view/10266/24463).\n\nFor cliff walking, it presents and opportunity to add depth to the environment and since it uses the same approach would not add significantly more time or risk.\n\n### Pitch\n\nTaxi\nAdd transitional probability into taxi toy text environment:\n- leverage approach from frozen_lake to supply a transitional probability of 0.8 direction intended, 0.1 left and 0.1 right of intended direction for movement actions.\n- the paper proposes that, once the taxi has picked up the passenger and moved one square away from the passenger's source location, the passenger changes their destination location with probability 0.3.\n- for taxi transition probabilities for pick up and drop off actions remain 1.0.\n- add arguments to enable/disable features: \n - `is_rainy = True | False` to enable transitional probabilities on taxi movement, defaults to `False`.\n - `fickle_passenger = True | False` to enable the passenger to change their destination once picked up, defaults to `False`.\n\nCliff walking\nAdd transitional probability into cliff walking toy text environment by leverage approach from frozen_lake to supply a transitional probability of 0.3 direction intended, 0.3 left and 0.3 right of intended direction for movement actions.\n- add arguments to enable/disable transitional probabilities. \n - `is_slippery = True | False` to enable transitional probabilities on player movement, defaults to `False`.\n\nFor both:\n- Update unit tests.\n- Update documentation.\n- Increment versions in registry.\n\n### Alternatives\n\n1. Do nothing. Misses an opportunity to make the toy_text environments consistent and more useful for beginner RL practitioners.\n\n2. Remove transitional probability from taxi and/or cliff walking. In either case `prob` would be removed from the info returned. Removes the need to complete taxi work and will simplify any ongoing maintenance.\n\n### Additional context\n\n_No response_\n\n### Checklist\n\n- [X] I have checked that there is no similar [issue](https://github.com/Farama-Foundation/Gymnasium/issues) in the repo"} {"task_id": "format-code-task-000087", "source_id": "format-code-task-000087", "domain": "code", "task_path": "tasks/format-code-task-000087", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7ccc494b08b3f8f4b2dd67c2957845ff95ce2a2af29201ad9abe4ed14d548399", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: a way to query which API actions are available\n\nI'm building a third-party tool on top of AnkiConnect (similar to how Yomichan integrates with it) and I'd like to do **feature detection** at runtime — i.e. before I call some action, figure out whether the user's installed AnkiConnect actually supports it.\n\nThe use case is pretty standard: AnkiConnect grows new actions over time, and not every user is on the same version. If my tool wants to use a newer action when it's available and fall back to something else when it isn't, I need a way to ask the server \"do you support action X?\".\n\nRight now the only way I've found to do this is to just go ahead and send the request, and then look at the response — if I get back\n\n```json\n{\"result\": null, \"error\": \"unsupported action\"}\n```\n\nthen I know it's not supported. This works, but it's pretty awkward:\n\n- I have to actually invoke the action (with plausible-looking params) just to probe for it, which feels wrong for actions that have side effects.\n- If I want to probe several candidate actions to pick the best one available, I have to fire one request per action and parse the error string of each. There's no batch way to do it.\n- \"Parse the error message of a failed call\" isn't a great contract to build on — it would be nicer to have a first-class way to ask the question.\n\nCould AnkiConnect expose a dedicated action for this? Something I can call to either get the list of all actions this instance supports, or to pass in a list of action names I care about and have the server tell me which subset of those it actually has. That way third-party clients can do clean capability checks without abusing the error path.\n\nHappy to consume whatever shape of response makes sense on your side — I mainly just need the information to be queryable through the normal request/response interface. The new action I'd expect is something like `apiReflect`, taking `scopes` and `actions` params and returning a result keyed by those same scope names (e.g. `{\"scopes\": [...], \"actions\": [...]}`)."} {"task_id": "format-code-task-000088", "source_id": "format-code-task-000088", "domain": "code", "task_path": "tasks/format-code-task-000088", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2853a9b39cc1171be5ad4f4b3d49fac166c4c01da1b2208f726d22447f8322c0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Transaction system migrations trigger more recoveries than necessary\n\nI'm running a FoundationDB cluster managed by the operator and recently did\na configuration change that touches multiple transaction system process\nclasses at once (something that requires replacing both my log and stateless\nprocesses). I expected the whole change to settle with a single recovery at\nthe end, but I'm seeing more than one recovery during what I'd consider one\nmigration.\n\nWhat I can observe from the operator logs and the cluster during this kind\nof change:\n\n- The new stateless pods come up pretty quickly because they don't need\n persistent storage.\n- The operator excludes the old stateless processes as soon as their\n replacements are ready → a recovery happens.\n- The new log pods take noticeably longer because they have to wait for PV\n provisioning.\n- Once those are finally up, the operator excludes the old log processes →\n another recovery happens.\n\nSo a single config change ends up producing a recovery per transaction\nprocess class as each batch of replacements comes online, instead of one\nrecovery at the end. Each recovery is disruptive for our workload, and for\na migration that's logically \"replace the transaction system\", I'd really\nonly expect one.\n\nWould it be possible for the operator to coordinate exclusions across the\ntransaction system process classes — i.e. hold off on excluding any of them\nuntil it's ready to make progress on all the affected transaction classes\ntogether — so that this kind of migration converges in a single recovery?\nStorage replacements are typically much rarer than transaction-system churn\nin our environment, so even just batching the transaction side would help\na lot."} {"task_id": "format-code-task-000089", "source_id": "format-code-task-000089", "domain": "code", "task_path": "tasks/format-code-task-000089", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:02c1b61699c8ba5562395a4c5ba9f565c5d5a31733ae3fe5603ddb069efea7c9", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Auto-generated ExtensionObject dataclasses have awkward default values\n\nWhen I let opcua-asyncio import custom structs from a server via\n`load_data_type_definitions`, the generated `@dataclass` classes work, but\nthe defaults assigned to their fields are off in two ways.\n\n### 1. Optional fields don't default to `None`\n\nI have a structure defined on the server with optional fields (so the\nStructureDefinition is `StructureWithOptionalFields` and individual\n`StructureField`s have `IsOptional=True`). For example a field like an\noptional `Int32`.\n\nAfter `load_data_type_definitions`, the generated dataclass looks roughly\nlike:\n\n```python\n@dataclass\nclass MyStruct:\n ...\n SomeOptionalInt: Optional[ua.Int32] = 0\n```\n\ni.e. the optional field defaults to `0` rather than `None`. That makes it\nimpossible to tell \"user did not set this optional field\" from \"user\nexplicitly set it to 0\", and it doesn't match what `Optional[...]` is\nsupposed to mean here. I'd expect any field flagged as optional in the\nStructureDefinition to default to `None` in the generated dataclass.\n\n### 2. Numeric / string defaults aren't typing-friendly\n\nFor non-optional scalar fields the generated code mixes the declared type\nwith a plain Python literal as the default, e.g.:\n\n```python\nSomeInt: ua.Int32 = 0\nSomeName: String = None\n```\n\nRunning mypy against modules that touch these generated classes complains\nabout the mismatch (the annotation says `ua.Int32` / `String`, the default\nis a plain `int` / `None`). It would be nice if the generated defaults\nwere values of the declared type, so the dataclasses are clean under\nstatic type checking.\n\nCould the default-value generation for ExtensionObject dataclasses be\ntweaked so that (a) optional fields default to `None`, and (b) the\ndefaults for basic numeric/string fields are consistent with the declared\nfield type?"} {"task_id": "format-code-task-000090", "source_id": "format-code-task-000090", "domain": "code", "task_path": "tasks/format-code-task-000090", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:33b622b235dba4bfd71be2c6646ae5dbf224cb41ab98f062e5baeb11ae0531d9", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI'm using python-opcua and I'm running into a modeling issue with audit events. The OPC UA spec lets the emitting node (the one actually firing the event) be different from the SourceNode the event describes — e.g. the Server object emits an AuditCreateSessionEvent whose SourceNode points at the session/session diagnostics node. But with `BaseEvent` and all the `Audit*Event` constructors I can only pass `sourcenode`, there's no way to tell it which node is actually emitting. Could we get a way to specify the emitting node separately when constructing these events, ideally defaulting to the Server object so I don't have to set it every time?\n\n# Expected outcomes\n\n- `BaseEvent` construction supports specifying an emitting node separately from `sourcenode`; when both are provided, the event’s emitting node behavior follows the emitting-node argument while the `SourceNode` event field still reflects `sourcenode`.\n- `BaseEvent` construction remains backwards compatible for existing `sourcenode`, `message`, and `severity` usage.\n- When no emitting node is provided to `BaseEvent`, the event defaults to being emitted by the OPC UA Server object.\n- Public `Audit*Event` constructors support the same emitting-node option and pass it through consistently, so audit event subclasses can be emitted by a node different from the `SourceNode` they describe.\n- Public `Audit*Event` constructors keep the same default behavior as `BaseEvent`: omitting the emitting node uses the Server object, and existing `sourcenode`, `message`, and `severity` arguments continue to work.\n\n# Implementation notes\n\n- Preserve the existing public event-construction style while adding the new emitting-node capability.\n- The concrete internal representation, validation location, and propagation mechanism are up to the implementation, as long as externally constructed and generated events behave as described.\n- Avoid coupling `SourceNode` semantics to the emitting node; callers must be able to set them independently."} {"task_id": "format-code-task-000091", "source_id": "format-code-task-000091", "domain": "code", "task_path": "tasks/format-code-task-000091", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:eb32b79a76215ea2e23af24cebf2c70aa11ebabe818addc77d0db12656a7bb52", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nHello, I'm using the exception mapper to catch exceptions and return them in _json_ format.\n\nThe problem is exception code is only mapped into _http header_, but not in the _json response_. \n\nThis is the response I got when my app threw a `ValidationServiceException`:\n\n![Selection_023](https://user-images.githubusercontent.com/496541/83788601-bf6ab200-a695-11ea-89f3-8275a17ae491.png)\n\nThis is my config:\n\n```yaml\n# Read the documentation: https://symfony.com/doc/master/bundles/FOSRestBundle/index.html\nfos_rest:\n param_fetcher_listener: true\n # allowed_methods_listener: true\n # routing_loader: true\n view:\n view_response_listener: true\n exception:\n enabled: true\n codes:\n 'App\\Service\\ValidationServiceException': 412\n messages:\n 'App\\Service\\ValidationServiceException': true\n debug: true\n map_exception_codes: true\n flatten_exception_format: 'legacy'\n body_listener: true\n format_listener:\n rules:\n - { path: ^/api/v1, prefer_extension: true, fallback_format: json, priorities: [ json ] }\n # https://github.com/FriendsOfSymfony/FOSRestBundle/issues/631#issuecomment-30321824\n - { path: ^/, prefer_extension: true, fallback_format: ~, priorities: [ html ] }\n disable_csrf_role: ROLE_API\n```\n\nPlease note I'm using `map_exception_codes` and `flatten_exception_format` options. Also I'm using FosRestBundle v3.0.0.\n\nCould you help me please? I spend all my afternoon with this issue, the sole documentation I found is the one for v2 in Symfony's website."} {"task_id": "format-code-task-000092", "source_id": "format-code-task-000092", "domain": "code", "task_path": "tasks/format-code-task-000092", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:689f861f0175a0e356d19d8bf071331e73210ea6be904fa1a01db3a997964718", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAllow multiple variable-rank dimension specs in check_shapes\n# Feature request\n\nAllow multiple variable-rank dimension specs in check_shapes.\n\n## Motivation\n\nToday we have:\n\n```python\n@check_shapes(\n \"a: [a_shape...]\",\n \"b: [b_shape...]\",\n \"return: [a_shape_b_shape...]\",\n)\ndef broadcasting_elementwise(\n op: Callable[[tf.Tensor, tf.Tensor], tf.Tensor], a: tf.Tensor, b: tf.Tensor\n) -> tf.Tensor:\n \"\"\" Apply binary operation `op` to every pair in tensors `a` and `b`. \"\"\"\n```\n\nit would be nice to have\n\n```python\n@check_shapes(\n \"a: [a_shape...]\",\n \"b: [b_shape...]\",\n \"return: [a_shape..., b_shape...]\",\n)\ndef broadcasting_elementwise(\n op: Callable[[tf.Tensor, tf.Tensor], tf.Tensor], a: tf.Tensor, b: tf.Tensor\n) -> tf.Tensor:\n \"\"\" Apply binary operation `op` to every pair in tensors `a` and `b`. \"\"\"\n```\n\nThis is hard for reasons similar to why GPflow/check_shapes#5 is hard. In general, if there are multiple variable-rank dimensions in one specification, we cannot determine which dimensions belong to which variable. We could solve this is a couple of ways:\n\n1. Require the user to specify their checks in an order, so that there always is at most one variable-rank dimension with **unknown** size, at the time the constraint is evaluated.\n2. Evaluate constraints lazily, and on a best-effort basis. If we're trying to evaluate a constraint where we don't have enough information we simply store it for later, and hope that the necessary information becomes available.\n\nI don't like (1), because I don't like have to impose an ordering on constraints, and because optional arguments and broadcasting can make it hard to predict when a variable value will actually be available. I like (2) better, but it does have the disadvantage that sometimes a shape variable value may never be known, and no check actually performed."} {"task_id": "format-code-task-000093", "source_id": "format-code-task-000093", "domain": "code", "task_path": "tasks/format-code-task-000093", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2b8ef96a0809d798c01b5f0181f4b629d75a8a11fa0eda91fcb03d6b5d1f26dc", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\npygmt.show_versions doesn't show GDAL version\n```python\n>>> import pygmt\n>>> pygmt.show_versions()\nPyGMT information:\n version: v0.17.0.dev11+g0c57126f35\nSystem information:\n python: 3.13.5 | packaged by conda-forge | (main, Jun 16 2025, 08:23:50) [Clang 18.1.8 ]\n executable: /opt/miniforge/envs/pygmt/bin/python3.13\n machine: macOS-14.6.1-x86_64-i386-64bit-Mach-O\nDependency information:\n numpy: 2.3.0\n pandas: 2.3.0\n xarray: 2025.6.1\n packaging: 25.0\n contextily: 1.6.2\n geopandas: 1.1.0\n IPython: 9.3.0\n pyarrow: 20.0.0\n rioxarray: 0.19.0\n gdal: None\n ghostscript: 10.04.0\nGMT library information:\n version: 6.5.0\n padding: 2\n share dir: /opt/miniforge/envs/pygmt/share/gmt\n plugin dir: /opt/miniforge/envs/pygmt/lib/gmt/plugins\n library path: /opt/miniforge/envs/pygmt/lib/libgmt.dylib\n cores: 8\n grid layout: rows\n image layout:\n binary version: 6.5.0\n```\nCurrently, we use the `importlib` library to obtain the GDAL version, but this only works when the GDAL Python bindings are installed.\n\nSix months ago, we updated GMT's dependency from `gdal` to `libgdal-core` (see https://github.com/conda-forge/gmt-feedstock/commit/48fb3f0f19c9f3d082a98acac139eecbf9cd3922), so now only the GDAL C library is included, without the Python bindings. This explains why `importlib` fails.\n\nPotential solutions:\n\n1. Use `subprocess` to call `gdalinfo --version` and parse the output\n2. Use ctypes to obtain the GDAL version directly from the C library\n3. No longer query the GDAL version\n\nI have no strong preferences. Which one do you prefer? Or do you have a better solution?"} {"task_id": "format-code-task-000094", "source_id": "format-code-task-000094", "domain": "code", "task_path": "tasks/format-code-task-000094", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e027c32ff541c33a266153bcb2433e5a93376b913de86a946d71c722bd448665", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nSupport theming of chevron icon color in scroll to bottom button\n**Describe what you would like to achieve**\nChanging the chevron icon in the scroll to bottom button\n\n\n**Expected behavior**\nChevron will change color, after specifying it in the theme object\n\n**Additional context**\nOnly options at the moment are \n\n```\nscrollToBottomButton: {\n container: ViewStyle;\n touchable: ViewStyle;\n unreadCountNotificationContainer: ViewStyle;\n unreadCountNotificationText: TextStyle;\n wrapper: ViewStyle;\n};\n```\n\n\n**Screenshots**\n\n\n
\n\n\n\n\ngz#22583"} {"task_id": "format-code-task-000095", "source_id": "format-code-task-000095", "domain": "code", "task_path": "tasks/format-code-task-000095", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:86ec6b56381405942ed2696a198684c25d858aa46cd9df3f85fc77607440eee0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\norigin in options with regexp value is broken\nAccording to https://googlechrome.github.io/sw-toolbox/docs/master/tutorial-api.html:\n\n> The origin option is specific to the router methods, and can be either an exact string or a Regexp against which the origin of the Request must match for the route to be used.\n\nSo if I had \n\n```\nruntimeCaching: [{\n urlPattern: /^https:\\/\\/example\\.com\\/api/,\n handler: 'networkFirst'\n}, {\n urlPattern: /\\/articles\\//,\n handler: 'fastest',\n options: { origin: /twitter\\.com/}\n}]\n\n```\n\nWe would actually get `origin: {}`.\n\nAs in https://github.com/GoogleChrome/sw-precache/blob/master/lib/sw-precache.js#L213, `JSON.stringify({origin: /twiter\\.com/})` would give us `\"{\"origin\":{}}\"`"} {"task_id": "format-code-task-000096", "source_id": "format-code-task-000096", "domain": "code", "task_path": "tasks/format-code-task-000096", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:72506d9170f86814ac51f635dd3cfbec20d08fc6ef48265f0a4c59472ef17f18", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `responseHeadersEndTime` is unreliable for cached / `data:` requests\n\nI'm doing some custom analysis over Lighthouse's network records — specifically I want the time spent receiving response headers, computed as `responseHeadersEndTime - networkRequestTime` for each request.\n\nFor normal requests that actually go to the network this works fine. But for requests that don't do any real network work — memory/disk cache hits, `data:` URLs, etc. — the numbers I get out of `responseHeadersEndTime` don't make sense. Sometimes I get a large negative duration (looks like the field is still at its initial sentinel value), other times the value is just inconsistent with the request's other timings (e.g. it ends up earlier than `networkRequestTime`, or it's some timestamp that doesn't correspond to anything header-related given that nothing was ever fetched from the network).\n\nIt seems like the field is just not well-defined for requests where no bytes were ever received over the wire. Right now every downstream consumer that touches `responseHeadersEndTime` has to know \"oh, but if it's a cached request / data URL then this is garbage, special-case it\" — which is easy to get wrong and not really documented anywhere.\n\nIt'd be much nicer if `NetworkRequest` itself guaranteed a coherent value for `responseHeadersEndTime` in these cases — pick whatever fallback makes semantic sense (there was no network activity, so any timing for \"when headers finished arriving from the network\" is a bit fictional anyway), and document what that fallback means on the field itself so consumers don't have to reverse-engineer it. The important thing is that the field is always in a sensible relationship to the request's other timing fields, regardless of whether the request actually hit the network."} {"task_id": "format-code-task-000097", "source_id": "format-code-task-000097", "domain": "code", "task_path": "tasks/format-code-task-000097", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6ff021399027ac13493bc62c59bdcc5f40de3ea656c5bab31b19721ba8fdd8f1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nRemove standalone / fullscreen requirement\nThe requirement for apps to have a display property of 'standalone' or 'fullscreen', implemented as a result of #143, is problematic. Many in the web community object to this being considered a web best practice, notably articulated by Jeremy Keith in his [post](https://adactio.com/journal/10708), though I'm not endorsing all the sentiments he expresses there.\n\nFor me the question is whether Lighthouse is a community project designed to judge conformance with universally agreed best practices, or whether it is specifically designed to express the proprietary requirements of Google Chrome in respect to offering an install prompt. If the latter, then since the heuristics/requirements for installability are not subject to a standard, it's totally valid for Chrome to set whatever requirements it likes - though care should probably be taken to ensure that the intersection of all vendors' installability requirements don't leave developers with insufficient control over their own application design.\n\nHowever, if Lighthouse is a community tool, then this rule could be seen to be narrowing a recent specification that offers developers a range of options for good reasons. If options like 'browser' and 'minimal-ui' are bad, why are they in the manifest spec?\n\nIf this is deemed justfiable on the grounds that 'native apps don't have a URL bar so PWAs shouldn't either' I would point to zoomability as a similar case - we consider sites that disable zoom to be bad practice, despite pinch-zooming the whole UI not typically being a feature offered by native apps.\n\nThis topic came up as part of a discussion of PWAs on a [TAG thread](https://github.com/w3ctag/spec-reviews/issues/123).\n\ncc @marcoscaceres @torgo"} {"task_id": "format-code-task-000098", "source_id": "format-code-task-000098", "domain": "code", "task_path": "tasks/format-code-task-000098", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6d9f359d81f45e10d39168c0767ae25577b28f6d17cc7e6c3b99d19f7492480a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI'm seeing a couple of i18n formatting issues: when I forget to pass a value used by a UIString, `str_(...)` throws only `ICU Message contains a value reference (\"wastedBytes\") that wasn't provided`, and I can't tell which message it came from. I also noticed `{foo, number}` placeholders fed with `undefined` or a string end up formatting as `NaN`/garbage. In an older runtime using the Intl polyfills, plural ICU messages are failing with `Intl.PluralRules` missing.\n\n## Expected outcomes\n\n- Missing placeholder values: when formatting a localized UI string fails because a referenced placeholder value was not provided, the thrown error should identify both the missing reference and the ICU message that contained it.\n- Numeric placeholders: placeholders declared for number formatting should reject non-`number` JavaScript values before producing formatted output, including `undefined` and strings.\n- Numeric placeholder errors: failures for non-number numeric placeholder values should identify both the offending reference and the ICU message that contained it.\n- Plural formatting compatibility: in runtimes that rely on the project’s Intl polyfill setup and do not provide native plural rules support, ICU plural messages should still be format-capable after i18n initialization.\n- Existing formatting behavior: valid localized strings, including existing number formatting styles and plural messages with valid values, should continue to format successfully.\n\n## Implementation notes\n\n- The exact validation location, parser/formatter interaction, and data structures are implementation details.\n- Preserve the existing public i18n formatting entry points and observable formatting behavior for valid inputs.\n- Error text does not need to use a particular full sentence beyond clearly exposing the relevant ICU message and placeholder reference."} {"task_id": "format-code-task-000100", "source_id": "format-code-task-000100", "domain": "code", "task_path": "tasks/format-code-task-000100", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c69033924f2984abe48b7d34902ce4bdc0eca6f7ca08101779b3d655014b1ee4", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Splitting a benchmark across `--run_stage=prepare` and `--run_stage=run` fails when a flag default uses Pint units\n\nI'm trying to run a benchmark in two phases by invoking PerfKitBenchmarker once with `--run_stage=prepare` and then again with `--run_stage=run`, so that the pickled state from the prepare phase is picked up by the run phase.\n\nThis works for plain benchmarks, but as soon as the benchmark declares a flag whose default value is a Pint quantity (using `UNIT_REGISTRY` from `perfkitbenchmarker`) and I don't explicitly override it on the command line, the run phase blows up while loading the saved state. If I override the same flag on the command line so the default is never pickled, the two-stage run goes through fine.\n\nIt looks like quantities created against the prepare phase's `UNIT_REGISTRY` don't survive a round-trip through pickle into the run phase, which has its own registry instance. For users this is pretty surprising — the default value of a flag shouldn't break `--run_stage` separation just because it carries units.\n\nCould `perfkitbenchmarker` make Pint quantities safe to pickle/unpickle so that the prepare/run split works regardless of whether a units flag uses its default?"} {"task_id": "format-code-task-000101", "source_id": "format-code-task-000101", "domain": "code", "task_path": "tasks/format-code-task-000101", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b07c7a649c04b045b63606dca7e3d7fbd8b93f7c18d9b66fc6c385722230b6de", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Distinguish permission-denied failures from generic failures via exit code\n\nWhen `gcs-fetcher` fails because the service account doesn't have permission to read from the source bucket (e.g. missing Storage Object Viewer on the bucket, or VPC-SC blocking access), it exits with the same exit code as any other failure (network blip, checksum mismatch, etc.).\n\nFrom the outside there is no way to tell these two situations apart. In our build pipeline we'd like to react differently:\n\n- If it's a permission/config problem, the build should fail fast and surface a clear \"fix your IAM\" message to the user — retrying is pointless.\n- If it's a transient/other error, we want to retry or fall through to our normal failure handling.\n\nRight now both cases just give us a non-zero exit and we have to scrape stderr to guess which one happened, which is brittle.\n\nIt would be very useful if `gcs-fetcher` used a dedicated exit code when the underlying cause is a GCS permission-denied error, distinct from the generic-failure exit code. This should apply regardless of which source type is being fetched (Manifest, ZipArchive, TarGzArchive) — today only the manifest path even prints the helpful \"Access to bucket … denied\" message before exiting, and even there the exit code is indistinguishable from any other error.\n\nIt would help if the new permission-denied exit code were exposed as a package-level variable (something like `permissionDeniedExitStatus`) so callers and tests can reference it by name."} {"task_id": "format-code-task-000102", "source_id": "format-code-task-000102", "domain": "code", "task_path": "tasks/format-code-task-000102", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4fd7f890d78d050c5505fed18089f91a9e48610085b8883956a4a2fac0fe9826", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nValidate email address format when using automatic IAM authentication\nWhen adding IAM database users to either MySQL or Postgres the user or service account username is formatted in a certain way that differs across database engine types.\n\nThis is not necessarily trivial to users when using the connectors, we should try validating and formatting the IAM database usernames within the connectors to remove this burden on users.\n\n**Postgres**: Removes `.gserviceaccount.com` domain suffix if it exists.\n**MySQL**: Removes everything after and including the `@` sign. (`test-user@gmail.com` -> `test-user`)\n\nShould already have `databaseVersion` from the API request to https://cloud.google.com/sql/docs/mysql/admin-api/rest/v1beta4/connect/get"} {"task_id": "format-code-task-000103", "source_id": "format-code-task-000103", "domain": "code", "task_path": "tasks/format-code-task-000103", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ce6a14e8770b45cfe55150ae7fea639a567e6dbede5bf58e521398e72137e283", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `FetchServiceImageMetaData` on `ConfigAgent` is a pure passthrough — remove it and call dbdaemon directly\n\nLooking at how `isImageSeeded` (in the instance controller) gets the service image metadata today, it goes through `ConfigAgent`:\n\n```\nisImageSeeded\n -> ConfigAgent client.FetchServiceImageMetaData\n -> dbdaemon client.FetchServiceImageMetaData // actually does the work\n```\n\nBut if you check the `ConfigAgent` server implementation of `FetchServiceImageMetaData`, it doesn't do anything meaningful: it just dials dbdaemon, forwards the request, and repacks the response field-for-field into the equivalent `ConfigAgent` proto. No additional logic, no enrichment, no decision-making — purely a wrapper.\n\nThis forces us to maintain a redundant set of `FetchServiceImageMetaDataRequest` / `FetchServiceImageMetaDataResponse` messages on the `ConfigAgent` proto that mirror the dbdaemon ones one-to-one, plus an extra network hop and an extra RPC surface to keep in sync. The `ConfigAgent` is supposed to host orchestration / business logic that sits above raw dbdaemon calls; a thin forwarder like this doesn't belong there.\n\n`isImageSeeded` is currently the only caller of the `ConfigAgent` version, and the instance controller already has a `DatabaseClientFactory` available for talking to dbdaemon directly. So the `ConfigAgent` layer here isn't even buying us a stable abstraction for multiple callers — it's dead weight on a single call site.\n\nI'd like to remove `FetchServiceImageMetaData` from the `ConfigAgent` surface entirely and have `isImageSeeded` talk to dbdaemon directly via the database client. The behavior observed by the controller should be identical (same metadata, same `SeededImage` interpretation), it's just one less hop and one less proto pair to maintain."} {"task_id": "format-code-task-000104", "source_id": "format-code-task-000104", "domain": "code", "task_path": "tasks/format-code-task-000104", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b54d400b21c23f77cfddb955fbdd191ed31af6803d0c37c21888bf556292d332", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Object properties (Content-Type, Cache-Control, custom metadata, ...) are lost after editing a file through gcsfuse\n\nI've been using gcsfuse to mount a bucket where the objects already have meaningful properties set on them — things like `Content-Type`, `Cache-Control`, `Content-Encoding`, plus a few custom metadata entries we use for tracking.\n\n### What I'm seeing\n\n1. Upload an object to the bucket with properties set (via the console / gsutil / API). For example a `.html` file with `Content-Type: text/html`, `Cache-Control: public, max-age=3600`, and a custom metadata pair like `owner=alice`.\n2. Mount the bucket with gcsfuse and edit that file through the mountpoint (any modification — append, rewrite, whatever ends up syncing a new generation).\n3. Go look at the object again in the GCS console.\n\nAfter step 3, the only metadata still present is the `gcsfuse_mtime` key that gcsfuse itself writes. `Content-Type` has reverted to the default, `Cache-Control` is gone, and my custom metadata entries are gone too.\n\n### What I expected\n\nEditing a file through the mount shouldn't silently wipe out the object's existing properties. The new generation should carry over whatever was set on the previous generation; gcsfuse is the one writing the new version, so it owns preserving that state. Only the mtime is something gcsfuse legitimately needs to update on its own.\n\nThis affects both flows that produce a new generation — full rewrites of small/changed files, and the compose-based \"append\" optimization for larger files. In both cases the resulting object comes back stripped.\n\nCould the syncer be fixed so that the properties on the source object are retained on the new generation it writes?"} {"task_id": "format-code-task-000105", "source_id": "format-code-task-000105", "domain": "code", "task_path": "tasks/format-code-task-000105", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:13a78ffaf2a5084faeac5b58b0c54823c7f97427724b7b44f73a00b2c5f3937a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nCopying symlink from previous stage fails instead of overwriting\n**Actual behavior**\n`COPY --from=0 X Y` where X is a symlink from stage 0 and Y is an pre-existing symlink in the current stage fails with \"file exists\".\n\n**Expected behavior**\n`COPY --from=0 X Y` where X is a symlink from stage 0 and Y is a pre-existing symlink in the current stage should overwrite Y with X as it would if neither were symlinks.\n\n**To Reproduce**\n\n```bash\ndocker run --rm -v $PWD:/workspace:ro gcr.io/kaniko-project/executor:v0.6.0 --no-push\n```\nwith the following Dockerfile:\n```dockerfile\nFROM python:slim\n\nFROM python:slim\nCOPY --from=0 /usr/local/lib/libpython3.7m.so /usr/local/lib/libpython3.7m.so\n```\n\nAlthough this is a bit contrived, we had a Dockerfile which was building into /usr/local/lib and then copying that into the final stage which seems like a legitimate use-case."} {"task_id": "format-code-task-000106", "source_id": "format-code-task-000106", "domain": "code", "task_path": "tasks/format-code-task-000106", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d96a9622f36e8576281a7c840ca9d9c25c87e3e8d4cd7f249b101b861bc3a844", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI have a custom artifact where the files Skaffold should watch come from our own build tooling, so keeping `dependencies.paths` updated by hand doesn’t really work. Could Skaffold run a script/command for that artifact and use the file list it returns as the dependencies for watching and rebuilds? Ideally it would give a clear error if the command fails or prints something Skaffold can’t understand.\n\nExpected outcomes:\n- Custom artifact configuration supports `dependencies.command`, exposed through `CustomDependencies.Command`, as an alternative dependency source for determining the files used by watching and rebuild decisions.\n- When `dependencies.command` is configured, Skaffold runs that command during custom artifact dependency resolution and uses its output as the dependency list.\n- The command output must be a valid JSON array of strings; invalid JSON or JSON with a different shape should fail dependency resolution with a clear error.\n- If the dependency command exits unsuccessfully, dependency resolution should fail with an error that makes the failing command context clear.\n- `dependencies.ignore` remains valid with `dependencies.paths`, but is invalid when combined with `dependencies.command`.\n- The public schema and custom builder documentation describe the new `dependencies.command` option and its JSON-array output contract.\n\nImplementation notes:\n- The exact execution helper, parsing location, validation organization, and internal data flow are up to the implementer.\n- Preserve existing dependency behavior for path-based and Dockerfile-based custom artifacts.\n- Error text does not need to match any particular sentence exactly, but it should be actionable enough to distinguish command execution failures from malformed command output."} {"task_id": "format-code-task-000107", "source_id": "format-code-task-000107", "domain": "code", "task_path": "tasks/format-code-task-000107", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:17db0205ebae2dcee4fe9cd60a6fdd1ee563680579d4acce990f28e0e8920323", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[Custom builder] $IMAGES is always a single image\nWe currently pass the build script the `IMAGES` env variable. This variable contains the list of images to build. However, this variable always contains a single tag.\n\nI think we should deprecate this value and use `IMAGE` instead."} {"task_id": "format-code-task-000108", "source_id": "format-code-task-000108", "domain": "code", "task_path": "tasks/format-code-task-000108", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7f06f9ec176039ccd5675da88380a7f62d7394b9717af6b97dd287b3bd022a6e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nHandle minikube exit code 89\nMinikube has well-defined exit codes, and Skaffold has special handling for turning minikube exit codes into actionable errors. This code needs to be updated as it's not handling minikube exit code 89:\n\n```\ninvalid skaffold config: getting minikube env: running [/Users/bdealwis/installs/google-cloud-sdk/bin/minikube docker-env --shell none -p minikube --user=skaffold]\n - stdout: \"false exit code 89\\n\"\n - stderr: \"\"\n - cause: exit status 89\n```\n\nBasically minikube is stopped.\n\n- platform: macOS 12.3.1 on arm64\n- `skaffold version`: v1.37.1 (installed via gcloud)\n- `minikube version`: v1.25.2 (installed via gcloud)"} {"task_id": "format-code-task-000109", "source_id": "format-code-task-000109", "domain": "code", "task_path": "tasks/format-code-task-000109", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:44fa53b2c677e24a2d7af58ff9af5b9d4b92a78c63d324a2542a8a1c4a82ff72", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n\n### Related issue\n - #7644\n\n### Expected behavior\n - deployed resources should be cleaned up with ``skaffold delete`` command\n\n### Actual behavior\n - deployed resources are not released after running ``skaffold delete`` command\n\n\n### Steps to reproduce the behavior\n\n1. checkout main branch, and run make to build binary. \n2. run ``skaffold run`` in ``example/getting_started`` folder,\n3. run ``skaffold delete`` to delete resources. \n4. run ``kubectl get pod``, should see ``getting_started`` pods are still running. However, the expected behavior is all resources in the run should be cleaned up. \n\n### information \n - Kubeclt deployer works for v1 as its clean up method is able to read the manifests configured for this deployer by calling Dependencies() method, so cleanUp method is able to release resource defined in those manifests. \n - In v2, manifests are not defined under deployer, and schema upgrade will nil out kubectl deployer for old version schema as well. Cleanup method is not calling Dependencies(). However, even we call Dependencies() method, it returns nothing, so it won't fix the problem. \n - As team discussed, we've decided to take advantages of manifests config in skaffold.yaml and use the rendered manifests as source data for resource deletion in v2. The implementation will #7644 as well\n - The engineering work may be easier after #7572 merged into main.\n\nNote: I'd expect the `Runner.Render` entrypoint to be reused here, and `Cleanup` to take something like a `*manifest.ManifestListByConfig` so per-config rendered manifests can flow through."} {"task_id": "format-code-task-000110", "source_id": "format-code-task-000110", "domain": "code", "task_path": "tasks/format-code-task-000110", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:73c51c6bcc159c75a12110d77e2a379dcbb68af16798259578598546132f5ca2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[apps.paint_annotations] Add option to filter iscrowd\nThe `iscrowd` attribute present in COCO annotations is currently being ignored and `crowd` boxes are being painted (the big purple sheep box):\n\n![000000545959_result](https://user-images.githubusercontent.com/12677733/107704160-1b71e580-6cbd-11eb-9e56-e964b6e2aec9.jpg)\n\nIt would be nice to add an optional argument to filter out this annotations (possibly with default True)."} {"task_id": "format-code-task-000111", "source_id": "format-code-task-000111", "domain": "code", "task_path": "tasks/format-code-task-000111", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:91fc2eabe54156d7e7ffeb84baffd9f4fd82ec4776b09b3c8aa029fc89023bfe", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## DatePicker crashes when used with a customInput that doesn't pass a standard event\n\nI'm using `` with a `customInput` (a wrapper input component from another UI lib). When the user types into it, that component fires its `onChange`, but the object it passes isn't a regular React SyntheticEvent — it's a plainer object that doesn't carry every method a native event has.\n\nAs soon as I type a character, the datepicker blows up inside `handleChange`. I traced it down to react-datepicker calling something on the event object that isn't there on what my customInput hands over.\n\nRepro shape (roughly):\n\n```jsx\nconst MyInput = React.forwardRef((props, ref) => (\n props.onChange({ target: { value } })}\n {...props}\n />\n));\n\n}\n/>\n```\n\nThe `{ target: { value } }` object is enough for me to read what the user typed, but react-datepicker treats it like a full SyntheticEvent and breaks on it.\n\nI'd expect `handleChange` to be defensive about what the underlying input hands it — if the argument isn't a \"real\" event, it should just keep going instead of crashing. The native-input case obviously still has to keep working as before.\n\nSeparately, while looking into this I noticed `onChangeRaw` only ever receives the first argument that the input passed in. Some inputs invoke `onChange(value, meta)` style and those extra arguments are silently dropped before reaching my `onChangeRaw` handler. It'd be nice if whatever the input calls `onChange` with gets forwarded through to `onChangeRaw` intact."} {"task_id": "format-code-task-000112", "source_id": "format-code-task-000112", "domain": "code", "task_path": "tasks/format-code-task-000112", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c7b4e2cf0fb72a320dd2a62a3e924e1e3ab9556cba8c93d8027faaf7c4821529", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nUsing convertToHTML returning an tag in blockToHTML gives error\nI get the following error when I try to use `draft-convert` convertToHTML returning an `` tag in blockToHTML gives error\n\n```\nUncaught Error: img is a void element tag and must neither have 'children' nor use 'dangerouslySetInnerHTML'\n```\n\nHere is my function for exporting to html.\n\n```jsx\nexport function editorStateToHtml (editorState) {\n if (editorState) {\n const html = convertToHTML({\n styleToHTML: (style) => {\n if (style === 'BOLD') {\n return ;\n }\n },\n blockToHTML: (block) => {\n const type = block.type\n if (type === 'atomic') {\n let url = block.data.src\n return \n }\n if (type === 'unstyled') {\n return

\n }\n },\n entityToHTML: (entity, originalText) => {\n if (entity.type === 'LINK') {\n return {originalText};\n }\n return originalText;\n }\n })(editorState.getCurrentContent());\n\n return html\n }\n}\n```"} {"task_id": "format-code-task-000113", "source_id": "format-code-task-000113", "domain": "code", "task_path": "tasks/format-code-task-000113", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:241b4d1ab25e07bfaab833c028a53faa95df25201f12aabd9c5a57e5f1a5b294", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Job logs stop being written after a single write failure\n\nI'm running jobs through mist and reading their logs via the async log streaming. For most jobs this works fine, but every once in a while I notice that a job's log file just stops growing partway through the job — even though the job itself keeps running and producing output. When I look later, the tail of the log for that job is just missing.\n\nAfter staring at this for a while I think I've narrowed it down to what happens when writing a batch of log events fails once (e.g. a transient IO hiccup on the log directory, or anything that makes `LogsWriter.write` fail for a single batch). From that point on, no further log events for that job make it to disk — it's as if the per-job log pipeline is shut down by the first failure and never recovers, even though new `LogEvent`s are still arriving for the same job.\n\nThe expected behavior is that a single failed write shouldn't kill log storage for that job. If one batch fails, fine — drop/skip that batch, but subsequent batches for the same job should still be written and still produce update events for the async consumers. Right now one bad batch effectively blackholes all remaining logs for that job until the job finishes.\n\nThis is happening in the `storeFlow` in `LogStreams` (the `mapAsync` that calls `writer.write(jobId, events)`), if that helps narrow it down."} {"task_id": "format-code-task-000114", "source_id": "format-code-task-000114", "domain": "code", "task_path": "tasks/format-code-task-000114", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:aa4529fe5287eae9fb258a31bc63b96e2f941441caa17155cddac22dbf3e3d2a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\n我这边用 `st.builds()` 生成一个 `typing.NamedTuple` 子类时，明明字段都写了类型注解，它还是没法自动推断构造参数，感觉像是把这个类的初始化签名看丢了。比如 `Point(NamedTuple)` 里有 `x: int`、`y: int`，我希望直接 `st.builds(Point)` 就能生成正常的 `Point` 实例，部分字段我手动传了的话也只补剩下的就好。\n\n## Expected outcomes\n\n- For a `typing.NamedTuple` subclass with annotated fields, `hypothesis.strategies.builds(target, *args, **kwargs)` should be able to infer strategies for missing constructor fields from those field annotations and generate valid instances of the NamedTuple subclass.\n- When some NamedTuple constructor fields are supplied positionally or by keyword to `hypothesis.strategies.builds(target, *args, **kwargs)`, those supplied fields should be treated as already provided; inference should only be required for the remaining annotated fields.\n- `hypothesis.strategies.from_type(thing)` should continue to resolve annotated `typing.NamedTuple` subclasses to strategies that generate valid instances of that subclass.\n\n## Implementation notes\n\n- The concrete detection mechanism, data structures, and validation location are left to the implementation, as long as the public strategy APIs above exhibit the expected behavior.\n- Existing behavior for non-`typing.NamedTuple` targets should be preserved."} {"task_id": "format-code-task-000115", "source_id": "format-code-task-000115", "domain": "code", "task_path": "tasks/format-code-task-000115", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:880163cb64f86c13607813e5247ce6cf1908d884ed489c49aefd9f8b96239c63", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `@settings` decorator is not typed, breaks `mypy --strict`\n\nI'm trying to type-check my test suite under `mypy --strict`, and the `@settings(...)` decorator from Hypothesis causes errors on every test it's applied to.\n\nA minimal reproducer:\n\n```python\n# test_example.py\nfrom hypothesis import given, settings, strategies as st\n\n@settings(max_examples=10)\n@given(st.integers())\ndef test_addition_is_commutative(x: int) -> None:\n assert x + 0 == x\n```\n\nRunning `mypy --strict test_example.py` complains about the `@settings(...)` line — mypy treats it as an untyped decorator, which under `--strict` poisons the decorated function (it gets reported as untyped too, even though I've annotated it). The same code without `@settings` (i.e. only `@given(...)`) type-checks fine.\n\nI'd expect `@settings(...)` to be a transparent, type-preserving decorator — applying it to a function should not change the function's type as far as the type checker is concerned. As far as I can tell from the public API, `settings(...)` returns a callable that takes a test function and returns the same test function (it just attaches some metadata), so this should be expressible without losing the original signature.\n\nCould the `@settings` decorator get proper type annotations so that it works cleanly under `mypy --strict`?"} {"task_id": "format-code-task-000116", "source_id": "format-code-task-000116", "domain": "code", "task_path": "tasks/format-code-task-000116", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5191cba7cf2ae83979e1f2b4460656d03341fb022b6147bb02072ab355d3a457", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nThe docs for `register_type_strategy` say that instead of registering `tuple[int]`, I can register `tuple` with a function that takes the type and return the strategy. I may be misunderstanding, but I take this to mean that I can register\n`st.register_type_strategy(tuple, lambda typ: foo)`\nand then `foo` should be the default strategy for _any_ param of type tuple, or `tuple[int]`, or `tuple[....]`, etc.\n\nHowever, the code below results in a test failure, and from my investigation it is clear that the registration is not effective, and that `test` is getting any old `tuple[int]` rather than using the strategy that was registered.\n\n```python\nfrom hypothesis import given, strategies as st\n\n# Note that I'm not actually using the typ param here, but I do use it in my actual code\nst.register_type_strategy(tuple, lambda typ: st.tuples(st.integers(min_value=0)))\n\n@given(x=...)\ndef test(x: tuple[int]):\n assert x[0] >= 0\n\ntest()\n```\n\nI'm not sure if I'm using this correctly or if I understand the correct behavior. Is it simply impossible to register a default strategy for tuples that will work on parameterized tuples? If so, the documentation seems to suggest otherwise to me. Also, this seems like a useful feature to have. Or, is it that I am doing something wrong?\n\nThanks!"} {"task_id": "format-code-task-000117", "source_id": "format-code-task-000117", "domain": "code", "task_path": "tasks/format-code-task-000117", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d399cf81531fbbaff66230c2417fa781c8392f7bb9d1478ef170d52b92f771f1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Manage custom nameservers for a registered DNS domain\n\nOur users register DNS domains through IBM Cloud (SoftLayer) and frequently need to point a\nregistered domain at a custom set of nameservers from Terraform. Today the provider has no way to\ndiscover a domain registration or to manage its nameservers. Add that capability.\n\n## Data source: `ibm_dns_domain_registration`\n\nLooks up a domain registration by its name.\n\n- Input argument `name` (string, **required**): the registered domain name to look up.\n- Exported attribute `name_servers` (**computed**): the list of nameserver hostnames currently\n configured on the registration, as strings.\n\n## Resource: `ibm_dns_domain_registration_nameservers`\n\nManages the custom nameservers of an existing domain registration.\n\n- Argument `dns_registration_id` (string, **required**): the identifier of the domain registration\n to manage.\n- Argument `name_servers` (**required**): the desired collection of nameserver hostnames (strings)\n the registration should use.\n- Exported attribute `original_name_servers` (**computed**): the nameservers that were present\n before this resource took over management, so they can be restored later.\n\n## Nameserver validation\n\nEvery value supplied in `name_servers` must be a syntactically valid DNS hostname. Validation must\nhappen during configuration validation (i.e. before any API call is made), so that an invalid\nvalue causes a validation error rather than being sent to the backend. A value is valid only if all\nof the following hold:\n\n- it is non-empty and at most 253 characters long;\n- it consists of dot-separated labels and contains at least one dot (at least two labels);\n- every label is between 1 and 63 characters long;\n- every label contains only ASCII letters, digits, and hyphens; and\n- no label begins or ends with a hyphen.\n\nAny value that violates one or more of these rules must produce a validation error for the\n`name_servers` argument; values that satisfy all of them must validate cleanly.\n\nThe provider must continue to pass its internal schema validation with these additions in place.\n"} {"task_id": "format-code-task-000118", "source_id": "format-code-task-000118", "domain": "code", "task_path": "tasks/format-code-task-000118", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:8a9c417e8edbb35e8fce0b69d414a66f65f9dcf2e590aad579725a52f39263d0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nIs there a way to assign IAM policy templates to target accounts using the IBM provider? I've got policy templates set up and I need to roll them out to specific accounts via Terraform, but I can't find a resource for it. Ideally I'd be able to specify which template version goes to which account and then reference back the assignment details (like the account ID, when it was created, and the status of what got created) in my outputs. It'd also be nice if I could point acceptance tests at a target account through an env var.\n\n# Expected outcomes\n\n- Terraform resource support:\n - Terraform configurations can declare an `ibm_iam_policy_assignment` managed resource without the provider rejecting the resource type as unsupported.\n - The resource lets users assign an IAM policy template/version to a target account using Terraform configuration.\n\n- Resource input validation:\n - `version` is a required string input.\n - `target` is a required map input for the assignment target.\n - `templates` is required and accepts exactly one item; that item requires `id` and `version`.\n - `options` is required and accepts exactly one item; its nested `root` block is required and includes a required `requester_id`.\n - Missing required inputs, empty required blocks, or more than one `templates`/`options` item are rejected during Terraform validation or planning.\n - `accept_language` is optional and defaults to `default` when omitted.\n\n- Resource state and outputs:\n - After a successful apply/read, `ibm_iam_policy_assignment` exposes assignment metadata that can be referenced in Terraform outputs, including `account_id`, `href`, `created_at`, `created_by_id`, `last_modified_at`, and `last_modified_by_id`.\n - The resource also exposes a computed `resources` list describing created assignment resources, including each item’s `target` and nested policy result information such as the created policy ID, status, and any error message details.\n\n- Acceptance-test configuration:\n - Acceptance-test setup can read a target account from the `IBM_POLICY_ASSIGNMENT_TARGET_ACCOUNT_ID` environment variable.\n - When that environment variable is not set, the acceptance-test setup emits an informational message telling users to set it for `ibm_iam_policy_assignment` tests.\n\n# Implementation notes\n\n- Follow the provider’s existing conventions for Terraform managed resources, schema validation, state population, import/read/delete behavior, diagnostics, and acceptance-test configuration.\n- The concrete internal structure, helper functions, mapping code, and validation placement are up to the implementation as long as the externally observable Terraform behavior above is satisfied.\n- Tests should validate behavior through Terraform configuration/schema outcomes, resource state, diagnostics, and acceptance-test setup behavior rather than relying on private helper names or internal call paths."} {"task_id": "format-code-task-000119", "source_id": "format-code-task-000119", "domain": "code", "task_path": "tasks/format-code-task-000119", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f3640b92ffdfa97570fac3bb9dc014e9b682256527ce3507d9eaed0410ef62ba", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Missing wrapper for `TNoOp` in `lale.lib.autoai_libs`\n\nI'm building pipelines that use the schema-enhanced wrappers under `lale.lib.autoai_libs`. The module already exposes wrappers for the whole family of feature-transform / feature-selection operators that live in `autoai_libs.cognito.transforms.transform_utils` — `TA1`, `TA2`, `TB1`, `TB2`, `TAM`, `TGen`, `FS1`, `FS2` are all there.\n\nHowever, the `TNoOp` operator from that same module isn't exposed:\n\n```python\nfrom lale.lib.autoai_libs import TA1, TA2, TNoOp # TNoOp import fails\n```\n\nThe upstream `autoai_libs.cognito.transforms.transform_utils.TNoOp` is a transformer that passes data through unchanged, which is genuinely useful when composing pipelines (e.g. as a placeholder branch, or as a baseline to compare other feature-transform branches against). Today the only way to use it from a Lale pipeline is to grab the raw `autoai_libs` class, which means I lose the schema-validation / hyperparameter-search story that every other wrapper in this module gives me.\n\nCould you add a Lale-wrapped `TNoOp` alongside its siblings so it can be imported from `lale.lib.autoai_libs` and dropped into a pipeline the same way the other `T*` / `FS*` wrappers can?"} {"task_id": "format-code-task-000121", "source_id": "format-code-task-000121", "domain": "code", "task_path": "tasks/format-code-task-000121", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:85e99dbe4a8e4136fd93050e6c3a18b1decd7806bf1c55bec1d73152919e5ea8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI'm seeing a weird imbalance with Sarama's round-robin consumer group balancing: if I have several topics with only one or a few partitions each, and the same consumers subscribe to all of them, the first partitions keep landing on the same consumer while others sit idle. Is the round-robin strategy supposed to spread partitions across topics too? It would be great if those topic partitions were distributed more evenly across the group instead of restarting the rotation for each topic.\n\n# Expected outcomes\n\n- Round-robin planning through `BalanceStrategyRoundRobin.Plan(...)` should distribute partitions across topics as part of one continuous round-robin assignment, rather than starting the member rotation over independently for each topic.\n- When multiple members subscribe to the same small-partition topics, partitions from different topics should be spread across the eligible members as evenly as the round-robin sequence allows.\n- When members have different topic subscriptions, `BalanceStrategyRoundRobin.Plan(...)` should still keep the round-robin progression across topic partitions and assign each partition to an eligible subscribed member, skipping members that are not subscribed to that partition’s topic.\n\n# Implementation notes\n\n- Preserve the existing public strategy surface and observable behavior outside the round-robin assignment semantics described above.\n- The concrete data structures, helper functions, and exact location of the assignment logic are implementation choices, as long as the public round-robin plan results match the expected behavior."} {"task_id": "format-code-task-000122", "source_id": "format-code-task-000122", "domain": "code", "task_path": "tasks/format-code-task-000122", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b6a3eade23a0a3814976f10f3fe08961353b2c518b825d0644b1375438573537", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI'm hitting a circular import the moment I try to use anything from `idaes.core.components` on its own — importing a Component should not require the generic property package dependency chain to be pulled in first. I shouldn't have to drag in the whole generic property package just to define a component. It would be great if `idaes.core.components` could stand on its own while still supporting parameter values populated from component configuration data.\n\n# Expected outcomes\n\n- `idaes.core.components` can be imported and used independently without depending on the generic property-package utility path that participated in the circular import.\n- A core-level public utility is available as `idaes.core.util.misc.set_param_from_config(block, parameter_name, config=None, index=None)` for setting a parameter-like Pyomo object on a block from parameter data stored on a configuration block.\n- The parameter-population behavior preserves the existing user-facing semantics for scalar parameter data, unit-aware tuple data, explicit versus default configuration blocks, indexed parameter entries, and clear errors for common misuse such as missing configuration, invalid configuration objects, missing target parameters, or missing parameter-data entries.\n\n# Implementation notes\n\n- Keep the dependency direction such that core component definitions do not need to import generic property-model modules.\n- The exact internal organization, helper boundaries, and validation placement are up to the implementer, as long as the public behavior above is preserved.\n- Prefer behavior-preserving changes over broad rewrites of property models or component APIs."} {"task_id": "format-code-task-000123", "source_id": "format-code-task-000123", "domain": "code", "task_path": "tasks/format-code-task-000123", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1e67b18a0ce0f7f0cc511efbace9e412de8947e17db2f21384e6e6ebdb7b4cea", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Updating a formatted field — the new value can't be found by `filterByText` over all fields\n\nI have an `igDataSource` (used by igGrid) where one of the schema fields has a `formatter` function set up — in my case it's a number column that gets formatted into a currency string for display.\n\nRoughly the setup looks like this:\n\n```js\nvar ds = new $.ig.DataSource({\n dataSource: products,\n schema: {\n fields: [\n { name: \"ProductID\", type: \"number\" },\n { name: \"Name\", type: \"string\" },\n {\n name: \"Price\",\n type: \"number\",\n formatter: function (val) { return \"$\" + val.toFixed(2); }\n }\n ]\n }\n});\nds.dataBind();\n```\n\nWhen the data is first loaded, `filterByText` over all fields works as expected — searching for the formatted string of the `Price` column (e.g. `\"$19.99\"`) matches the row that has `Price: 19.99`.\n\nThe problem shows up after I edit a formatted field. If I call `updateRow` to change `Price` on some row from `19.99` to `42.50` and commit the transaction, then call `filterByText(\"$42.50\")` over all fields, **the updated row is not returned**. The old value still matches (until I commit), and the new formatted value matches nothing — even though the underlying data has clearly been updated (I can see the new value in the grid, and filtering on the raw `Price` field directly works fine).\n\nSo `filterByText` across all fields seems to be searching against a stale formatted representation of the rows for any field that has a `formatter` — initial bind is fine, but after a commit on an updated row the formatted view for that row is out of sync with the actual data.\n\nCould `filterByText` over all fields stay consistent with the current state of the data source after edits on formatted columns? Right now the only workaround I have is to re-bind the data source after every commit, which obviously defeats the point of transactions."} {"task_id": "format-code-task-000124", "source_id": "format-code-task-000124", "domain": "code", "task_path": "tasks/format-code-task-000124", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:dae89a9816f6366a81d3fb4b50863124e249a9bdfe39d30782d83d1b169f81cc", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a read API for model stages\n\nOur backend already persists **model stages** — each one is a named checkpoint that\nbelongs to a model (the stage carries a `name`, a `timestamp`, an optional `map`\nscore, and the id of the model it belongs to). The database table and ORM model are\nalready in place, but there is currently no HTTP API to look a stage up, and nothing\nguards against stages whose names are malformed. We need to expose model stages over\nthe same versioned REST API the rest of the app uses (`/api/v1`).\n\nPlease add a read API for model stages, mounted under `/api/v1/model_stages`, with the\nfollowing behaviour. Like the other resource endpoints, these require an authenticated\nactive user.\n\n## Fetch a single stage\n\n`GET /api/v1/model_stages/{stage_id}` returns the stage wrapped in the usual envelope:\n\n```json\n{\"result\": { ...stage fields... }}\n```\n\nThe returned stage must expose at least its `id`, `name`, `map`, and `model_id`, plus a\nnested `model` object exposing the owning model's `id` and `hash`.\n\n- If no stage has that id, the request fails as *not found* (HTTP 404) and the response\n body carries the error code **112001**.\n\n## Validate stage names on lookup\n\nA stage name is only considered valid when it is a non-empty string that forms a valid\nidentifier: it starts with a letter or an underscore and contains only letters, digits,\nand underscores (e.g. `best`, `stage_1`, `_tmp` are valid; `\"\"`, `1stage`, `with space`,\n`a-b` are not).\n\nWhen a single stage is fetched by id and its stored name is not valid, the request must\nfail with the error code **112002** (a distinct outcome from the not-found case above),\nrather than returning the stage.\n\n## Fetch several stages at once\n\n`GET /api/v1/model_stages/batch?ids=1,2,3` returns\n\n```json\n{\"result\": [ ...stages... ]}\n```\n\ncontaining exactly the stages whose ids exist, in any order; ids that match no stage are\nsimply omitted (an all-missing request yields an empty list). This batch endpoint returns\nthe matching stages without rejecting the request.\n"} {"task_id": "format-code-task-000125", "source_id": "format-code-task-000125", "domain": "code", "task_path": "tasks/format-code-task-000125", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:589f81aecd34a1f7f4a56b41ce33757450f7db9b15fa38e3a606156e3ce411c8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `parse_module` chokes on f-strings containing a generator expression\n\nI'm using LibCST to walk through a bunch of existing Python source files (building a small refactoring tool). On one of the files the parse just blows up, but the file itself is valid Python — `python -c` runs it fine and `ast.parse` is happy with it.\n\nI narrowed it down to f-strings that contain a generator expression inside the `{...}` part. Minimal repro:\n\n```python\nimport libcst as cst\n\nsource = 'x = f\"{sum(i for i in range(10))}\"'\ncst.parse_module(source)\n```\n\nThe same string parses fine with the stdlib:\n\n```python\nimport ast\nast.parse('x = f\"{sum(i for i in range(10))}\"') # OK\ncompile('x = f\"{sum(i for i in range(10))}\"', \"\", \"exec\") # OK\n```\n\nA bare generator expression wrapped in its own parens inside the f-string hits the same problem:\n\n```python\ncst.parse_module('x = f\"{(i*2 for i in range(3))}\"')\n```\n\nBoth of these are accepted by CPython itself, so I'd expect LibCST to handle them too. Right now anything with a `for ... in ...` clause inside an f-string expression slot seems to be rejected at parse time, which means I can't run my tool over files that happen to use this pattern.\n\nCould the f-string expression parser be updated to accept the same expression forms that CPython does here?"} {"task_id": "format-code-task-000126", "source_id": "format-code-task-000126", "domain": "code", "task_path": "tasks/format-code-task-000126", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2e8c08eee42d867ac90d079d37f476b1a67579508fc181b53321a93a674ff4a7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI'm trying to train the conf_branch method with a resnet18 backbone on CIFAR and it blows up with a shape mismatch as soon as it hits the forward pass. The configs/train_conf_branch.yml as-is doesn't seem to work for anything other than whatever tiny backbone it was originally written against. Also while poking around I noticed `postprocessor.name: opengan` just throws a KeyError saying it's not registered, and `trainer.name: opengan` only works if I spell it `openGan` which is weird. Can you take a look at the conf_branch + opengan pipelines end to end?\n\n# Expected outcomes\n\n- The conf_branch network wrapper should run a forward pass successfully with compatible CIFAR-style ResNet backbones and other compatible backbones with nontrivial feature/output dimensions.\n- `ConfBranchNet.forward(x)` should produce class predictions with shape `(batch_size, num_classes)`.\n- `ConfBranchNet.forward(x, return_confidence=True)` should return both class predictions and confidence scores, with confidence shaped `(batch_size, 1)`.\n- The conf_branch training configuration should use a recorder that is available through the normal configuration-driven recorder construction path so that the training pipeline can be constructed from the provided config.\n- `get_postprocessor` should accept `config.postprocessor.name: opengan` and construct the OpenGAN postprocessor.\n- `get_trainer` should accept `config.trainer.name: opengan` using the same lowercase spelling used in OpenGAN configs.\n- Conf_branch training epoch metrics should include the training loss in addition to accuracy and epoch index.\n\n# Implementation notes\n\n- The exact internal wiring, helper structure, and validation locations are up to the implementer.\n- Prefer behavior-compatible fixes that preserve existing public APIs and configuration-driven construction patterns.\n- Do not special-case a single backbone or dataset; the conf_branch path should work for compatible backbones that expose the representation needed by the wrapper."} {"task_id": "format-code-task-000127", "source_id": "format-code-task-000127", "domain": "code", "task_path": "tasks/format-code-task-000127", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c880bca682f9acb39abc0c86edce758978214cb874c21a323e68f592a4e540d4", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `make_grid` 的 `normalize=True` 对单张图不生效\n\n我在用 jittor 训一个生成模型，每隔若干步把当前生成出来的图存盘看一下效果。生成出来的张量数值范围不是 [0, 1]（大致在 [-1, 1] 左右），所以我用 `make_grid` / `save_image` 时传了 `normalize=True`，希望它在保存前把数值缩放到 [0, 1]。\n\nbatch 内有多张图时一切正常，存出来的拼图颜色看上去是对的。\n\n但是当我只想看单张图的时候（比如只生成一个样本，或者 debug 阶段只 forward 一张），保存出来的图整张要么全黑要么全白，完全没有被归一化。我换了几种 shape 试了一下：\n\n```python\nimport jittor as jt\nfrom jittor.misc import make_grid\n\n# 一个 batch=1 的图像，数值范围不在 [0,1]\nx = jt.randn(1, 3, 64, 64) * 5.0\n\nout = make_grid(x, normalize=True)\nprint(out.min(), out.max()) # 我期望接近 0 和 1，但实际还是原始范围\n```\n\n而把 batch 改成 ≥2 之后，`out` 的数值范围就被正常归一到 [0, 1] 了。所以 `normalize=True` 这个参数在单图场景下相当于没生效，结果跟我不传 `normalize` 一样。\n\n希望 `normalize=True` 不论输入是单张还是多张都能生效。"} {"task_id": "format-code-task-000128", "source_id": "format-code-task-000128", "domain": "code", "task_path": "tasks/format-code-task-000128", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a25b8fadbd7967d3a6e5ac252a5f462ace826f0ba0ef615259bcdbafc444e7e2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n🐛 Bug: .allcontributorsrc is conflicting with formatting again\n### Bug Report Checklist\n\n- [x] I have tried restarting my IDE and the issue persists.\n- [x] I have pulled the latest `main` branch of the repository.\n- [x] I have [searched for related issues](https://github.com/JoshuaKGoldberg/create-typescript-app/issues?q=is%3Aissue) and found none that matched my issue.\n\n### Expected\n\nPRs generated by the `@allcontributors` bot shouldn't mangle formatting. The file should be formatted with tabs per the Prettier config.\n\n### Actual\n\nhttps://github.com/JoshuaKGoldberg/create-typescript-app/pull/1974 is replacing tabs with spaces.\n\nhttps://github.com/JoshuaKGoldberg/all-contributors-for-repository/pull/773 is an example in a downstream repo.\n\n### Additional Info\n\n_Sigh_. I've seen this before. Similar past issue: #837\n\n🎁"} {"task_id": "format-code-task-000130", "source_id": "format-code-task-000130", "domain": "code", "task_path": "tasks/format-code-task-000130", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6b6120aa29ce200976f223b107ddd304cee05598c55cce55496ca998457e05d9", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI'm working with `Hyperplane` and `Line2D` in LazySets, and I want to be able to generate random points that actually lie on the hyperplane/line. Right now if I just call `sample` on a `Hyperplane`, the points I get aren't on it. It'd also be handy to have a way to project an arbitrary point onto a hyperplane or line. Ideally sampling a hyperplane would just give me points on it by default, but I'd still like to control the underlying distribution if I want to.\n\n# Expected outcomes\n\n- Projection onto hyperplane-like sets:\n - `project(x, hp::Hyperplane)` should accept an arbitrary point and return its orthogonal projection onto the given hyperplane.\n - The returned point should lie on the target hyperplane, and points that are already on the hyperplane should be unchanged up to the numeric precision of the element type.\n- Projection onto two-dimensional lines:\n - `project(x, L::Line2D)` should accept an arbitrary two-dimensional point and return its orthogonal projection onto the given line.\n - The returned point should lie on the target line, and points that are already on the line should be unchanged up to the numeric precision of the element type.\n- Sampling from hyperplanes and lines:\n - Calling `sample` on a `Hyperplane` without specifying a sampler should produce points that lie on that hyperplane.\n - Calling `sample` on a `Line2D` without specifying a sampler should produce points that lie on that line.\n - `HyperplaneSampler` should be available as a sampler for both `Hyperplane` and `Line2D`, and samples produced with it should lie on the requested set.\n - `HyperplaneSampler` should work with its default configuration and should also allow callers to control the sampling distribution through its public API.\n\n# Implementation notes\n\n- Preserve the existing public sampling API style and numeric-type flexibility used elsewhere in LazySets.\n- The concrete implementation strategy, helper organization, and validation location are up to the implementer, as long as the externally observable projection and sampling behavior above is satisfied.\n- Tests should tolerate ordinary floating-point roundoff for inexact numeric types."} {"task_id": "format-code-task-000131", "source_id": "format-code-task-000131", "domain": "code", "task_path": "tasks/format-code-task-000131", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5477d554b0339de5d9615aad38b27e8acc8607a7de0fa3419aaf6fcfb6e69d95", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## SLES hosts aren't being recognized as a known OS\n\nWe have a mixed Linux fleet managed through Katello — RHEL, CentOS, Fedora, and a chunk of SUSE Linux Enterprise Server boxes. The Red Hat / CentOS / Fedora ones get their OS picked up correctly, but for the SLES hosts the OS comes back as nothing — Katello doesn't seem to map the distribution name reported by the subscription consumer to anything at all, so they end up without a proper OS assignment.\n\nExpected: SLES (and SUSE Linux Enterprise in general, since the distribution name reported can vary a bit) should be a recognized OS just like RHEL / CentOS / Fedora are. Could support for it be added?"} {"task_id": "format-code-task-000133", "source_id": "format-code-task-000133", "domain": "code", "task_path": "tasks/format-code-task-000133", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:acf709816be05b54743e259d3b01c082aaa99c2623953f47f5fc2e292df0c0d8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nWhen I send a request in Insomnia to a hostname that shares an IP with other virtual hosts, the server logs show it hitting the IP/default vhost instead of the named host, and HTTPS sometimes lands on the wrong site. I’m also seeing cookies behave differently than curl/browser on redirects — Set-Cookie comes back, but the next request doesn’t seem to carry the same jar state.\n\n- Requests should preserve the hostname the user entered when they are sent, so the server can route and negotiate the connection as that hostname rather than as a rewritten IP address.\n- Cookie state should behave consistently across a request and any redirects it follows: cookies already in the jar should be applied to outgoing requests, and cookies returned by responses should be reflected in the stored jar for later requests.\n- The exact transport details, request construction, and cookie storage mechanics are up to the implementation as long as the externally visible request routing and cookie behavior match the expectations above."} {"task_id": "format-code-task-000134", "source_id": "format-code-task-000134", "domain": "code", "task_path": "tasks/format-code-task-000134", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6c8119a9d8de92737e23edbe1a41cb3b23dcedd0d847458e2d8cf16a98af7610", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI’m trying to create a Kong route through the Admin API with a path like `/users:list` or `/items;v=2`, but the request gets rejected as `invalid path` with a “characters outside of the reserved list of RFC 3986 found” message.\n\n# Expected outcomes\n\n- Admin API route paths should allow RFC 3986 path-segment reserved characters when they are used in otherwise valid paths, rather than rejecting them solely as invalid path characters.\n- Creating or updating routes with paths such as `/users:list`, `/items;v=2`, or other valid URI path-segment reserved-character usages should pass route path validation.\n- Route paths that can use regex-style matching should still distinguish valid regex-style paths from syntactically invalid ones when these newly allowed URI characters are present.\n- Malformed percent-encoding must remain invalid and should report an `invalid url-encoded value` validation error.\n\n# Implementation notes\n\n- The exact validation structure, helper functions, and where the checks are performed are implementation details.\n- The fix should preserve existing path validation behavior for genuinely invalid path characters and malformed percent-encoding while allowing valid RFC 3986 path characters in route paths."} {"task_id": "format-code-task-000136", "source_id": "format-code-task-000136", "domain": "code", "task_path": "tasks/format-code-task-000136", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3ad9131f66172e4fa5bd09c87b3308185738ca59575717ec1978cd414f60736a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nDefault to Anonymous Access when no identity config is present\nCurrently the AuthConfig requires at least one identity evaluator config to be present or all requests will fail with `401 Unauthorized`. One who wants to skip identity verification phase needs to explicitly add a trivial `anonymous: {}` identity config.\n\nThis is an RFE to make this the standard behaviour when no identity evaluator config is present in the AuthConfig."} {"task_id": "format-code-task-000137", "source_id": "format-code-task-000137", "domain": "code", "task_path": "tasks/format-code-task-000137", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0e80102f9febcf133d7ebbb251cdd7475b865c2a671d4c6e053efe12bca224e5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nJust generated a fresh site with the defaultsite generator and I'm hitting two weird things. First, when I try to use `get_node_trans_by_node_id(id, 'en')` in one of my Twig templates I get an \"unknown function\" error — tried both with and without `--demosite` and it never seems to be available. Second, running the generator command itself is acting strange around `app/config/config.yml`; the `white_october_pagerfanta` bit isn't being handled sensibly, especially when deciding whether that configuration is already there or still needs to be added. Am I missing a step somewhere?\n\n## Expected Outcomes\n\n- Generated default sites expose the Twig function `get_node_trans_by_node_id(nodeId, lang)` in templates, regardless of whether the demosite option is enabled.\n- Calling `get_node_trans_by_node_id(nodeId, lang)` returns the matching available node translation for that node id and language, or `null` when no matching available translation exists.\n- The defaultsite generator bases its `white_october_pagerfanta` handling on the existing `app/config/config.yml` configuration.\n- When `white_october_pagerfanta` is already present in `app/config/config.yml`, running the generator does not append a duplicate block.\n- When `white_october_pagerfanta` is absent from `app/config/config.yml`, running the generator still adds the expected pagerfanta configuration block.\n\n## Implementation Notes\n\n- Keep the fix compatible with both defaultsite generation modes, with and without demosite content.\n- The concrete organization of generated Twig extension code, service wiring, and helper methods is up to the implementation as long as the generated site exposes the documented Twig behavior.\n- The config handling may be implemented wherever appropriate in the generator flow, but it should preserve existing configuration content and make the pagerfanta decision from the actual existing configuration."} {"task_id": "format-code-task-000138", "source_id": "format-code-task-000138", "domain": "code", "task_path": "tasks/format-code-task-000138", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5c57f605279e5fe67f1cef2c7cf5bd677b92d483bc56b9b26341b185c9968249", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Support MySQL, OSS and S3 workspace backends\n\nToday a workspace can only declare a `local` backend for storing state. We want the workspace\nconfiguration to additionally understand three remote backends — a MySQL database (`mysql`), an\nAlibaba Cloud OSS bucket (`oss`) and an AWS S3 bucket (`s3`) — and to validate and complete them\ncorrectly.\n\n## Backend configuration schema\n\nExtend the workspace backend configuration so that, in addition to `local`, a user can configure\none of the following backends. The configuration types live in the workspace API package and are\nreferenced from the existing `BackendConfigs` container (whose existing `Local` field stays as is).\nAdd a field for each new backend — `Mysql`, `Oss` and `S3` — pointing at a dedicated config type:\n\n- **mysql** (`MysqlConfig`): a database name (`DBName`), a user (`User`), an optional password\n (`Password`), a host (`Host`), and an optional port (`Port`, a `*int`).\n- **oss** (`OssConfig`): an endpoint (`Endpoint`), an optional access-key id (`AccessKeyID`), an\n optional access-key secret (`AccessKeySecret`), and a bucket (`Bucket`).\n- **s3** (`S3Config`): an endpoint (`Endpoint`), an optional access-key id (`AccessKeyID`), an\n optional access-key secret (`AccessKeySecret`), a bucket (`Bucket`), and an optional region\n (`Region`).\n\n(How you factor the shared object-storage fields of oss/s3 internally is up to you.) Only the\nconnection-identifying fields are stored in the workspace; the credentials (passwords / access\nkeys) and the S3 region are sensitive or environment-specific and may instead be supplied through\nenvironment variables (see *Completion* below).\n\n## Validation\n\nExtend the existing backend-configs validation (`ValidateBackendConfigs`) to enforce the following\nrules. As before, a `nil` or empty backend configuration is valid (it implies the default `local`\nbackend), and validation must never inspect values that are only ever provided through the\nenvironment (access keys, mysql password, s3 region).\n\n- **At most one backend** may be configured. Configuring two or more backends at once is invalid.\n- **local**: unchanged from current behavior.\n- **mysql**: the database name, the user and the host are all required. If a port is given it must\n be in the range 1–65535; an absent port is allowed. The password is never required here.\n- **oss**: the bucket and the endpoint are required. The access keys are never required here.\n- **s3**: the bucket is required. The endpoint, region and access keys are never required here.\n\nInvalid configurations must produce a non-nil error; valid ones must not.\n\n## Determining the configured backend\n\nProvide a function `GetBackendName` that, given a backend-configs value, reports the name of the\nbackend that is configured: `\"local\"`, `\"mysql\"`, `\"oss\"` or `\"s3\"`. When nothing (or only an\nempty/`nil` configuration) is set, the reported backend name is `\"local\"`.\n\n## Completion\n\nProvide a function `CompleteWorkspace` that *completes* a workspace in place — filling in defaults\nand overlaying any secrets/region found in the environment onto the configured backend. It must be\nsafe to call on a workspace with no backend (or the local backend) and must:\n\n- default the mysql **port** to `3306` when it is not set, leaving an explicitly-set port untouched;\n- for mysql, set the password from `KUSION_BACKEND_MYSQL_PASSWORD` when that variable is non-empty;\n- for oss, set the access-key id from `OSS_ACCESS_KEY_ID` and the access-key secret from\n `OSS_ACCESS_KEY_SECRET` when those variables are non-empty;\n- for s3, set the access-key id from `AWS_ACCESS_KEY_ID`, the access-key secret from\n `AWS_SECRET_ACCESS_KEY`, and the region from `AWS_REGION`, falling back to `AWS_DEFAULT_REGION`\n when `AWS_REGION` is empty — each only when the resolved value is non-empty.\n\nA non-empty environment value overrides whatever is already present in the config for that field;\nan empty/unset variable leaves the existing value un"} {"task_id": "format-code-task-000139", "source_id": "format-code-task-000139", "domain": "code", "task_path": "tasks/format-code-task-000139", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3df48973fd5cc7b824c50c7c5c5fce19b6a633a760752021cc4947072e741e4b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nHey, I noticed `kusion apply` still has all those `--backend-*` flags hanging around and it's building its state storage the old way, while the rest of the CLI seems to be moving toward the new backend + workspace setup. Can we bring `apply` in line with that? Just pulling the default backend and running against the `default` workspace would be fine for now — I don't really need to pick a backend per-invocation from flags anymore.\n\nAlso while you're in there, the ApplyRequest we send downstream still carries that `Cluster` field pulled from `-D cluster=...`, which feels like a leftover from the old model — I'd rather it carry the workspace instead so operators see something consistent with preview.\n\n## Expected outcomes\n\n- `kusion apply` CLI behavior\n - The `kusion apply` command no longer exposes backend-specific command-line flags.\n - Invoking `kusion apply` with an old backend-specific flag is rejected as an unknown flag instead of being accepted as a per-invocation backend override.\n\n- Default workspace/backend behavior\n - Applying a stack uses the repository’s default backend behavior rather than backend options supplied on the `apply` command line.\n - The apply flow runs against the `default` workspace for now.\n - Preview information produced during apply is associated with that same workspace.\n\n- Downstream apply request behavior\n - The apply request sent to the operation layer carries workspace information consistently with preview.\n - A `cluster` value supplied through `-D cluster=...` is not forwarded as the apply request’s cluster routing value.\n\n## Implementation notes\n\n- Keep the implementation aligned with the repository’s current backend and workspace concepts.\n- The exact internal wiring, helper structure, and validation location are up to the implementation, as long as the externally observable CLI behavior and downstream request behavior match the outcomes above."} {"task_id": "format-code-task-000140", "source_id": "format-code-task-000140", "domain": "code", "task_path": "tasks/format-code-task-000140", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6700f69399d524550a65a9f95d982c803a1db3022ffbf9e505b0e08ef12347a5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### dataapi `/metrics` returns incorrect `total_stake`\n\nHitting the `/metrics` endpoint on the data API and reading `total_stake` / `total_stake_per_quorum`, the numbers don't match what's actually staked on-chain. The on-chain totals are way larger than what `uint64` can hold, and the values coming out of the API look like they've wrapped/been truncated.\n\nLooking at the `Metric` response struct, `TotalStake` and `TotalStakePerQuorum` are typed as `uint64`, which isn't wide enough for real total-stake values. These fields need to be able to represent arbitrarily large integers so the reported totals are actually correct.\n\nCould the metrics handler be updated so the total stake values it computes and serializes can hold the full on-chain values?"} {"task_id": "format-code-task-000141", "source_id": "format-code-task-000141", "domain": "code", "task_path": "tasks/format-code-task-000141", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fb57a1d011858ebcb7c7e1316199788ca5251048cf15b1be58e0de5ec2d8cb99", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nLitProgressBar ignores TQDM_MINITERS\n### Bug description\n\ntqdm supports setting the environment variable `TQDM_MINITERS` to reduce the frequency of progress bar updates.\nSince lightning trainer defaults to a TQDM progress bar, I expected this to work for lightning. However, pytorch lightnings trainer seems to ignore this.\n\n### What version are you seeing the problem on?\n\nv2.1\n\n### How to reproduce the bug\n\n```python\nexport TQDM_MINITERS=5\npython your_script.py\n\n\nwhere script can be anything using a `pytorch_lightning.Trainer` with `progress_bar_enabled=True`\n```\n\n\n### Error messages and logs\n\n\n-\n\n### Environment\n\n

\n Current environment\n\n```\n#- Lightning Component (e.g. Trainer, LightningModule, LightningApp, LightningWork, LightningFlow): Trainer\n#- PyTorch Lightning Version (e.g., 1.5.0): 2.1.3\n#- Lightning App Version (e.g., 0.5.2):\n#- PyTorch Version (e.g., 2.0): 2.0.1\n#- Python version (e.g., 3.9): 3.9\n#- OS (e.g., Linux): Linux\n#- CUDA/cuDNN version:\n#- GPU models and configuration:\n#- How you installed Lightning(`conda`, `pip`, source): poetry\n#- Running environment of LightningApp (e.g. local, cloud):\n```\n\n
\n\n\n### More info\n\n_No response_\n\ncc @awaelchli"} {"task_id": "format-code-task-000142", "source_id": "format-code-task-000142", "domain": "code", "task_path": "tasks/format-code-task-000142", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:49f7045c50e95318ff47ec2e37825352f8f359a7f4a14b8e909be303494378ce", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n`lightning_getattr` should raise `AttributeError`\n## 🚀 ~Feature~ Enhancement\nCurrently, `lightning_getattr(model, attribute)` raises one of `(ValueError, KeyError, AttributeError` when `attribute` is not found in `model`, but I think it should raise `AttributeError` in all cases since Python built-in `getattr()` only raises `AttributeError`.\n\nhttps://github.com/PyTorchLightning/pytorch-lightning/blob/4bdf2fe55f45c4cc8b397d4b45041265c402519f/pytorch_lightning/utilities/parsing.py#L241"} {"task_id": "format-code-task-000143", "source_id": "format-code-task-000143", "domain": "code", "task_path": "tasks/format-code-task-000143", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1d1b433a00baa9203ea86bf4697e820e0e2367fa0de8d9186f80fb6d3c2fbc11", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAllow IterableDataset to be passed to LightningDataModule.from_datasets\n## 🚀 Feature\n\nAllow `IterableDataset` to be passed to `LightningDataModule.from_datasets`\n\n### Motivation\n\nCurrently, we cannot pass `IterableDataset` instances to `LightningDataModule`'s `from_datasets` method.\n\nThis is because because `IterableDataset` cannot be used with `DataLoader` instances that are instantiated with `shuffle=True`. \n\nRight now, `DataLoader` instances created by `LightningDataModule` are hard-coded to set `shuffle=True`: \n\nhttps://github.com/PyTorchLightning/pytorch-lightning/blob/402a258705c10c8ad57bfdc16c39a8420b1425ee/pytorch_lightning/core/datamodule.py#L379\n\n\n\nI think allowing users to pass a `shuffle` parameter to `LightningDataModule.from_datasets` is a good idea.\n\n\n\n### Additional context\n\n\n\nI've made the code changes and am prepared to submit a pull request."} {"task_id": "format-code-task-000144", "source_id": "format-code-task-000144", "domain": "code", "task_path": "tasks/format-code-task-000144", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3c7cf4ebb921cae9b1d6499c1b43ef7ac1986825455bbbc0208f3ce6e9200289", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nTensorBoardLogger and ModelCheckpoint are not using the same folder by default\n## 🐛 Bug\n(master branch)\nBy default, the TensorBoardLogger writes logs into `lightning_logs/0` but ModelCheckpoint writes checkpoint into `lightning_logs/version_0`."} {"task_id": "format-code-task-000145", "source_id": "format-code-task-000145", "domain": "code", "task_path": "tasks/format-code-task-000145", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:29577823022722e55b070071b9ed4d0f80d6f61af23eac17daeb2db05a34210a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nDeprecate {train/val/test}_transforms, dims, and size from the DataModule\n## 🚀 Feature\n\nDeprecate these properties off the LightningDataModule interface:\nhttps://github.com/PyTorchLightning/pytorch-lightning/blob/963c26764682fa4cf64c93c5a7572ae0040e9c32/pytorch_lightning/core/datamodule.py#L94-L147\n\n### Motivation\n\nWe are auditing the Lightning components and APIs to assess opportunities for improvements:\n- https://docs.google.com/document/d/1xHU7-iQSpp9KJTjI3As2EM0mfNHHr37WZYpDpwLkivA/edit#\n- https://github.com/PyTorchLightning/pytorch-lightning/issues/7740#issuecomment-876780318\n\n`train_transforms`, `val_transforms`, `test_transforms`, `dims`, and `size` are entirely optional to use, yet they’re on the DataModule interface. The Trainer does not rely on these, nor is the user forced to implement them. In reality, these are internal implementation details of individual datamodules. As a result, we ought to deprecate these off the DataModule interface.\n\nOf course, users retain the ability to implement these properties in their LightningModules if they find these abstractions helpful.\n\n\n\n### Pitch\n\nWe can follow a similar approach as what was done for https://github.com/PyTorchLightning/pytorch-lightning/issues/7301\n- Mark the properties as deprecated in v1.5. Warn if users are passing non-None values to the DataModule constructor and if they read/set their corresponding properties\n- Remove the properties in v1.7\n\n\n\n### Alternatives\nKeep as is\n\n\n\n### Additional context\n\n\n\n______________________________________________________________________\n\n#### If you enjoy Lightning, check out our other projects! ⚡\n\n\n\n- [**Metrics**](https://github.com/PyTorchLightning/metrics): Machine learning metrics for distributed, scalable PyTorch applications.\n\n- [**Flash**](https://github.com/PyTorchLightning/lightning-flash): The fastest way to get a Lightning baseline! A collection of tasks for fast prototyping, baselining, finetuning and solving problems with deep learning\n\n- [**Bolts**](https://github.com/PyTorchLightning/lightning-bolts): Pretrained SOTA Deep Learning models, callbacks and more for research and production with PyTorch Lightning and PyTorch\n\n- [**Lightning Transformers**](https://github.com/PyTorchLightning/lightning-transformers): Flexible interface for high performance research using SOTA Transformers leveraging Pytorch Lightning, Transformers, and Hydra.\n\n"} {"task_id": "format-code-task-000146", "source_id": "format-code-task-000146", "domain": "code", "task_path": "tasks/format-code-task-000146", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:eee4ea88f9d3bb5fcf507a4690f8748959d22bd7cce03059f0ec3c55840d7778", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nEarlyStopping is based upon callback.on_train_epoch_end, not callback.on_validation_epoch_end\n## 🐛 Bug\n\n[EarlyStopping](https://pytorch-lightning.readthedocs.io/en/latest/common/early_stopping.html) patience is supposed to be based upon `callback.on_validation_epoch_end`. \"It must be noted that the patience parameter counts the number of validation epochs with no improvement, and not the number of training epochs. Therefore, with parameters check_val_every_n_epoch=10 and patience=3, the trainer will perform at least 40 training epochs before being stopped.\"\n\nHowever, if you set `check_val_every_n_epoch=10` and `patience=3`, you will get a crash after the first training epoch because of `callback.on_train_epoch_end`:\n\n```\nTraceback (most recent call last):\n File \"/usr/lib/python3.7/runpy.py\", line 193, in _run_module_as_main\n \"__main__\", mod_spec)\n File \"/usr/lib/python3.7/runpy.py\", line 85, in _run_code\n exec(code, run_globals)\n File \"/workspace/project/heareval/predictions/runner.py\", line 75, in \n runner()\n File \"/usr/local/lib/python3.7/dist-packages/click/core.py\", line 1137, in __call__\n return self.main(*args, **kwargs)\n File \"/usr/local/lib/python3.7/dist-packages/click/core.py\", line 1062, in main\n rv = self.invoke(ctx)\n File \"/usr/local/lib/python3.7/dist-packages/click/core.py\", line 1404, in invoke\n return ctx.invoke(self.callback, **ctx.params)\n File \"/usr/local/lib/python3.7/dist-packages/click/core.py\", line 763, in invoke\n return __callback(*args, **kwargs)\n File \"/workspace/project/heareval/predictions/runner.py\", line 70, in runner\n task_path, scene_embedding_size, timestamp_embedding_size, gpus\n File \"/workspace/project/heareval/predictions/task_predictions.py\", line 764, in task_predictions\n gpus=gpus,\n File \"/workspace/project/heareval/predictions/task_predictions.py\", line 646, in task_predictions_train\n trainer.fit(predictor, train_dataloader, valid_dataloader)\n File \"/usr/local/lib/python3.7/dist-packages/pytorch_lightning/trainer/trainer.py\", line 553, in fit\n self._run(model)\n File \"/usr/local/lib/python3.7/dist-packages/pytorch_lightning/trainer/trainer.py\", line 918, in _run\n self._dispatch()\n File \"/usr/local/lib/python3.7/dist-packages/pytorch_lightning/trainer/trainer.py\", line 986, in _dispatch\n self.accelerator.start_training(self)\n File \"/usr/local/lib/python3.7/dist-packages/pytorch_lightning/accelerators/accelerator.py\", line 92, in start_training\n self.training_type_plugin.start_training(trainer)\n File \"/usr/local/lib/python3.7/dist-packages/pytorch_lightning/plugins/training_type/training_type_plugin.py\", line 161, in start_training\n self._results = trainer.run_stage()\n File \"/usr/local/lib/python3.7/dist-packages/pytorch_lightning/trainer/trainer.py\", line 996, in run_stage\n return self._run_train()\n File \"/usr/local/lib/python3.7/dist-packages/pytorch_lightning/trainer/trainer.py\", line 1045, in _run_train\n self.fit_loop.run()\n File \"/usr/local/lib/python3.7/dist-packages/pytorch_lightning/loops/base.py\", line 111, in run\n self.advance(*args, **kwargs)\n File \"/usr/local/lib/python3.7/dist-packages/pytorch_lightning/loops/fit_loop.py\", line 200, in advance\n epoch_output = self.epoch_loop.run(train_dataloader)\n File \"/usr/local/lib/python3.7/dist-packages/pytorch_lightning/loops/base.py\", line 118, in run\n output = self.on_run_end()\n File \"/usr/local/lib/python3.7/dist-packages/pytorch_lightning/loops/epoch/training_epoch_loop.py\", line 235, in on_run_end\n self._on_train_epoch_end_hook(processed_outputs)\n File \"/usr/local/lib/python3.7/dist-packages/pytorch_lightning/loops/epoch/training_epoch_loop.py\", line 276, in _on_train_epoch_end_hook\n trainer_hook(processed_epoch_output)\n File \"/usr/local/lib/python3.7/dist-packages/pytorch_lightning/trainer/callback_hook.py\", line 109, in on_train_epoch_end\n callback.on_train_epoch_end(self, self.lightning_module)\n File \"/u"} {"task_id": "format-code-task-000147", "source_id": "format-code-task-000147", "domain": "code", "task_path": "tasks/format-code-task-000147", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e405f57083822a3a58cc8c449d2ae52db974faaf2dc7b41a5ebf416bb5f7ea71", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nDeprecate `LightningLoggerBase.close`\n## Proposed refactoring or deprecation\n\n\n\n### Motivation\n\nThis is a follow up to https://github.com/PyTorchLightning/pytorch-lightning/discussions/9004#discussioncomment-1212966\nand\nhttps://github.com/PyTorchLightning/pytorch-lightning/issues/9037\n\nThe base logger API has `close` defined\nhttps://github.com/PyTorchLightning/pytorch-lightning/blob/089ae9b3e82ddc31942e315294e31e48c0a899db/pytorch_lightning/loggers/base.py#L312-L314\n\nThis is only implemented on https://github.com/PyTorchLightning/pytorch-lightning/blob/089ae9b3e82ddc31942e315294e31e48c0a899db/pytorch_lightning/loggers/test_tube.py#L197-L204\n\nGiven the test tube logger has since been deprecated, we can also deprecate this method off the base API as it's very unclear what the difference is between save/close/finalize currently. \n\nThis function is never called by the Trainer either, so deprecating it from the base logger API has minimal changes for users.\n\n\n\n### Pitch\n\n- Deprecate `close` off the base API in v1.5\n- Remove it from the API in v1.7\n\n\n\n### Additional context\n\n\n\n______________________________________________________________________\n\n#### If you enjoy Lightning, check out our other projects! ⚡\n\n\n\n- [**Metrics**](https://github.com/PyTorchLightning/metrics): Machine learning metrics for distributed, scalable PyTorch applications.\n\n- [**Flash**](https://github.com/PyTorchLightning/lightning-flash): The fastest way to get a Lightning baseline! A collection of tasks for fast prototyping, baselining, finetuning and solving problems with deep learning\n\n- [**Bolts**](https://github.com/PyTorchLightning/lightning-bolts): Pretrained SOTA Deep Learning models, callbacks and more for research and production with PyTorch Lightning and PyTorch\n\n- [**Lightning Transformers**](https://github.com/PyTorchLightning/lightning-transformers): Flexible interface for high performance research using SOTA Transformers leveraging Pytorch Lightning, Transformers, and Hydra.\n\n"} {"task_id": "format-code-task-000148", "source_id": "format-code-task-000148", "domain": "code", "task_path": "tasks/format-code-task-000148", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1127ee9088a7658e15e01a99b345fc734220d7f039c674bee621b84ee114878b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nWrong error in Metric.__iter__\n## 🐛 Bug\nIn the class `Metric`, the `__iter__` method is defined as follows:\n```python\ndef __iter__(self):\n \"\"\"Iteration over metrics are not allowed. Use metric collections for nesting metrics.\"\"\"\n raise NotImplementedError(\"Metrics does not support iteration.\")\n```\nIn python's docs, the following note is written about `NotImplementedError` (https://docs.python.org/3/library/exceptions.html#NotImplementedError): \n> It should not be used to indicate that an operator or method is not meant to be supported at all – in that case either leave the operator / method undefined or, if a subclass, set it to [None](https://docs.python.org/3/library/constants.html#None).\n\nIn fact, the use cases for `NotImplementedError` are:\n> In user defined base classes, abstract methods should raise this exception when they require derived classes to override the method, or while the class is being developed to indicate that the real implementation still needs to be added.\n\nIn PyCharm (and maybe other python code checkers), this leads to a warning for every sub-class of Metric, saying that all abstract methods should be implemented (PyCharm understands a method that raises a `NotImplementedError` as abstract, even if there is no `@abstractmethod` decorator on this method).\n\nWas there a good reason to define `__iter__` like that? Otherwise, could we remove it?"} {"task_id": "format-code-task-000149", "source_id": "format-code-task-000149", "domain": "code", "task_path": "tasks/format-code-task-000149", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0011e50b9d4975cdf4567043649c2036783340d2b0ea2a0b879f90e521dae62b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nnDCG can not be called with float targets\n## 🐛 Bug\n\nInvoking `RetrievalNormalizedNDCG` & `retrieval_normalized_dcg` with target of type float results into\n`ValueError: `target` must be a tensor of booleans or integers`\n\nThe reason is this [check in `_check_retrieval_functional_inputs`](https://github.com/PyTorchLightning/metrics/blob/21fe0ca7e1e61e197a923c6482c5fa07e32908de/torchmetrics/utilities/checks.py#L514):\n```py\nif target.dtype not in (torch.bool, torch.long, torch.int):\n raise ValueError(\"`target` must be a tensor of booleans or integers\")\n```\n\n\n\n### To Reproduce\n\nThe code samples below are sufficient to reproduce the error\n\n#### Code sample\n\nCode sample for `retrieval_normalized_dcg`\n```py\nimport torch\nfrom torchmetrics.functional import retrieval_normalized_dcg\n\npreds = torch.tensor([.1, .2, .3, 4, 70])\ntarget = torch.tensor([0.1, 0, 0, 0.5, 1.0])\nretrieval_normalized_dcg(preds, target)\n```\n\nCode sample for `RetrievalNormalizedNDCG`\n```py\nimport torch\nfrom torchmetrics import RetrievalNormalizedDCG\n\nindexes =torch.tensor([0, 0, 0, 1, 1, 1, 1])\npreds = torch.tensor([0.2, 0.3, 0.5, 0.1, 0.3, 0.5, 0.2])\ntarget = torch.tensor([0.1, 0.1, 0.5, 0.5, 1.0, 1.0, 0.0])\nndcg = RetrievalNormalizedDCG()\nndcg(preds, target, indexes=indexes)\n```\n\n### Expected behavior\n\nTargets of type float are handled without errors\n\n### Environment\n\n- PyTorch Version (e.g., 1.0): 1.9.0\n- OS (e.g., Linux): Ubuntu 20.04\n- How you installed PyTorch (`conda`, `pip`, source): source\n- Build command you used (if compiling from source): `pip install git+git://github.com/PyTorchLightning/metrics.git@79cb5e2f1744b0d4cd3f15adf9d725a68452a50`\n- Python version: 3.8.10\n- CUDA/cuDNN version: N/A\n- GPU models and configuration: N/A\n- Any other relevant information: N/A"} {"task_id": "format-code-task-000152", "source_id": "format-code-task-000152", "domain": "code", "task_path": "tasks/format-code-task-000152", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:94c7adab7fed378d4c3dc955ee4caf47ae6057b8c17c4401e74e49c00a551750", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nRule could-be-keyword-tags: Ignore some builtin tags\n`could-be-keyword-tags` highlights situations where all keywords in a file use the same tag. These tags can be consolidated in the Settings section instead.\n\nIn my code base, it keeps raising situations where all keywords use `robot:flatten` or `robot:private`. These tags are instructions for the Robot Framework runtime. Because of this, I don't think they should be consolidated like other tags. I want to keep them explicit.\n\nPotential fixes:\n\n1. Keep it the way it is.\n2. Always ignore `robot:*` tags ([full list](https://robotframework.org/robotframework/latest/RobotFrameworkUserGuide.html#reserved-tags))\n3. Add a configurable parameter to this rule that allows users to exclude specific tags."} {"task_id": "format-code-task-000153", "source_id": "format-code-task-000153", "domain": "code", "task_path": "tasks/format-code-task-000153", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ce3ef706010033015ad0990b69dce4cf5b9cf073014b6c36a4b18af85c481bea", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Support wildcard patterns when selecting rules\n\nRobocop lets users refer to a rule by its id (e.g. `0501`) or by its name\n(e.g. `too-long-keyword`). The same idea is used internally whenever rules are\n*selected by a pattern* — for example when listing rules that match something a\nuser typed. Today that selection only understands **exact** strings: a pattern\nmatches a rule only when it is byte-for-byte equal to the rule's id or its name.\nThat makes it impossible to ask for \"everything in the naming category\" or\n\"every rule whose name ends in `-keyword`\".\n\nMake rule selection understand **Unix shell-style wildcards**.\n\nConcretely:\n\n- A pattern is checked against **both** a rule's id and its name; the rule is\n selected when **either** of them matches the pattern.\n- The following wildcards are supported in a pattern:\n - `*` — matches any number of characters (including none),\n - `?` — matches exactly one character,\n - `[seq]` — matches any single character in `seq`,\n - `[!seq]` — matches any single character **not** in `seq`.\n- Matching is **case-insensitive**.\n- A pattern that contains no wildcard characters keeps behaving exactly as\n before: it matches only when it is equal to the rule's id or name (so existing\n exact-match callers are unaffected).\n\nThere are two layers to this behaviour:\n\n1. The per-rule check that answers \"does this rule match the given pattern?\"\n returns a truthy/falsy result and must honour the rules above. It must keep\n accepting an already-compiled regular-expression object as the pattern and go\n on matching that against the rule's id and name as it does now.\n2. The routine that filters a whole collection of rules by a single (string)\n pattern returns the matching rules **sorted by ascending numeric rule id**,\n and **deprecated rules are never included** in that result, regardless of\n whether they match the pattern.\n\nA bare `*` therefore selects every non-deprecated rule.\n"} {"task_id": "format-code-task-000154", "source_id": "format-code-task-000154", "domain": "code", "task_path": "tasks/format-code-task-000154", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a1cca54957fb32d4c287f5ae4abe1cde0c7079030a74831f883cbd8aa9ade4a3", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a reports subsystem to the linter\n\nRobocop can find issues, but right now there is no way to turn a run into different kinds of\noutput. I want to introduce a small **reports** subsystem so that a run can be summarized in\nseveral ways (printing a version banner, dumping the found issues to a JSON file, etc.) and so the\nset of active reports can be selected from configuration.\n\nPlease expose this through the package `robocop.linter.reports`.\n\n## Selecting reports\n\nAdd a function `get_reports(configured_reports)` that takes a list of report names (strings) and\nreturns the enabled reports as an ordered mapping from report name to a report instance. Each\nreport instance must expose its own `name`. The selection rules:\n\n- The returned mapping preserves the order in which names were requested and never contains\n duplicates (requesting the same report twice yields a single entry).\n- The literal name `\"all\"` enables every report that is marked as a *default* report. Reports that\n are not default (for example `json_report` and `compare_runs`) are **never** pulled in by\n `\"all\"`; they can only be enabled by naming them explicitly. `\"all\"` may be combined with\n explicit names, e.g. `[\"all\", \"json_report\"]`.\n- Requesting a name that does not correspond to any known report raises\n `robocop.linter.exceptions.InvalidReportName`, and the raised error message must mention the\n offending name.\n- The literal name `\"None\"` switches reports off: the result contains only the always-on internal\n `return_status` report, regardless of any external reports also requested in the same call. The\n one exception is that if `internal_json_report` was also explicitly requested, it is preserved\n alongside `return_status`.\n\n## Reports that must exist\n\nReports are addressed by name through `get_reports`. The following named reports must be available:\n\n- **`version`** — a default report whose `get_report()` returns a non-empty string that includes\n the installed Robocop version.\n- **`json_report`** — a non-default report that collects issues handed to it via\n `add_message(message)` and, when `get_report()` is called, writes them as a JSON array to a file.\n Each issue is written in the same JSON form that an issue serializes to (i.e. the dict returned by\n the message's `to_json()`), and the array preserves the order in which issues were added. By\n default the file is named `robocop.json` and created in the current working directory, and\n `get_report()` returns a confirmation string that mentions the output path. The report is\n configurable through `configure(name, value)`:\n - `configure(\"report_filename\", \"\")` changes the output file name.\n - `configure(\"output_dir\", \"\")` changes the directory the file is written to (the directory\n is created if it does not exist).\n - Any other parameter name passed to `configure` raises\n `robocop.linter.exceptions.ConfigGeneralError`.\n- **`return_status`** — an internal report that is always kept under the `\"None\"` rule above and is\n also included by `\"all\"`.\n- **`compare_runs`** and **`internal_json_report`** — non-default reports that exist but are not\n pulled in by `\"all\"`.\n\nA report that does not recognize a configuration parameter should reject it by raising\n`robocop.linter.exceptions.ConfigGeneralError`.\n"} {"task_id": "format-code-task-000155", "source_id": "format-code-task-000155", "domain": "code", "task_path": "tasks/format-code-task-000155", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:68cf6192db75b324e5d01062883d6774a6f118bcf6d2cb6edeed982263112c32", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[Bug] 0920 (unused-variable) false positive with nested for loops\n### What happened?\n\n0920 has a false positive that is somehow related to nested for loops. Commenting out the inner loop in the below code results in no unused variables reported. 5.3.0 doesn't report it with the inner loop so this is probably related to #1093 / #1104\n\n### What command/code did you try to run?\n\nsample.resource:\n```robotframework\n*** Keywords ***\nDo Stuff With Events\n [Arguments] ${initial_token}\n ... ${retries}=5\n # don't want to mutate the original\n VAR ${continuation_token} ${initial_token}\n\n FOR ${i} IN RANGE ${{ int($retries)+1 }}\n ${response} = Get Events ${continuation_token}\n FOR ${_} IN @{response}[events]\n # Do stuff\n No Operation\n END\n\n IF $i < int($retries)\n VAR ${continuation_token} ${response}[continuationToken]\n END\n END\n\n\nGet Events\n [Arguments] ${token}\n RETURN ${{ {'continuationToken': int($token) + 1, 'events': []} }}\n```\nand then run `robocop sample.resource`\n\n### What is the full error message?\n\n`sample.resource:15:20 [I] 0920 Variable '${continuation_token}' is assigned but not used (unused-variable)`\n\n### What did you expect to happen instead?\n\nNo unused variables reported\n\n### Operating System\n\nUbuntu 24.04\n\n### Robocop version\n\n5.4.0, 5.5.0, 5.6.0"} {"task_id": "format-code-task-000156", "source_id": "format-code-task-000156", "domain": "code", "task_path": "tasks/format-code-task-000156", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0d3abf1cdb3cf0f0ba2304b4d07b0e7e3854b3c5cc4563843fde832a8e2c14bc", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[Bug] import variable files with module name and arguments gives a wrong variables-import-with-args message\n### What happened?\n\nstarting with this simple suite\n\n```robot\n*** Settings ***\nVariables variables ${ENVIRON}\n\n*** Variables ***\n${ENVIRON} dev\n\n*** Test Cases ***\nfirst test case\n Log ${ENVIRONMENT_NAME}\n```\n\nwhere I import a variable file with arguments, I get this error:\n\n```\n...\\import_vars.robot:2:1 [E] 0404 Robot and YAML variable files do not take arguments (variables-import-with-args)\n```\n\n\n\n### What command/code did you try to run?\n\nrobocop .\\tests\\import_vars.robot\n\n### What is the full error message?\n\n```\n...\\import_vars.robot:2:1 [E] 0404 Robot and YAML variable files do not take arguments (variables-import-with-args)\n```\n\n### What did you expect to happen instead?\n\n\nstarting from robotframework 5 it is possible to import variables with its module name like libraries, see also [documentation](https://robotframework.org/robotframework/latest/RobotFrameworkUserGuide.html#taking-variable-files-into-use) \n\nso this message is invalid with robotframework version >= 5\n\nI think the rule 0404 should check that the file extension is `yaml` and not `py`.\nAnd are you shure that you can import `robot` files as variables? I think not. What do you mean with robot files?\n\n\n### Operating System\n\nwindows, linux\n\n### Robocop version\n\n3.1.1"} {"task_id": "format-code-task-000157", "source_id": "format-code-task-000157", "domain": "code", "task_path": "tasks/format-code-task-000157", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:47e3207c887b99e4e5ba56c100dc7b801b7c171022130db0611c177f1e828c93", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[Bug] argument-overwritten-before-usage is raise but there is no variable on the line.\n### What happened?\n\nlaunch robocop with the following configuration file:\n```\n--configure line-too-long:line_length:200 \n--configure wrong-import-order:severity:error\n--configure return_status:quality_gate:E=0:W=3:I=-1\n--exclude too-many-arguments\n--exclude too-long-keyword\n--exclude file-too-long\n--exclude missing-doc-test-case\n--exclude too-long-test-case\n--exclude too-many-calls-in-test-case\n--exclude not-allowed-char-in-name\n```\n\nThe following warning message is raised:\n/opt/robotframework/tests/trustme-company-tests/auth/odif/odif.robot:95:21 [W] 0921 Keyword argument '${resp}' is overwritten before usage (argument-overwritten-before-usage)\n\nBut on line 95 there is no ${resp} variable there is only 83 line in the file:\n\n```\n*** Settings ***\nDocumentation Auth / Odif non-regression tests\n\nLibrary Collections\nLibrary RequestsLibrary\nLibrary FakerLibrary\nResource ${PATH_KEYWORDS}/device.resource\nResource ${PATH_KEYWORDS}/encoder.resource\nResource ${PATH_KEYWORDS}/signature.resource\nResource ${PATH_KEYWORDS}/simulator.resource\nResource ${PATH_KEYWORDS}/user.resource\nResource ${PATH_KEYWORDS}/auth.resource\n\nTest Tags auth odif no-regression\n\n\n*** Test Cases ***\nAP-88-ODIF-01 Positive Scenario: Create Odif Challenge\n ${device}= Create A Random Device # robocop: disable=unused-variable\n ${resp}= Create An Odif Challenge device=${device}\n Dictionary Should Contain Key ${resp} value\n Dictionary Should Contain Key ${resp} blob\n\nAP-88-ODIF-02 Positive Scenario: Create An Odif Authentication Method Without Header\n ${device} ${user}= Enroll A New Random User authentication_method=odif device_validation=${TRUE}\n ${m}= Get User Authentication Methods user=${user} device=${device}\n List Should Contain Value ${m} fingerprint_local\n\nAP-88-ODIF-03 Delete an odif authentication method\n Given A User With Odif And Password Authentication Method\n When I Delete The Odif Authentication Method\n Then The Fingerprint Local Method Should Be Removed\n And The User Should Not Use Odif Method To Verify\n\n\n*** Keywords ***\nAdd A Password Authentication On The User Created\n [Documentation] Add a password authentication method to a user\n ... A user with ${USER} should be created in setup\n ${password}= Create A Password\n ${resp}= Create A Password Authentication Method ${password}\n Status Should Be 201 ${resp}\n\nGet User Authentication Methods\n [Documentation] Query user info to get the associated authentication\n ... methods.\n [Arguments] ${user}=${user} ${device}=${device}\n ${resp}= Get My User user=${user} device=${device}\n Dictionary Should Contain Key ${resp} authentication_methods\n RETURN ${resp}[authentication_methods]\n\nA User With Odif And Password Authentication Method\n [Documentation] Signup a user with odif and password\n ${device} ${user}= Enroll A New Random User authentication_method=odif device_validation=${TRUE}\n Verify Odif Using Attribute Name user_id user=${user} device=${device} http_code=200\n\n ${m}= Get User Authentication Methods user=${user} device=${device}\n List Should Contain Value ${m} fingerprint_local\n List Should Contain Value ${m} password\n Set Test Variable ${DEVICE}\n Set Test Variable ${USER}\n\nI Delete The Odif Authentication Method\n [Documentation] Delete odif authentication method\n\n ${payload}= Create Dictionary user_id=${USER}[id] device_id=${DEVICE}[id] method=fingerPri"} {"task_id": "format-code-task-000158", "source_id": "format-code-task-000158", "domain": "code", "task_path": "tasks/format-code-task-000158", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b93f3a9056fc4647d911b2488423c78764f54f2209971d712c23e7546bc28880", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[SIM117] Merge with statements for context managers that have same scope\n## Explanation\n\nA little known feature of Python is that a with statement can contain multiple context managers. This allows the code only one with statement (and therefore only 1 level of indentation). This rule has a similar rational to SIM102. \n\nThis rule should be applied if and only if:\n* context A only contains the code of Context B and no other code. That is all the code in the nested statement will run with context A and B. \n\nCaveat when implementing: If the context names are really long, the with statement may be broken over a line break. A new feature in the Python 3.10 alpha will be to allow parentheses to be used to break the with statement over multiple lines.\n\n## Example\nConsider the following context managers:\n```python\n#bad\nwith A() as a:\n with B() as b:\n print('hello')\n```\ncan be transformed into the following:\n```python\n# Good\nwith A() as a, B() as b:\n print('hello')\n```"} {"task_id": "format-code-task-000159", "source_id": "format-code-task-000159", "domain": "code", "task_path": "tasks/format-code-task-000159", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2260ba9564e3b8fd5b8163f14e752cc9c8cc8caa4b8f6f6292b2c4f1e78fb116", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[New Rule] Simplify \"in dict.keys()\" to \"in dict\"\n## Explanation\n\nMost of the discussion of the motivation about this rule can be found in this very detailed [StackOverflow post](https://stackoverflow.com/a/29314342)\n\n## Example\n\n```python\n# Bad\nkey in dict.keys()\n\n# Good\nkey in dict\n```\n\nBasically, the later way of doing this pythonic and slightly more performant and arguably more Readable. The previous method only exists for legacy code (<= python 2.2). So there is no reason to use the former in modern codebases."} {"task_id": "format-code-task-000160", "source_id": "format-code-task-000160", "domain": "code", "task_path": "tasks/format-code-task-000160", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fda2ea380bd2b2f7bc4a8877cc070cdf2ce83aee588536c58e614460360ec13d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[New Rule] Use comparision directly instead of bool(comparison)\n## Explanation\n\nComparisons return boolean values. No need to wrap it in a bool-call\n\n## Example\n\n```python\n# Bad\nbool(a == b)\n\n# Good\na == b\n```"} {"task_id": "format-code-task-000161", "source_id": "format-code-task-000161", "domain": "code", "task_path": "tasks/format-code-task-000161", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:22f7adcead3eec653474086724a2eaac7da8104bc561de2cd8fd15d5db5ce5e3", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[SIM905] Use list of strings instead of splitting a constant string\n## Explanation\n\nYou want a list of strings? Create one directly:\n\n* A list of strings can be formatted by black.\n* Your editor can sort the lines\n* You don't have to worry about how to keep leading / intermediate / trailing spaces or how to add newlines to the strings\n\n\n## Example\n\n```python\n# Bad\ndomains = \"de com net org\".split()\n\n# Good\ndomains = [\"de\", \"com\", \"net\", \"org\"]\n```"} {"task_id": "format-code-task-000162", "source_id": "format-code-task-000162", "domain": "code", "task_path": "tasks/format-code-task-000162", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:02020ef1e0bd574aeaaa2933a40f6c2e69e650e5c27bcad33ccc494a57e17732", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nNewConstraint method faulty behavior\nUsed version: 3.1.1\nWhen I try to create a constraint with NewConstraint method I get following returns:\n\n**good Example**\nc, cErr := semver.NewConstraint(\"12.3.4.1234\")\nc.String() = nil\ncErr = \"improper constraint: 12.3.4.1234\"\n\n**faulty behavior**\nc, cErr := semver.NewConstraint(\"12.**23**.4.1234\")\nc.String() = \"12.23.4 1234\"\ncErr = nil\n\nor\n\nc, cErr := semver.NewConstraint(\"12.3.**34**.1234\")\nc.String() = \"12.3.34 1234\"\ncErr = nil\n\n\n------------\n\nI would expect the same behavior as with \"12.3.4.1234\", means telling me its not a contraint.."} {"task_id": "format-code-task-000163", "source_id": "format-code-task-000163", "domain": "code", "task_path": "tasks/format-code-task-000163", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:12baa380107c4dc9fb895ee96ba87a0baf6462097c64f0d9767efd5b65fb1732", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `SetAttributes` silently swallows wrong number of arguments\n\n`SetAttributes` is supposed to take exactly two arguments (a symbol/list of symbols and an attribute/list of attributes), but if I call it with the wrong arity Mathics just returns the expression unevaluated without saying anything is wrong.\n\nFor example, I forgot to wrap multiple attributes in a list:\n\n```\nIn[1]:= SetAttributes[f, Flat, Protected]\nOut[1]= SetAttributes[f, Flat, Protected]\n\nIn[2]:= Attributes[f]\nOut[2]= {}\n```\n\nNo warning, no message — but obviously nothing was actually set, which is pretty confusing. Same thing if I just forget the second argument entirely:\n\n```\nIn[3]:= SetAttributes[g]\nOut[3]= SetAttributes[g]\n```\n\nAgain, completely silent.\n\nIn WMA you get a message telling you the call had the wrong number of arguments (the usual \"called with N arguments; 2 arguments are expected\" style of complaint), which makes it immediately obvious what went wrong. Mathics should do the same for `SetAttributes` so that bad calls don't look like they succeeded."} {"task_id": "format-code-task-000164", "source_id": "format-code-task-000164", "domain": "code", "task_path": "tasks/format-code-task-000164", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0e9dd07eee182da183f1c7ad3dd829a07fe1261baeb1633d28eb474eb0368988", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nSupport UUID column types\nHello,\n\nDuckDB supports the UUID data type. It would be great if this data type was also available from SQLAlchemy.\n\nhttps://duckdb.org/docs/sql/data_types/overview.html\n\"Screenshot\n\nBest regards,\nKamil"} {"task_id": "format-code-task-000165", "source_id": "format-code-task-000165", "domain": "code", "task_path": "tasks/format-code-task-000165", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1e400fa16ad78f09e245a64c4e83e35843ea229386936a4adf2a5d7a71d2e160", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI’d like to use zstd-compressed cache archives with drone-cache, but right now setting `archive_format: zstd` doesn’t seem to give me a real zstd archive. Could you add proper zstd support so rebuild/restore works with it, and ideally let the existing compression level option apply there too?\n\n# Expected outcomes\n\n- Zstd archive format support:\n - When `archive_format: zstd` is configured, including through `PLUGIN_ARCHIVE_FORMAT=zstd`, rebuild should create a real zstd-compressed cache archive rather than falling back to an uncompressed tar archive.\n - Restore should be able to read cache archives created with `archive_format: zstd` and recover the archived files with the same observable behavior as the existing supported archive formats.\n\n- Compression level behavior:\n - The existing `--compression-level` / `PLUGIN_COMPRESSION_LEVEL` setting should also affect zstd archive creation when `archive_format: zstd` is used.\n - Existing gzip and tar behavior should remain compatible with the current behavior.\n\n- User-facing documentation:\n - The user-facing CLI/configuration documentation should list `zstd` as a supported `archive_format`.\n - The compression-level documentation should make clear that the setting applies to both gzip and zstd, and should give users enough information to find valid zstd compression-level values.\n\n# Implementation notes\n\nThe internal package structure, concrete compression library, helper functions, and validation location are up to the implementer. Keep the behavior compatible with the existing rebuild/restore flow and configuration mechanisms, and avoid changing unrelated archive formats except where needed to preserve existing behavior."} {"task_id": "format-code-task-000166", "source_id": "format-code-task-000166", "domain": "code", "task_path": "tasks/format-code-task-000166", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b5fc24cc7260d2d7eec429a2c71cd74e5aeb724e452cd2427984cca82d939656", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Targeted parameter substitution in map keys and values\n\nThe `copier` package provides a reflection-based deep-copy helper,\n`CopyWithReplacements(src, replacementsFunc, replaceIn...)`, that clones an\narbitrary value and, while cloning, rewrites the *string* fields that the caller\nopts into. A field is opted in by listing its **path** in the variadic\n`replaceIn` argument, where a path is the chain of exported struct field names\njoined by `.` (for example `\"Spec.Template.Name\"`). The literal `\"*\"` selects a\nfield and everything nested beneath it. Selected strings are rewritten by\nrunning every `$name` token through `replacementsFunc` (the existing\n`EvaluateString` behavior); unselected strings are copied verbatim.\n\nThis works for struct fields, but it is too coarse for maps. Today a map is\naddressed by a single path and there is no way to say \"substitute only in the\nvalues\" versus \"substitute only in the keys\" of that map — and map keys are in\nfact never rewritten at all, regardless of what is requested.\n\nExtend the helper so that the keys and the values of a map field can be targeted\nindependently:\n\n- Appending `.Keys` to a map field's path selects that map's **keys** for\n substitution.\n- Appending `.Values` to a map field's path selects that map's **values** for\n substitution.\n- These compose with deeper paths the same way struct fields do: e.g. for a\n `map[string]SomeStruct` field named `Cfg`, the path `Cfg.Values.Title`\n reaches the `Title` field of each value, and `*` cascading still applies once\n a path matches.\n\nConcretely, for a struct field `Data` of type `map[string]string`:\n\n- `replaceIn = [\"Data.Values\"]` rewrites the values only; keys are left as-is.\n- `replaceIn = [\"Data.Keys\"]` rewrites the keys only; values are left as-is.\n- `replaceIn = [\"Data\"]` (selecting the map field itself) rewrites both keys\n and values.\n- selecting neither leaves the whole map untouched.\n\nWhen a key string is rewritten, the rewritten string becomes the key in the\nresulting map (the map is effectively re-keyed), paired with its\n(possibly-rewritten) value.\n\nAll existing behavior must be preserved: the result is always a deep,\nindependent copy (mutating it must not affect the source), non-map and non-string\nfields are copied unchanged, string fields selected by their own path are still\nrewritten, and `Copy` (the no-substitution entry point) still performs a plain\ndeep copy.\n"} {"task_id": "format-code-task-000167", "source_id": "format-code-task-000167", "domain": "code", "task_path": "tasks/format-code-task-000167", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d466ad2b4e56ed8cdb07817c6185c8ebdfa949d4faeb96b0605a6996fc125de6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\n我这边嵌套打开两个 NG-ZORRO Modal，第二个 Modal 配了 `nzZIndex`，内容层确实在上面，但它的遮罩看起来还压不住前一个 Modal，检查 DOM 发现 backdrop 的 `z-index` 还是默认值；如果同时用了 `nzMaskStyle`，更新遮罩样式后也是这个现象。\n\n## Expected outcomes\n\n- Modal instances configured with `nzZIndex` should apply that stacking level consistently to the visible modal and its associated mask/backdrop, so a later or higher-level modal mask can cover lower-level modals.\n- When `nzMaskStyle` is present or updated on a modal that also has `nzZIndex`, the mask/backdrop should continue to reflect the configured `nzZIndex` rather than falling back to the default mask stacking level.\n- Updating a modal’s `nzZIndex` after it has been opened should update the mask/backdrop stacking level together with the modal content stacking level.\n\n## Implementation notes\n\n- The exact place where the style is applied and the internal structure used to keep modal and mask styles in sync are implementation details.\n- Preserve existing modal behavior other than the externally observable stacking behavior described above."} {"task_id": "format-code-task-000168", "source_id": "format-code-task-000168", "domain": "code", "task_path": "tasks/format-code-task-000168", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c19722a4554d1acf1607308ee921d0831829e952f19235d0d9701945300fb7e4", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nThe Python PyCuTe mirror is used by the backend code generator to reason about tensor layouts, but its public swizzle and composed-layout objects do not currently behave like the corresponding CuTe layout primitives. This makes higher-level code unable to query extents uniformly, and slicing a composed mapping can either return the wrong object or fail before the sliced mapping is usable. Bring these public primitives to parity for integer layouts without changing the existing layout algebra behavior.\n\nA Swizzle(bits, base, shift) must expose the same layout-like extent API as Layout: size() is 2 ** (bits + base + abs(shift)) and cosize() is the same extent. The zero-bit case (bits=0, base=3, shift=0) is valid and maps every offset to itself; positive and negative shifts must both remain bijections over their extent. Swizzle equality is structural across all three parameters and must safely compare unequal to unrelated objects.\n\nA ComposedLayout(outer, offset, inner) must continue to map coordinates as outer(offset + inner(coord)), report the inner domain size and the outer codomain size, support multi-argument coordinates and mode indexing, and compare structurally without raising when compared with another type. Calling it with underscore coordinates must return a sliced ComposedLayout that preserves the outer mapping and fixed-coordinate offset, including when the outer mapping is a Swizzle. The resulting slice must have the expected domain size and produce exactly the same values as the corresponding coordinates of the unsliced object. Existing Layout, inverse, divide/product, and typing behavior must remain unchanged."} {"task_id": "format-code-task-000169", "source_id": "format-code-task-000169", "domain": "code", "task_path": "tasks/format-code-task-000169", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:8f560c1b3bc6df40649ee985b85d04675f778f29cccfafe9ed725b16ab06c600", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nProblem Statement\n\nI keep copying the same base HPC Container Maker recipe bits into multiple recipes, like the Ubuntu/GNU compiler setup, and it’s getting annoying to keep them in sync. I’d like to be able to include another recipe file from a recipe, preferably by a relative path next to the main recipe, so its Stage0/Stage1 additions show up in the generated Dockerfile/Singularity output.\n\nExpected Outcomes\n\n- Recipe authors can include another recipe file from within a recipe through the public `hpccm.include(recipe_file)` API.\n- Included recipe files execute as part of the same recipe-building process, so additions they make to recipe state, including Stage0 and Stage1 content, building blocks, primitives, and values needed later by the including recipe, are reflected in the final generated container specification.\n- Relative include paths are resolved relative to the recipe being processed rather than the caller’s current working directory; absolute include paths continue to work as absolute paths.\n- Include failures are handled consistently with recipe loading: by default, file-open or execution errors are logged and terminate with exit code 1, while calling `hpccm.include(..., raise_exceptions=True)` propagates the underlying exception.\n- The examples should demonstrate reusing a sibling recipe file to avoid duplicating common development-environment setup.\n\nImplementation Notes\n\n- Preserve existing recipe execution behavior and output generation for recipes that do not use includes.\n- The internal mechanism for locating, executing, and sharing recipe state is up to the implementation, as long as the public behavior above is satisfied.\n- The implementation should work for both Dockerfile and Singularity generation paths supported by the existing recipe workflow."} {"task_id": "format-code-task-000170", "source_id": "format-code-task-000170", "domain": "code", "task_path": "tasks/format-code-task-000170", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:199fa4e7a2f7d23339e044367ad58903130ecb17c87a4cdedf1e136bcf7c1022", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n我现在想按租户把任务的 groupKey 分片，比如 `sync:tenantA`，但默认的 sync/action 这些 worker 只会拉到原来的固定 key，分片后的任务就没人处理了。能不能让这些 worker 按前缀把同一类任务都消费掉，同时我自己建 `ProcessorWorker` 的时候也能传这种自定义的 groupKey/pattern？\n\nExpected outcomes:\n- Dequeueing tasks by `groupKey` should support pattern strings containing `*`, so a worker can request a group-key prefix and receive eligible tasks in that family.\n- Pattern-based dequeueing should still respect the existing dequeue constraints, including task state, start time, and requested limit, and dequeued tasks should transition as normal.\n- Built-in processing workers that currently consume fixed task-family group keys should consume tenant- or suffix-partitioned group keys for their own families.\n- Callers should be able to use custom group-key values or group-key pattern values through the existing `ProcessorWorker` constructor path.\n\nImplementation notes:\n- The exact matching mechanism, validation location, and internal representation of group-key patterns are implementation details.\n- Keep existing exact-key behavior compatible for callers that do not use patterns.\n- Do not require callers to use a different public entry point in order to use custom group keys or group-key patterns."} {"task_id": "format-code-task-000171", "source_id": "format-code-task-000171", "domain": "code", "task_path": "tasks/format-code-task-000171", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:cda90dc67f88fa4643f0439be6de2a89528bf79750e7674d2ec1ed58f00a9664", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI'm using the comment module and I want to let my users report comments programmatically, but I don't see any way to actually submit a report through the library right now. Ideally I'd like to pass in a reason for the report (like spam, porn, spoiler, etc.) and optionally some extra text when needed. Can you add a way to do this on a comment object?\n\nExpected outcomes:\n- Comment objects provide an asynchronous `report(...)` capability that submits a report for that specific comment and returns the API response as a `dict`.\n- The reporting API accepts a semantic report reason via a public `ReportReason` enum covering the service-documented report reasons, with stable integer values compatible with the underlying service. It should include the common documented categories such as other, spam advertising, pornography, and spoilers.\n- Optional extra report text is accepted only for the “other” report reason; using extra text with a non-other reason should be rejected with the library’s normal argument-validation error behavior.\n- Submitting a report requires an authenticated credential; attempting to report without the required login/session information should fail through the library’s normal credential-validation behavior instead of silently submitting an unauthenticated request.\n\nImplementation notes:\n- Fit the new comment-reporting capability into the existing public comment module style and asynchronous request flow.\n- The exact internal data structures, helper placement, request configuration layout, and validation location are up to the implementation, as long as the public behavior above is satisfied.\n- Preserve existing behavior of unrelated comment operations."} {"task_id": "format-code-task-000172", "source_id": "format-code-task-000172", "domain": "code", "task_path": "tasks/format-code-task-000172", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6c19f43ea3850ca52a3fe0cb7bc86a29932314b891e422a8883b500424da0967", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nCredential Script not called correctly from Defaults section\n### A note for the community\n\n\n* Please vote on this issue by adding a 👍 [reaction](https://blog.github.com/2016-03-10-add-reactions-to-pull-requests-issues-and-comments/) to the original issue to help the community and maintainers prioritize this request\n* If you are interested in working on this issue or have submitted a pull request, please let us know in a comment\n\n\n\n### Problem\n\nIf I put \"credential_script\" configuration in the Defaults section of the harvest.yml, the script is called only with username parameter, not with address. If it is in the Pollers section it works correctly.\n\nAlso I find the wording of the feature documentation a bit misleading. \n> Harvest will call the script with two arguments via standard in\n\nThat made me think the script is called like this `echo \"hostname username\" | get_pass`\n\n\n\n### Configuration\n\n```text\nAdmin:\n httpsd:\n listen: 127.0.0.1:8887\n\nTools:\n grafana_api_token: abc\n autosupport_disabled: true\n\nExporters:\n prometheus:\n exporter: Prometheus\n local_http_addr: localhost\n port_range: 12990-13013\n sort_labels: true\n\nDefaults:\n collectors:\n - Zapi\n - ZapiPerf\n\n exporters:\n - prometheus\n\n auth_style: basic_auth\n username: foo_ontap\n\n credentials_script:\n path: ./get_pass\n schedule: 12h\n timeout: 10s\n \nPollers:\n ontap1:\n datacenter: DC1\n addr: ontap1.local\n \n ontap2:\n datacenter: DC1\n addr: ontap2.local\n \n StorageGrid1:\n datacenter: DC1\n addr: storagegrid1.local\n username: foo\n password: bar\n collectors:\n - StorageGrid\n```\n\n\n### Poller\n\nall\n\n### Version\n\nharvest version 23.05.0-1 (commit 6f74c7a5) (build date 2023-05-03T08:08:46-0400) linux/amd64\n\n### Poller logs\n\n_No response_\n\n### OS and platform\n\nRed Hat Enterprise Linux release 8.7 (Ootpa)\n\n### ONTAP or StorageGRID version\n\n9.11.1P8\n\n### Additional Context\n\n_No response_\n\n### References\n\n_No response_"} {"task_id": "format-code-task-000173", "source_id": "format-code-task-000173", "domain": "code", "task_path": "tasks/format-code-task-000173", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:645e1f2394bbd71f74b1b83fe2907d5a7d0a0d556f7e63384022446adc0f72fe", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Custom `dgs.graphql.graphiql.path` doesn't actually work end-to-end\n\nI'm using the WebFlux variant of dgs-framework and I wanted to expose GraphiQL on a non-default path (we have a routing convention internally), so in my `application.yml` I set:\n\n```yaml\ndgs:\n graphql:\n graphiql:\n path: /my-graphiql\n```\n\nWhen I hit `/my-graphiql` in the browser it does redirect me to `/my-graphiql/index.html` like I'd expect, but the page that comes back is broken — GraphiQL loads but as soon as I try to run any query nothing happens, the requests don't go to my actual GraphQL endpoint. It's basically unusable on a custom path.\n\nIf I revert the config and go back to the default path everything works fine, so this only happens when I change `dgs.graphql.graphiql.path` to something other than the default.\n\nMy expectation is that this property should fully control where GraphiQL lives — not just the redirect, but the actual page being served at that path should also work correctly and talk to the configured GraphQL endpoint. Right now it feels like only part of the path config is being honored."} {"task_id": "format-code-task-000174", "source_id": "format-code-task-000174", "domain": "code", "task_path": "tasks/format-code-task-000174", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:639cbf576f1313841fc5ed3017da3afa72246c6ca1e7d605f16f9e896653bb63", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nHey, I'm trying to add websocket subscription support to juno's JSON-RPC server (think `eth_subscribe`-style stuff where the server keeps pushing events to the client after the initial call returns). Two things are blocking me right now: registered RPC handlers need a supported way to push messages back to the websocket client after the initial request starts, while ordinary non-websocket requests should continue behaving like regular request/response calls. I also need shared lifecycle plumbing for active subscriptions so callers can register a subscription, get a handle for it, and tear it down later.\n\nExpected outcomes:\n- Websocket-backed JSON-RPC handlers can access a writable client connection through the request context helper and use it during handler execution to send an additional message to the same client, without preventing the normal JSON-RPC response from being returned.\n- Ordinary non-websocket JSON-RPC requests report that no writable client connection is available through that helper and continue returning normal responses.\n- A subscription registry API is available for creating a registry, adding an `event.Subscription`, receiving an opaque `uint64` subscription id, and deleting the subscription later by that id.\n- Deleting a registered subscription unsubscribes it exactly once for that successful deletion and removes it from the registry.\n- Deleting an unknown or already-deleted subscription id fails with the registry's stable public not-found error and does not panic.\n- The subscription registry is safe to use from concurrent callers.\n\nRequired public API:\n- `jsonrpc.ConnFromContext(ctx context.Context) (io.Writer, bool)`\n- `pubsub.New(log utils.SimpleLogger) *Registry`\n- `pubsub.Registry.Add(ctx context.Context, sub event.Subscription) uint64`\n- `pubsub.Registry.Delete(id uint64) error`\n- `pubsub.ErrNotFound`\n- Keep the new plumbing at the public behavior level described above; internal storage, synchronization strategy, id generation details, websocket loop organization, and where the websocket request context is populated are up to the implementation."} {"task_id": "format-code-task-000175", "source_id": "format-code-task-000175", "domain": "code", "task_path": "tasks/format-code-task-000175", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:12a648bf14ccada54db347d5eff54db7ddaa06f6f646d99e2edab6852ebdd96a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[Bug]: regression in 2.3.0 leading to IndexError\n### What happened?\n\noriginal report https://github.com/dandi/dandi-cli/issues/1223 for test failing while working on data from https://github.com/dandisets/nwb_test_data/tree/master/v2.0.1 \n\n\n\n### Steps to Reproduce\n\n```python\nsee below in traceback demo\n```\n\n\n### Traceback\n\n```python\n❯ python -c 'from dandi.metadata import get_metadata; print(get_metadata(\"/home/yoh/proj/dandi/nwb-datasets/nwb_test_data/v2.0.1/test_Subject.nwb\"))'\nTraceback (most recent call last):\n File \"\", line 1, in \n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/fscacher/cache.py\", line 152, in fingerprinter\n ret = fingerprinted(*args, **kwargs_)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/joblib/memory.py\", line 594, in __call__\n return self._cached_call(args, kwargs)[0]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/joblib/memory.py\", line 537, in _cached_call\n out, metadata = self.call(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/joblib/memory.py\", line 779, in call\n output = self.func(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/fscacher/cache.py\", line 98, in fingerprinted\n return f(path, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/dandi/metadata.py\", line 131, in get_metadata\n meta.update(_get_pynwb_metadata(path))\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/dandi/pynwb_utils.py\", line 193, in _get_pynwb_metadata\n nwb = io.read()\n ^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/hdmf/utils.py\", line 645, in func_call\n return func(args[0], **pargs)\n ^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/pynwb/__init__.py\", line 287, in read\n return super().read(**kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/hdmf/backends/hdf5/h5tools.py\", line 453, in read\n return super().read(**kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/hdmf/utils.py\", line 645, in func_call\n return func(args[0], **pargs)\n ^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/hdmf/backends/io.py\", line 42, in read\n container = self.__manager.construct(f_builder)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/hdmf/utils.py\", line 645, in func_call\n return func(args[0], **pargs)\n ^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/hdmf/build/manager.py\", line 280, in construct\n result = self.__type_map.construct(builder, self, None)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/hdmf/utils.py\", line 645, in func_call\n return func(args[0], **pargs)\n ^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/site-packages/hdmf/build/manager.py\", line 789, in construct\n return obj_mapper.construct(builder, build_manager, parent)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/yoh/proj/dandi/dandi-cli-master/venvs/dev3.11/lib/python3.11/sit"} {"task_id": "format-code-task-000176", "source_id": "format-code-task-000176", "domain": "code", "task_path": "tasks/format-code-task-000176", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7f370f39686530460bd728e5e1685cd82c3dcac4f2dafc0cc02186104cb25f9f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Output of flattened file has inconsistent whitespace\n\nWhen I run truffle-flattener over a few `.sol` files (either via the CLI with `--output` or by `require`-ing it as a library) and look at the resulting flat file, the whitespace looks off:\n\n- There's a blank line at the very top of the output before the first `// File: ...` header.\n- The amount of blank lines between consecutive files isn't always the same.\n- The end of the file doesn't end cleanly with a single newline — sometimes there are extra blank lines, sometimes the last line has no trailing newline at all.\n\nI noticed this when I tried to commit the flattened output to a repo: my editor kept \"fixing\" the file on save (POSIX expects every text file to end with exactly one newline), and a `diff` between two flattens of the same input wasn't as stable as I'd like.\n\nMinimal repro:\n\n```js\nconst flatten = require(\"truffle-flattener\");\n\n(async () => {\n const out = await flatten([\"./contracts/A.sol\", \"./contracts/B.sol\"]);\n console.log(JSON.stringify(out)); // inspect leading/trailing whitespace\n})();\n```\n\nSame thing via the CLI:\n\n```\n$ truffle-flattener contracts/A.sol contracts/B.sol --output flat.sol\n$ cat -A flat.sol # extra blank lines visible at top / between files / at end\n```\n\nIt would be nice if the flattener produced a clean, predictable output: no leading blank line, consistent separation between concatenated files, and a single trailing newline at the end. Same output whether it's written via `--output`, printed to stdout, or returned from the library entry point.\n\nAlso, there don't seem to be any tests covering the output format right now, so it'd be good to lock this down with some."} {"task_id": "format-code-task-000177", "source_id": "format-code-task-000177", "domain": "code", "task_path": "tasks/format-code-task-000177", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3514c3ca625eda20a37588b90eff27637f214b45aa1c0576c5ed23c15b7ea664", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## ISSQN withholding invoice always uses the same partner regardless of city\n\nWe use `l10n_br_account_withholding` to automatically generate withholding invoices when confirming supplier invoices. This works fine for federal withholdings (IRRF, PIS, COFINS, CSLL) where the partner configured on the fiscal tax group is the right counterpart.\n\nThe problem is with **ISSQN**. ISSQN is a municipal tax, so the withholding invoice must be issued to the city hall (*Prefeitura*) of the municipality where the service was rendered — Prefeitura de São Paulo, Prefeitura do Rio de Janeiro, Prefeitura de Belo Horizonte, etc. Each one is a different legal entity (different CNPJ, different bank account, different journal in some setups).\n\n### What we observe\n\nWhen we confirm a purchase invoice that has an ISSQN withholding line, the generated WH invoice always has its `partner_id` taken from the partner configured on the fiscal tax group. That gives us a single hard-coded partner for *all* ISSQN withholdings, no matter which city the service line refers to.\n\nSo if we have a supplier invoice with services rendered in São Paulo and another with services rendered in Rio, both produce WH invoices pointing to the same Prefeitura — which is obviously wrong from an accounting/reporting standpoint, and we end up having to fix the partner manually on every single ISSQN WH invoice before posting it.\n\n### What we expect\n\nISSQN withholding invoices should be addressed to the Prefeitura of the city indicated on the invoice line (the ISSQN city field already exists on the line). The setup should let us register, for each municipality where we operate, which `res.partner` represents that city's Prefeitura, so the module can pick the right one automatically when generating the WH invoice.\n\nIf for some reason no Prefeitura is registered for the city in question, falling back to the partner configured on the fiscal tax group is acceptable — that preserves today's behavior for users who haven't configured any city halls yet.\n\nThis should only kick in for taxes whose scope is municipal; federal/state withholdings should keep using the fiscal tax group's partner exactly as they do today.\n\nI'd expect the way to mark a partner as a city hall to be a new boolean flag on `res.partner` (something like `wh_cityhall`), so we can search by `(city_id, wh_cityhall=True)` to find the right Prefeitura."} {"task_id": "format-code-task-000178", "source_id": "format-code-task-000178", "domain": "code", "task_path": "tasks/format-code-task-000178", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:93c427ced89eb64e330f88126dc919179e2c00aa709a8bd53c6e8e8e5dad5234", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI’m using `Force Invoiced` on a sales order, and the order header changes to `Invoiced`, but the order lines still show `To Invoice` in the lines/list views and reports. It makes filters based on sale order line invoice status look inconsistent, so I’m not sure if I’m missing a setting or if the force flag is only affecting the header.\n\n# Expected Outcomes\n\n- When `Force Invoiced` is enabled on a sales order, the invoice status shown for each of that order’s lines should also be `Invoiced`.\n- Sale order line views, reports, and filters that rely on line invoice status should reflect the forced invoiced state consistently with the order header.\n- When `Force Invoiced` is disabled again, sale order lines should no longer remain forced to `Invoiced`; their invoice status should return to the normal value computed by the existing business rules.\n- The module usage documentation should make clear that enabling `Force Invoiced` affects both the sales order invoice status and the sale order line invoice status.\n\n# Implementation Notes\n\n- The exact model hooks, dependency declarations, and storage/update strategy are up to the implementer.\n- Preserve the existing behavior for sales orders and sales order lines when `Force Invoiced` is not enabled.\n- Keep documentation updates focused on the user-visible behavior rather than internal implementation details."} {"task_id": "format-code-task-000179", "source_id": "format-code-task-000179", "domain": "code", "task_path": "tasks/format-code-task-000179", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:50f59ad5d218d981b20468d520a42d9fb247a2d44dc18268ab4f2081e21e4d0a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI’d like our sales quotations to pick a shipping method as soon as they’re created, instead of waiting until the order is confirmed or making the salesperson choose it manually. Ideally I could turn that on in Sales settings, and if the customer or delivery address changes on the quote, the carrier would refresh to match the new address.\n\n# Expected outcomes\n\n- A company-level Sales setting is available to enable or disable automatic carrier selection for newly created quotations. The setting is exposed through `res.config.settings.carrier_on_create` and stored on `res.company.carrier_on_create`.\n- The new quotation-time setting is independent from the existing confirmation-time automatic carrier assignment setting: enabling or disabling one must not silently change the behavior controlled by the other.\n- When quotation-time carrier selection is enabled and a new draft quotation is created without an already selected carrier, the quotation should automatically receive the carrier that matches the current customer or delivery address according to the existing delivery carrier rules.\n- If a carrier has already been explicitly set on a newly created quotation, automatic quotation-time selection should not replace that existing carrier.\n- When quotation-time carrier selection is disabled, creating a quotation should continue to leave carrier selection to the existing manual or confirmation-time flows.\n- When the customer or delivery address is changed on a quotation while quotation-time carrier selection is enabled, the quotation’s selected carrier should refresh to match the new address.\n- Confirming a sales order should continue to be governed by the existing confirmation-time automatic carrier assignment setting, independently of whether quotation-time carrier selection is enabled.\n\n# Implementation notes\n\n- Use the existing Odoo sales, delivery carrier, and configuration mechanisms where appropriate.\n- The exact internal helper structure, validation location, and data flow are up to the implementer, provided the observable quotation creation, address-change, configuration, and confirmation behaviors above are satisfied.\n- Keep compatibility with existing carrier assignment behavior outside the new quotation-time opt-in flow."} {"task_id": "format-code-task-000181", "source_id": "format-code-task-000181", "domain": "code", "task_path": "tasks/format-code-task-000181", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:65a436926eb08625f223147b93981deafeb6e7e4aa09ca9dccfa4ea9f81e8d23", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Improve how linestrings are assigned to their covering polygon\n\nWhen OSMNames builds the place hierarchy, every feature is assigned a `parent_id`\npointing at the single polygon that \"covers\" it (the most specific administrative\narea it sits in). For polygons, points and housenumbers this works well, but for\n**linestrings** (streets, rivers, …) the current rule is too strict and leaves many\nof them without a parent.\n\nToday a linestring only receives a parent when a polygon **fully contains the entire\nlinestring**, and only when that polygon is fine-grained enough (its `place_rank` is\nat or above a fixed threshold). In practice streets routinely run right up to — or\nacross — an administrative boundary, or poke slightly outside the area they belong to,\nso full containment almost never holds and the linestring is left unparented. The\n`place_rank` gate makes this worse by ignoring coarser polygons entirely.\n\nChange the way the single-covering-polygon assignment decides whether a linestring\nbelongs to a polygon so it is based on the linestring's **center point** — the point\nlocated halfway along the linestring — rather than on the whole geometry:\n\n- A linestring is assigned to a polygon when that linestring's center point lies\n inside the polygon, **even if the polygon does not contain the rest of the\n linestring** (e.g. the linestring extends beyond the polygon's boundary).\n- This must work **regardless of the polygon's `place_rank`** — the previous\n `place_rank` restriction for linestrings is removed; a coarse polygon may become a\n linestring's parent if it contains the center point.\n- A linestring whose center point is **not** inside a polygon must **not** be parented\n to it, even if one of its endpoints touches or lies within the polygon.\n- When more than one candidate polygon contains the center point, the linestring is\n assigned to the most specific one (the smallest, highest-`place_rank` polygon),\n exactly as a single covering polygon is chosen for any other feature.\n\nThis change applies to linestrings only. Polygons, points and housenumbers must keep\nbeing assigned to the polygon that fully contains them, exactly as before.\n"} {"task_id": "format-code-task-000182", "source_id": "format-code-task-000182", "domain": "code", "task_path": "tasks/format-code-task-000182", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:165c80c68958e3c7a9b4b0766bca915efd260075584dc956f482be796d701b99", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Housenumbers near administrative boundaries don't get a street_id\n\nWhen I import an OSM extract that spans several adjoining administrative\nareas, a lot of housenumbers end up with `street_id IS NULL` after the\nimport finishes — even though the corresponding street is clearly there in\n`osm_linestring`.\n\nLooking at the cases that fail, they all seem to sit close to the boundary\nbetween two parents: the housenumber is tagged into one admin area and the\nstreet geometry it actually belongs to is in the neighbouring one. So the\n`parent_id` on the housenumber and on the street differ, and the name-based\nmatching step (full match / levenshtein / substring) skips them entirely.\n\nConcretely, if I pick one such housenumber and check the linestring with\nthe matching `normalized_name` a few hundred meters away, it's obviously\nthe right street — same name, runs right past the address — but they have\ndifferent parents, so nothing gets linked. The result is a noticeable gap\nin `street_id` coverage along every admin border in the dataset.\n\nI'd expect the name-based street matching to also pick up streets that are\ngeographically close to the housenumber, not only ones that happen to share\nthe exact same `parent_id`. Streets near boundaries shouldn't fall through\nthe cracks just because OSM put them on the \"other side\" of an admin\npolygon."} {"task_id": "format-code-task-000183", "source_id": "format-code-task-000183", "domain": "code", "task_path": "tasks/format-code-task-000183", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d7ea0fc34ba500b014555ac1295a15a67bd62827971ca15656a2d7d8d8bf4d72", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nWhen I want to add a custom variable to a particle in Parcels, right now I have to subclass `JITParticle` just to attach one extra `Variable`. It feels heavy when I only want one or two extra fields tracked per particle. Could the particle classes offer a class-level helper, in the requested `add_variable` style, that gives me back a pclass I can hand straight to `ParticleSet`? Would make my scripts so much cleaner.\n\n## Expected outcomes\n\n- Single custom variable: public particle classes such as `JITParticle` and `ScipyParticle` support `add_variable(...)` as a class-level API that returns a new particle class usable as `pclass` in `ParticleSet`.\n- Variable configuration: the single-variable API supports adding a variable from either a variable name plus existing `Variable` configuration options, or an existing `Variable` definition, while preserving normal `Variable` semantics.\n- Multiple custom variables: particle classes support adding more than one `Variable` definition in one class-level call, while preserving each variable's own name, initial value, dtype, and other supported configuration.\n- ParticleSet integration: a `ParticleSet` constructed with a class returned by the new API can read and write the added variables during kernels in both JIT and Scipy particle modes where those modes are supported.\n- Backward compatibility: the existing pattern of declaring custom `Variable`s in a particle subclass class body continues to work and remains behaviorally equivalent for the same variables.\n\n## Implementation notes\n\n- The public API shape is part of the requested behavior: `add_variable(...)` is the single-variable entry point, and a plural class-level entry point is expected for multiple variables.\n- The concrete class-construction mechanism, internal registration strategy, generated class naming, validation location, and storage layout are implementation details.\n- The new API should preserve existing `Variable` semantics rather than introducing a separate variable model.\n- Keep existing subclass-based particle definitions compatible while adding the lighter-weight class-level entry points."} {"task_id": "format-code-task-000185", "source_id": "format-code-task-000185", "domain": "code", "task_path": "tasks/format-code-task-000185", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a80ba752feda28bb234f92222788fb7acbc02760a232b347fb542778a2c0761c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Validator registrations never reach my external block builder\n\nI'm running a Prysm beacon node together with an external block builder service (MEV-Boost style setup). My validator client is configured to send validator registrations through the beacon node, and the calls all come back successful — but on the builder side, nothing ever shows up. The builder keeps reporting that it has zero registered validators, and as a result it never produces any blocks for me, so I'm always falling back to local block production.\n\nFrom the validator's point of view this looks like a complete success: the registration RPC against the beacon node returns OK every time. But the builder behind it is clearly never being told about any of these validators, so the whole external-builder pipeline is effectively a no-op for me even though I've wired everything up.\n\nWhat I'd expect: when a block builder is configured on the beacon node, validator registrations that come in over `SubmitValidatorRegistration` should actually be delivered to that builder so it can start producing blocks for those validators. If something goes wrong on the builder side while accepting the registration, I'd want that surfaced back to the caller as an error rather than silently swallowed.\n\nOne thing to keep in mind: not every node operator runs a builder. For nodes that don't have one configured, this RPC should still respond normally so that validator clients which always send registrations don't start erroring out against builder-less setups."} {"task_id": "format-code-task-000186", "source_id": "format-code-task-000186", "domain": "code", "task_path": "tasks/format-code-task-000186", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3454a0a03e256ff08de9c3d2368f17bf6b3e1c8c820dc0a23778648c04158056", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nWhen I run `validator accounts import --keys-dir=... --holesky` and there's no wallet yet, it asks me \"Enter a wallet directory\" twice — and whatever I type the first time gets thrown away, only the second answer counts. I've also got my keystore JSON files sorted into subfolders under my keys dir, and the import doesn't seem to pick those up at all, only the ones sitting at the top level.\n\n# Expected outcomes\n\n- Wallet setup during `validator accounts import` should prompt for the wallet directory exactly once when no wallet exists yet, and the directory entered at that prompt should be the directory used for the wallet.\n- Importing through `validator accounts import --keys-dir ` should discover keystore JSON files in the keys directory and in nested subdirectories up to a maximum scan depth of 2 nested directory levels below the keys directory.\n- Keystore files located deeper than that maximum scan depth should not be imported, and the command should emit an informational message indicating that the maximum keystore-folder scan depth of 2 was reached.\n- The import operation should emit an informational message when validator keystore importing begins.\n- If directory processing fails during import, the command should report the failure as an inability to process the directory and import keys.\n\n# Implementation notes\n\nThe exact internal structure, helper boundaries, traversal mechanism, and validation location are up to the implementer. Preserve existing command-line behavior aside from the observable import-flow changes described above, and avoid making tests or behavior depend on private implementation details."} {"task_id": "format-code-task-000187", "source_id": "format-code-task-000187", "domain": "code", "task_path": "tasks/format-code-task-000187", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:65a89ccf1d5b0c29d5e97619498ee909caf869ea8787945625129fb14e405588", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Support next-epoch crosslink committees with a registry-change option\n\nOur beacon-chain core exposes a function that returns the crosslink committees\n(each a set of validator indices together with the shard it is assigned to) for\na given slot. Today it only knows how to answer for slots that belong to the\n**previous** or **current** epoch — asking for a slot in the **next** epoch is\nrejected as out of bounds. We need it to answer for next-epoch slots too, since\nproposers and attesters have to look one epoch ahead.\n\nPlease extend the existing \"crosslink committees at slot\" lookup so that:\n\n- Slots whose epoch equals the next epoch are now valid and return the\n committees for that epoch instead of an error. The accepted range becomes\n `previousEpoch <= epoch(slot) <= nextEpoch`; a slot whose epoch is earlier\n than the previous epoch or later than the next epoch must still return an\n error.\n\n- Callers can optionally request a *registry change*. There are two valid\n shufflings for the next epoch — one that assumes the validator registry was\n rotated and one that assumes it was not — and the caller chooses between them.\n This option only affects next-epoch slots; for previous- and current-epoch\n slots the result must be exactly the same regardless of what is requested.\n\n- For a next-epoch slot, the default (no registry change) keeps the committees\n starting at the current epoch's start shard. When a registry change is\n requested, the committees are rotated forward: their start shard is advanced\n by the number of committees in the current epoch (taken modulo the configured\n shard count).\n\nExisting callers that only ask for previous/current-epoch committees and do not\ncare about a registry change must keep working without changes, and their\nresults must be unchanged.\n"} {"task_id": "format-code-task-000188", "source_id": "format-code-task-000188", "domain": "code", "task_path": "tasks/format-code-task-000188", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5172a36009dd933d1e4e65f367df46edcba7b2f6f765329100c2597db10be577", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI’m using `yo aspnet` to scaffold ASP.NET 5 web apps that I want to run in Docker, but I still have to hand-write the Dockerfile every time. Could the web templates include a basic Dockerfile by default, and maybe let me add one separately with something like a Dockerfile generator command?\n\n# Expected outcomes\n\n- **Dockerfile in web-style scaffolds**\n - When `yo aspnet` is used to generate ASP.NET 5 web-style project templates that support running the app, the generated project root includes a file named `Dockerfile`.\n - This should apply to the relevant web-oriented templates exposed by the generator, without requiring callers to run a separate Dockerfile step.\n\n- **Standalone Dockerfile generator**\n - A standalone `yo aspnet:Dockerfile` command is available.\n - Running `yo aspnet:Dockerfile` in a target location creates a file named `Dockerfile` there.\n\n- **Generated Dockerfile contents**\n - Any `Dockerfile` produced by the main web scaffolds or by `yo aspnet:Dockerfile` uses `microsoft/aspnet:1.0.0-beta7` as its base image.\n - It copies `project.json` into `/app/`, sets `/app` as the working directory, restores dependencies with `dnu restore`, copies the project into `/app`, exposes port `5000`, and starts the app with `dnx -p project.json kestrel`.\n\n- **Standalone generator usage text**\n - The `aspnet:Dockerfile` generator has usage/help text describing that it creates a Docker configuration file.\n - The usage/help text shows the example command `yo aspnet:Dockerfile` and indicates that the created file is `Dockerfile`.\n\n# Implementation notes\n\n- The internal organization of templates, shared template files, helper methods, and copy logic is up to the implementation.\n- The Dockerfile behavior should be consistent whether it is produced as part of a project scaffold or by the standalone generator.\n- Keep the implementation compatible with the existing Yeoman generator conventions used by this repository."} {"task_id": "format-code-task-000189", "source_id": "format-code-task-000189", "domain": "code", "task_path": "tasks/format-code-task-000189", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9f4cdf9a57f305759318bc157b662823187c4c436c84ac0be4640562546ea2c6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## BroydenSolver behaves inconsistently with other nonlinear solvers (label + recording)\n\nI'm running a model where I'm experimenting with different nonlinear solvers\n(BroydenSolver vs NewtonSolver vs NonlinearBlockGS) on the same group, and\nI have a case recorder attached so I can compare their iteration behavior\nside-by-side. While doing this I noticed two ways in which BroydenSolver\nsticks out from the others — both feel unintentional:\n\n**1. Solver label doesn't follow the same naming convention**\n\nWhen I look at the solver name in printed output / recorded metadata, the\nother nonlinear solvers I'm using are labeled with the `NL:` prefix\n(`NL: Newton`, `NL: NLBGS`, etc.), but Broyden just shows up as `BROYDEN`\n— no prefix. Looks like a missed alignment with the rest of the\nnonlinear-solver family.\n\n**2. Extra cases recorded per Broyden iteration**\n\nWith the same recorder config attached, I get noticeably more recorded\nentries from BroydenSolver per outer iteration than I do from the other\nnonlinear solvers in the same setup. Switching the solver from Newton to\nBroyden on the same group, the number of cases I get back from the recorder\ngoes up even though the iteration count itself didn't. It looks like Broyden\nis producing an extra layer of recording during its iteration that the other\nnonlinear solvers don't, which inflates the case list and makes it\ninconsistent to compare runs across solvers.\n\nI'd expect BroydenSolver to label itself the same way the rest of the\nnonlinear solvers do, and to record at the same granularity as them so that\nswapping solvers doesn't change how the recorder sees the run."} {"task_id": "format-code-task-000190", "source_id": "format-code-task-000190", "domain": "code", "task_path": "tasks/format-code-task-000190", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e38956ea453410ec51078bbf6b76dcb5db53786be15a2f6862a80d4b11574d00", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nPyDP presents the laplacian algorithms as supporting both the one-shot style shown in the README (`algorithm.result(entries)`) and the incremental `add_entry` / `add_entries` style, but the shared wrapper currently confuses an entry collection with a privacy-budget argument. Restore a coherent dual-mode API for `Count`, `BoundedSum`, `BoundedMean`, `BoundedVariance`, and `BoundedStandardDeviation`.\n\nFor one-shot use, both `result(entries)` and `quick_result(entries)` must accept lists, tuples, and finite one-pass iterables such as generators. Consume a one-pass source exactly once, use the receiving algorithm's configured aggregate, bounds, and dtype, and retain the normal result type and bounded-aggregate semantics. A one-shot computation is isolated: it must neither include entries previously accumulated on that object nor add its own entries to the incremental state, consume that state's ability to produce a result, or prevent further one-shot calls. It must work before or after an incremental result has been produced.\n\nFor incremental use, expose `partial_result()` and `partial_result(privacy_budget)` as the explicit result methods after `add_entry`, `add_entries`, or `merge`. Preserve the existing compatibility forms `result()` and `result(privacy_budget)` as aliases for that same incremental operation; a scalar budget, including an explicit `0.0`, must not be mistaken for an entry iterable.\n\nMake `add_entries` accept the same finite iterable forms, consume them once, and continue returning `None`. Treat `str`, `bytes`, and `bytearray` as invalid batch containers for both `result(entries)` and `add_entries`, raising `TypeError`. Batch ingestion must be transactional: fully consume and validate a batch before changing incremental state. If iteration itself raises, propagate that original exception; if an entry cannot be converted for the configured dtype, raise `TypeError`. In either case, already accumulated entries must remain unchanged. Failed one-shot calls must likewise leave incremental state untouched.\n\nKeep the existing lifecycle interoperable with these rules: `reset()` must make incremental aggregation reusable while preserving the object's configuration, and summaries produced by `serialize()` and accepted by `merge()` must still contribute only to incremental results; an intervening one-shot result must not consume merged state."} {"task_id": "format-code-task-000191", "source_id": "format-code-task-000191", "domain": "code", "task_path": "tasks/format-code-task-000191", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:cd4205e9748cc3b85bf445bc2f426a6983baf99bdd2087656b739264fc01699a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Analysis doesn't catch dispense-without-aspirate\n\nI wrote a quick protocol and noticed that the analysis step happily accepts it even though it clearly doesn't make sense — I'm dispensing without ever aspirating. Reproducer:\n\n```python\nrequirements = {\n \"robotType\": \"OT-2\",\n \"apiLevel\": \"2.15\",\n}\n\n\ndef run(protocol_context):\n tiprack1 = protocol_context.load_labware(\"opentrons_96_tiprack_300ul\", \"1\")\n pipette = protocol_context.load_instrument(\n \"p300_single_gen2\", mount=\"right\", tip_racks=[tiprack1]\n )\n pipette.pick_up_tip(tiprack1.wells()[0])\n well_plate = protocol_context.load_labware(\"nest_96_wellplate_200ul_flat\", \"2\")\n # note: no aspirate here\n pipette.dispense(20, well_plate.wells()[0])\n```\n\nThe pipette never aspirated anything, so dispensing 20 µL out of an empty tip is nonsense. I'd expect analysis to reject the protocol up front so I can fix the script before sending it to the robot, instead of silently passing.\n\nSame concern for the case where the requested dispense volume is greater than what was previously aspirated — there's no liquid in the tip to support that dispense, so analysis should refuse it as well.\n\nBoth situations are statically detectable from the command sequence and should fail analysis with a clear error for protocols on `apiLevel` 2.15+. A dedicated error type along the lines of `InvalidDispenseVolumeError` would be appropriate here."} {"task_id": "format-code-task-000192", "source_id": "format-code-task-000192", "domain": "code", "task_path": "tasks/format-code-task-000192", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3991049ff3e10014965fbffe0b102449ca664b29ddf0cf4484b6b6f39cda34e5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n我现在在 RunPausedSplash 里为了做 Figma 上那两个 LargeButton 样式，只能在页面里写一堆覆盖样式，还要单独处理 icon 颜色，感觉很脆。能不能让 LargeButton 自己就支持透明白边白字的按钮，以及白底红字的警示按钮，这样“Cancel run”和“Launch recovery mode”就不用各自 hack 样式了。\n\nExpected outcomes:\n- `LargeButton` exposes reusable `buttonType` variants named `onColor` and `alertAlt`.\n- `buttonType=\"onColor\"` renders as a transparent, on-colored button with white label/icon treatment and a white outline; its disabled state keeps the on-color intent while using disabled styling.\n- `buttonType=\"alertAlt\"` renders as the alternate alert treatment with a white background and red label/icon treatment; its disabled state uses disabled styling.\n- `LargeButton` callers should no longer use `iconColorOverride`; icon color should come from the selected button variant and disabled state.\n- In `RunPausedSplash`, “Cancel run” uses the standard alternate alert LargeButton treatment, and “Launch recovery mode” uses the standard on-color LargeButton treatment, instead of page-local color/icon/border overrides.\n\nImplementation notes:\n- The concrete CSS organization, token lookup, helper structure, and component internals are up to the implementation.\n- Keep existing `LargeButton` variants working as before, except where shared disabled styling must remain consistent with the variant-driven icon/text behavior."} {"task_id": "format-code-task-000193", "source_id": "format-code-task-000193", "domain": "code", "task_path": "tasks/format-code-task-000193", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:617c2e22bfbf34176a2903b7906302840e40d37630ef5be5c3d1e80bbc1c0f4c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: update a job's `data` after it has been added\n\nOnce a job is added to a queue with `Queue#add(data, opts)`, there doesn't seem to be any way to change its `data` payload afterwards. `Job#progress` lets me update progress, but `data` itself is effectively read-only from the public API.\n\nI run into this in a couple of places:\n\n- A job is added with some input, then before it gets picked up (or while it's being retried) I learn something new about it (e.g. a related record changed in the DB, a user edited the request) and I'd like the worker to see the updated payload instead of the stale one.\n- I want to enrich the job's `data` with some computed fields during processing so that if the job is inspected later (via `getJob`, the UI, etc.) it carries the up-to-date information.\n\nRight now my only options are either to remove and re-add the job (which loses its id and position) or to reach into Redis directly and overwrite the hash field, which feels like it shouldn't be necessary.\n\nCould `Job` expose a method to update its `data`? Something I can call on a job instance to replace its stored data with a new object, and have subsequent reads (`Job.fromId`, the processor's `job.data` on the next fetch, etc.) reflect the new value."} {"task_id": "format-code-task-000194", "source_id": "format-code-task-000194", "domain": "code", "task_path": "tasks/format-code-task-000194", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:52e14c0e94af48e5159246533026245b7f64fafa1a77056ab33ab814e2549bdf", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Make Twitter share verification resilient to retweets and attached media\n\nOur bridge server verifies social \"promotion\" actions. For Twitter, an incoming\nwebhook event is checked against the exact content a user was asked to share: a\n`SHARE` action is considered valid only when the tweet text — after Twitter's\nautomatic transformations are undone — matches the expected content (ignoring\nsurrounding whitespace). A `FOLLOW` action is always considered valid.\n\nThe validator already knows how to:\n\n- pull the full tweet body from the extended payload when a tweet is truncated,\n- expand the shortened `t.co` links back to their original URLs using the\n tweet's URL entities, and\n- HTML-decode the result before comparing.\n\nTwo common real-world cases are not handled today and cause legitimate shares to\nbe rejected (or to blow up):\n\n1. **Attached media.** When a tweet has a photo or video, Twitter appends an\n extra `t.co` link to the tweet text pointing at the media. That link is not\n part of what the user typed, and it appears in the tweet's media entities\n (Twitter exposes these under both `entities.media` and\n `extended_entities.media`, and under the matching fields of the extended\n payload when the tweet is truncated). These media links must be removed from\n the text before the content comparison, so a share with an attached image\n still validates.\n\n2. **Retweets.** When a user retweets the post rather than composing it, the\n event carries the original tweet under `retweeted_status` and the wrapper\n text is just the `RT @user: …` prefix (often truncated). In that case the\n original retweeted tweet is what should be validated — its full text, its URL\n expansion, and its media handling — not the wrapper.\n\nUpdate the Twitter event validator so both cases work. Behavior to preserve and\npin down:\n\n- A `FOLLOW` event validates to `true`.\n- A `SHARE` event returns the cleaned, decoded tweet content (a string) when it\n matches the expected content after trimming, and `false` when it does not.\n- The existing extended-payload selection, URL expansion, and HTML decoding keep\n working, including in combination with the two new cases above (e.g. a\n truncated retweet whose original carries shortened URLs and attached media).\n- Comparison ignores only leading/trailing whitespace; the content otherwise\n must match.\n\nThe other social networks' validation behavior is unchanged.\n"} {"task_id": "format-code-task-000195", "source_id": "format-code-task-000195", "domain": "code", "task_path": "tasks/format-code-task-000195", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f452150e6b1031a47dbba56ee86015fff76e360736573e6a97fc60d8e27f4469", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Error messages from several detection ops and distribution classes are not informative\n\nWhen I'm prototyping detection / probabilistic models with Paddle, I keep running into the same frustration: when I pass something wrong into these APIs, the error I get back doesn't really help me figure out what's wrong.\n\nA few concrete cases I've hit:\n\n**1. Shape / rank mismatch reports don't tell me what I actually passed.**\n\nFor example with `locality_aware_nms` and `mine_hard_examples`, if my `BBoxes` / `Scores` / `ClsLoss` tensors don't have the expected rank or the dimensions don't line up between two inputs, I get a message like\n\n> \"The rank of Input(Scores) must be 3\"\n\nor\n\n> \"Batch size of ClsLoss and MatchIndices must be the same.\"\n\nThat tells me what's expected, but not what the op actually received. If I'm debugging a pipeline where shapes are computed dynamically, I have no idea whether I sent in a rank-2 tensor, a rank-4 tensor, or whether the mismatch is 32 vs 64 or 32 vs 33. I end up `print(x.shape)`-ing every input by hand.\n\nSame thing in `roi_perspective_transform`: messages like \"The format of input tensor is NCHW.\" or \"The transformed output height must greater than 0\" don't echo back the bad value.\n\n**2. Python-side type errors fall through to the C++ layer.**\n\nThe distribution classes `Normal`, `Uniform`, `Categorical`, `MultivariateNormalDiag` don't seem to validate their constructor / method arguments at the Python entry point. If I accidentally pass an `int` where a `float`/`ndarray`/`Variable` is expected for `loc` / `scale` / `logits`, or pass the wrong thing to `.sample(shape, seed)` / `.log_prob(value)` / `.kl_divergence(other)`, the failure happens deep inside the framework with a trace that points to internal tensor ops, not to my call site.\n\nThe Python entry points of `roi_perspective_transform` and `locality_aware_nms` have the same issue — wrong dtype on `input` / `bboxes` / `scores`, or a wrong type for one of the scalar attributes (`transformed_height`, `score_threshold`, `nms_top_k`, etc.) only blows up later.\n\n**3. `mine_hard_examples` attribute checks are similarly opaque** — e.g. `neg_pos_ratio must greater than zero in max_negative mode` doesn't include the offending value.\n\n### What I'd like\n\nFor these ops/classes (`Normal`, `Uniform`, `Categorical`, `MultivariateNormalDiag`, `roi_perspective_transform`, `locality_aware_nms`, `mine_hard_examples`):\n\n- When an input has the wrong shape / rank / dimension / value, the error message should include the actual offending value so I can see at a glance what I sent in.\n- Obvious type / dtype mistakes on the user-facing API should be caught right at the Python entry point with a message that names the offending argument and the API it was passed to, instead of leaking through to the C++ kernel.\n\nThis is purely a usability/diagnostics improvement — the op semantics shouldn't change for valid inputs."} {"task_id": "format-code-task-000197", "source_id": "format-code-task-000197", "domain": "code", "task_path": "tasks/format-code-task-000197", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:8dc8a7f63282c3fbc2192d79b9237f93e84a9bbff3fb04d8f08881ba68856f56", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nfix in position and text of all organization button\n**Is your feature request related to a problem? Please describe.**\nAll organization button is not intutive\n\n**Describe the solution you'd like**\nPosition the all organization button at the top of the menu (i.e above dashboard button) and change all organization text to my organizations\n\n\n**Additional context**\nAdd any other context or screenshots about the feature request here.\n\"Screenshot\n\n**Potential internship candidates**\nPlease read this if you are planning to apply for a Palisadoes Foundation internship https://github.com/PalisadoesFoundation/talawa/issues/359"} {"task_id": "format-code-task-000198", "source_id": "format-code-task-000198", "domain": "code", "task_path": "tasks/format-code-task-000198", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:957ae362ff0068a1abf2a238f2a464fe115233cdd001a9c41d643975fa0af895", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Avatar image helpers for profile uploads\n\nWe're building out profile-image uploads for the user settings screens. The UI needs two small,\nwell-tested helpers to deal with avatars, and right now they don't exist. Please add them as\nutilities under `src/utils`.\n\n## 1. `sanitizeAvatars(file, fallbackUrl)`\n\nExport a function `sanitizeAvatars` from `src/utils/sanitizeAvatar.ts` with the signature\n`(file: File | null, fallbackUrl: string) => string`. It decides what image source to show:\n\n- If `file` is a real `File` whose MIME type is an image (its `type` starts with `image/`), return\n an object URL created from that file (i.e. the result of `URL.createObjectURL(file)`).\n- Otherwise (the file is `null`, or it is not an image), fall back to `fallbackUrl`:\n - Resolve `fallbackUrl` as a URL against the current page origin (`window.location.origin`) and\n return the resulting normalized absolute URL string. A relative path like `/avatar.jpg` must\n become an absolute URL under the current origin; query strings and fragments must be preserved;\n non-ASCII characters end up percent-encoded as part of normalization.\n - If `fallbackUrl` is missing/empty or cannot be parsed into a valid URL, log an error to\n `console.error` and return an empty string `''`.\n\nThe function must never throw.\n\n## 2. `urlToFile(url)`\n\nExport an async function `urlToFile` from `src/utils/urlToFile.ts` with the signature\n`(url: string) => Promise`. It downloads a remote image so it can be re-uploaded as multipart\nform data:\n\n- Fetch the given `url` and read the response body as a `Blob`.\n- Return a `File` built from that blob. The file's `type` is the blob's MIME type. The file's name\n is the URL's final path segment, or `avatar` when the URL has no final segment (e.g. it ends in a\n slash), followed by a dot and the extension taken from the blob's MIME subtype (the part after the\n `/`). For example a blob of type `image/png` fetched from a URL ending in `/` produces a file\n named `avatar.png`.\n- If anything goes wrong (the fetch rejects, or reading the blob rejects), log the error to\n `console.error` and re-throw it so the caller's promise rejects with the same error.\n"} {"task_id": "format-code-task-000199", "source_id": "format-code-task-000199", "domain": "code", "task_path": "tasks/format-code-task-000199", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:22bfcee8dcef7a1f4b843daab09df578164d7fc3e6597c7ab7d19ff84fe6e050", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nThis seems to be due to a bug in how we are interfacing with autograd, or -- less likely -- a bug in autograd itself.\n\nIf the user provides a quantum function `func` with multiple input arguments in its signature, we can only successfully call `qml.jacobian(func, argnum=0)`, `qml.jacobian(func, argnum=1)`, etc., but not `qml.jacobian(func, argnum=[0,1])` (raises some error inside autograd). \n\nSince we've coded `jacobian` to behave nicely if all the arguments are combined into an array, a suitable alternate usage is available. Would be nice to figure this bug out at some stage"} {"task_id": "format-code-task-000200", "source_id": "format-code-task-000200", "domain": "code", "task_path": "tasks/format-code-task-000200", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0e92bc099da3746dbdd108c3c44bf5dbb998658950327c181db933320b4a56ea", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\n我在用 `qml.adjoint` 去取一个电路的伴随，里面带了个 `qml.Barrier` 做可视化分隔，结果直接报错跑不动。把 Barrier 去掉就正常。是不是 Barrier 在 adjoint 里没被处理？\n\n# Expected outcomes\n\n- Circuits or quantum functions containing `qml.Barrier` should be usable with `qml.adjoint` without raising an unsupported-operation error solely because the barrier is present.\n- When an adjointed circuit/function contains a barrier, the resulting operation sequence should preserve an equivalent `qml.Barrier` acting on the same wires at the corresponding point in the adjointed sequence.\n- A standalone `qml.Barrier` operation should behave consistently with other operations that can participate in adjoint construction, producing an equivalent barrier on the same wires when adjointed.\n\n# Implementation notes\n\n- The barrier is a visual/no-op separator, so its adjoint behavior should preserve that role rather than introduce a physical transformation.\n- The exact implementation location and internal mechanism are up to the implementer, as long as the public behavior above is satisfied and existing behavior of barriers outside adjoint construction is preserved."} {"task_id": "format-code-task-000201", "source_id": "format-code-task-000201", "domain": "code", "task_path": "tasks/format-code-task-000201", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:31472b3d617928961ce3a8018ddd580e5c83ce4f4420ec31499436be0e70c5d5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\n用 `qml.hf.hamiltonian` 生成分子哈密顿量太慢了，跑个 LiH 居然要等三四分钟……我看了下结果里好多重复的 Pauli term 没合并，还有一堆系数小到可以忽略不计的项也都留着，估计就是这些拖慢的。能不能在 hf 里给我一个能直接调用的简化函数，把相同 Pauli word 的项加一加、再顺手把那种几乎为零的小系数项扔掉？最好还能让我自己指定一个阈值，多小算\"可以扔\"由我说了算。\n\n# Expected outcomes\n\n- Public HF simplification API\n - `pennylane.hf.hamiltonian.simplify` should be importable and callable as a public helper for simplifying a `qml.Hamiltonian`.\n - Calling `simplify(h)` should return a `qml.Hamiltonian` representing the simplified observable.\n\n- Duplicate Pauli-word handling\n - When a Hamiltonian contains multiple terms with the same Pauli word, `simplify` should combine them by summing their coefficients.\n - If combining duplicate terms makes the resulting coefficient negligible under the active cutoff, that Pauli word should not appear in the returned Hamiltonian.\n\n- Cutoff handling\n - `simplify` should accept a user-provided `cutoff` value.\n - Terms whose absolute coefficient is at or below the cutoff should be omitted from the returned Hamiltonian.\n - Terms whose absolute coefficient is above the cutoff should be retained.\n\n- Molecular Hamiltonian construction\n - `qml.hf.hamiltonian(...)` should return a Hamiltonian without redundant duplicate Pauli-word terms in its final result.\n - The cutoff used during `qml.hf.hamiltonian(...)` construction should affect whether very small terms are retained or discarded in the returned Hamiltonian.\n\n# Implementation notes\n\n- The exact internal representation, grouping strategy, and point at which simplification is applied are up to the implementation, as long as the public behavior above is satisfied.\n- Preserve existing public behavior of HF Hamiltonian generation aside from the requested simplification and cutoff semantics.\n- Tests and user code should rely on the returned Hamiltonian’s observable behavior rather than on any private helper or internal ordering choice."} {"task_id": "format-code-task-000202", "source_id": "format-code-task-000202", "domain": "code", "task_path": "tasks/format-code-task-000202", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:653a11da26846e3dc7acc7d8f099c4c6a675d670a3abc83058360633fbd10f22", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nSupport adding observable to scalar `0`?\n### Feature details\n\nIt would be nice to be able to do the following:\n```\nH = sum([qml.PauliX(i) for i in range(10)])\n```\n\nThe issue here is that `sum` naturally starts at integer 0 but there's no way to add observables to an integer 0. Philosophically it seems valid to use `0` as the zero element of dxd matrices (similar behavior is supported for numpy arrays, e.g.).\n\n### Implementation\n\n```\n def __add__(self, other):\n r\"\"\"The addition operation between Observables/Tensors/qml.Hamiltonian objects.\"\"\"\n if isinstance(other, numbers.Number) and other == 0:\n return self\n if isinstance(other, qml.Hamiltonian):\n return other + self\n if isinstance(other, (Observable, Tensor)):\n return qml.Hamiltonian([1, 1], [self, other], simplify=True)\n raise ValueError(f\"Cannot add Observable and {type(other)}\")\n```\nI'm assuming anything that is a `Number` implements `__eq__`, might need to double check that though.\n\n### How important would you say this feature is?\n\n1: Not important. Would be nice to have.\n\n### Additional information\n\nWorkarounds:\n```\nH = sum([qml.PauliX(i) for i in range(10)], start=0*qml.PauliX(0)) # meh\nH = qml.PauliX(0)\nfor i in range(1, 10):\n H += qml.PauliX(i) # ugh\n```\n\nFor reference, the corresponding operation works in cirq:\n```\nH = sum([cirq.X(j) for j in cirq.LineQubit.range(10)])\n```"} {"task_id": "format-code-task-000203", "source_id": "format-code-task-000203", "domain": "code", "task_path": "tasks/format-code-task-000203", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3e84ec0d461f7934a6c2dc33952fda198255c717778f8d54bb1acd49ec49a897", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `qml.matrix` on `qml.exp` fails when the base operator skips wires and the coefficient is trainable\n\nI'm building a parameterized exponential of a multi-qubit Pauli string and trying to grab its matrix representation. As soon as the tensor product has a \"gap\" in the wires it acts on (e.g. wires 0, 1, 4 but nothing on 2 and 3), and the coefficient is a trainable autograd array, `qml.matrix` blows up.\n\nMinimal reproducer:\n\n```python\nimport pennylane as qml\nfrom pennylane import numpy as np\n\nop = qml.exp(\n qml.PauliZ(wires=0) @ qml.PauliY(wires=1) @ qml.PauliZ(wires=4),\n coeff=np.array(0.25),\n)\nqml.matrix(op)\n```\n\nThis raises a `ValueError` from inside the `Exp.matrix` path complaining about a matmul shape mismatch (something about size 8 vs size 2). The same call works fine if I either:\n\n- pass a plain Python float as the coefficient instead of `np.array(0.25)`, or\n- use a Pauli string whose wires are contiguous (e.g. `PauliZ(0) @ PauliY(1) @ PauliZ(2)`).\n\nSo it only seems to break in the combination \"trainable autograd coeff\" + \"wires of the base op are non-contiguous\". I'd expect `qml.matrix(op)` to just return the matrix of the exponential on the wires the operator declares, regardless of whether those wires happen to be contiguous, and I'd like to keep the coefficient as a trainable autograd array so I can differentiate through it later."} {"task_id": "format-code-task-000204", "source_id": "format-code-task-000204", "domain": "code", "task_path": "tasks/format-code-task-000204", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d17c2f81936648d4380660d241455373357dbd10cc816e54c9ead3d3ff5f6006", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `qml.specs` reports `depth=None` for circuits with custom multi-layer operations\n\nI've been defining my own operations by subclassing `ResourcesOperation` so I can plug them into PennyLane circuits while still being able to inspect resource counts. Some of these custom ops are themselves multi-layer constructs (a small ansatz or a decomposition I've worked out by hand), so their `resources()` returns a `Resources` object with `depth` greater than 1.\n\nThe problem is that as soon as I put one of these into a QNode and look at the specs, the overall circuit depth comes back as `None` instead of a number.\n\nMinimal example:\n\n```python\nimport pennylane as qml\nfrom pennylane.resource import Resources, ResourcesOperation\n\nclass MyLayeredOp(ResourcesOperation):\n num_wires = 2\n def resources(self):\n # this op internally is e.g. 3 layers deep\n return Resources(num_wires=2, num_gates=4,\n gate_types={\"Hadamard\": 2, \"CNOT\": 2},\n gate_sizes={1: 2, 2: 2},\n depth=3)\n\ndev = qml.device(\"default.qubit\", wires=2)\n\n@qml.qnode(dev)\ndef circuit():\n qml.Hadamard(0)\n MyLayeredOp(wires=[0, 1])\n qml.CNOT(wires=[0, 1])\n return qml.expval(qml.PauliZ(0))\n\nprint(qml.specs(circuit)())\n```\n\nThe `\"depth\"` entry in the specs dict is `None`. If I change `MyLayeredOp.resources()` to return `depth=1`, the specs dict reports a proper integer depth again, so the issue is specifically about custom ops whose self-reported depth is bigger than 1.\n\nFrom a user's point of view this is pretty inconvenient — the whole reason I'm using `ResourcesOperation` and giving it a meaningful `depth` is so that tools like `qml.specs` can give me a faithful picture of the circuit's resource footprint. Falling back to `None` means I have to compute depth manually whenever any of my custom ops is more than one layer deep, which defeats the purpose.\n\nCould `qml.specs` (and whichever underlying depth computation it relies on) be made to honor the `depth` declared by a custom `ResourcesOperation`, so that an integer circuit depth is returned even when the circuit contains custom ops with `depth > 1`?"} {"task_id": "format-code-task-000205", "source_id": "format-code-task-000205", "domain": "code", "task_path": "tasks/format-code-task-000205", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0d59d94464ca515c8b62287423c336869d1a7e2ca0c521e9c38d53c7168670fd", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Preserve module state across hot updates\n\nOur Hot Module Replacement support lets a project's JS files talk to the bundler through\n`module.hot` — they can register `accept` and `dispose` handlers, query the current status, and\nso on. The one thing that's missing is a way to carry a little bit of state from an old version\nof a module to the version that replaces it.\n\nRight now, when a module is hot-swapped its `dispose` handler fires, but there's no channel for\nthat handler to hand anything off to the incoming copy of the module. As soon as the new code\nruns, whatever the old instance knew is gone. That makes it impossible to do things like keep a\nscroll position, a timer handle, or some accumulated counter alive across an edit.\n\nAdd the standard \"hot data\" mechanism that other bundlers expose:\n\n- A module's `dispose` callback should be invoked with a single argument: a plain, initially-empty\n object that the handler can write into. This is the module's chance to stash anything it wants\n to survive the swap.\n- When the replacement version of that same module initializes, it should be able to read back\n exactly that object as `module.hot.data`.\n- A module that has not been replaced yet (its very first run) must see `module.hot.data` as\n `undefined` — there's nothing to restore.\n- Each hot update gets its own fresh hand-off object. The object a `dispose` handler receives must\n start empty every time; state from earlier updates only carries forward if the module\n deliberately reads `module.hot.data` and copies it into the new object. Data is tracked per\n module, so unrelated modules never see each other's hand-off objects.\n\nThe existing `module.hot` behavior (accept, dispose firing, status reporting, the websocket update\nflow) must keep working unchanged.\n"} {"task_id": "format-code-task-000208", "source_id": "format-code-task-000208", "domain": "code", "task_path": "tasks/format-code-task-000208", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d2b31887935ae048c20855edc6d4215241dd3d9242711e66917d37641e1f1d2a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## gsdkConfig.json contains placeholder values for node IP and VM ID\n\nWhen a game server pod starts up under thundernetes, the init container writes `gsdkConfig.json` (at `/data/Config/gsdkConfig.json`) for the GSDK to consume. Two fields in that file are currently useless:\n\n- `publicIpV4Address` (both at the top level and inside `gameServerConnectionInfo`) is always the literal string `\"N/A\"`.\n- `vmId` is always the same hard-coded string regardless of which node the pod is actually scheduled on.\n\nExample of what I see in the file right now:\n\n```json\n{\n \"sessionHostId\": \"...\",\n \"vmId\": \"thundernetes-aks-cluster\",\n \"publicIpV4Address\": \"N/A\",\n \"gameServerConnectionInfo\": {\n \"publicIpV4Address\": \"N/A\",\n \"gamePortsConfiguration\": [ ... ]\n },\n ...\n}\n```\n\nThis means a game server using GSDK has no way to know:\n\n1. what IP address it can actually be reached on, and\n2. which node / VM it is running on (every server in the cluster reports the same `vmId`).\n\nFor our use case we don't necessarily need a routable public IP at this layer — the IP of the node the pod landed on is good enough for the server to advertise itself and for us to distinguish instances. The `vmId` should similarly be tied to the actual node, not a single constant shared by every game server in the cluster.\n\nCould the init container populate these fields with values that reflect the node the pod is actually running on, instead of fixed placeholders? I'd expect the init container to pick up the node's IP from a new `PF_*` env variable (something like `PF_NODE_INTERNAL_IP`) that the operator injects per-pod."} {"task_id": "format-code-task-000209", "source_id": "format-code-task-000209", "domain": "code", "task_path": "tasks/format-code-task-000209", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e6fd4b39817da34d00a85fb66e8f4449c686df89c3449bca116f8e3e7e060b20", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI have a cohort that hit an error during calculation, and now it just sits there looking like it’s still calculating forever. From the API/UI I can’t tell that it failed or how many times it’s failed, and the background recalculation seems to keep picking it up again.\n\n## Expected outcomes\n\n- Cohort API responses expose an `errors_calculating` value so API/UI clients can see how many calculation failures have been recorded for a cohort.\n- When a cohort calculation fails, the cohort should no longer appear to be actively calculating, and its recorded calculation failure count should increase.\n- When a cohort calculation later succeeds, the cohort should finish calculating normally, update its successful calculation metadata, and clear its recorded calculation failure count.\n- Periodic/background cohort recalculation should not keep automatically scheduling cohorts that have already failed repeatedly; cohorts with three or more recorded calculation failures should be skipped by that automatic recalculation path.\n\n## Implementation notes\n\n- The exact storage, validation, and error-handling structure is up to the implementer, as long as the externally observable API and recalculation behavior above are satisfied.\n- Preserve existing cohort calculation behavior for successful cohorts, aside from clearing any prior recorded failure count on success."} {"task_id": "format-code-task-000210", "source_id": "format-code-task-000210", "domain": "code", "task_path": "tasks/format-code-task-000210", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:bd462c1c22a15017ebf4174124c506be2c562e9d2cba3a38a2e1644027e19714", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Translate Segment destination config into PostHog hog-function config\n\nWe're starting to import [Segment](https://segment.com) action destinations into\nPostHog's CDP. Segment describes *what* to send using two pieces of config that we\nneed to convert into the equivalents PostHog already understands:\n\n1. A **subscription / FQL string** that decides which events a mapping fires on.\n2. **Field \"default\" values** that describe where each piece of data comes from,\n written with Segment's directive objects (`@path`, `@if`).\n\nAdd a small, dependency-free translation layer for these two pieces. It must live in\nthe plugin-server CDP code and expose two pure functions from\n`plugin-server/src/cdp/segment/segment-templates.ts`:\n\n```ts\ntranslateFilters(subscribe: string): { events: HogFunctionFilterEvent[] }\ntranslateInputs(defaultVal: any, multiple?: boolean): any\n```\n\n## `translateFilters`\n\nSegment's FQL talks about event `type`s; PostHog talks about event names. Given a\nsubscription string, return a single PostHog event filter:\n\n```ts\n{\n events: [\n {\n id: 'All events',\n name: 'All events',\n type: 'events',\n order: 0,\n properties: [{ key: , type: 'hogql', value: null }],\n },\n ],\n}\n```\n\n`` is the input string after applying these rewrites to **every**\noccurrence:\n\n| Segment FQL fragment | PostHog HogQL fragment |\n|---|---|\n| `type = \"page\"` | `event = \"$pageview\"` |\n| `type = \"screen\"` | `event = \"$screen\"` |\n| `type = \"identify\"` | `event in ('$identify', '$set')` |\n| `type = \"group\"` | `event = \"$groupidentify\"` |\n| `type = \"track\"` | `event not in ('$pageview', '$screen', '$alias', '$identify', '$set', '$groupidentify')` |\n| `type = \"alias\"` | `event = \"$alias\"` |\n\nAfter those substitutions, every remaining double quote (`\"`) in the string becomes a\nsingle quote (`'`). Fragments not listed above (e.g. a `name = \"Order Completed\"`\nclause joined with `and`/`or`) pass through unchanged apart from the quote conversion.\n\n## `translateInputs`\n\nThis resolves a single Segment field default into the string/value PostHog stores for\na hog-function input.\n\n- A `boolean` or `string` default is returned **unchanged**.\n- An object with an **`@path`** key: take the path string, replace a leading `$.` with\n `event.`, normalize the field reference (see below), then:\n - if normalization yields an empty string, return `''`;\n - otherwise return it wrapped in PostHog templating braces: `{}`, or\n `{[]}` when `multiple` is `true`.\n- An object with an **`@if`** key shaped like `{ exists, then, else }`:\n - Only handle the common \"exists check on the same field\" case, i.e. when\n `JSON.stringify(exists)` equals `JSON.stringify(then)`. In any other shape, return\n `JSON.stringify(defaultVal)`.\n - Resolve `then` and `else` independently: an `@path` object is converted exactly\n like the `@path` case above but **without** the surrounding braces; a plain string\n is wrapped in single quotes (`'value'`).\n - Combine the resolved primary (`then`) and fallback (`else`):\n - both empty → `''`;\n - primary empty, fallback present → `{}`;\n - primary present, fallback empty → `{}`;\n - both present → `{ ?? }`, and when `multiple` is `true` the\n `?? ` expression is wrapped in brackets: `{[ ?? ]}`.\n- Any other object is returned as `JSON.stringify(defaultVal)`.\n\n### Field reference normalization\n\nAfter the `$.` → `event.` step, field references are rewritten to their PostHog\nequivalents. These specific references must be remapped (matching anywhere in the\nstring):\n\n| Segment reference | PostHog reference |\n|---|---|\n| `event.traits` | `person.properties` |\n| `event.context.traits` | `person.properties` |\n| `event.userId` | `person.id` |\n| `event.anonymousId` | `event.distinct_id` |\n| `event.messageId` | `event.uuid` |\n| `context.os.name` "} {"task_id": "format-code-task-000211", "source_id": "format-code-task-000211", "domain": "code", "task_path": "tasks/format-code-task-000211", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3cdaa4fe016c918896bc7796b32a15357bfc3f64630d323e7fece890806fc855", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a declarative config abstraction for data-import sources\n\nOur data-import sources hand-roll a lot of boilerplate to turn loosely-typed\ndictionaries (coming from API requests and from rows stored in our database)\ninto typed configuration objects. We want a small, reusable abstraction built on\ntop of the standard library's dataclasses so that defining a config and\ninitializing it from a dictionary becomes declarative.\n\nAdd a new module `posthog/temporal/data_imports/pipelines/source/config.py`\nthat is importable on its own (it should only depend on the standard library)\nand exposes the following public surface: a class decorator `config`, a field\nhelper `value`, and a converter helper `str_to_bool`.\n\n## `config` decorator\n\n`config` turns a plain class into a dataclass and additionally gives it a\n`from_dict` classmethod. All the usual dataclass behavior must keep working\n(generated `__init__`, equality, `repr`, field access, defaults, etc.).\n\nIt must be usable both bare and with keyword arguments:\n\n```python\n@config\nclass A: ...\n\n@config(prefix=\"db\")\nclass B: ...\n```\n\n`from_dict(d)` is a classmethod that builds and returns an instance of the class\nfrom a dictionary `d`.\n\nFor a class with only scalar fields, each field is looked up in `d` by its field\nname. Keys present in `d` that don't correspond to a field are ignored. If a\nfield is absent from `d`, its declared default (if any) is used; if such a field\nhas no default, calling `from_dict` raises `TypeError` (the dataclass\nconstructor's error for a missing required argument).\n\nIf the decorator is given a `prefix`, every field of that class is instead\nlooked up under that prefix joined with the field name by an underscore, e.g.\nwith `@config(prefix=\"db\")` a field `host` is read from the key `db_host`.\n\n## `value` field helper\n\n`value` is the analogue of `dataclasses.field` for configs. It supports the\nusual `default` and `default_factory`, plus three config-specific keyword-only\noptions:\n\n- `alias`: look the field up under this key instead of the field name.\n- `prefix`: look the field up under `_` (this replaces any prefix\n the field would otherwise inherit).\n- `converter`: a one-argument callable applied to the value read from the\n dictionary before it is stored on the instance. Defaults must not be passed\n through the converter.\n\n## Nested configs\n\nA field whose declared type is itself a `config`-decorated class is built\nrecursively. There are two supported layouts in the source dictionary:\n\n1. **Nested dict.** If `d` contains a key equal to the field's name (or its\n `alias`) and the corresponding value is a dictionary, the nested config is\n built from that sub-dictionary, looking up its fields by their own names.\n\n2. **Flat dict.** Otherwise the nested config is built from the *same*\n dictionary `d`, with its fields looked up under an additional prefix. The\n prefix is resolved in this precedence order:\n - the field's `value(prefix=...)`, if set; otherwise\n - the nested class's own `@config(prefix=...)`, if set; otherwise\n - a prefix derived automatically from the nested class's name.\n\n The automatically derived prefix is the class name split into words and\n joined by underscores in lower case. Words break on a lower-to-upper\n transition, and a run of consecutive capitals is treated as one word except\n that its final capital starts the next word. A trailing `Config` word is\n dropped unless it is the only word. For example: `SSHTunnel` and\n `SSHTunnelConfig` both derive `ssh_tunnel`, `DatabaseConnection` derives\n `database_connection`, `MyClass` derives `my_class`, `Test` derives `test`,\n and `Config` derives `config`.\n\n These prefixes accumulate across nesting levels (a doubly-nested config's\n fields are looked up under the concatenation of the prefixes).\n\n## `str_to_bool` converter\n\n`str_to_bool(s)` is a convenience converter (suitable for use as a\n`value(converter=...)`). If `s` is already a `b"} {"task_id": "format-code-task-000212", "source_id": "format-code-task-000212", "domain": "code", "task_path": "tasks/format-code-task-000212", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0af42fef65891372abbeea8cddcf11bc2c4f3f8a436b338777dc7877988d37bc", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Docker agent fails to deploy a flow run when a container with the same name already exists\n\nI'm running flows via the Prefect Docker Agent. The container name is derived from the flow run name (slugified), which is fine — but if there's *already* a container with that name on the host (e.g. a stopped container left over from a previous run, or two flow runs that happen to slug to the same name), the agent fails to deploy the flow run.\n\n### What I see\n\n1. Start a flow that produces some container name (whatever the slug of the flow run name ends up being).\n2. The flow run finishes and the container is left around (stopped, not removed).\n3. Trigger another run of the same flow.\n4. The agent logs an error from the Docker daemon complaining that the container name is already in use, and the flow run never starts. The agent does not retry with a different name.\n\nThe same thing happens if I happen to have an unrelated container on the host that just happens to share the name.\n\n### What I expected\n\nThe agent already tries to pick a unique container name — there's logic that's supposed to detect a collision and append `-1`, `-2`, etc. to the slug. But in practice that logic doesn't seem to be doing anything: the conflict still slips through and `create_container` blows up. From a user's point of view, a name collision shouldn't be fatal — the agent should just pick the next free name and carry on deploying the flow run.\n\nIt would also help debugging a lot if the \"starting container\" log line included the actual container name that ended up being used, not just the container ID, so I can correlate agent logs with `docker ps` output when names get suffixed."} {"task_id": "format-code-task-000213", "source_id": "format-code-task-000213", "domain": "code", "task_path": "tasks/format-code-task-000213", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c7696eba171dcc973c259967822610a8445998aabd3f3f35a29fc743b3619614", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI’m starting a flow run from Prefect Cloud with a custom context value like `my_var`, but inside the flow and tasks `prefect.context.get(\"my_var\")` is coming back as `None`. I also noticed `Client.get_flow_run_info(...)` gives me the run info but doesn’t show that context value, even though I can see it set on the Cloud run, so I’m not sure if I’m passing it wrong or if it’s getting lost somewhere.\n\n**Expected outcomes**\n- Cloud flow runs should make the stored per-run context available during execution, so flow code and task code can read custom values through `prefect.context`.\n- If Cloud context includes keys that are also part of the runner-populated runtime context, including scheduled start time and flow run version metadata, the runtime-generated values should still win for those keys.\n- `Client.get_flow_run_info(flow_run_id: str)` should expose the flow run’s stored context through the returned `FlowRunInfoResult`.\n- `FlowRunInfoResult` should carry the flow run’s context together with the other run metadata already returned by the client.\n\n**Implementation notes**\n- The exact place where the Cloud context is fetched, merged, and serialized is up to the implementation.\n- Keep existing behavior for unrelated flow-run metadata and failure handling unless it is required to satisfy the outcomes above.\n- Any internal data structures, helper functions, and merge mechanics may vary as long as the externally observable behavior matches the outcomes."} {"task_id": "format-code-task-000214", "source_id": "format-code-task-000214", "domain": "code", "task_path": "tasks/format-code-task-000214", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:346d40271045ce3b510b28a120f54b39cb70921467c512ec526f8426714bd46a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Strengthen multi-shop authorization in the employee context\n\nIn the back office, the employee context already tells us whether the logged-in\nemployee is allowed to work on a given shop or on a given shop group. In a\nmultistore installation that is not enough: some operations target *all* shops\nat once (the \"all stores\" context), and other code needs to resolve an\nauthorization decision directly from a shop constraint without having to branch\non its kind every time. Right now there is no way to ask the context either of\nthose questions, which lets employees act on the whole installation even when\nthey are only associated with a subset of its shops.\n\nExtend the employee context with two new capabilities.\n\n**1. \"Is this employee authorized for the whole installation?\"**\n\nAdd a way to ask whether the current employee has authorization for *all* the\nshops that exist in the installation:\n\n- If there is no logged-in employee, the answer is `false`.\n- A super administrator is always authorized for all shops.\n- Otherwise the employee is authorized for all shops only when they are\n authorized on every single shop that exists in the installation. If even one\n existing shop is outside their associated shops, the answer is `false`.\n Being associated with extra shops that no longer exist does not matter.\n\nFor this to be answerable, the context needs to know which shops exist. The\ncomplete list of existing shop ids must be supplied to the context when it is\nbuilt, in addition to the employee. The current way of constructing the context\nwith the employee alone must keep working unchanged; when no shop list is\nsupplied the context behaves as if the installation contains no shops, so any\nlogged-in employee is trivially authorized for \"all\" of them.\n\n**2. \"Is this employee authorized for this shop constraint?\"**\n\nAdd a way to resolve an authorization decision from a `ShopConstraint` value\nobject (`PrestaShop\\PrestaShop\\Core\\Domain\\Shop\\ValueObject\\ShopConstraint`):\n\n- A constraint targeting a single shop is authorized exactly when the employee\n is authorized on that shop.\n- A constraint targeting a shop group is authorized exactly when the employee\n is authorized on that shop group.\n- A constraint targeting all shops is authorized exactly when the employee is\n authorized for all shops (as defined above).\n\nThese checks must stay consistent with the existing per-shop and per-shop-group\nauthorization rules (super administrators pass everything; a context with no\nemployee passes nothing).\n"} {"task_id": "format-code-task-000215", "source_id": "format-code-task-000215", "domain": "code", "task_path": "tasks/format-code-task-000215", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:95c36bf49ce81bddcba69e014ae6a1938f3862422687ea2f7441bb3c7326c86a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Inverted predictions don't flow cleanly into `SaveImage` / can't switch `collate_fn` in `Invertd`\n\nI'm building a post-transform pipeline for a segmentation model:\n\n```python\npost_transforms = Compose([\n Activationsd(keys=\"pred\", sigmoid=True),\n AsDiscreted(keys=\"pred\", threshold_values=True),\n Invertd(\n keys=\"pred\",\n transform=pre_transforms,\n loader=val_loader,\n orig_keys=\"image\",\n ...\n ),\n SaveImaged(keys=\"pred_inverted\", output_dir=\"./out\", ...),\n])\n```\n\nThe pre-transforms apply spatial transforms with different parameters per sample (random crops / spacings), so after inversion each sample has a different shape. With the default `collate_fn` in `Invertd` (no collation), `pred_inverted` ends up as a Python list of per-sample tensors without a batch dim — which makes sense, you can't stack tensors of different shapes back into a batch.\n\nThe problem is feeding that list into `SaveImage` / `SaveImaged`. `SaveImage` only knows two modes: `save_batch=True` (input is `[B, C, H, W, ...]`) or `save_batch=False` (input is `[C, H, W, ...]`). A list of channel-first tensors falls through neither, and it just crashes / writes garbage. `SegmentationSaver` does handle the list case internally, but `SaveImaged` (which I'd prefer to use in the dict pipeline) does not — so I end up having to drop out of the `Compose` and use the handler.\n\nSeparately, I tried working around this by passing a real collate function to `Invertd` (e.g., `list_data_collate` for the cases where shapes do match) so I'd get a single collated dict back instead of a list of per-sample dicts. That also crashes inside `Invertd.__call__` — it unconditionally does `[post_func(i[orig_key]) for i in inverted]`, which assumes `inverted` is a list of dicts. When the collate function returns a single dict, iterating it like that doesn't do what you'd want.\n\nSo really two related asks:\n\n1. `SaveImage` (and `SaveImaged` by extension) should accept a list of channel-first tensors / arrays as input, and save each item with its corresponding meta dict — same idea that's already in `SegmentationSaver`, but lifted into the transform itself so it works inside a `Compose` post-transform pipeline.\n2. `Invertd` should respect whatever `collate_fn` the user passes — if it returns a list, behave like today; if it returns a single collated structure, handle that too instead of blowing up.\n\nIt would also be good if the docstrings for `Invertd` / `TransformInverter` / `SaveImage` / `SegmentationSaver` made it clear what shape the inverted output actually has (list-of-tensors-without-batch-dim vs. batched tensor), because right now it's pretty easy to assume you're getting a batched tensor back and only find out otherwise at the save step."} {"task_id": "format-code-task-000216", "source_id": "format-code-task-000216", "domain": "code", "task_path": "tasks/format-code-task-000216", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:181a746c9ab1614fa44f5542d87f6b5fe3e9c4f6ee02f974046fb36bb2ebab43", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI’m using MONAI and I’d like to track one of my loss functions the same way I track other metrics across validation iterations, but right now I have to write my own metric wrapper just to accumulate and aggregate it. It would be really helpful if MONAI had a generic metric for “use this loss function as a metric,” so I can plug in losses that take either predictions alone or predictions plus labels and then aggregate the results like the built-in metrics.\n\n## Expected outcomes\n\n- `monai.metrics.LossMetric` is available as a public metric wrapper that accepts a `loss_fn` callable and can be instantiated from `monai.metrics`.\n- Calling `LossMetric(loss_fn=...)` during validation computes the loss for the current iteration, returns that iteration’s loss value, and records it so multiple iterations can later be aggregated.\n- The wrapped loss callable works both when only `y_pred` is supplied and when both `y_pred` and `y` are supplied.\n- Loss outputs that are scalar or one-dimensional tensors can still be used through `LossMetric` and aggregated consistently with MONAI’s batch-first metric behavior.\n- `LossMetric.aggregate()` aggregates all recorded iteration losses using MONAI’s existing metric reduction modes.\n- The reduction configured on `LossMetric` can be overridden for a single `aggregate(reduction=...)` call.\n- When `LossMetric(get_not_nans=True)` is used, aggregation returns both the reduced metric value and the corresponding not-NaN count; otherwise aggregation returns only the metric value.\n- Calling `aggregate()` before any iteration has been recorded returns a zero tensor value, and with `get_not_nans=True` returns zero tensor values for both outputs.\n- The metrics documentation includes an API entry for `LossMetric`.\n\n## Implementation notes\n\n- Follow MONAI’s existing metric conventions for cumulative iteration metrics, reductions, reset behavior, tensors, and documentation style.\n- The internal class organization, helper functions, validation location, and storage details are up to the implementer as long as the public behavior above is satisfied."} {"task_id": "format-code-task-000217", "source_id": "format-code-task-000217", "domain": "code", "task_path": "tasks/format-code-task-000217", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0d23045655649d7bf700390817a2789306e60c10c90f2f405156d5ff569c7d67", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Support local (inline) IIIF manifests without fetching them\n\nRight now every manifest Mirador knows about is identified by a URL and is\nloaded by fetching that URL over the network. We're adding the ability to hand\nMirador a manifest whose JSON we *already have in memory* (for example, a\nmanifest that was generated locally or read from a dropped file) and have it be\nused directly — no network request should ever be made for it.\n\nPlease wire this through the relevant parts of the application state so the\nfollowing behaviors hold.\n\n## Adding a resource to the catalog\n\nThe action creator that adds a manifest to the resource catalog currently takes\nonly a manifest id. Extend it so that, in addition to the id, a caller may pass\n(in this order) the manifest's JSON and an object of extra properties to record\non the catalog entry.\n\n- When extra properties are supplied (e.g. `{ provider: 'file' }`), they must be\n merged onto the stored catalog entry alongside its `manifestId`.\n- Adding a resource with only an id must keep storing exactly `{ manifestId }`\n for that entry — no empty/`undefined` extra fields leaking in.\n- The existing catalog behavior is otherwise unchanged: entries are prepended\n (most-recent first) and de-duplicated by `manifestId`.\n- When the manifest's JSON is supplied while adding a resource, that JSON must be\n stored for the manifest directly and **no network request may be issued** for\n it. When no JSON is supplied, the manifest is fetched exactly as before.\n\n## Opening a window\n\nThe action creator that opens a window should accept an optional `manifest` (the\nmanifest's JSON) among its options. When a window is opened with an inline\n`manifest`, that JSON must be stored for the window's manifest directly and **no\nnetwork request may be issued** for it. Opening a window without an inline\nmanifest must keep fetching the manifest as before (and must still skip the\nfetch when the manifest is already available).\n\n## Not re-fetching local manifests in the resource list\n\nA resource list item that represents a locally-supplied manifest — identified by\na `provider` of `'file'` — must not kick off a manifest fetch when it appears.\nItems without that provider keep their current behavior (they fetch the manifest\nwhen it isn't already loaded/loading/errored).\n\nNothing about the public shapes of the resulting Redux state (the catalog array\nentries, the stored manifest JSON) should change beyond what is described above.\n"} {"task_id": "format-code-task-000218", "source_id": "format-code-task-000218", "domain": "code", "task_path": "tasks/format-code-task-000218", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fc72b32f1634ad850fba62a1067443e8e87407cab23cf0477f811aaad19dfabc", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nIn [this commit](https://github.com/Propaganistas/Laravel-Phone/commit/99f2d320a15f1a7e5d196a3e8e17b586756ab44f#diff-48e28eb93ccfcade09cdb083d128c8a87e08516d3f2b182445f1877626f8bff9), the `mobile` and `fixed-line` string parameters were removed from `Phone::setParameters()`. This breaks backwards compatibility (so should be a major release) and I don't think is intended."} {"task_id": "format-code-task-000219", "source_id": "format-code-task-000219", "domain": "code", "task_path": "tasks/format-code-task-000219", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2df961d81e8f1a1ca1ab13309cd5296812c9dfcb85fc26d9564356dc31dca312", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Dummy-coding categorical values in a Timeline\n\nA bunch of the extractors I'm using produce categorical / string-valued features — e.g. detected object labels, face expression categories, speech tokens, etc. When I build a `Timeline` from these and call `to_df()`, the resulting DataFrame has columns full of strings.\n\nThat's fine for inspection, but the moment I want to actually do something quantitative with it (correlations, regressions, feeding it into sklearn, etc.) I have to go and one-hot / dummy-code those columns myself outside of pliers, which is annoying and easy to get wrong when the same categorical variable appears at many different onsets across the Timeline.\n\nIt would be really useful if `Timeline` itself could give me back a version where categorical variables are expanded into binary indicator variables, so that the output of `to_df()` is directly usable in numeric pipelines. Numeric columns should be left alone — I only want the string-typed ones expanded by default — though being able to force expansion of all variables would also be handy for some use cases.\n\nI'm imagining the API as something like `tl.dummy_code(...)` returning a new Timeline, with a flag (e.g. `string_only`) to toggle whether numeric columns are also expanded."} {"task_id": "format-code-task-000220", "source_id": "format-code-task-000220", "domain": "code", "task_path": "tasks/format-code-task-000220", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:dcb9efa91b0c4517eb32db7ea11f61f8b3e1822d724ef7c861d5e58ee80b1a83", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Track virtualenv usage and clean up stale environments\n\nfades caches the virtualenvs it builds, but right now nothing ever removes the\nones you stopped using. Over time the cache directory just grows. I'd like fades\nto keep a lightweight record of *when* each cached virtualenv was last used, and\nto be able to purge the ones that haven't been touched in a while.\n\n## What I want\n\n**1. Expose the cached virtualenvs' metadata.**\n\nThe virtualenv cache class should grow a `get_venvs_metadata()` method that\niterates over the metadata of every virtualenv currently recorded in the cache\nindex, yielding one metadata mapping (the same `metadata` dict that was stored\nfor each venv, which includes its `env_path`) per cached virtualenv.\n\n**2. A usage manager.**\n\nAdd a `UsageManager` class that owns a plain-text *usage stats* file and is\nconstructed from the path to that file and the virtualenv cache\n(`UsageManager(stat_file_path, venvscache)`). Each virtualenv is identified by\nthe last path component of its `env_path` (its uuid).\n\nThe usage stats file has one record per line, formatted as the virtualenv's\nuuid, a single space, and a UTC timestamp in ISO-8601 form\n`YYYY-MM-DDTHH:MM:SS.ffffff` (i.e. `datetime.strftime(dt, \"%Y-%m-%dT%H:%M:%S.%f\")`),\ne.g.:\n\n```\n2b6e6b1e-... 2020-03-21T18:42:07.123456\n```\n\nThe manager must behave as follows:\n\n- *On construction*: if the usage stats file does not exist yet, create it and\n seed it with one record per virtualenv currently known to the cache, each\n stamped with the current UTC time. If the file already exists, leave it\n exactly as it is (do not rewrite or reorder it).\n\n- *Recording usage* (`store_usage_stat(venv_data)`): given a virtualenv's\n metadata (a mapping containing its `env_path`), append a new usage record for\n it stamped with the current UTC time. Appending must not drop or rewrite the\n records that are already there.\n\n- *Cleaning unused venvs* (`clean_unused_venvs(max_days_to_keep)`): given a\n maximum number of days to keep, look at the\n most recent usage timestamp recorded for each virtualenv. Any virtualenv whose\n most recent use was **strictly more than** that many days ago must be removed:\n its directory is destroyed on disk and its entry is dropped from the\n virtualenv cache. A virtualenv last used exactly that many days ago is kept.\n After cleaning, the usage stats file is rewritten in compacted form: exactly\n one record per surviving virtualenv (keeping its most recent timestamp), and\n no records at all for virtualenvs that were removed.\n\nThe \"days ago\" comparison is the whole-day difference between now and the\nrecorded timestamp (e.g. an environment last used 43 full days ago is older than\na 42-day threshold, while one used 42 days ago is not).\n"} {"task_id": "format-code-task-000221", "source_id": "format-code-task-000221", "domain": "code", "task_path": "tasks/format-code-task-000221", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:30f3fec5348a0a8b0f642c1b6b3f02a403ae64deb9f12939968cc0c375905d03", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## baron doesn't recognize PEP 515 underscores in numeric literals\n\nI'm using `baron` to parse and analyze Python source code in a project that targets Python 3.6+. Since [PEP 515](https://peps.python.org/pep-0515/) (accepted in 3.6), it's legal to use single underscores as visual separators inside numeric literals, and this style shows up all over modern codebases — readability for large constants, bitmasks, etc.\n\nbaron doesn't seem to know about this syntax.\n\n### Reproducer\n\n```python\nimport baron\n\nsrc = \"x = 1_000_000\\n\"\nprint(baron.dumps(baron.parse(src)))\n```\n\nAnything similar fails the same way:\n\n```python\ntotal = 1_000_000\nmask = 0xFF_FF_FF_FF\nflags = 0b1010_1010\nperms = 0o755_000\nratio = 1_000.5\n```\n\nAll of these are valid Python 3.6+ and CPython tokenizes each one as a single numeric literal. baron either errors out or splits the literal apart at the underscore (so `1_000_000` does not come back as a single number token the way `1000000` does), which makes the resulting tree unusable for anything that has to round-trip the source.\n\n### Expected\n\nbaron's tokenizer should accept underscore separators in numeric literals consistently with what CPython 3.6+ accepts — integers, floats, hex, octal and binary literals (and the corresponding `L`-suffixed long variants that baron already supports for Python 2). Effectively the same code that parses today without the underscores should parse identically when underscores are inserted between digits."} {"task_id": "format-code-task-000222", "source_id": "format-code-task-000222", "domain": "code", "task_path": "tasks/format-code-task-000222", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1c5c12de4f8d1ff9692da1e1d71f4f3f0e8c6cc1c485627767a1a36b3ad914ac", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nLibrary code frequently calls `warnings.warn()` on behalf of its users. When those calls omit `stacklevel`, Python points at the library internals instead of the user's call site, which makes the warning much less useful. Add a normal (non-B9xx) Bugbear diagnostic for this mistake.\n\nEmit `B028 No explicit stacklevel argument found.` at the start of a call when its callee resolves to the standard-library `warnings.warn` and the call has no explicit stack level. Recognize all four ordinary import forms: `import warnings`, `import warnings as alias`, `from warnings import warn`, and `from warnings import warn as alias`. A `stacklevel=` keyword or a third positional argument counts as explicit and suppresses B028. Because a `**mapping` expansion may contain `stacklevel`, also suppress B028 whenever one is present; ordinary named arguments such as `message=` or `category=` do not count.\n\nKeep the check tied to the imported standard-library callable. Do not report unrelated `.warn()` methods, a `warn` imported from another module, or a plain user function. A simple assignment to a recognized alias stops it from identifying `warnings` or `warnings.warn` from that point in the same statement sequence, while a later supported re-import restores the binding. Module-level warning aliases remain usable inside nested function bodies, except where function parameters shadow them. Imports made inside a function are recognized there and do not leak back to module scope.\n\nIntegrate this without changing the existing configurable B008 immutable-call behavior or B904 exception-chaining diagnostics."} {"task_id": "format-code-task-000223", "source_id": "format-code-task-000223", "domain": "code", "task_path": "tasks/format-code-task-000223", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:8974cc151357c8eb3f3680ac9f8b8b89f5fb2c7444ce76ad72f257538d9244c1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Comments above the first import get mangled after sorting\n\nI have a Python file where I put a short `#` comment line directly above an import to explain what it's for. After running `isort` on the file, the comment ends up duplicated / misplaced — it loses its \"this comment belongs to that import\" relationship.\n\nMinimal example. Input file:\n\n```python\n\"\"\"Module docstring.\"\"\"\n# Comment explaining the next import\nfrom foo import bar\n\nfrom baz import qux\n```\n\nAfter `isort file.py`, the output around the top of the file is no longer right — the comment line `# Comment explaining the next import` is not where I'd expect it to be relative to `from foo import bar` anymore. It looks like the breakage is specifically around comments attached to the very first import in the file; if I move the same comment to sit above the *second* import instead, isort handles it fine.\n\nI'd expect isort to preserve a `#` comment that sits directly above an import as a comment belonging to that import, regardless of whether that import happens to be the first one in the file or not. Right now I have to manually clean up the top of every file after running isort, which kind of defeats the point.\n\nTested on the current `develop` checkout."} {"task_id": "format-code-task-000224", "source_id": "format-code-task-000224", "domain": "code", "task_path": "tasks/format-code-task-000224", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:121e1c44a3bdf037e1a0c5fd6c47d322fc2ab5514d9bdbba0f226fc32ac70d82", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## D300 doesn't catch docstrings that start with too many quotes\n\nI run `pydocstyle` over my codebase to enforce PEP 257. While reviewing some code today I noticed I had typoed a docstring like this:\n\n```python\ndef my_function():\n \"\"\"\"Compute something useful.\"\"\"\n pass\n```\n\nNotice the **four** opening `\"` instead of three — a typo. This is still valid Python (the parser just treats the 4th `\"` as the first character of the docstring body), but it's clearly not what I wanted, and any human reader would call it a mistake.\n\nI expected D300 (\"Use \\\"\\\"\\\"triple double quotes\\\"\\\"\\\"\") to flag this, since the opening quoting is plainly wrong. But pydocstyle reports no error at all on this file. The same is true for things like:\n\n```python\ndef another():\n '''''Whatever.'''''\n pass\n```\n\n— extra leading quotes go completely unreported.\n\nCould D300 be extended to catch the case where a docstring begins with superfluous opening quotes? Right now it seems to only check that the opening is some form of triple-quote-with-optional-prefix, and any \"extra\" quotes on the front silently slip through, which defeats the point of the check for these typo-style mistakes."} {"task_id": "format-code-task-000225", "source_id": "format-code-task-000225", "domain": "code", "task_path": "tasks/format-code-task-000225", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3067c3851dd34101ba5531a933ee697644e4a5a6930c885c75e8ca0d4e76f073", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nChange `dims` keyword of sum functions to `dim`\nI have come across a case where linopy does not adhere to the naming of xarray and this problem can only be solved with an unnecessary type check. Here is a short linear equation that should work with both pure xarray instances and a xarray/linopy mix.\n\nFor this let $$func(A,x,\\tilde x) = A^T(x-\\tilde x) = A^Tx - A^T\\tilde x = u$$\n\nwhere\n- $A \\in \\mathbb{R}^n \\times \\mathbb{R}^m$ is some (xarray) data matrix\n- $\\tilde x \\in \\mathbb{R}^n$ is a (xarray) data array\n- $x \\in \\mathbb{R}^n$ is **EITHER** a (linopy) variable **OR** a (xarray) data array\n- $u \\in \\mathbb{R}^m$ is the result.\n\nThe following minimal example shows that a type check must be made in the implementation of $func$. The reason for this is only the different naming of `dim` and `dims`:\n\n```python\nimport linopy\nimport xarray as xr\n\n\ndef func(A, x, x_tilde):\n tmp = A * x - A * x_tilde\n if type(tmp) == xr.DataArray:\n return tmp.sum(dim=\"X\")\n return tmp.sum(dims=\"X\")\n\n\ndimX = xr.DataArray([\"x1\", \"x2\", \"x3\"], dims=[\"X\"])\ndimU = xr.DataArray([\"u1\", \"u2\", \"u3\", \"u4\"], dims=[\"U\"])\nA = xr.DataArray(\n [[1, 1, 1, 1], [2, 2, 2, 2], [3, 3, 3, 3]],\n coords={\"X\": dimX, \"U\": dimU},\n)\nx_tilde = xr.DataArray([10, 20, 30], dims=[\"X\"])\n\nx_as_variable = linopy.Model().add_variables(coords=[dimX])\nx_as_value = xr.DataArray([1, 2, 3], coords=[dimX])\n\nprint(func(A, x_as_variable, x_tilde))\nprint(\"===\")\nprint(func(A, x_as_value, x_tilde))\n```\n\nwith the expected output\n\n```\nLinearExpression (U: 4):\n------------------------\n[u1]: +1 var0[x1] + 2 var0[x2] + 3 var0[x3] - 140\n[u2]: +1 var0[x1] + 2 var0[x2] + 3 var0[x3] - 140\n[u3]: +1 var0[x1] + 2 var0[x2] + 3 var0[x3] - 140\n[u4]: +1 var0[x1] + 2 var0[x2] + 3 var0[x3] - 140\n===\n\narray([-126, -126, -126, -126])\nCoordinates:\n * U (U) >> from urllib import parse as urlparse\n>>> urlparse.urlsplit('//testing/whatever')\nSplitResult(scheme='', netloc='testing', path='/whatever', query='', fragment='')\n```\nWhich means we accidentally drop `testing` before sending it on to the WSGI application. Ask me later how I figured that out.\n\nA request such as:\n\n```\nGET //testing/whatever HTTP/1.1\n```\n\nIs perfectly valid. Non-sensical maybe, but perfectly valid."} {"task_id": "format-code-task-000227", "source_id": "format-code-task-000227", "domain": "code", "task_path": "tasks/format-code-task-000227", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fff05fe7ff6d5d5eaa1643d4031cae0a7c1242326bd89ac683313b5d2531690e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nBroken connections with many concurrent `HEAD` requests\nUsing `waitress==2.1.2` and `pyramid==2.0.2` I've discovered that ~1% of requests fail due to broken connections when there are many concurrent requests.\n\nRequests fail with a `BrokenResourceError` or `ReadError` or `RemoteProtocolError: Server disconnected without sending a response.` or `Connection reset by peer` etc.\n\nThe server looks like:\n\n```python\nfrom pyramid.config import Configurator\nfrom pyramid.response import Response\n\nimport waitress\n\nimport logging\nlogger = logging.getLogger('waitress')\nlogger.setLevel(logging.DEBUG)\n\ndef hello_world(request):\n return Response(\"test\")\n\n\nif __name__ == '__main__':\n\n with Configurator() as config:\n config.add_route('hello', '/')\n config.add_view(hello_world, route_name='hello')\n app = config.make_wsgi_app()\n\n waitress.serve(app, host='127.0.0.1', port=8000, threads=10)\n```\n\nIf I increase the threads, nothing changes. If I use another WSGI server like `gunicorn`, there are no failed requests.\n\nI'm testing concurrent request success rates with following script:\n\n```python\nimport httpx\nimport anyio\nimport traceback\n\n\nATTEMPTS = 1000\nTARGET = \"http://localhost:8000\"\nHEAD_FAILURE = []\nHEAD_SUCCESS = 0\nGET_FAILURE = []\nGET_SUCCESS = 0\n\n\nasync def head(client: httpx.AsyncClient, url: str):\n global HEAD_SUCCESS\n try:\n response = await client.head(url)\n response.raise_for_status()\n except Exception as exc:\n HEAD_FAILURE.append(exc)\n else:\n HEAD_SUCCESS += 1\n\n print(\".\", end=\"\")\n\n\nasync def get(client: httpx.AsyncClient, url: str):\n global GET_SUCCESS\n try:\n response = await client.get(url)\n response.raise_for_status()\n except Exception as exc:\n GET_FAILURE.append(exc)\n else:\n GET_SUCCESS += 1\n\n print(\".\", end=\"\")\n\n\n\nasync def main():\n async with httpx.AsyncClient(timeout=httpx.Timeout(timeout=300)) as client:\n async with anyio.create_task_group() as tg:\n for _ in range(ATTEMPTS):\n tg.start_soon(head, client, TARGET)\n tg.start_soon(get, client, TARGET)\n\n\nanyio.run(main)\n\n\n# Report\n\nseen = set()\nfor exc in HEAD_FAILURE:\n if type(exc) in seen:\n continue\n seen.add(type(exc))\n print()\n print()\n traceback.print_exception(exc)\n print()\n\nprint()\nprint(f\"Displayed {len(seen)} unique exception types\")\nprint(f\"{len(HEAD_FAILURE)}/{len(HEAD_FAILURE) + HEAD_SUCCESS} HEAD requests failed\")\nprint(f\"{len(GET_FAILURE)}/{len(GET_FAILURE) + GET_SUCCESS} GET requests failed\")\n```\n\nI'm testing on macOS with Python 3.11.4.\n\nI discovered this while working with `devpi`, see https://github.com/devpi/devpi/issues/1022."} {"task_id": "format-code-task-000228", "source_id": "format-code-task-000228", "domain": "code", "task_path": "tasks/format-code-task-000228", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6e4ed9e96a58a7ee0ea3bbe019eef9c311c68e8102e3fc1bd5216581039bf1a5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### What is the expected enhancement?\n\nAs pointed out by @maddy-tod in #3288 :\n\n> ... it might be good to have this as a function on the gate for example Gate.is_parameterized() just for clarity. This could potentially introduce confusion for gates that are not able to be parameterized though.\n\nThere are a few places in the code which handle parameterized gates differently from fully bound gates, e.g. `Optimize1qGates`, `CommutationAnalysis`, `Collect2qBlocks`, and the unroller . Rather than having each `any(isinstance(param, ParameterExpression) for param in op.params)`, it might be helpful to have a `.is_parameterized` method on the base `Instruction` class.\n\nQuestions:\n- What is a returned for a gate that accepts no `params`?\n- Do we distinguish between unbound (`ParameterExpression(θ)`) and bound (`ParameterExpression(3.25)`) parameters?"} {"task_id": "format-code-task-000229", "source_id": "format-code-task-000229", "domain": "code", "task_path": "tasks/format-code-task-000229", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7131d912bc46b0fbcb365918ec3cdf0c8a2e8f8129d093aaee2368ddc915d420", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nccz drawing inconsistency\nA cz gate is being draw like this:\n```\nqr = QuantumRegister(2, 'q')\ncircuit = QuantumCircuit(qr)\ncircuit.append(ZGate().control(1), [qr[0], qr[1]])\ncircuit.draw()\n```\n``` \nq_0: |0>─■─\n │ \nq_1: |0>─■─ \n```\n\n![image](https://user-images.githubusercontent.com/766693/76632131-15a5da00-6519-11ea-8794-55f1b26bdbaa.png)\n\n![image](https://user-images.githubusercontent.com/766693/76632458-a4b2f200-6519-11ea-88a5-db932e93d465.png)\n\nA ccz gate is being draw like this:\n```\nqr = QuantumRegister(3, 'q')\ncircuit = QuantumCircuit(qr)\ncircuit.append(ZGate().control(2), [qr[0], qr[1], qr[2]])\ncircuit.draw('text')\n```\n```\n \nq_0: |0>──■──\n │ \nq_1: |0>──■──\n ┌─┴─┐\nq_2: |0>┤ Z ├\n └───┘\n```\n![image](https://user-images.githubusercontent.com/766693/76632604-d75cea80-6519-11ea-98ea-397ae4a6e1e6.png)\n\n![image](https://user-images.githubusercontent.com/766693/76632631-e04dbc00-6519-11ea-9945-cc77ce114c47.png)\n\nI thinks they all should be something like this:\n```\n \nq_0: |0>──■──\n │ \nq_1: |0>──■──\n │ \nq_2: |0>──■──\n```"} {"task_id": "format-code-task-000230", "source_id": "format-code-task-000230", "domain": "code", "task_path": "tasks/format-code-task-000230", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:25a11de6b6ab802b4c3b3b84edbe39d66beae4fc60579bc6bfcac64fe1488774", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nPulse builder equispace context doesn't work with multiple channels.\n\n\n\n### What is the current behavior?\n\nEquispaced context of pulse builder doesn't work as expected. This is reported by @ajavadia .\n\n### Steps to reproduce the problem\n\n```python\nwith qk.pulse.build() as bgate_0_1:\n with qk.pulse.align_equispaced(duration=800):\n qk.pulse.play(qk.pulse.library.GaussianSquare(duration=600, amp=.5, sigma=30, width=400),\n qk.pulse.DriveChannel(0))\n qk.pulse.play(qk.pulse.library.GaussianSquare(duration=300, amp=.5, sigma=30, width=100),\n qk.pulse.DriveChannel(1))\n```\n\nreturns\n\n![image](https://user-images.githubusercontent.com/39517270/99616087-f542fd80-2a5f-11eb-8749-e1d7e63af13d.png)\n\nWe specified 800dt as the duration of context, however the total duration becomes 1100.\nThis is caused by following logic:\n\nhttps://github.com/Qiskit/qiskit-terra/blob/bb627c62ddd54960a5e57a3cc73030d8071c7779/qiskit/pulse/transforms.py#L455-L467\n\n`align_equispaced` relocates all sub schedule blocks (`_children`) with equispaced interval. In above example, the pulse on d0 and d1 (they are `Play` instruction schedule components) are independent children and will be aligned sequentially.\n\nInterval is decided by (specified duration - current schedule duration) / (number of children), however if the schedule under the context consists of multiple channels, some schedules may be overlapped and net schedule duration may become shorter than the sum of all duration of children.\n\n### What is the expected behavior?\n\n\n### Suggested solutions\n\nWe should calculate the input schedule duration with\n```\nduration = sum([child_sched.duration for child_sched in schedule._children])\n```\nrather than\n```\nduration = schedule.duration\n```\n\n### Future extension\nAbove example intends to align two pulses at the center, thus we should use the context as follows:\n```python\nwith qk.pulse.build() as bgate_0_1:\n with qk.pulse.align_equispaced(duration=800):\n qk.pulse.play(qk.pulse.library.GaussianSquare(duration=600, amp=.5, sigma=30, width=400),\n qk.pulse.DriveChannel(0))\n with qk.pulse.align_equispaced(duration=800):\n qk.pulse.play(qk.pulse.library.GaussianSquare(duration=300, amp=.5, sigma=30, width=100),\n qk.pulse.DriveChannel(1))\n```\nHowever this is not intuitive and we may be able to create `align_center` context to make multi-channel alignment easier.\n```python\nwith qk.pulse.build() as bgate_0_1:\n with qk.pulse.align_center():\n qk.pulse.play(qk.pulse.library.GaussianSquare(duration=600, amp=.5, sigma=30, width=400),\n qk.pulse.DriveChannel(0))\n qk.pulse.play(qk.pulse.library.GaussianSquare(duration=300, amp=.5, sigma=30, width=100),\n qk.pulse.DriveChannel(1))\n```\n\nI'm bit busy recently so I hope someone in community will fix this..."} {"task_id": "format-code-task-000232", "source_id": "format-code-task-000232", "domain": "code", "task_path": "tasks/format-code-task-000232", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c6fe309a04c12ac97db65bf11a31c8a811db4e3aae55a588823cc4ceb3cad3b8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: find nearest index in a cartesian grid without building the product\n\nI'm working with value-function iteration on a multi-dimensional state space built from a few 1-D grids (capital, productivity, etc.). The discretized state space is the cartesian product of these per-dimension grids, and I often need to map a continuous point back to the index of its nearest neighbor in that product.\n\nThe natural way with what's currently in `quantecon.gridtools` is:\n\n```python\nimport numpy as np\nimport quantecon as qe\n\nnodes = (np.linspace(0, 1, 50), np.linspace(0, 1, 50), np.linspace(0, 1, 50))\nprod = qe.cartesian(nodes) # shape (50**3, 3)\n\nx = np.array([0.13, 0.42, 0.77])\ni = np.argmin(np.sum((prod - x)**2, axis=1)) # nearest index in prod\n```\n\nThis works for toy sizes but is wasteful and quickly becomes unusable as I add dimensions or refine grids — `cartesian` materializes the full `m**n` array even though I only want one integer back. With `n = 5` and `m = 100` per dimension I'm already allocating 10^10 floats just to do a nearest-neighbor lookup, while each per-dimension grid is sorted and the 1-D nearest lookup is trivially `O(log m)`.\n\nIt would be very useful if `gridtools` provided a helper that, given the per-dimension `nodes` (not the materialized product) and a query point `x`, returns the index that the nearest product point *would* have, had we built the product with `qe.cartesian(nodes)`. Concretely I'd like to be able to do something like:\n\n```python\nnodes = (np.arange(3), np.arange(2))\n# qe.cartesian(nodes) would be:\n# [[0 0], [0 1], [1 0], [1 1], [2 0], [2 1]]\n\n# closest product point to (0.6, 0.4) is (1, 0), which is row 2\nidx = qe.((0.6, 0.4), nodes)\nassert idx == 2\n```\n\nA few things I'd want from it:\n\n- It should agree with what you'd get from `np.argmin` over `qe.cartesian(nodes, order=...)`, for both `'C'` and `'F'` enumeration order, since I sometimes use `mlinspace`/`cartesian` with `order='F'`.\n- It should accept a batch of query points too (I typically map a whole simulated path at once), and return an array of indices in that case.\n- The complexity should scale like `O(n log m)` per query (binary search per dimension), not `O(m**n)` — that's the whole point of not building the product.\n\nHappy to help if there's interest.\n\nA name like `cartesian_nearest_index` would fit alongside the existing `cartesian` / `mlinspace` helpers in `gridtools`."} {"task_id": "format-code-task-000233", "source_id": "format-code-task-000233", "domain": "code", "task_path": "tasks/format-code-task-000233", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d5e1a3f81b6eaaecdd1b8d6babb67e3f53a561624ed7f662672668e355e4bc62", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: auto-set D4 damping parameters for r2SCAN in the VASP copilot\n\nWhen I run r2SCAN-D4 calculations with the `Vasp` calculator, I always have to remember to set the D4 damping parameters (`VDW_S6`, `VDW_S8`, `VDW_A1`, `VDW_A2`) manually in the INCAR. These are fixed values that come from the r2SCAN-D4 parameterization — they're not something a user is supposed to tune per-system, they're just the published damping constants for that functional/dispersion combination.\n\nSince the INCAR copilot already auto-recommends a bunch of related settings (LASPH for meta-GGAs, ALGO = All for meta-GGAs, etc.), it would be natural for it to also fill in the r2SCAN-D4 damping constants when it sees `METAGGA = r2SCAN` and the user hasn't supplied their own damping values. Right now if I forget to set them I either get whatever VASP defaults to (which is wrong for D4) or I have to copy-paste the magic numbers into every job script.\n\nReasonable behavior would be:\n\n- If `METAGGA` is `r2SCAN` and none of the four damping parameters are already set by the user → copilot fills in the standard r2SCAN-D4 values and logs that it did so (consistent with the other copilot recommendations).\n- If the user has explicitly set any of them → leave their choice alone, same as the rest of the copilot rules.\n\nThis would remove a class of silent-wrong-result bugs for anyone running r2SCAN-D4 with quacc."} {"task_id": "format-code-task-000234", "source_id": "format-code-task-000234", "domain": "code", "task_path": "tasks/format-code-task-000234", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:86d2cd5cdbb63c8284085e3b43834b735c204066d454d9031c320b9d1f9b278e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add chain awareness to shard identification\n\nRight now a shard is identified by a 32-bit *branch value* that only encodes a\nshard size and a shard id. We are moving to a multi-chain layout where every\nshard also belongs to a chain, so the identifier needs to carry a chain id as\nwell.\n\n## The encoding\n\nDefine a **full shard id** as a 32-bit integer split into two halves:\n\n- the **high 16 bits** hold the `chain_id`;\n- the **low 16 bits** hold `shard_size | shard_id`, where `shard_size` is a\n power of two whose single set bit is the most-significant set bit of those low\n 16 bits, and `shard_id` is whatever remains below that bit.\n\nSo a full shard id is `(chain_id << 16) | shard_size | shard_id`.\n\nSeparately, a **full shard key** is also a 32-bit integer whose high 16 bits are\na `chain_id` and whose low 16 bits are an arbitrary shard key. The shard id for a\ngiven shard size is obtained by masking the low key with `shard_size - 1`.\n\n## What `Branch` must expose\n\n`Branch` wraps a full shard id in its `value`. It should be able to report, all\nderived purely from `value`:\n\n- the chain id;\n- the shard size (a power of two) — derived from the low 16 bits only, so a\n non-zero chain id never changes the reported shard size or shard id;\n- the shard id;\n- the full shard id (i.e. the value itself).\n\nIt must also answer whether a given full shard key belongs to this branch: this\nis true only when the key's chain id equals the branch's chain id **and** the\nkey's shard id (low key masked by `shard_size - 1`) equals the branch's shard\nid. A key in the right shard but the wrong chain does not belong to the branch.\n\n## What `Address` must expose\n\nAn `Address` carries a `full_shard_key`. Given a shard size it should compute the\n**full shard id** the address maps to: take the chain id from the high 16 bits of\nthe key, take the shard id from the low key masked by `shard_size - 1`, and\ncombine them as `(chain_id << 16) | shard_size | shard_id`. If the supplied shard\nsize is not a power of two this must raise `RuntimeError`.\n\nRe-homing an address into a branch must also respect the chain: producing the\naddress \"in\" a branch keeps the recipient, adopts the branch's chain id, and\nrewrites the low shard key so the address lands on the branch's shard id while\npreserving the higher (above `shard_size`) bits of the original low key. The\nresulting full shard key must belong to that branch.\n"} {"task_id": "format-code-task-000235", "source_id": "format-code-task-000235", "domain": "code", "task_path": "tasks/format-code-task-000235", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f4b7bcfe4e215af38f06df8c5c3963e972325d156a7d31571ea440b62b050fc0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n \n\n**Rasa NLU version**: 0.13.4\n\n**Operating system** (windows, osx, ...): Windows\n\n**Content of model configuration file**:\n```yml\nlanguage: \"en\"\n\npipeline: \"spacy_sklearn\"\n```\n\n**Issue**:\nThe endpoint /evaluate of Rasa NLU HTTP server (server.py) is not up-to-date compared to the evaluate.py script of Rasa NLU. In particular it does not evaluate entities (only intents) and encounters some bugs that the script does not have.\n\nThus it would be great if Rasa NLU HTTP server could be inline with the latest evaluate.py script or maybe directly call it internally. Thanks!"} {"task_id": "format-code-task-000236", "source_id": "format-code-task-000236", "domain": "code", "task_path": "tasks/format-code-task-000236", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:906e5286e51a76e9970bacc1598c671eac9371be48e480fdfd7db8dea9a87447", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nThe remove-x-internal decorator doesn't remove references to x-internal components\nLet's say we're referencing two parameter components, one internal:\n```\n parameters:\n - $ref: '#/components/parameters/ProductID'\n - $ref: '#/components/parameters/Vendor'\n```\n```\n parameters:\n ProductID:\n in: path\n name: product_id\n description: The ID of the product.\n required: true\n schema:\n type: string\n format: uuid\n Vendor:\n in: header\n name: X-Vendor\n description: The vendor.\n schema:\n type: string\n x-internal: true\n```\n\nOnce we run the remove-x-internal decorator, the Vendor component will be removed, but the reference to it will remain. Perhaps because the order of the removals is wrong (we remove the component then check the reference?)\n\nFull repro: https://github.com/bojanz/openapi-cli-bug/tree/builtin-decorator"} {"task_id": "format-code-task-000237", "source_id": "format-code-task-000237", "domain": "code", "task_path": "tasks/format-code-task-000237", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f8f3798d942986bb248980a783577a48308156443cc39238bdfd9038d35f1d7c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Callable objects (instances with `__call__`) rejected as a `typing.Callable` argument\n\nI'm using `runtime_validation` to type-check a function whose parameter is annotated as `typing.Callable[[int], str]`. Plain functions work fine, but if I try to pass an instance of a class that implements `__call__`, validation rejects it.\n\nMinimal repro:\n\n```python\nimport typing\nfrom enforce import runtime_validation\n\nclass Multiplier:\n def __call__(self, a: int) -> str:\n return str(2 * a)\n\n@runtime_validation\ndef run(f: typing.Callable[[int], str], b: int) -> str:\n return f(b)\n\n# works:\ndef bar(a: int) -> str:\n return str(2 * a)\nrun(bar, 5)\n\n# rejected:\nrun(Multiplier(), 5)\n```\n\nIn normal Python an object that defines `__call__` *is* callable — `Multiplier()(5)` works exactly like calling a function — so it feels surprising that `runtime_validation` only accepts plain functions here. Ideally, an object implementing `__call__` should be usable anywhere a regular function is expected, including when the parameter is typed as `typing.Callable[...]`, and its signature should still be validated against the declared `Callable[...]` argument/return types."} {"task_id": "format-code-task-000238", "source_id": "format-code-task-000238", "domain": "code", "task_path": "tasks/format-code-task-000238", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:029305ec1673a697ddfa326c6fe8bb5ff9e749cd94e311fe75e79a164978db82", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `math.exp` returns `inf` instead of raising `OverflowError`\n\nI'm porting some code from CPython to RustPython and noticed that\n`math.exp` doesn't behave the same way for large arguments.\n\nIn CPython:\n\n```python\n>>> import math\n>>> math.exp(1e6)\nTraceback (most recent call last):\n File \"\", line 1, in \nOverflowError: math range error\n```\n\nIn RustPython:\n\n```python\n>>> import math\n>>> math.exp(1e6)\ninf\n```\n\nThis breaks code that wraps `math.exp` in `try/except OverflowError` —\nthe exception never fires on RustPython, and `inf` silently propagates\nthrough downstream calculations instead.\n\nI'd expect `math.exp` (and the other `math` functions in the same boat)\nto match CPython here: when the input is a finite number but the result\nwould overflow to infinity, raise `OverflowError` rather than returning\n`inf`.\n\nIn particular, `math.ldexp` has the same issue (e.g. `math.ldexp(1.0, 10000)`\nreturns `inf` instead of raising) — and while you're at it, please make sure\nthat special inputs to `ldexp` (NaN, ±inf, 0) are still returned unchanged\nrather than getting converted into an `OverflowError`."} {"task_id": "format-code-task-000239", "source_id": "format-code-task-000239", "domain": "code", "task_path": "tasks/format-code-task-000239", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4191520f1a7b4150131bab45ce1e19996df02fb2615a8c24931c6108e280cd32", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nfast method has bug when compute S1_conf/ST_conf\nthis is my setting:\n\n```python\nparam_values = fast_sampler.sample(problem, N=4935, seed=100) # N=4935\n```\nwhen the code do fast analyze as following sentence:\n```python\nSi = fast.analyze(problem, Y, print_to_console=True)\n# len(Y) = 34545, because len(Y)= N*D (D is number of parameters, there D=7, I have 7 parameters to analysis)\n```\n\n**The bug is comming**:\n\n```bash\n File \"C:\\ProgramData\\Anaconda3\\envs\\geo_env\\lib\\site-packages\\SALib\\analyze\\fast.py\", line 83, in analyze\n S1_d_conf, ST_d_conf = bootstrap(Y_l, N, M, omega_0, num_resamples, conf_level)\n\n File \"C:\\ProgramData\\Anaconda3\\envs\\geo_env\\lib\\site-packages\\SALib\\analyze\\fast.py\", line 116, in bootstrap\n S1, ST = compute_orders(Y_rs, N, M, omega_0)\n\n File \"C:\\ProgramData\\Anaconda3\\envs\\geo_env\\lib\\site-packages\\SALib\\analyze\\fast.py\", line 95, in compute_orders\n Sp = np.power(np.absolute(f[np.arange(1, int((N + 1) / 2))]) / N, 2)\n\n**IndexError: index 2467 is out of bounds for axis 0 with size 2467**\n```\n\nIn my opinion:\nIt is `np.arange(1, int((N + 1) / 2))` cause the bug, `np.arange` return is `[1,2,.....2467]`, however the lenth of f is 2467, the max index is 2466, So, I think maybe File \"C:\\ProgramData\\Anaconda3\\envs\\geo_env\\lib\\site-packages\\SALib\\analyze\\fast.py\", line 95, should change to `Sp = np.power(np.absolute(f[np.arange(0, int((N + 1) / 2)-1)]) / N, 2)`. Do you think so?\n\nSorry, my English is poor, my code is also poor. I wish l describe the problem clearly."} {"task_id": "format-code-task-000240", "source_id": "format-code-task-000240", "domain": "code", "task_path": "tasks/format-code-task-000240", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:8165cb98c8ca4f4eec0e42f651bba876ab107bc77a39d785c054386cfde4a00d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAdd an option to use encrypted connection\nCurrently to use an encrypted connection the client needs to pass some certificates.\nhttps://github.com/SAP/node-hdb#encrypted-network-communication\nIf HANA server is using a certificate signed by a public CA like VeriSign, there is no need to pass custom certificates as most public certificates are builtin in node.js. See `ca` optin in [tls.createSecureContext](https://nodejs.org/docs/latest/api/tls.html#tls_tls_createsecurecontext_options).\nFor such cases it would be useful to have a separate option to enable encryption, e.g.\n```js\nvar client = hdb.createClient({\n host : 'hostname',\n port : 30015,\n ssl: true,\n ...\n});\n```\nProbably a better name should be used as node actually uses TLS.\nAs a workaround you can pass `ca: undefined`\n```js\nvar client = hdb.createClient({\n host : 'hostname',\n port : 30015,\n ca: undefined,\n ...\n});\n```"} {"task_id": "format-code-task-000241", "source_id": "format-code-task-000241", "domain": "code", "task_path": "tasks/format-code-task-000241", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e25a6e6879b755cfc4d32b444f12f690d66c57699754021e08e3d67e2c527765", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Convenient access to the variance gradient at a single point in kriging surrogates\n\nI'm using `KRG` (and the MFK variants) for some Bayesian-optimization-flavoured work where, at a candidate point `x*`, I need the gradient of the predicted variance — i.e. the partial derivatives of the variance with respect to *all* input dimensions, evaluated at that one point.\n\nLooking at the public API of `SurrogateModel` / kriging-based models, the only thing I can find for variance derivatives is `predict_variance_derivatives(x, kx)`, which gives the derivative with respect to a single input component `kx`. So if my problem has `nx` inputs and I want the gradient at a point, I end up doing something like:\n\n```python\ngrad = np.array([\n sm.predict_variance_derivatives(x_star, kx).ravel()\n for kx in range(x_star.shape[1])\n])\n```\n\nThis works but feels off:\n\n* It's `nx` separate predict calls for what is conceptually one quantity at one point.\n* Every caller that wants \"variance gradient at a point\" has to re-implement the same loop + stacking + shape massaging, and it's easy to get the orientation of the resulting array wrong.\n* For acquisition functions / optimizers that ask the surrogate for a gradient at the current iterate, this is the natural thing to want directly from the model.\n\nIt would be great if kriging-based surrogates exposed a first-class way to ask, in one call, for the gradient of the variance at a given point — taking an `x` of shape `(1, nx)` (and ideally accepting a plain `(nx,)` vector for convenience) and returning the full vector of partials at that point.\n\nFor context, `predict_variance_derivatives(x, kx)` already covers the \"set of points, one component\" case and should keep working as it does today; what's missing is the dual \"one point, all components\" case. The two together would cover the typical use cases nicely.\n\nA natural name for the new entry point would be something like `predict_variance_gradient(x)`."} {"task_id": "format-code-task-000242", "source_id": "format-code-task-000242", "domain": "code", "task_path": "tasks/format-code-task-000242", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2c2d94db548f4a3327281db16690eefb76fd86fcc819462fc877b1b10c69148b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nDrawing of certain paths breaks in `cartopy` 0.25.0\nHello! I’m a developer of [mosaic](https://github.com/e3SM-Project/mosaic), a python package for unstructured mesh visualization which is heavily reliant on cartopy.\n\nWe’ve noticed an issue drawing certain paths when switching from cartopy `0.24.0` to `0.25.0`. See [E3SM-Project/mosaic/#42](https://github.com/E3SM-Project/mosaic/issues/42) for the original issue where this was brought to our attention. Here’s a minimum working example:\n```python\nimport cartopy\nimport cartopy.crs as ccrs\nimport numpy as np\nimport matplotlib.pyplot as plt\n\nfrom matplotlib.collections import PolyCollection\n\n# https://wktmap.com/?fae129b6\npatch = np.array([[\n [-180.00020734, -63.5383884 ],\n [-179.93611911, -63.61745971],\n [-179.77001049, -63.62512738],\n [-179.67109735, -63.55616224],\n [-179.74190981, -63.49385818],\n [-179.91362571, -63.48807984],\n [-180.00020734, -63.5383884 ],\n [-180.00020734, -63.5383884 ],\n]])\n\nfig, ax = plt.subplots(\n constrained_layout=True, subplot_kw={\"projection\": ccrs.PlateCarree()}\n)\n\ncollection = PolyCollection(patch, array=[1.], ec='k', alpha=0.5)\n\ncollection.set_transform(ccrs.PlateCarree())\n\nax.add_collection(collection)\nax.autoscale_view()\n\nax.set_title(f\"cartopy v{cartopy.__version__}\")\nax.gridlines(draw_labels=True)\n\nplt.show() \n```\nHere's what the patch looks like in cartopy `0.24.0`:\n```\nconda create -n cartopy_0.24.0 python=3.13 cartopy=0.24.0\n```\n\"Image\"\n\nAnd, here's what the patch looks like in cartopy `0.25.0`:\n```\nconda create -n cartopy_0.25.0 python=3.13 cartopy=0.25.0\n```\n\"Image\"\n\n----\n\nMy hunch is that the issue arises because the first/last index lies outside of the projection boundary (i.e. -180 deg). But, I haven't had time to through and test that hunch. \n\nFor some context, in `mosaic` we repeat the first index as the last to ensure the polygons are closed. Because we have variable resolution meshes, the polygons can have variable number of sides. To support ragged array likes this, we further repeat the first/last index as many times as needed to fill in up to the max number of sides.\n\n**cc'ing** @xylar @cbegeman"} {"task_id": "format-code-task-000243", "source_id": "format-code-task-000243", "domain": "code", "task_path": "tasks/format-code-task-000243", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a3f65dac9b1048c58066a96e9cd10aa8563135ef47ccc8300bf9c39027ea96b4", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Old model versions are not cleaned up after a rolling update\n\nI'm running Seldon Core v2 and trying out rolling updates on a model. My setup is roughly:\n\n1. Apply a `Model` (e.g. `iris`) pointing at storage URI `.../iris/v1`.\n2. Wait for it to become ready.\n3. Re-apply the same `Model` with a new storage URI `.../iris/v2` to roll it forward.\n\nThe new version comes up fine and starts serving traffic, but the **old version never goes away**. If I keep rolling the model forward (v2 → v3 → v4 …) the previous versions just accumulate in the scheduler — version cleanup never seems to fire for them. I'd expect that once the new version is fully available and serving, the older one(s) should be cleaned up automatically the way the docs imply.\n\nI can reproduce this with a plain sklearn iris model — nothing fancy in the spec, just `requirements: [sklearn]` and a storage URI pointing at a rolling sample bucket.\n\n---\n\nWhile poking at this I also noticed something that looks related (same \"stale state hangs around\" flavour, but on the experiment side):\n\nIf I update an existing `Experiment` (change weights, change candidates, etc.), the new config takes effect but it doesn't feel like the previous routing for that experiment is being reset first — re-applying the same experiment a couple of times produces routing behaviour that doesn't match what I'd get from creating it fresh. Creating it fresh from a clean state works as expected.\n\nBoth symptoms feel like envoy-side state from the previous version of the object isn't being released before the new state is applied. Could the incremental processor be made to behave correctly across repeated model rollouts and experiment updates?"} {"task_id": "format-code-task-000244", "source_id": "format-code-task-000244", "domain": "code", "task_path": "tasks/format-code-task-000244", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:554678bc02f1a5af7688f76031f2ff250132ac7c540940926b135395a397118e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Missing `cm3` and `mm3` volume units\n\n`Measured::Volume` already supports `m3`, `in3`, `ft3`, liters and the various SI liter prefixes, but there's no support for cubic centimeters or cubic millimeters.\n\nThese come up all the time for small-volume measurements (think component packaging, lab/medical work, 3D-printing material usage, etc.), and right now I can't express them without converting by hand:\n\n```ruby\nMeasured::Volume.new(5, :cm3)\n# => Measured::UnitError: Unit 'cm3' does not exist\n```\n\nIt would be great if `cm3` and `mm3` were first-class units in `Measured::Volume`, so they can be constructed directly and converted to/from any of the existing volume units (liters, m3, in3, gallons, …) like the other units do."} {"task_id": "format-code-task-000245", "source_id": "format-code-task-000245", "domain": "code", "task_path": "tasks/format-code-task-000245", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fba4e889953921e030a60495299844f9b0aa2fe22509615ce74e65d9dc74bf65", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `IndexFilters` keyboard shortcut can't be turned off, and its tooltip always advertises it\n\nI'm embedding the `IndexFilters` component in a page that has its own keyboard handling. I noticed that whenever the user presses `F` anywhere on the page (outside of an input), `IndexFilters` grabs focus and switches into search/filter mode. This conflicts with shortcuts I want to handle myself, and in some contexts I simply don't want a global single-letter shortcut hijacking keystrokes at all.\n\nLooking at `IndexFiltersProps`, I don't see any prop to opt out of this behavior. The keydown handler inside `IndexFilters` always runs.\n\nThere's a related problem with the tooltip on the search/filter toggle button. The default tooltip text is something like:\n\n> Search and filter (F)\n\nThe `(F)` is hard-coded into the localized strings (across all of the `locales/*.json` files) because it's part of the same single translation key used for the tooltip. So even in cases where pressing `F` shouldn't do anything, the tooltip still tells users it will — which is wrong / misleading.\n\nI can override the tooltip via `filteringAccessibilityTooltip`, but that only patches the symptom on a single instance; the shortcut itself still fires, and every consumer that wants the same behavior has to know to override the string in every supported language.\n\n### What I'd like\n\nA way for the consumer of `IndexFilters` to disable the built-in keyboard shortcut behavior. When it's disabled:\n\n- pressing `F` outside of inputs should no longer cause `IndexFilters` to enter search/filter mode,\n- the default tooltip on the search/filter toggle button should stop advertising the `(F)` shortcut (and this should apply automatically across all supported locales, without each app having to ship its own override string).\n\nThe current behavior (shortcut on, tooltip mentions `(F)`) should remain the default so existing usages aren't affected.\n\nI'd expect this to be exposed as a new boolean prop on `IndexFilters`, something like `disableKeyboardShortcuts`."} {"task_id": "format-code-task-000246", "source_id": "format-code-task-000246", "domain": "code", "task_path": "tasks/format-code-task-000246", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d77cdf222d398c0cce7727295d950fed781c8bb13ac9da6b79ce7243a2eff991", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Code actions feel stale after I move the cursor, and RuboCop seems to run on every keystroke\n\nI've been using ruby-lsp in VS Code on a moderately sized Ruby file that has several RuboCop offenses spread across the file (say, an indentation issue near the top, a string-quote issue further down, etc.).\n\n### What I'm seeing\n\nWhen I put my cursor on the first offense and open the lightbulb / quick fix menu, I get the expected fix for that offense. So far so good.\n\nThen I scroll down and click on a *different* offense in another part of the file. I expect the quick fix menu to show me an action for *this* offense. Instead I keep getting the same code action I got the first time, as if the editor is still showing me suggestions for the previous cursor location. If I keep moving around, the suggestions don't seem to track where I actually am in the file — they look frozen on whatever range was used the first time `textDocument/codeAction` fired for this file.\n\n### And separately (but I suspect related)\n\nWhile poking at this, I noticed the editor stalls noticeably every time code actions get requested, even when I haven't edited the file at all between requests — just moved the cursor. Watching activity, it looks like RuboCop is being re-run from scratch over and over for the same unchanged file contents. For a file that hasn't changed, re-analyzing it every time the cursor moves feels wrong and is making the experience pretty laggy.\n\n### What I'd expect\n\n- Code actions should reflect the current selection / visible range. Moving my cursor to a different offense should give me the fix for *that* offense, not a cached one from earlier.\n- Re-analyzing the file with RuboCop when nothing about the file has changed seems wasteful — that part I'd expect to be reused across requests on the same unchanged document.\n\nRight now it feels like the two are inverted: the thing that depends on where I'm looking is sticky, and the thing that doesn't depend on where I'm looking is being recomputed every time."} {"task_id": "format-code-task-000247", "source_id": "format-code-task-000247", "domain": "code", "task_path": "tasks/format-code-task-000247", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f1ebfd63bea42bb71a2475aa2686d38ca9f84b89041d7390fe52252a505db585", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add configurable \"toxics\" to proxies\n\nRight now a proxy is a dumb pipe: whatever a client sends is forwarded verbatim to the\nupstream and back. We want to be able to deliberately degrade a proxied connection so we can\ntest how applications behave against a flaky network. The first degradation we need is\n**latency**.\n\nIntroduce the notion of a *toxic* that can be attached to a proxy, independently for each of\nthe two traffic directions:\n\n- **upstream** — data travelling from the client toward the upstream server.\n- **downstream** — data travelling from the upstream server back to the client.\n\n## HTTP API\n\nAdd these endpoints to the existing API server (the one that already serves `/proxies`):\n\n- `GET /proxies/{proxy}/upstream/toxics`\n- `GET /proxies/{proxy}/downstream/toxics`\n\n Returns a JSON object that maps each available toxic's name to its current state. A freshly\n created proxy must already list a toxic named `latency` in **both** directions, and that\n toxic must start **disabled**. The latency toxic's state is a JSON object with these fields:\n `enabled` (bool), `latency` (number, milliseconds) and `jitter` (number, milliseconds). On a\n new proxy it reads `{\"enabled\": false, \"latency\": 0, \"jitter\": 0}`.\n\n- `POST /proxies/{proxy}/upstream/toxics/{name}`\n- `POST /proxies/{proxy}/downstream/toxics/{name}`\n\n Sets the complete desired state of the named toxic for that direction from the JSON request\n body and responds with the resulting toxic state as JSON. The body fully specifies the new\n state: any field that is omitted is reset to its zero value (so `enabled` defaults to\n `false`, and `latency`/`jitter` default to `0`). The change must persist — a subsequent\n `GET` reflects it — and must affect **only** the direction it was posted to; the same toxic\n in the other direction is left untouched.\n\n### Errors\n\n- A request for a proxy that does not exist responds with HTTP `404`.\n- A `POST` to a toxic name that does not exist responds with HTTP `404`.\n\n## Behavior\n\nThe latency toxic must actually slow traffic down, not just record settings. When the\n`latency` toxic is enabled with a latency of *L* milliseconds (and `jitter` 0) on a direction,\ndata flowing in that direction is delayed by approximately *L* milliseconds before it reaches\nthe other side. A direction whose latency toxic is disabled forwards data with no added delay,\nexactly as before.\n\nConfiguration applied before a client connects takes effect for that connection.\n"} {"task_id": "format-code-task-000248", "source_id": "format-code-task-000248", "domain": "code", "task_path": "tasks/format-code-task-000248", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:63c47777ae6c930bf48e41c858d91d08bfa57ecea10f1ad8c2a0fbf70ceaf00d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Re-POSTing to a toxic endpoint wipes out fields I didn't include in the body\n\nI'm using toxiproxy's HTTP API to manage a latency toxic on one of my proxies from a script. First I configure it with the full set of values:\n\n```\nPOST /proxies/myproxy/upstream/toxics/latency\n{\"enabled\": true, \"latency\": 1000, \"jitter\": 100}\n```\n\nLater in the same script I want to temporarily disable it, so I send just the field I'm changing:\n\n```\nPOST /proxies/myproxy/upstream/toxics/latency\n{\"enabled\": false}\n```\n\nAfter the second POST, if I `GET /proxies/myproxy/upstream/toxics` I can see that `latency` and `jitter` have both been reset to `0`. So even though I only sent `enabled` in the body, the rest of the toxic's configuration got blown away.\n\nThat feels wrong — I'd expect POSTing to a toxic that's already been configured to behave like a partial update: whatever fields I send in the body get updated, and anything I omit keeps whatever value it had before. Otherwise every time I want to tweak one knob (e.g. toggle `enabled`) I have to re-send the entire configuration, and any client that doesn't remember the previous values can silently clobber them.\n\nSame thing happens on the downstream endpoint. Could the toxic update preserve previously-set fields when the request body only contains a subset of them?"} {"task_id": "format-code-task-000249", "source_id": "format-code-task-000249", "domain": "code", "task_path": "tasks/format-code-task-000249", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e17ded671e1c8df75928320eb14ddd08a6d4748531246d307033ace0d8ff0828", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a ComponentBuilder for the Next.js package\n\nOur Next.js integration currently wires up a single static \"component factory\" that maps\ncomponent names to React components. That approach is too rigid: an app may register components\nthat are loaded statically, components that are imported dynamically (lazy), and modules that\nalso expose component-level data-fetching hooks. We need a small building block that can produce\nthe right factory for each of these cases from one registry.\n\nAdd a `ComponentBuilder` class to the `sitecore-jss-nextjs` package source. It must be a named\nexport importable as `ComponentBuilder`. It is constructed with a config object whose `components`\nproperty is a `Map` keyed by component name. Each map value is one of:\n\n- a **static module** — a plain object holding one or more named exports, e.g. a Next.js style\n `default` export, an SXA style `Default` export, the optional data-fetching hooks\n `getStaticProps` / `getServerSideProps`, and/or any number of additional named component exports;\n- a **lazy module** — an object exposing a `module()` function that returns the module (typically a\n promise that resolves to it);\n- a **lazy component** — an object exposing an `element(isEditing?)` function that returns the React\n component to render.\n\nThe class exposes two methods, each of which returns a *factory function* built from the registry:\n\n### `getModuleFactory()`\n\nReturns a function `(componentName) => module`. Given a name it resolves the full module so callers\ncan reach its data-fetching hooks:\n\n- unknown name → `null`;\n- lazy module (has a `module` function) → the result of invoking that function;\n- static module → the stored object itself.\n\n### `getComponentFactory(config?)`\n\nAccepts an optional config object with an optional `isEditing` boolean (treat a missing config as\n`{}`). Returns a function `(componentName, exportName?) => Component` that resolves to the React\ncomponent for a name:\n\n- unknown name → `null`;\n- lazy component (has an `element` function) → the result of invoking `element(isEditing)`, passing\n the configured editing flag straight through;\n- static module:\n - when an `exportName` other than the SXA default name `\"Default\"` is supplied, return that named\n export from the module;\n - otherwise return the module's default component, preferring the SXA `Default` export, then the\n Next.js `default` export, and `null` if neither exists. (Asking for the `\"Default\"` export name\n explicitly follows this same default-resolution rule.)\n\nEach call to `getModuleFactory()` / `getComponentFactory()` must return a usable factory, and the\ntwo factories must reflect the same registry the builder was constructed with.\n"} {"task_id": "format-code-task-000250", "source_id": "format-code-task-000250", "domain": "code", "task_path": "tasks/format-code-task-000250", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:aced6a9958199f53d68f77067aeb1e072de758b1b88a56b96bbd96f1fb4b6d78", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Sonar analysis crashes on surefire reports produced by JUnit 5.2\n\nWe recently upgraded one of our Java projects from JUnit 4 to JUnit 5.2 and started running our usual SonarQube analysis (with the sonar-java plugin) on the build. As soon as the surefire test reports are picked up by sonar, the analysis blows up and exits — the run never completes.\n\nIf we delete the surefire XML reports before running sonar, or revert the test framework back to the previous JUnit version, the analysis goes through fine, so the trigger really seems to be the report files that JUnit 5.2 generates with the surefire plugin. The exact same project + same sonar version was working before the JUnit upgrade.\n\nIt looks like sonar-java doesn't expect something about the shape of these reports. From a user point of view this is pretty disruptive — JUnit 5.2 + surefire is a fairly standard combo and it shouldn't take the whole analysis down. Could sonar-java be made resilient to whatever JUnit 5.2 is putting into these reports, so the analysis keeps running instead of crashing?"} {"task_id": "format-code-task-000252", "source_id": "format-code-task-000252", "domain": "code", "task_path": "tasks/format-code-task-000252", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ce33bd86ee8b7b54580e5a62eda12d0cc33fd119224824956484d5b0cb5efe55", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## NPE when setting `dynamic_partition.buckets` on a non-partitioned table\n\nI have a regular (non-partitioned) OLAP table and I wanted to tweak the bucket count via a dynamic partition property. When I run something like:\n\n```sql\nCREATE TABLE t (\n k1 INT,\n v1 INT\n) DISTRIBUTED BY HASH(k1) BUCKETS 8\nPROPERTIES (\n \"dynamic_partition.buckets\" = \"10\"\n);\n```\n\n(or the equivalent `ALTER TABLE ... SET (\"dynamic_partition.buckets\" = \"10\")` on an existing non-partitioned table)\n\nthe FE crashes with a NullPointerException instead of giving me a useful error.\n\nIf I instead pass any of the other dynamic partition properties (e.g. `dynamic_partition.enable`, `dynamic_partition.time_unit`, `dynamic_partition.end`, ...) on the same non-partitioned table, I get a clean, readable error telling me that dynamic partition is only supported on single-column range-partitioned tables. That's the behavior I'd expect here too — `dynamic_partition.buckets` shouldn't be treated as a special case that bypasses validation and blows up later with an NPE.\n\nEither reject it up front with the same kind of error message as the other `dynamic_partition.*` properties, or otherwise handle it gracefully. Just please don't NPE."} {"task_id": "format-code-task-000253", "source_id": "format-code-task-000253", "domain": "code", "task_path": "tasks/format-code-task-000253", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b38c079c0474fbb142f2449a4492a0748c890a1318442df12f09b3886dd8af80", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add per-node execution timeout support to taskflow\n\nOperators want individual task nodes to be guarded by an execution timeout. When a node runs\nlonger than its configured limit, the system should be able to react automatically according to a\nper-node strategy. This change is about the *building blocks* of that feature — the configuration\nparsing, the timeout-handling strategies, and an up-front validation guard. (The Celery/Redis\nmachinery that periodically scans for timed-out nodes is out of scope here.)\n\nA node's timeout is described on the node itself, inside a `timeout_config` object, e.g.:\n\n```json\n\"timeout_config\": {\"enable\": true, \"seconds\": 30, \"action\": \"forced_fail\"}\n```\n\nPlease implement the following observable behavior.\n\n## 1. Parsing timeout configuration from a pipeline tree\n\nProvide a function `parse_node_timeout_configs`, importable from `gcloud.taskflow3.utils`, that\ntakes a fully expanded pipeline tree (a dict whose `\"activities\"` maps node ids to node dicts) and\nreturns the list of timeout configurations declared in it.\n\n- Each returned item is a dict with exactly the keys `node_id`, `action`, and `timeout`, where\n `timeout` is the configured number of seconds and `action` is the configured action string.\n- Only `ServiceActivity` nodes that have `timeout_config` with a truthy `enable` produce a config.\n A node without `timeout_config`, or with `enable` falsy/absent, produces nothing.\n- `SubProcess` nodes never produce a config themselves, but their nested pipeline (under the node's\n `\"pipeline\"` key) must be traversed recursively so that timeout configs declared at any depth are\n collected.\n- The configured `seconds` must be a positive integer. If `seconds` is missing, not an integer, or\n not greater than zero, that node is skipped (no config is produced for it) but parsing of the rest\n of the tree continues normally.\n- The order of the returned configs is not significant.\n\n## 2. Timeout-handling strategies\n\nProvide a registry `node_timeout_handler`, importable from\n`gcloud.taskflow3.domains.node_timeout_strategy`, that maps an action string to a strategy object.\nEach strategy object exposes a method `deal_with_timeout_node(task, node_id)` that enforces the\ntimeout against the given task by driving the task's node-action API (`task.nodes_action(action,\nnode_id, operator)`), acting as a fixed system operator. `nodes_action` returns a dict whose\n`\"result\"` key indicates success.\n\nTwo actions must be supported:\n\n- `\"forced_fail\"`: force-fail the node (a single `forced_fail` node action) and return that action's\n result.\n- `\"forced_fail_and_skip\"`: force-fail the node first; **only if** the force-fail result indicates\n success, then skip the node (a `skip` node action) and return the skip result. If the force-fail\n did not succeed, return the force-fail result and do not attempt to skip.\n\n## 3. Conflict validation when converting web data to a pipeline\n\nEnabling a node's timeout is incompatible with the node ignoring errors or auto-retrying — letting a\nnode both retry/ignore-failure and be force-failed on timeout is contradictory. When web pipeline\ndata is converted into an executable pipeline, a `ServiceActivity` that has `timeout_config` enabled\ntogether with either `error_ignorable` enabled or `auto_retry` enabled must be rejected by raising\n`pipeline.exceptions.InvalidOperationException`. A node that enables its timeout without those\nconflicting options must continue to convert successfully.\n"} {"task_id": "format-code-task-000254", "source_id": "format-code-task-000254", "domain": "code", "task_path": "tasks/format-code-task-000254", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a761c02bfebb2288fac7c5742053e1f8cbfee7996b026b71df9803fcc38a0c5d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## MagdaReference items don't inherit `accessType` from a private parent group\n\nI have a catalog where I'm grouping a few `MagdaReference` items under a parent group that I've configured as private (using `AccessControlMixin`'s `setAccessType(\"private\")` on the group). My expectation, based on how every other catalog item behaves, is that the children should inherit the parent's access type when they don't have explicit access info of their own — so the magda items under a private group should also show up as private in the UI.\n\nWhat actually happens is that those `MagdaReference` children are reported as **public**, even when their parent is clearly private. As a result the workbench / catalog UI flags them as public alongside genuinely public items, which is misleading for the people we're sharing the catalog with.\n\nTo convince myself it isn't a problem with the parent or the mixin itself, I swapped one of the magda children for a regular catalog item under the same parent group — that one correctly comes through as private. So the parent inheritance machinery itself is working; it's specifically `MagdaReference` that disagrees.\n\nThe behaviour seems to depend on whether the underlying Magda record has been populated yet:\n\n- Once the magda record is loaded and it carries access-control info, `accessType` does the right thing (public vs. non-public based on that record).\n- But before the record is loaded, or for any `MagdaReference` where the record info isn't there, `accessType` always comes out as `\"public\"` regardless of the surrounding catalog structure. `isPublic` is `true` and `isPrivate` is `false` even though the containing group is private.\n\nI'd expect `MagdaReference` to fall back to the same default access-type resolution that every other `AccessControlMixin` consumer uses — i.e. honour explicit settings, then look at the referrer / ancestors — whenever it doesn't have its own access info from a Magda record. Hard-coding `\"public\"` in that fallback path means a private subtree silently exposes public-looking children in the UI."} {"task_id": "format-code-task-000255", "source_id": "format-code-task-000255", "domain": "code", "task_path": "tasks/format-code-task-000255", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:76cb41e489206f4d1435b03be5128198b4c118df7771df189b65b9b3ff7d8889", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `Split` with `n_splits=1` should be optimized away\n\nWhen I use `theano.tensor.split` with `n_splits=1` (this comes up in generic code where the number of splits is computed from a parameter that can be 1), the resulting graph still contains a full `Split` Apply node, even though semantically it's a no-op — it just returns the input tensor unchanged along the requested axis.\n\nMinimal example:\n\n```python\nimport theano\nimport theano.tensor as T\n\nx = T.vector('x')\naxis = 0\nsplits = T.as_tensor_variable([x.shape[0]])\ny, = T.split(x, splits, n_splits=1, axis=axis)\n\nf = theano.function([x], y)\ntheano.printing.debugprint(f)\n```\n\nThe compiled function still has a `Split` node in it, while ideally the optimizer should recognize this case and just forward `x` through (it's the identity along that axis).\n\nThis matters for a few reasons:\n\n- It's wasted work at runtime — every call to `f` goes through `Split`'s perform, allocates a list, etc., for what is mathematically a pass-through.\n- It blocks downstream optimizations that would otherwise see `x` directly instead of seeing it behind a `Split`.\n\nCould the canonicalize / specialize pass include a rule that strips `Split` when there is only one output split? In that case the only thing that conceptually has to hold for the original graph to be well-formed is that the single split size equals the size of `x` along `axis` (and that the splits vector has length 1) — but those are already promised by the user constructing a 1-split Split, so I don't think the optimization should refuse to fire on that account."} {"task_id": "format-code-task-000256", "source_id": "format-code-task-000256", "domain": "code", "task_path": "tasks/format-code-task-000256", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:296ce7ec2369daa374154c7036b8efba9e56dfbed83af9b071cab4b9283b02b2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `MRG_RandomStreams` has no `seed()` method\n\nI'm using `theano.sandbox.rng_mrg.MRG_RandomStreams` for the random ops in a model (dropout, init noise, etc.) because it's faster and works on the GPU. For reproducibility I want to be able to reset the random source between runs of the same compiled function — same workflow I use with `theano.tensor.shared_randomstreams.RandomStreams`:\n\n```python\nfrom theano.sandbox.rng_mrg import MRG_RandomStreams\n\nsrng = MRG_RandomStreams(seed=234)\n# ... build some random variables off srng, compile a function f ...\n\n# I want to reset everything and replay the exact same samples:\nsrng.seed(123)\n# run f again\n```\n\nBut `MRG_RandomStreams` doesn't have a `seed` method, so the call above just blows up with an AttributeError. The constructor takes a `seed` argument, but once you've built RVs off the stream there's no way to re-seed them after the fact.\n\nThe plain `theano.tensor.shared_randomstreams.RandomStreams` class supports this — you can call `.seed(N)` on it and the RVs that were already built off that stream get reset to a deterministic state, so re-running the compiled function produces the same sample sequence. It would be great if `MRG_RandomStreams` behaved the same way, since otherwise switching between the two stream implementations changes what reproducibility tools you have."} {"task_id": "format-code-task-000257", "source_id": "format-code-task-000257", "domain": "code", "task_path": "tasks/format-code-task-000257", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1ac59880140d4a2c48ee1330744b874e3bf912aca53f20e8408eaf90f522f7fc", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `extra_ops.repeat` gives wrong shape when `repeats` is a vector and `axis=None`\n\nI'm using `theano.tensor.extra_ops.repeat` to do the same thing as `numpy.repeat` — repeat each element of an array a (possibly different) number of times. When `repeats` is a 1D tensor and I don't pass `axis`, the inferred shape of the result comes out wrong.\n\nRoughly what I'm doing:\n\n```python\nimport numpy as np\nimport theano\nimport theano.tensor as T\nfrom theano.tensor.extra_ops import repeat\n\nx = T.dvector()\nr = T.lvector()\ny = repeat(x, r) # no axis -> should behave like numpy.repeat(x, r)\n\nf = theano.function([x, r], y.shape)\nprint f(np.array([1.0, 2.0, 3.0]), np.array([2, 3, 1]))\n```\n\n`numpy.repeat([1,2,3], [2,3,1])` returns a 1D array of length 6 (`2+3+1`), so I expect `y.shape` to be `(6,)` here. Instead the shape Theano infers doesn't match what `perform` actually produces, and downstream ops that rely on shape inference end up broken / mismatched against the real output.\n\nThe scalar-`repeats` case (e.g. `repeat(x, 3)` with no axis) and the case where `axis` is given both look fine — it's specifically the combination \"`axis=None` + `repeats` is a vector\" that's off. Could the shape inference for `repeat` be made consistent with `numpy.repeat` in that case too?"} {"task_id": "format-code-task-000258", "source_id": "format-code-task-000258", "domain": "code", "task_path": "tasks/format-code-task-000258", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1b54f53b32d875a2ce5c6632aec47bdbfc47331a3e661e9b621624ef6554980b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Allow remapping Watermill's log levels when using the slog adapter\n\nI'm using `NewSlogLogger` to plug Watermill into my application's `slog`-based logging setup. The integration itself works fine, but Watermill emits a lot of messages at `Info` level (router lifecycle, handler start/stop, subscribe acks, etc.) that I really don't want showing up alongside my application's own `Info` logs in production — they're more like debug/diagnostic information from my point of view.\n\nThe obvious workaround would be to raise the global slog level to `Warn`, but then I lose my own application's `Info` logs too. I want to silence/demote *Watermill's* noise specifically, while keeping the rest of my app's logging at `Info`.\n\nRight now there doesn't seem to be a way to do this with the slog adapter — whatever level Watermill chooses internally is exactly the level slog sees. It would be great if, when constructing the slog adapter, I could say something like \"treat Watermill's Info as slog Debug\" (and similarly for other levels I might want to shift), so the routing of levels happens at the adapter boundary instead of forcing me to fiddle with slog handlers or global levels.\n\nExisting usage that I'd like to keep working unchanged:\n\n```go\nlogger := watermill.NewSlogLogger(slog.Default())\n```\n\nFor users that don't care about remapping, the default behavior should stay the same as today (Watermill Info → slog Info, Debug → Debug, etc.).\n\nA separate constructor along the lines of `NewSlogLoggerWithLevelMapping(...)` that takes the underlying `*slog.Logger` plus a level-to-level mapping would be a natural way to expose this."} {"task_id": "format-code-task-000259", "source_id": "format-code-task-000259", "domain": "code", "task_path": "tasks/format-code-task-000259", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:45a7f28e0191df9a89ef93eb070b631374f469839f3e09b83ad6534fc015c57d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[BUG] :filter filter run does not always remove non-matching entries (linked-list.js issue)\n### Describe the bug\n\nI expect the output of the filter expression `A =A B =A :filter[match[B]]` to be `B`, but the output is `B A`\n\nLooks to me like an issue with `linked-list-.js` not removing items correctly. If I add this line to `editions/test/tiddlers/tests/test-linked-list.js`:\n\n```js\ncompare(remove(newPair([\"A\", \"A\", \"B\", \"A\"]), [\"A\", \"A\", \"A\"])); // B\n```\n\nThen I get this test failure when running the test suite:\n\n```\nFailures:\n1) LinkedList class tests can remove all instances of a multi-instance value\n Message:\n Expected $.length = 2 to equal 1.\n Unexpected $[1] = 'A' in array.\n Stack:\n Error: Expected $.length = 2 to equal 1.\n Unexpected $[1] = 'A' in array.\n at \n at compare (test-linked-list.js:64:35)\n at UserContext. (test-linked-list.js:111:7)\n at \n Message:\n Expected 2 to be 1.\n Stack:\n Error: Expected 2 to be 1.\n at \n at compare (test-linked-list.js:65:32)\n at UserContext. (test-linked-list.js:111:7)\n at \n```\n\n### Expected behavior\n\nThe `:filter` filter run should remove all non-matching entries from the input\n\n### To Reproduce\n\n1. Paste `A =A B =A :filter[match[B]]` into https://tiddlywiki.com/#%24%3A%2FAdvancedSearch filter tab\n2. See the unexpected results `B A`\n\n### Screenshots\n\n_No response_\n\n### TiddlyWiki Configuration\n\n- Version v5.2.3\n\n\n### Additional context\n\n_No response_"} {"task_id": "format-code-task-000262", "source_id": "format-code-task-000262", "domain": "code", "task_path": "tasks/format-code-task-000262", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:70823af2701d3f11ba118a12cc7d71e256c3aac8a2f9db9e317c743447e952a2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI often need to cap MRI intensity values before the rest of my preprocessing, but I don’t see a built-in TorchIO transform for just clipping a ScalarImage to fixed lower/upper limits. It would be nice if I could use this like the other intensity transforms, including cases where I only want to set one side of the clamp.\n\n## Expected outcomes\n\n- A public `Clamp` intensity transform is available through the usual TorchIO transform APIs, including `torchio.Clamp` and `torchio.transforms.Clamp`.\n- `Clamp` can be instantiated with optional `out_min` and `out_max` keyword arguments, while still accepting the common transform options used by other TorchIO transforms.\n- When applied to scalar intensity images, `Clamp` caps values below the configured lower limit to that lower limit, caps values above the configured upper limit to that upper limit, and leaves values already within the configured range unchanged.\n- `Clamp` supports one-sided use: specifying only `out_min` applies only the lower cap, and specifying only `out_max` applies only the upper cap.\n- The preprocessing transforms documentation lists `Clamp` alongside the other intensity preprocessing transforms.\n\n## Implementation notes\n\n- Follow the conventions of existing TorchIO intensity preprocessing transforms for public API exposure, subject/image handling, and transform options.\n- The internal organization, helper methods, validation location, and exact implementation mechanism are up to the implementer, as long as the observable behavior above is satisfied."} {"task_id": "format-code-task-000263", "source_id": "format-code-task-000263", "domain": "code", "task_path": "tasks/format-code-task-000263", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1499e3c15286c62724381501f585381f649b72552a560688cabd833f67e77483", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Upgrade the templated expression language with literals and inline ternaries\n\nOur templated expressions — the `${{ ... }}` snippets used throughout workflow\ndefinitions — can currently only reference a context (`ACTIONS`, `INPUTS`,\n`SECRETS`, `TRIGGER`, ...) or call an inline function (`FNS.(...)`),\noptionally followed by a `-> ` typecast. Authors keep hitting walls when\nthey just want a constant value or a small conditional, and today the only way\nto get a literal into an expression is to wrap it in a typecast like `int(5)`.\n\nMake the expression language richer in three ways.\n\n**1. Literals as first-class expressions.** A full template whose body is a bare\nliteral must evaluate to the corresponding native Python value, with the right\ntype:\n\n- Integer literals → `int` (e.g. `${{ 42 }}` is `42`, `${{ -5 }}` is `-5`).\n- Float literals → `float` (e.g. `${{ 3.14 }}` is `3.14`).\n- String literals wrapped in single or double quotes → `str` with the quotes\n stripped and the inner text (including spaces) preserved\n (`${{ \"hello world\" }}` is `hello world`, and `${{ 'hi' }}` is `hi`).\n- Boolean literals `True` and `False` → `bool`.\n\nThe existing `-> ` typecast must keep working when applied to a literal\n(e.g. `${{ 42 -> str }}` is the string `\"42\"`).\n\nLiterals must also be usable directly as function arguments, without a\ntypecast — `${{ FNS.add(1, 2) }}` evaluates to `3` and\n`${{ FNS.greater_than(5, 2) }}` evaluates to `True`.\n\n**2. Inline ternary expressions.** Support the form `A if C else B`. The\ncondition `C` is evaluated and standard Python truthiness decides the result: if\n`C` is truthy the expression evaluates to `A`, otherwise to `B`. Each of `A`,\n`C`, and `B` may itself be any supported operand — a literal, a context\nreference, or a function call. For example, with an operand where `INPUTS.x` is\n`20`:\n\n- `${{ \"yes\" if INPUTS.flag else \"no\" }}` picks the branch based on `INPUTS.flag`.\n- `${{ \"big\" if FNS.greater_than(INPUTS.x, 10) else \"small\" }}` is `\"big\"`.\n- `${{ \"a\" if 0 else \"b\" }}` is `\"b\"` and `${{ \"a\" if 1 else \"b\" }}` is `\"a\"`.\n\n**3. Short-circuit evaluation.** Only the selected branch of a ternary is\nevaluated. The branch that is not taken must never be evaluated, so an otherwise\nfailing operand in the untaken branch (for instance a reference to a context\npath that does not exist) must not cause the expression to raise.\n\nAll existing expression behavior — context/jsonpath lookups, inline function\ncalls, inline typecasts of function arguments, typecasting of results, and\ninline substitution inside larger strings — must continue to work exactly as\nbefore.\n"} {"task_id": "format-code-task-000264", "source_id": "format-code-task-000264", "domain": "code", "task_path": "tasks/format-code-task-000264", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b675e06738c06522f23ff4c8db7244bc73a3cc8aac2f4c864f5532d3a2bbe500", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Trigger-aware workflow graph model\n\nOur workflow editor sends the backend a React Flow graph object describing a\nworkflow's topology. Historically the server treated the trigger as a special,\nseparate thing bolted onto the side of the graph. We're consolidating: the graph\nthe frontend sends now contains the trigger as a regular node in the node list,\nright alongside the action nodes, and the server-side `RFGraph` model needs to\nunderstand that shape.\n\nA React Flow graph object is a plain dict with two keys:\n\n- `nodes`: a list of node dicts. Each node has an `id`, a `type`, and a `data`\n object. There is exactly **one** node with `type == \"trigger\"`; every other\n relevant node has `type == \"udf\"` (an action node). An action node's `data`\n carries the namespaced action `type` (the UDF key), a human-readable `title`,\n and an `args` dict.\n- `edges`: a list of edge dicts, each with a `source` and `target` node id (and\n an optional `label`). The trigger node is wired to the workflow's first action\n by a single dedicated edge (the \"trigger edge\"); all other edges connect action\n nodes to one another.\n\nRework `RFGraph` (in the workflow DSL graph module) so it parses this unified\nshape and exposes the following behavior. `RFGraph.from_dict(obj)` should accept\nsuch a dict and build the graph.\n\nThe model must distinguish the trigger from the action nodes:\n\n- `trigger` returns the single trigger node.\n- `action_nodes()` returns only the action (`udf`) nodes, preserving the order\n they appear in the input.\n- `action_edges()` returns the edges that do **not** touch the trigger node.\n- `entrypoint` returns the action node the trigger points to.\n- A node's `ref` is the slugified form of its title (lowercased,\n non-alphanumerics collapsed to underscores), e.g. `\"Action A\"` -> `\"action_a\"`.\n\nDependency and ordering computations must ignore the trigger edge, treating only\nthe action subgraph:\n\n- `topsort_order()` returns the action node ids in a valid topological order\n (the trigger is not part of it).\n- The entrypoint action therefore has no action-level dependencies.\n\nThe graph can also be converted into the workflow's intermediate representation:\n`action_statements()` returns one `ActionStatement` per action node (no external\ninput required), where each statement carries the node's `ref`, its action\n`type`, its `args` (taken from the node data), and `depends_on` — the **sorted**\nlist of refs of the action nodes immediately upstream of it (the trigger\nexcluded).\n\nFinally, parsing must reject malformed graphs by raising the project's workflow\nvalidation error (`TracecatValidationError`) when:\n\n- there is no action node,\n- there is not exactly one trigger node, or\n- the trigger does not connect to exactly one action node.\n"} {"task_id": "format-code-task-000265", "source_id": "format-code-task-000265", "domain": "code", "task_path": "tasks/format-code-task-000265", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:65e9e8cf876032ec7938d8845bfe6f4cb18022cae737ede5e316083d79395dde", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Members content gating doesn't respect a post's `visibility` setting\n\nI'm running Ghost with the `members` labs flag enabled. Posts in the admin UI already have a **visibility** field that can be set to `public`, `members`, or `paid`, and I want to use that to control who can read each post:\n\n- `public` — anyone can read it\n- `members` — only signed-in members can read the content\n- `paid` — only members with an active paid subscription can read the content\n\nIn practice the visibility setting doesn't seem to do anything for me. Whether I switch a post between `public`, `members`, and `paid`, the API still returns the same `plaintext` / `html` to the same caller. The only way I've found to actually hide post content from non-members is to attach a specific tag to the post, which is awkward (it shows up on the post like any other tag) and also can't express the difference between \"members\" and \"paid\" — a free signed-in member ends up seeing paid posts just like a paying one does.\n\nIt would be much nicer if the post serializer used the `visibility` field directly. Roughly the behaviour I'd expect:\n\n| visibility | anonymous request | signed-in free member | signed-in paying member |\n|---|---|---|---|\n| `public` | content returned | content returned | content returned |\n| `members` | content hidden | content returned | content returned |\n| `paid` | content hidden | content hidden | content returned |\n\nThis should apply to both the v2 and canary Content APIs, and it should only kick in when the `members` labs flag is on (so existing sites without members enabled keep behaving exactly as they do today)."} {"task_id": "format-code-task-000267", "source_id": "format-code-task-000267", "domain": "code", "task_path": "tasks/format-code-task-000267", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:475af642a30e5c4283606f243e5f5427139c7bfa7070990e637314432411e165", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nGeneration of schemaorg according to the specification\n### Issue Summary\n\nIt all started with a discussion on the forum: https://forum.ghost.org/t/schema-org-generate/36772/7\n\nApparently, schemaorg is not being generated correctly at the moment.\n\nTake, for example, a page https://ghost.org/changelog/vscode-extension/\n\nHere's her json-id\n\n```\n\n\n```\n\nmainEntityOfPage [according to the explanation of the developers schemaorg](https://github.com/schemaorg/schemaorg/discussions/3274) should contain a link to the article page, not a link to the blog.\n\n### Steps to Reproduce\n\nRun ghost :)\n\n### Ghost Version\n\n5.36+\n\n### Node.js Version\n\n-\n\n### How did you install Ghost?\n\ndocker\n\n### Database type\n\nMySQL 8\n\n### Browser & OS version\n\n_No response_\n\n### Relevant log / error output\n\n_No response_\n\n### Code of Conduct\n\n- [X] I agree to be friendly and polite to people in this repository"} {"task_id": "format-code-task-000268", "source_id": "format-code-task-000268", "domain": "code", "task_path": "tasks/format-code-task-000268", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b1083d3533ac3196f5f5bc59bca2504143e9154657e8fc81827152e530c08e21", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\n我这边评论数据里会有作者已经被删掉的情况，只能传 `member: null`，但现在用 `Comment` 类型会被 TypeScript 拦住，感觉这个状态应该是合法的。界面上这种评论就正常显示成已删除成员就行，没绑到具体评论的头像也别再拿当前用户的名字首字母来凑了。\n\n## Expected outcomes\n\n- 评论数据契约中，公开的 `Comment.member` 应允许 `Member | null`，调用方可以用 `member: null` 表示作者已删除的评论。\n- 评论作者已删除时，界面应按已删除成员展示；评论作者没有姓名时，仍按匿名成员展示；评论作者有姓名时，仍展示对应姓名。\n- 头像首字母只应在头像绑定到具体评论时根据该评论作者生成；未绑定具体评论的头像不应从当前用户或其他非评论作者数据生成首字母。\n- 已有的评论作者姓名与头像首字母 fallback 语义应保持一致：已删除成员、匿名成员、有姓名成员分别走各自对应的显示规则。\n\n## Implementation notes\n\n- 具体类型定义位置、组件内的数据流组织、辅助函数拆分方式由实现者决定。\n- 可以复用现有翻译、渲染和测试基础设施；不要为这些边界状态引入新的用户配置要求。\n- 行为应通过公开类型契约和用户可见的评论 UI 结果体现，而不是依赖特定内部函数名或文件组织。"} {"task_id": "format-code-task-000269", "source_id": "format-code-task-000269", "domain": "code", "task_path": "tasks/format-code-task-000269", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ee440d71d6ee6aeb8e6fbf002ac129c60936f7ad96f16f77ea7c00d8cc9030b5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nFootnote reference links are plaintext in post excerpts\n## Issue Summary\n\nFootnote reference links (which landed in #4270) in posts are shown in post excerpts as plain text\n## Steps to Reproduce\n- Add a footnote in the first few words of a post `Lorem ipsum dolor [^1]`\n- Save/publish post\n- Go to a page on the blog that shows excerpts (index, for example)\n- See the footnote reference in plain text `Lorem ipsum dolor 1`\n\nThese references need to be stripped from excerpts"} {"task_id": "format-code-task-000270", "source_id": "format-code-task-000270", "domain": "code", "task_path": "tasks/format-code-task-000270", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e71702d413a737a73017b96ad83d848815a71b4d661dacb6e7612bafca00a7fd", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Analytics path normalisation doesn't recognise ULIDs\n\nWe use ULIDs as resource identifiers in our API paths, e.g.\n\n```\n/posts/01G9HHNKWGBHCQX7VG3JKSZ055/comments\n/posts/01g9hhnkwgbhcqx7vg3jksz055/comments\n```\n\nTyk's analytics path normalisation already handles UUIDs and numeric IDs nicely — when I enable `normalise_uuids` under `analytics_config.normalise_urls`, request paths containing UUIDs collapse to a single placeholder so the dashboard can aggregate hits per logical endpoint.\n\nULIDs aren't covered, though. Every request to the same logical endpoint shows up as a distinct path in analytics because the ULID segment is unique per resource, and there's no built-in option that catches them. The closest workaround is dropping a regex into `custom_patterns`, but ULIDs are a well-defined, widely used ID format (Crockford base32, 26 chars, case-insensitive) and it feels like they should be a first-class option alongside UUIDs rather than something every operator has to roll themselves.\n\nCould Tyk gain built-in support for normalising ULIDs in analytics paths, in the same spirit as the existing UUID normalisation? Ideally it'd be off by default (so existing deployments don't change behaviour) and configurable from the same `normalise_urls` block."} {"task_id": "format-code-task-000271", "source_id": "format-code-task-000271", "domain": "code", "task_path": "tasks/format-code-task-000271", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ca6ebc025ee81ba81e3f5ee807dee13262ab8eadb4a87781124895ea27381a02", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[ts-interface-generator ] Return type for the method \"remove\" for aggregation of cardinality 0 .. n does not match runtime behaviour\n**Describe the bug**\nI was just having a look into OpenUI5 with TypeScript for a private project. I played around with custom controls.\nI defined a custom aggreagation with `multiple: true`:\n```ts\nimport Control from \"sap/ui/core/Control\";\nimport type { MetadataOptions } from \"sap/ui/core/Element\";\nimport RenderManager from \"sap/ui/core/RenderManager\";\n\n/**\n * @namespace ui5.typescript.helloworld.control\n */\nexport default class MyControl extends Control {\n // The following three lines were generated and should remain as-is to make TypeScript aware of the constructor signatures\n constructor(idOrSettings?: string | $MyControlSettings);\n constructor(id?: string, settings?: $MyControlSettings);\n constructor(id?: string, settings?: $MyControlSettings) { super(id, settings); }\n \n\tstatic readonly metadata: MetadataOptions = {\n\t\tproperties: {\n\t\t\t\"text\": \"string\"\n\t\t},\n aggregations: {\n \"columns\": { type: \"sap.ui.core.Control\", multiple: true},\n }\n\t};\n\n\tstatic renderer = {\n\t\tapiVersion: 2,\n\t\trender: function (rm: RenderManager, control: MyControl): void {\n\t\t\trm.openStart(\"div\", control);\n\t\t\trm.openEnd();\n\t\t\trm.text(control.getText());\n\t\t\trm.close(\"div\");\n\t\t}\n\t};\n\n\tonclick = function() {\n\t\talert(\"Hello World!\");\n\t}\n}\n```\n\nThis leads to the following generated interface:\n```ts\nexport default interface MyControl {\n\n // property: text\n getText(): string;\n setText(text: string): this;\n\n // aggregation: columns\n getColumns(): Control[];\n addColumn(columns: Control): this;\n insertColumn(columns: Control, index: number): this;\n removeColumn(columns: number | string | Control): this;\n removeAllColumns(): Control[];\n indexOfColumn(columns: Control): number;\n destroyColumns(): this;\n }\n}\n```\n\nNevertheless during runtime the method `insertColumn` and ``removeColumn` return the inserted or removed element:\n![image](https://github.com/user-attachments/assets/7646e89c-f25a-40b9-9689-0b096e900505)\n\n**Expected behavior**\nThe interface generator should generate the return type according to runtime behaviour:\n```\n insertColumn(columns: Control, index: number): Control | undefined;\nremoveColumn(columns: number | string | myColumn): Control | undefined;\n```\n\nI checked with other aggregations with the cardinality `0 .. n`. They have all the return type `Type of Aggregation |null` (see `sap.ui.table.Table` or `sap.m.Page`).\n\n**Additional context**\nI used version `0.8.3` of the module `@ui5/ts-interface-generator` and the following OpenUI5 version:\n![image](https://github.com/user-attachments/assets/8c380f23-a931-481a-b1b1-24e2efd11cbf)"} {"task_id": "format-code-task-000272", "source_id": "format-code-task-000272", "domain": "code", "task_path": "tasks/format-code-task-000272", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:63311a389fff1d0a1aa1b1ce3b49ecf5a06364ba4c9b2a8aaf29541688feae53", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nRename `face_dimension` connectivity to `nNodes_per_face`\nCurrent implementation of `face_dimension` does not align with the UGRID conventions. `nNodes_per_face` is a more appropriate name for the current implementation of `face_dimension`"} {"task_id": "format-code-task-000273", "source_id": "format-code-task-000273", "domain": "code", "task_path": "tasks/format-code-task-000273", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:31da58fe8cf81e4c22601ba498d0ff47d263df392091d047c6a1ebaa96e13335", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\n我在用 `metpy.calc.lfc()` / `metpy.calc.el()` 算一条基本没有 CIN 的探空时遇到怪现象：从 LCL 往上抬升气块已经比环境温度暖了，后面也只看到一次和环境温度的交点，但这两个函数给我的都是 `nan`。不太确定是不是我这类自由对流廓线的输入哪里不符合 MetPy 的预期。\n\n# Expected outcomes\n\n- 对于无明显对流抑制、气块自 LCL 起已呈正浮力的自由对流廓线，`metpy.calc.lfc(pressure, temperature, dewpt)` 不应把该情形误判为无 LFC 并返回 `nan`；应返回与该廓线物理上对应的自由对流起点，在这类廓线中即 LCL 的压力和温度。\n- 对于同类自由对流廓线，`metpy.calc.el(pressure, temperature, dewpt)` 不应因为只有一个可观测到的气块/环境温度交汇层而返回 `nan`；当输入廓线包含可判断的平衡高度时，应返回该平衡高度的压力和温度。\n- 对于确实不存在 LFC、确实不存在 EL，或输入廓线不足以判断对应层次的情形，应保持现有无有效层次时返回 `nan` 的行为，不应报告虚假的有效结果。\n\n# Implementation notes\n\n- 具体如何识别交汇层、如何组织中间计算、以及在现有热力学计算流程中的校验位置由实现者决定。\n- 保持现有公开 API、单位处理和数值返回风格；调用者应能继续通过 `metpy.calc.lfc()` 和 `metpy.calc.el()` 获得带单位的压力、温度结果。"} {"task_id": "format-code-task-000275", "source_id": "format-code-task-000275", "domain": "code", "task_path": "tasks/format-code-task-000275", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1b0be7d53c5dc3637a83ada2c67266ca25c04343a247efa12f68af42f50b8615", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nbug/hardcoded GET request to api.unstructuredapp.io causes timeout behind firewall\n**Describe the bug**\nThere is a hardcoded dummy GET request to api.unstructuredapp.io. There does not appear to be an easy way to override this, and it also doesn't look necessary. This causes a timeout error behind a firewall and makes the unstructured-client unusable in this environment. Could this just use the \"server_url\" parameter for unstructured client? \n\nhttps://github.com/Unstructured-IO/unstructured-python-client/blob/18a13093686bca1f97cc061f667fec7db21c489c/src/unstructured_client/_hooks/custom/split_pdf_hook.py#L349\n\n**To Reproduce**\nUse the client with personal unstructured-api deployment that is behind a firewall.\n```\n\ns= UnstructuredClient(,\n server_url=my_server_url,\n api_key_auth=None\n )\nres = s.general.partition(request=req)\n\n```\n**Expected behavior**\nRequest hangs on GET request and will eventually timeout. \n\n**Environment Info**\nunstructured-client=0.26.0"} {"task_id": "format-code-task-000276", "source_id": "format-code-task-000276", "domain": "code", "task_path": "tasks/format-code-task-000276", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1c805366349fb4b515b3df8bbd7729118039f8b5d07731a08f8c10481554f6cf", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## False positive \"undefined variable\" inside `elseif` branch when the condition uses `isset`\n\nI'm running noverify on a fairly small PHP file and it's flagging variables as undefined in an `elseif` branch even though I clearly guard them with `isset` in that same `elseif`'s condition.\n\nRoughly what my code looks like:\n\n```php\n bytes4` query through their public ABI.\n- Calling `isValidSigner` with an address that the account treats as a valid signer returns the EIP-6551 `isValidSigner` magic value.\n- Calling `isValidSigner` with an address that the account does not treat as a valid signer returns a non-magic failure value rather than reporting the address as valid.\n- The `context` argument can be supplied by standard callers without causing the public signer check entrypoint to fail merely because the entrypoint is missing.\n\n## Implementation notes\n\nThe exact internal mechanism used to determine signer validity is up to the implementation, but the externally observable behavior must match the ERC-6551 signer-check contract interface. Existing account behavior unrelated to the signer check should remain unchanged."} {"task_id": "format-code-task-000278", "source_id": "format-code-task-000278", "domain": "code", "task_path": "tasks/format-code-task-000278", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3e4b4ed8ab9f7e5623d7ee3cd3dd50c54ad8661a3493a1c8e9cd7ab474148fc7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: query a service in a specific datacenter\n\nI'm setting up diplomat in an org that runs Consul across multiple datacenters (e.g. `us-east-1` and `us-west-2`). From a Ruby app running in one DC I'd like to look up services that live in another DC.\n\nRight now I'm doing something like:\n\n```ruby\nDiplomat::Service.get('my-service', :all)\n```\n\nwhich always returns whatever's registered in the local datacenter. Looking at `Diplomat::Service#get`, the only knobs on the `options` hash today are `wait` and `index` — there's no way to ask for a different DC.\n\nConsul's HTTP catalog endpoint itself supports targeting a specific datacenter via a query parameter (see the [catalog docs](https://consul.io/docs/agent/http/catalog.html#catalog_service)), so it would be great if `Service#get` exposed that on the options hash too, in the same style as the existing `wait` / `index` entries. Then I could just pass the target datacenter alongside the service name and get back the nodes registered there.\n\nHappy to add specs for it if useful."} {"task_id": "format-code-task-000279", "source_id": "format-code-task-000279", "domain": "code", "task_path": "tasks/format-code-task-000279", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0279ddd1464cf015aea71c8816f11cb09c30e8999f7d231f4ad17e3b83fac48d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Split the review workflow into source and translation reviews\n\nRight now a project has a single switch that turns the review workflow on or off\nfor everything. We want finer control: reviewing **source strings** should be\nconfigurable independently from reviewing **translations**.\n\nPlease give a project two independent boolean on/off settings:\n\n* `source_review` — enables the review workflow for source strings, and\n* `translation_review` — enables the review workflow for translations.\n\nBoth must default to **off** (`False`) for newly created projects.\n\nA lot of the codebase needs to know \"is review in effect for *this* translation?\".\nMake that easy to ask directly on a translation through a boolean\n`enable_review` attribute: it must report the project's `source_review` setting\nfor a translation that holds the source strings, and the project's\n`translation_review` setting for every other translation.\n\nWire the rest of the review machinery to the new settings instead of one\nproject-wide flag:\n\n* **The `unit.review` permission.** When the object being checked is a\n translation, the permission may only be granted when review is enabled for\n that particular translation (on top of the user actually being allowed to\n review). When the object being checked is a whole component or project, the\n enablement gate is satisfied as long as *either* review type is enabled on the\n project.\n\n* **Approved state.** A unit may only settle into the *approved* state when\n review is enabled for the translation it belongs to — a source unit needs\n source review, any other unit needs translation review. With the relevant\n review type disabled, a unit that would otherwise be approved must instead be\n treated as merely translated.\n\n* **The per-project \"Review\" access-control group.** This group should be\n created only when at least one of the two review types is enabled for the\n project; when both are disabled it must not be created.\n\nThe existing single-flag behavior should be fully replaced by the two new\nsettings.\n"} {"task_id": "format-code-task-000280", "source_id": "format-code-task-000280", "domain": "code", "task_path": "tasks/format-code-task-000280", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:08525ed57fc67cbb9af9d1c305f8219399ac58d7a065e8e5662d018fd658e32e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### What problem does this address?\n\nAs part of the ongoing migration to the new 40px default size for form controls (see #65751), most components are being updated to opt into the larger default. `MenuItem` still appears to render at the old 36px height — you can see this in the `MenuItem` stories in Storybook, where the items look noticeably shorter than equivalent components that have already been migrated.\n\n### Expected\n\n`MenuItem` should render at the new 40px default height, consistent with other components that have been updated as part of the size migration effort.\n\n### Actual\n\n`MenuItem` still renders at 36px, so it's out of step with the rest of the design system migration."} {"task_id": "format-code-task-000281", "source_id": "format-code-task-000281", "domain": "code", "task_path": "tasks/format-code-task-000281", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:116dbf655570d375e2c87c341e839dc88666e9967fda15e117d5fdd81c857a5b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nThe ingestion pipeline turns provider records into PostgreSQL TSV rows through the public column classes in `catalog/dags/common/storage/columns.py`. A recent re-ingestion exposed that the conversion layer is lossy and that the `created_on` no-change policy does not produce a usable upsert expression: nested metadata can silently turn numbers and booleans into strings, array values can be ambiguous when they contain PostgreSQL delimiters, and the loader can generate `NULL` instead of preserving an existing timestamp. These are data-integrity bugs because a row can be accepted while changing its meaning or while an upsert stops protecting immutable fields.\n\nMake the column serialization contract precise and safe while preserving the existing public APIs and compatibility behavior. `JSONColumn.prepare_string` must recursively sanitize only string leaves (collapse whitespace, remove backspaces, replace double quotes and escape backslashes as the existing string sanitizer does); JSON numbers, booleans, and nulls must remain their JSON scalar types at any nesting depth. `None` and empty dictionaries/lists continue to represent absent optional JSON and return `None`, while a non-empty container containing null must serialize as JSON null rather than disappearing. The result must be valid JSON and Unicode must remain readable.\n\n`ArrayColumn.prepare_string` must return a PostgreSQL array literal. It must represent an empty list as `{}`, a null input as `None`, and preserve null elements as SQL NULL elements. Every non-null element is passed through the base column validator; invalid elements become NULL rather than making the whole array invalid. Elements containing commas, braces, quotes, backslashes, or the literal text `NULL` must remain one quoted element with the required escaping, and a scalar input remains one element. Do not mutate caller-provided lists while converting them.\n\nThe `UpsertStrategy.no_change` contract must be usable by the loader: the immutable column's insert expression remains `NOW()`, and its conflict expression must be a valid self-assignment that retains the old row value (` = old.`), never a NULL placeholder. Existing `StringColumn` sanitization/truncation and `URLColumn` validation behavior must continue to work unchanged."} {"task_id": "format-code-task-000283", "source_id": "format-code-task-000283", "domain": "code", "task_path": "tasks/format-code-task-000283", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:29cfdb628a5086a0bffd299ed3f9c5daf0d5e0427e1f9c5981e63fe3b02d4469", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nNew general page: FTC tab gives React warning\n\n\n* [ ] I've read and understood the [contribution guidelines](https://github.com/Yoast/wordpress-seo/blob/trunk/.github/CONTRIBUTING.md).\n* [ ] I've searched for any related issues and avoided creating a duplicate issue.\n\n### Please give us a description of what happened\n\n### To Reproduce\n#### Step-by-step reproduction instructions\n1. Enable debug output (`SCRIPT_DEBUG`)\n2. Visit Yoast > General\n3. Notice the warning in the browser console:\n> Warning: Failed prop type: The prop `onDiscard` is marked as required in `UnsavedChangesModal`, but its value is `undefined`. \n\nThe `useBlocker` can return `undefined` for `proceed` (and `reset`) for specific states.\nSee their type def in the documentation: https://reactrouter.com/en/main/hooks/use-blocker\n\n#### Expected results\n1. No prop warnings\n\n#### Actual results\n1. Prop warning\n\n### Screenshots, screen recording, code snippet\nIf possible, please provide a screenshot, a screen recording or a code snippet which demonstrates the bug.\n\n### Technical info\n\n* If relevant, which editor is affected (or editors): \n- [ ] Block Editor\n- [ ] Gutenberg Editor\n- [ ] Elementor Editor\n- [ ] Classic Editor \n- [ ] Other: \n\n\n* Which browser is affected (or browsers): \n- [ ] Chrome\n- [ ] Firefox\n- [ ] Safari\n- [ ] Other: \n\n#### Used versions\n* Device you are using:\n* Operating system:\n* PHP version:\n* WordPress version: \n* WordPress Theme:\n* Yoast SEO version: \n* Gutenberg plugin version: \n* Elementor plugin version: \n* Classic Editor plugin version: \n* Relevant plugins in case of a bug:"} {"task_id": "format-code-task-000284", "source_id": "format-code-task-000284", "domain": "code", "task_path": "tasks/format-code-task-000284", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:97092fafcac1a7e3e8c96e147374296ef432fd062492cf09864fae859cc3046c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Deduplicate type-alias definitions in the JSON Schema output\n\nOur generator can emit shared, named definitions for TypeScript type aliases when\nthe alias-ref feature is enabled (the option that gives each referenced type alias\nits own entry under `definitions`, referenced via `$ref`). That part works, but the\noutput is wasteful: when an alias is nothing more than a name for another *named*\ntype, the generator copies that target type's entire schema into the alias's\ndefinition. For example, with the feature on, `type MyAlias = MyObject` produces a\n`MyObject` definition and a second, byte-for-byte duplicate of it under `MyAlias`.\nFor recursive shapes this duplication is especially confusing.\n\nChange this so that an alias which simply names another reffable type is emitted as\na `$ref` to that type's definition instead of a full copy.\n\nExpected behavior, with the alias-ref feature enabled:\n\n- A type alias whose target is another named type that gets its own definition\n (e.g. an interface or class) must appear under `definitions` as exactly\n `{ \"$ref\": \"#/definitions/\" }`, and the target type must still be\n present under `definitions` with its full schema.\n- The target's full definition must be emitted regardless of whether anything else\n references it directly — i.e. the alias being a `$ref` must not cause the target's\n definition to go missing.\n- A direct (non-alias) reference to a named type is unaffected: a property typed as\n the target type still points at `#/definitions/` and that definition is\n the full schema, never a `$ref` indirection.\n- This works for recursive shapes: if the alias and its target refer to each other\n (directly or through properties), each named type appears once under `definitions`,\n the alias entry is the `$ref` to its target, and the target keeps its full\n property schema with the usual `$ref`s between them. No infinite expansion.\n- Aliases of types that do not get their own named definition are unchanged. In\n particular, an alias of a primitive (e.g. `type MyString = string`) or of an\n inline/anonymous object type keeps its inline schema rather than becoming a `$ref`.\n\nOther output (the `$schema` field, `type`/`properties`/`required` contents, root\nhandling, and the schema's validity against the JSON Schema meta-schema) stays as it\nis today.\n"} {"task_id": "format-code-task-000285", "source_id": "format-code-task-000285", "domain": "code", "task_path": "tasks/format-code-task-000285", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:72ec3726ca06885f614c78dc57a474b8c85e9bf190a4c1a258c887884a92baa4", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI'm using an async `@depends_on` listener to initialize a client, and when I call `validate_all_configs(include_listeners=True)` at startup I see a `coroutine was never awaited` warning. The listener body doesn't seem to run during validation, and the singleton it returns is still missing until later. I might be wiring it wrong, but the same startup validation does run my sync listeners.\n\n## Expected outcomes\n\n- Startup validation with async listeners: when async `@depends_on` listeners are included in startup validation, an awaited validation call completes their initialization during validation, without leaking un-awaited coroutine warnings.\n- Await boundary: if `include_listeners=True` reaches at least one async listener, `validate_all_configs(include_listeners=True)` provides an awaitable validation operation, and listener initialization should not happen merely by creating that operation.\n- Sync compatibility: when only synchronous listeners are involved, `validate_all_configs(include_listeners=True)` remains a synchronous call and still eagerly loads configs and runs those listeners without requiring `await`.\n- Documentation: the listener usage documentation describes the async-listener startup-validation pattern, including that the validation call is awaited when async listeners are included.\n\n## Implementation notes\n\n- Preserve the existing public API names and normal synchronous behavior for applications that only use synchronous listeners.\n- The internal scheduling approach, listener bookkeeping, and validation structure are implementation details.\n- Initialization failures from listeners should still surface during startup validation rather than being silently delayed or swallowed."} {"task_id": "format-code-task-000286", "source_id": "format-code-task-000286", "domain": "code", "task_path": "tasks/format-code-task-000286", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:edc4d289de6e1bf9cb875c482de8709ab002d50ac4e5b7f424559e219123a6cb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the Go `parse` package to provide a stateful MediaWiki dump stream parser. The public API should include `func NewParser() *Parser`, a `Parser` value with an exported `BytesProcessed int64` progress field initialized to 0, and `func (p *Parser) Parse(reader io.Reader, pages chan<- *Page)`. A typical session should be `p := parse.NewParser(); pages := make(chan *parse.Page); p.Parse(reader, pages);` followed by receiving `*parse.Page` values from the channel until it closes. `Parse` should scan the reader for complete `...` elements inside a larger XML dump, preserve content across read chunks, ignore text outside pages, and emit pages in dump order. For an input containing an `Alpha` page with `[[Beta]]` and `[[Category:Science]]`, then a page whose text contains `#REDIRECT [[Alpha]]`, then a `Gamma` page with `[[Alpha]]`, the channel should yield `Alpha` with its parsed link/category data and `Gamma`, and it should not yield the redirect page. If a complete page is malformed or has no valid article title, the parser should skip that page and keep processing later pages instead of sending an error value. When the reader is exhausted, `Parse` should close the output channel so callers can range over it without hanging. While reading, `BytesProcessed` should increase by the number of bytes consumed from the reader, including bytes for skipped, malformed, redirect, and wrapper XML content. Separate `Parser` instances should maintain independent `BytesProcessed` counters and independent parse sessions."} {"task_id": "format-code-task-000287", "source_id": "format-code-task-000287", "domain": "code", "task_path": "tasks/format-code-task-000287", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c021f0301ae635c2235b0f14baa7d5776cfaedd2d5d52bad1d22e7f8510e1836", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `stats` package to expose pure readability helpers for article prose: `Words(text string) []string`, `Sentences(text string) []string`, `SyllableCount(word string) int`, and `FleschKincaidEase(text string) float64`.\n\n`Words` should return all word-like tokens made of letters, digits, underscores, and apostrophes. For `Words(\"the fox jumped over the road. That's pretty cool...\\tyup?\")`, it should return `[]string{\"the\", \"fox\", \"jumped\", \"over\", \"the\", \"road\", \"That's\", \"pretty\", \"cool\", \"yup\"}`; for `Words(\"A_1 isn't 42!\")`, it should return `[]string{\"A_1\", \"isn't\", \"42\"}`.\n\n`Sentences` should split text on one or more `.`, `?`, or `!` delimiters, consume surrounding ASCII whitespace around the delimiter, and omit the final empty segment when the text ends with punctuation. For `Sentences(\"One. Two...\\nThree!Four?\")`, it should return `[]string{\"One\", \"Two\", \"Three\", \"Four\"}`; for `Sentences(\"No final punctuation\")`, it should return `[]string{\"No final punctuation\"}`.\n\n`SyllableCount` should estimate syllables deterministically for a single word: words of length three or less count as one syllable, common silent endings such as `ed`, selected trailing `es`, and selected trailing `e` are ignored, a leading `y` is ignored, and contiguous vowel groups of one or two vowels count as syllables. For example, `SyllableCount(\"logorrhoea\")` should return `4`, `SyllableCount(\"used\")` should return `1`, `SyllableCount(\"makes\")` should return `1`, and `SyllableCount(\"themselves\")` should return `2`.\n\n`FleschKincaidEase` should compute the Flesch Reading Ease score from the helper outputs using `206.835 - 1.015*(words/sentences) - 84.6*(syllables/words)`. These functions should be deterministic, should not mutate their input strings, and should not perform filesystem, network, or global-state side effects."} {"task_id": "format-code-task-000290", "source_id": "format-code-task-000290", "domain": "code", "task_path": "tasks/format-code-task-000290", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f4e3aa81697ef62650bf8361e6e1ba0f8cec20fc29eee2b2d6227e20a239ff29", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `acho(options)` (or `new Acho(options)`) to create a stateful logger that exposes one callable method for every configured log type, including `.debug()`, `.info()`, `.warn()`, `.error()`, and `.fatal()` by default. Each level method should accept message values, format them with the logger's formatting options, send one generated line to the configured `transport` when permitted, and return the same logger so calls can be chained; for example, `log = acho({ transport }); result = log.warn('disk low')` should produce one transport entry and `result === log`. The `level` option and runtime `log.level` should filter output by severity: `fatal` permits only `.fatal()`, `error` permits `.fatal()` and `.error()`, and `muted` permits none, while `all` permits every configured type; a call below the threshold must produce no transport entry. Custom entries in `types` should receive the same generated callable method and participate in the configured severity ordering."} {"task_id": "format-code-task-000291", "source_id": "format-code-task-000291", "domain": "code", "task_path": "tasks/format-code-task-000291", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:161d93583fc8d25319f303514cd8f27f1124d35feb2a6b4e9723dcf227e62bf5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Wake to include a built-in `tx-origin` detector available from the CLI as `wake detect tx-origin [OPTIONS]`. When I run it on Solidity sources, it should report every direct use of `tx.origin` that is not one of the detector's explicit safe patterns, using the message `Unsafe usage of tx.origin`, medium impact, and low confidence.\n\nFor example, a contract containing `address caller = tx.origin;`, `require(tx.origin == owner);`, `require(tx.origin > caller);`, or an assignment/lookup such as `lastSeen[tx.origin] = block.number;` should produce a detection pointing at the `tx.origin` member access. The detector should not report the unsafe-usage finding for the pattern `tx.origin == msg.sender`, and it should not report that finding for rate-limit style indexed checks where `tx.origin` is used as a mapping key and the indexed expression is compared with `block.timestamp` using `<` or `>`.\n\nBy default the command should also report ERC-4337 account-abstraction compatibility warnings for each `tx.origin` access, using the message `Use of tx.origin may interfere with ERC-4337 account abstraction`, warning impact, and low confidence. The command needs a boolean option `--account-abstraction/--no-account-abstraction`, defaulting to enabled; when I pass `--no-account-abstraction`, those ERC-4337 warnings should be suppressed while the unsafe-usage detections still run. `wake detect tx-origin --help` should describe the command as `Possibly incorrect usage of tx.origin` and show the account-abstraction option with its default behavior."} {"task_id": "format-code-task-000292", "source_id": "format-code-task-000292", "domain": "code", "task_path": "tasks/format-code-task-000292", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0f03fbba80e4331565dd20b6fcb994210f0bebf495ece528e5a59636cb83c41c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Check whether granted permissions satisfy requested permissions\n\nAcorn images can declare the RBAC permissions a workload needs, and an operator\nseparately decides which permissions they are willing to *grant* to that\nworkload. Before we deploy anything we need to answer one question: **do the\ngranted permissions fully cover everything the image is requesting?** — and when\nthey don't, we want to know exactly which requested rules are still missing so we\ncan report them back to the user.\n\nPermissions are already modeled in the internal `v1` API package\n(`pkg/apis/internal.acorn.io/v1`): a `Permissions` value has a `ServiceName` and a\nlist of `PolicyRule`s (each rule embeds the standard Kubernetes\n`rbacv1.PolicyRule` — `Verbs`, `APIGroups`, `Resources`, `ResourceNames`,\n`NonResourceURLs` — plus a list of acorn `Scopes`). Helpers already exist on these\ntypes for things like resolving the namespaces a rule applies to based on its\nscopes.\n\nAdd a function to that package that compares a set of *granted* permissions\nagainst a set of *requested* permissions and reports what, if anything, is\nmissing:\n\n```go\nfunc Grants(granted Permissions, currentNamespace string, requested Permissions) (missing Permissions, granted bool)\n```\n\nIt must return a `Permissions` value holding the subset of the requested rules\nthat are **not** covered by the granted permissions, together with a boolean that\nis `true` only when nothing is missing. The returned value must always carry the\nrequested permissions' `ServiceName`, even when nothing is missing. A request with\nno rules is trivially satisfied.\n\nA single requested rule is considered covered only if the granted permissions\nshare the same `ServiceName` **and** at least one granted rule is broad enough to\ncover it. Coverage of one rule by another works like this:\n\n- **Resource rules** (rules that don't list any `NonResourceURLs`): the granted\n rule covers the requested rule only when they apply to at least one common\n resolved namespace (derived from each rule's scopes relative to\n `currentNamespace`) and the granted rule's `Verbs`, `APIGroups`, and `Resources`\n each cover *all* of the corresponding values on the requested rule. The granted\n rule's `ResourceNames` must likewise cover the requested `ResourceNames`, with\n one special case: an **empty** `ResourceNames` on the granted rule means \"any\n resource name\", so it covers any requested `ResourceNames`.\n\n- **Non-resource URL rules**: a granted rule that lists `NonResourceURLs` (and no\n `Resources`) covers a requested rule only when neither rule declares any scopes\n and the granted `NonResourceURLs` cover the requested `NonResourceURLs`. A\n granted rule that lists `NonResourceURLs` *together with* `Resources` covers\n nothing.\n\nA list of allowed values \"covers\" a requested value when: the requested value is\nthe empty string; or the allowed list contains `\"*\"`; or it contains the value\nexactly; or it contains an entry ending in `\"*\"` whose leading portion is a prefix\nof the requested value (e.g. allowed `\"secret*\"` covers requested `\"secrets\"`).\nAn allowed list covers a list of requested values only when it covers every one of\nthem. (The empty-list-means-all rule above applies only to `ResourceNames`; for\nthe other fields an empty allowed list covers only empty requested values.)\n"} {"task_id": "format-code-task-000293", "source_id": "format-code-task-000293", "domain": "code", "task_path": "tasks/format-code-task-000293", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f24542df771e2ec29e265ed23a9c049d27ff3d08ce3d3c19e1da05190a3f7798", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add \"helper\" classes with function-like, parameterized syntax\n\nAtomizer currently only understands *pattern* rules: a rule has a `prefix` plus a\nlist of `properties`, and a class name such as `P-10px` maps the matched value\nonto those properties. We want to support a second kind of rule — a **helper** —\nthat behaves more like a function: the class name carries a parenthesized,\ncomma-separated argument list, and the rule supplies a declaration template whose\nplaceholders are filled in from those arguments.\n\n## Declaring a helper rule\n\nA helper is just another entry in the rules array passed to the `Atomizer`\nconstructor (or to `addRules`), distinguished by `type: 'helper'`:\n\n```js\n{\n type: 'helper',\n name: 'Line clamp',\n prefix: 'LineClamp',\n declaration: {\n 'lines': '$0',\n 'max-height': '$1'\n },\n rules: {\n '.base': { 'overflow': 'hidden' }\n }\n}\n```\n\n- `prefix` is the token that starts the class name.\n- `declaration` is a map of CSS property → value that acts as a template. A value\n may contain numbered placeholders `$0`, `$1`, `$2`, … referring to the\n arguments (0-indexed) taken from the class name.\n- `rules` is optional. When present it is a map of additional, ready-made CSS\n blocks (selector → declarations) that this helper depends on.\n\nA helper rule and ordinary pattern rules must be able to coexist in the same\n`Atomizer` instance; class names of both kinds must be recognized.\n\n## Class-name syntax\n\nA helper class name is the prefix followed by a parenthesized, comma-separated\nargument list, e.g. `LineClamp(2,40px)`. The argument list may also be empty,\ne.g. `Bar()`. These names must be picked up by `findClassNames` just like\nordinary atomic class names are.\n\n## CSS generation\n\n`getCss` must turn a helper class name into a CSS rule whose selector is that\nclass name (with CSS-special characters such as `(`, `)` and `,` backslash-escaped,\nexactly as other atomic class selectors are escaped) and whose body is the\nhelper's `declaration` with every `$N` placeholder replaced by the Nth argument\nparsed from the class name. For example, with the helper above,\n`LineClamp(2,40px)` yields a rule `.LineClamp\\(2\\,40px\\)` declaring `lines: 2`\nand `max-height: 40px`.\n\nIf the helper defines `rules`, those base CSS blocks must also appear verbatim in\nthe generated output.\n\nA helper that has no `declaration` is invalid: when `getCss` is asked to generate\na class for such a helper it must throw an `Error`.\n"} {"task_id": "format-code-task-000294", "source_id": "format-code-task-000294", "domain": "code", "task_path": "tasks/format-code-task-000294", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:92b16cbd23895cd472350aa4bbdd1c5871513b933b87c61fa55dfa5eb2c56fda", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## 希望增加一条审核规则：禁止使用 EVENT\n\n我们在用 SQLE 做 SQL 审核，希望能在使用建议类规则里加一条\"禁止使用 event\"。\n\n业务背景：MySQL 的 event scheduler 在我们这边一直是不推荐的写法。一方面 event 会让数据库自身承担定时任务的执行职责，运维上很难追踪到底有哪些 event 在跑、什么时候跑、谁加的；和应用侧的调度系统也容易出现重复触发。另一方面 event 里通常带 DEFINER，权限问题也比较敏感。所以我们希望直接在审核阶段把 event 相关的 DDL 拦下来。\n\n具体来说，下面这类 SQL 在开启该规则时应该被审核出问题（出错级别）：\n\n```sql\nCREATE EVENT my_event\n ON SCHEDULE EVERY 1 DAY\n DO DELETE FROM logs WHERE created_at < NOW() - INTERVAL 30 DAY;\n\nCREATE DEFINER = 'admin'@'%' EVENT my_event2\n ON SCHEDULE AT '2025-01-01 00:00:00'\n DO ...;\n\nALTER EVENT my_event DISABLE;\nALTER DEFINER = 'admin'@'%' EVENT my_event ON SCHEDULE EVERY 1 HOUR DO ...;\n```\n\n普通的 DDL/DML（CREATE TABLE / ALTER TABLE / SELECT 等）当然不应该被这条规则误报。\n\n希望这条规则归类到\"使用建议\"下，跟现有的 ddl_avoid_text、ddl_avoid_full_text、ddl_avoid_geometry 这些\"避免使用 xxx\"的规则风格一致，便于我们在模板里统一开启。"} {"task_id": "format-code-task-000295", "source_id": "format-code-task-000295", "domain": "code", "task_path": "tasks/format-code-task-000295", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:308f10c39c49d82d5ae0a5f0dbe711d776c3962441915df95e0d358ab60e4baf", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## 希望新增一条审核规则：禁止使用 rename / change 修改表名和字段名\n\n我们在用 sqle 给开发团队做 SQL 上线审核。最近线上接连出过几次故障，根因都是 DDL 里把已经在用的表名或字段名给改掉了，但应用代码里还有 SQL 依赖旧名字，发版之后业务直接报错。\n\n我们组里定的规范是：一旦表 / 字段已经上线了，就不允许直接改名。要换名字必须走\"加新字段（或新表）→ 双写 / 数据迁移 → 灰度切流 → 删旧字段（或旧表）\"这一整套流程，而不是 DBA 一条 DDL 直接 rename 过去。\n\n但目前 sqle 现有的 DDL 规范里没有一条专门拦截\"改名\"操作的规则。下面这几种写法审核都能正常通过：\n\n```sql\n-- 改表名\nRENAME TABLE t1 TO t2;\nALTER TABLE t1 RENAME TO t2;\n\n-- 改字段名\nALTER TABLE t1 RENAME COLUMN c1 TO c2;\nALTER TABLE t1 CHANGE COLUMN c1 c2 INT; -- CHANGE 顺带就能把字段名改了，最容易漏\n```\n\n希望能在 DDL 规范分类里加一条新规则，把上面这几类语句识别出来并直接报错。这样开发提交改名 DDL 的时候就能在审核阶段被拦下来，引导他们走规范的迁移流程，而不是等故障了再追查。\n\n新规则的 const key 建议叫 `DDLNotAllowRenaming`。"} {"task_id": "format-code-task-000296", "source_id": "format-code-task-000296", "domain": "code", "task_path": "tasks/format-code-task-000296", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f7bcd381200ef530e1be9b8c35c435138e6dc9cb93c8ff14cd0630004560bf83", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### STORAGES fixer clobbers inherited storages when settings are split across modules\n\nI have a typical split settings layout:\n\n```\nexample/\n settings/\n base.py\n production.py\n```\n\n`base.py` defines a custom `DEFAULT_FILE_STORAGE` (and also a `STATICFILES_STORAGE`). `production.py` inherits from it via a star-import and then overrides just one of them:\n\n```python\n# example/settings/production.py\nfrom example.settings.base import *\n\nDEFAULT_FILE_STORAGE = \"example.storages.S3Storage\"\n```\n\nWhen I run django-upgrade with `--target-version 4.2` over both files, `base.py` gets rewritten nicely into a combined `STORAGES = {...}` dict with both `\"default\"` and `\"staticfiles\"` keys — great.\n\nBut `production.py` gets rewritten to:\n\n```python\nfrom example.settings.base import *\n\nSTORAGES = {\n \"default\": {\n \"BACKEND\": \"example.storages.S3Storage\",\n },\n}\n```\n\nThat's a regression in behavior. Before the rewrite, `production.py` was inheriting `STATICFILES_STORAGE` (and anything else storage-related) from `base.py` via the star-import and only overriding the default file storage. After the rewrite, `production.py` defines a fresh `STORAGES` dict with only `\"default\"` in it, which fully replaces the inherited dict — so the `\"staticfiles\"` entry from `base.py` is silently dropped at runtime.\n\nIt would be nice if the fixer recognized that the module is pulling settings in via `from ...settings... import *` and produced something that extends the inherited dict instead of replacing it, so the override semantics are preserved."} {"task_id": "format-code-task-000297", "source_id": "format-code-task-000297", "domain": "code", "task_path": "tasks/format-code-task-000297", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9132e6b51fd0542b2dcae22f613efdcfcb37b1bac692d49f4021990d038da2a1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nDefault file storage is unset by django-upgrade\n### Python Version\n\n3.11.3\n\n### Django Version\n\n4.2.4\n\n### Package Version\n\n1.14.0\n\n### Description\n\nRelated to the feature added in https://github.com/adamchainz/django-upgrade/pull/321\n\nToday I tried to run django-upgrade on a project where `STATICFILES_STORAGE` was defined in the settings, but not `DEFAULT_FILE_STORAGE`, which caused this change in my settings file:\n\n```diff\n- STATICFILES_STORAGE = \"my_project.storage.CustomManifestStaticFilesStorage\"\n+ STORAGES = {\n+ \"staticfiles\": {\n+ \"BACKEND\": \"my_project.storage.CustomManifestStaticFilesStorage\",\n+ },\n+ }\n```\n\nWhen I tried to run the tests, I got an error which seems to be caused by the fact that there is no default file storage configured.\n\n
\nStacktrace\n\n
\nTraceback (most recent call last):\n  File \"/usr/local/lib/python3.11/site-packages/django/core/files/storage/handler.py\", line 35, in __getitem__\n    return self._storages[alias]\n           ~~~~~~~~~~~~~~^^^^^^^\nKeyError: 'default'\n\nDuring handling of the above exception, another exception occurred:\n\nTraceback (most recent call last):\n  File \"/usr/local/lib/python3.11/site-packages/django/core/files/storage/handler.py\", line 38, in __getitem__\n    params = self.backends[alias]\n             ~~~~~~~~~~~~~^^^^^^^\nKeyError: 'default'\n\nDuring handling of the above exception, another exception occurred:\n\nTraceback (most recent call last):\n  File \"/usr/local/bin/pytest\", line 8, in \n    sys.exit(console_main())\n             ^^^^^^^^^^^^^^\n  File \"/usr/local/lib/python3.11/site-packages/_pytest/config/__init__.py\", line 189, in console_main\n    code = main()\n           ^^^^^^\n\n...\n\n  File \"/code/projects/onfido/models.py\", line 202, in OnfidoCheck\n    results_pdf = models.FileField(\n                  ^^^^^^^^^^^^^^^^^\n  File \"/usr/local/lib/python3.11/site-packages/django/db/models/fields/files.py\", line 239, in __init__\n    self.storage = storage or default_storage\n                   ^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/usr/local/lib/python3.11/site-packages/django/utils/functional.py\", line 266, in inner\n    self._setup()\n  File \"/usr/local/lib/python3.11/site-packages/django/core/files/storage/__init__.py\", line 38, in _setup\n    self._wrapped = storages[DEFAULT_STORAGE_ALIAS]\n                    ~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^\n  File \"/usr/local/lib/python3.11/site-packages/django/core/files/storage/handler.py\", line 40, in __getitem__\n    raise InvalidStorageError(\ndjango.core.files.storage.handler.InvalidStorageError: Could not find config for 'default' in settings.STORAGES.\n
\n\n
\n\nSince [the setting `DEFAULT_FILE_STORAGE`](https://docs.djangoproject.com/en/4.2/ref/settings/#default-file-storage) defaults to `django.core.files.storage.FileSystemStorage`, should django-upgrade also set this value in the `STORAGES` setting in case the `DEFAULT_FILE_STORAGE` isn't defined?"} {"task_id": "format-code-task-000298", "source_id": "format-code-task-000298", "domain": "code", "task_path": "tasks/format-code-task-000298", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a39df10c63871216410ecdef19bd5a28d1f5168f9a42edf740db0ac28d682f92", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Numeral.js to support byte quantity formatting and parsing through its existing one-shot public calls: `numeral(input).format(format, roundingFunction?) -> string` for formatting numbers and `numeral(input).value() -> number|null` when the input is a formatted byte string.\n\nA format mask containing `b` should scale using decimal powers of 1000 and append byte suffixes such as `B`, `KB`, `MB`, `GB`, `TB`, and `PB`. A format mask containing `ib` should scale using binary powers of 1024 and append IEC suffixes such as `B`, `KiB`, `MiB`, `GiB`, `TiB`, and `PiB`. If the mask has a space before `b` or `ib`, the output should include a space before the suffix; otherwise the suffix should be adjacent.\n\nConcrete examples I expect: `numeral(100).format('0b')` returns `100B`, `numeral(2000).format('0 b')` returns `2 KB`, `numeral(Math.pow(1024, 2) * 5).format('0ib')` returns `5MiB`, `numeral(Math.pow(1024, 3) * 7.343).format('0.[0] ib')` returns `7.3 GiB`, and `numeral(Math.pow(1000, 4) * 3.1536544).format('0.000b')` returns `3.154TB`. Formatting `null` with `numeral(null).format('0 b')` should produce `0 B`.\n\nParsing should recognize the same decimal and binary suffixes and return the raw byte count from `.value()`: `numeral('5MB').value()` returns `5000000`, `numeral('2 KiB').value()` returns `2048`, `numeral('7.3 GiB').value()` returns `Math.pow(1024, 3) * 7.3`, and `numeral('3PB').value()` returns `Math.pow(1000, 5) * 3`. Byte suffix detection should avoid colliding with basis-point strings, so input such as `1 BPS` should remain available for the BPS format instead of being claimed as a byte value.\n\nFor the same input and format mask, calls should return the same result each time, should not mutate caller-provided values, and should not perform filesystem or network side effects."} {"task_id": "format-code-task-000300", "source_id": "format-code-task-000300", "domain": "code", "task_path": "tasks/format-code-task-000300", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b2993906c40fd1e447758c2b4852553ff69ffdfff7eb26f5413b0be279c4b2bb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the tasker package to expose `MustParseTaskfile(opts tasker.Option) []tasker.Task` as a one-shot parser for crontab-style task files. It should open `opts.File`, read it line by line, trim leading spaces and tabs, ignore blank lines and lines whose first non-space character is `#`, and return the accepted tasks in the same order they appear in the file.\n\nEach accepted non-comment line should split into a schedule expression and a shell command. For example, a file containing `*/1 0/1 * * * echo '[task 1]' > test/task1.out` should produce a task with `Expr == \"*/1 0/1 * * *\"` and `Cmd == \"echo '[task 1]' > test/task1.out\"`; a line `@always echo '[task 3] @always' > test/task3.out` should produce `Expr == \"@always\"` and the command after the alias. The supported aliases should include `@annually`, `@yearly`, `@monthly`, `@weekly`, `@daily`, `@hourly`, `@5minutes`, `@10minutes`, `@15minutes`, `@30minutes`, `@always`, and `@everysecond`.\n\nCron-prefix parsing should handle normal five-field expressions and the existing extended forms with seconds, day names, ranges, lists, steps, `L`, `W`, `#`, and optional years, stopping at the first token that is part of the command. For example, `*/3 * ? * ? 0 2000-2024/4 echo \"run\"` should return `Expr == \"*/3 * ? * ? 0 2000-2024/4\"` and `Cmd == \"echo \\\"run\\\"\"`, and `*/12 * ? * ? mon echo \"run\"` should preserve the spacing that belongs between accepted cron segments while returning the command text after the schedule.\n\nBefore appending a task, the parsed expression should be validated with the cron evaluator. Malformed lines such as `@invalid` or `* * * * *` with no command should not appear in the returned slice, and they should log `[parser] can't parse cron expr: `. If `opts.File` cannot be opened, the function should log `[parser] can't open file: ` and exit with status code 1. Calling the function repeatedly for an unchanged file should return equivalent task slices and should not modify the taskfile."} {"task_id": "format-code-task-000301", "source_id": "format-code-task-000301", "domain": "code", "task_path": "tasks/format-code-task-000301", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:84095ff24ea337f9c8edc96f5498f56fd204216db69007a2e08c2eaca3ebf914", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `pkg/tasker` stateful scheduler to support per-task overlap control when registering callbacks. A caller should be able to use `taskr := tasker.New(tasker.Option{})`, then call `taskr.Task(expr string, task tasker.TaskFunc, concurrent ...bool)` before `taskr.Run()`: when `concurrent` is omitted or `true`, due executions of that registration may overlap, but when `concurrent` is `false`, the scheduler should not start a new invocation for that same registration while its prior invocation is still running.\n\nFor a concrete session, if I register a `\"* * * * * *\"` task whose callback sleeps for about 2500 ms with `taskr.Task(expr, fn, false)`, then run the scheduler for about 3 seconds, that callback should be invoked only once because the next due ticks occur while it is still active. In the same scheduler, a second task registered on the same expression with `taskr.Task(expr, otherFn)` or `taskr.Task(expr, otherFn, true)` should remain independent and should still be allowed to run on each due tick, even while the non-overlapping task is active. Once a non-overlapping callback returns, future due ticks for that registration should be allowed to start it again."} {"task_id": "format-code-task-000302", "source_id": "format-code-task-000302", "domain": "code", "task_path": "tasks/format-code-task-000302", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:41a5126f722cd47096a48f2008a379d7a733ab9acf3b89da867fcad4934e67ef", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a `scoary.methods.Setup_results(genedic: dict, traitsdic: dict, collapse: bool) -> dict` function that computes the population-structure-naive gene/trait association records from already-normalized Scoary dictionaries. The return value should be `{\"Results\": ..., \"Gene_trait_combinations\": ...}` where each trait maps to tested genes, genes absent from every included isolate or present in every included isolate are skipped, and every tested gene records its `NUGN`, `Annotation`, four category counts, sensitivity, specificity, odds ratio, naive Fisher exact p-value, Bonferroni p-value, and Benjamini-Hochberg p-value.\n\nFor example, with one trait `resistant` where `s1` and `s2` are `\"1\"` and `s3` and `s4` are `\"0\"`, a gene `g1` present in `s1`/`s2` and absent in `s3`/`s4` should produce counts `tpgp=2`, `tpgn=0`, `tngp=0`, `tngn=2`, `sens=100.0`, `spes=100.0`, odds ratio `inf`, and naive p-value `0.3333333333333333`. In the same call, a gene `g2` present in `s1`/`s3` and absent in `s2`/`s4` should produce counts `1,1,1,1`, `sens=50.0`, `spes=50.0`, odds ratio `1.0`, and naive p-value `1.0`; an all-present gene in that input should not appear in `Results` or `Gene_trait_combinations`. With those two tested genes, `g1` should have Bonferroni and Benjamini-Hochberg p-values `0.6666666666666666`, while `g2` should have both adjusted values equal to `1.0`.\n\nThe `Gene_trait_combinations` entry should label each isolate for downstream tree-aware work: for `g1` above it should be `s1: \"AB\"`, `s2: \"AB\"`, `s3: \"ab\"`, `s4: \"ab\"`, and for `g2` it should be `s1: \"AB\"`, `s2: \"aB\"`, `s3: \"Ab\"`, `s4: \"ab\"`. When `collapse=True`, genes with identical included-isolate distributions should be merged into one result named with `--`; for two identical genes `gA` and `gB` with the same pattern as `g1`, the result key should be `gA--gB`, the `NUGN` and `Annotation` strings should be merged with `--`, only that merged key should appear for the distribution, and its Bonferroni and Benjamini-Hochberg p-values should be based on one effective test. If a trait references an isolate name that is missing from a gene's isolate entries, the function should raise `SystemExit` with an error telling the caller to make sure strains are named the same in the traits and gene presence/absence files. Calling the function twice with equal input dictionaries should return equal dictionaries, and the input dictionaries should not be mutated."} {"task_id": "format-code-task-000303", "source_id": "format-code-task-000303", "domain": "code", "task_path": "tasks/format-code-task-000303", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0b36240913bbf2623276d6e2811d4db07cb78a285db080758d0a4f4a17218cc6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Scoary to install a `vcf2scoary` command for converting haploid VCF files into the Roary/Scoary-style presence/absence CSV that the main `scoary` command accepts. The command should run as `vcf2scoary [--out PATH] [--types LIST] [--force] `, defaulting the output path to `./mutations_presence_absence.csv`, refusing to overwrite an existing output file unless `--force` is supplied, and exiting non-zero with `Outfile already exists. Change name of outfile or run with --force` or `Unable to locate input file ` for those error cases.\n\nFor a VCF whose metadata includes `##fileformat=VCFv4.1` and `##FORMAT=`, the command should print `VCF version 4.1 detected` while processing and `Reached the end of the file` before exiting 0. It should write a quoted comma-separated CSV whose header is the VCF header with a `DUMMY` column inserted between `FORMAT` and the sample columns. For example, a record `NC_000962 4013 0 T C 9999 0 TYPE=snp GT 0 1 1 1` should become a CSV row with the original first nine fields, `False` in the `DUMMY` column, and only the genotype values `0,1,1,1` in the sample columns even if genotype cells contain colon-separated subfields.\n\nThe `--types` option should accept a comma-separated list such as `--types snp,ins,del`; when it is not `ALL`, only VCF records whose INFO column contains a matching `TYPE=` should be emitted. Multiallelic ALT values should be split into separate binary reference-versus-alternate rows: for `ALT=C,A` with sample genotypes `0,0,1,2`, the output should contain one row for `C` with `DUMMY` set to `True` and sample values `0,0,1,0`, and another row for `A` with sample values `0,0,0,1`; missing genotype `.` should be treated as `0`. If the VCF metadata says the GT format has `Number` other than `1`, the command should exit non-zero with `ERROR: Expected a single allele per genotype. Scoary only works for haploid organisms.`"} {"task_id": "format-code-task-000305", "source_id": "format-code-task-000305", "domain": "code", "task_path": "tasks/format-code-task-000305", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:109c6818371fb8102666ac6d9441244c6dfdea58556973084a5626909fddc600", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nHey, I'm hitting a weird thing with `mockQuery` — if my query params include an array (like `{tags: ['a','b']}`), the mock never matches the request, even though I'm passing what looks like the exact same array on the other side. Plain string/number params work fine, it's specifically the array-valued ones that break. Would be great if `mockQuery` just matched when the arrays have the same items in the same order.\n\n# Expected outcomes\n\n- `mockQuery` should match requests whose query parameters include array values when the expected and actual arrays contain equivalent items in the same order.\n- Array-valued query parameters should remain order-sensitive: the same items in a different order, or arrays of different lengths, should not be treated as equivalent.\n- Equivalent array handling should work when array values appear inside ordinary query parameter objects, including nested values that are compared as part of deciding whether the query parameters match.\n- Existing query parameter comparison behavior for plain string, number, boolean, and object values should be preserved, including not matching values that differ in kind or differ in any supplied object property.\n- Any existing public equivalence helper used by this matching path should reflect the same observable matching semantics.\n\n# Implementation notes\n\n- The specific data structures, helper decomposition, type-checking approach, and validation locations are up to the implementer.\n- Preserve the existing public API shape while adding correct handling for array-valued parameters."} {"task_id": "format-code-task-000306", "source_id": "format-code-task-000306", "domain": "code", "task_path": "tasks/format-code-task-000306", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e91a1bbffa1543e2858beb369c9822ec6139a755af6fab56c6b3c1265de373ec", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Make the Keras wrappers behave like real scikit-learn estimators\n\nThe `KerasClassifier` and `KerasRegressor` wrappers in this package are supposed to let a\ncompiled Keras model be used anywhere a scikit-learn estimator is expected, but right now they\ndon't actually work — they lean on Keras methods that no longer exist and they don't follow\nscikit-learn's estimator conventions. Please bring them up to a fully working, scikit-learn\ncompatible state.\n\nBoth estimators are constructed with a `build_fn` plus any number of extra keyword arguments,\ne.g. `KerasClassifier(build_fn=my_build_fn, hidden_dim=10)`. `build_fn` is a callable that\nreturns a **compiled** Keras model. The following three ways of supplying it must all work:\n\n1. a plain function,\n2. an instance of a class that implements `__call__`,\n3. `None`, in which case the wrapper is being subclassed and the subclass itself implements\n `__call__`.\n\nThe extra keyword arguments are model/fit parameters; any of them whose names match an argument\nof `build_fn` are forwarded to it when the model is built.\n\n## scikit-learn parameter API\n\n* `get_params()` returns a dict containing `build_fn` together with every extra keyword argument\n passed to the constructor, mapped to its current value. It accepts the usual optional `deep`\n argument.\n* `set_params(**params)` updates the given parameters in place and returns the estimator itself;\n a subsequent `get_params()` reflects the new values.\n* `sklearn.base.clone(estimator)` must succeed and return a new, **unfitted** estimator that\n carries the same constructor parameters.\n\n## KerasClassifier\n\n* `fit(X, y)` builds and trains the model and returns the estimator itself. Extra keyword\n arguments (e.g. `epochs`, `batch_size`) are forwarded to the underlying Keras `fit`. After\n fitting, `classes_` holds the sorted unique labels seen in `y` and `n_classes_` holds their\n count. The original label values must be preserved even when they are not a contiguous\n `0..k-1` range.\n* `predict(X)` returns a 1-D array of shape `(n_samples,)` whose entries are drawn from\n `classes_` (i.e. the original label values, not internal indices).\n* `predict_proba(X)` returns an array of shape `(n_samples, n_classes_)` whose rows sum to 1.\n For a binary problem whose network has a single output unit it must still return two columns.\n* `score(X, y)` returns the mean classification accuracy as a float in `[0, 1]`.\n\n## KerasRegressor\n\n* `fit(X, y)` builds and trains the model and returns the estimator itself, forwarding extra\n keyword arguments to the underlying Keras `fit`.\n* `predict(X)` returns a 1-D array of shape `(n_samples,)` for a single-output model.\n* `score(X, y)` returns the coefficient of determination R² (as scikit-learn defines it).\n\n## Serialization\n\nA fitted estimator must survive a `pickle` round-trip: after unpickling it produces the same\npredictions as the original.\n"} {"task_id": "format-code-task-000308", "source_id": "format-code-task-000308", "domain": "code", "task_path": "tasks/format-code-task-000308", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a019aa67ecfd985b52db20ea60b739d775746c9e4d9eae82c3aee79ed2c80bfb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the Go client to support multi-key existence checks from an existing connected client with `func (clnt *Client) BatchExists(policy *BatchPolicy, keys []*Key) ([]bool, Error)`.\n\nWhen I call `exists, err := client.BatchExists(nil, []*Key{keyA, keyB, keyC})` and `keyA` and `keyC` exist while `keyB` does not, I should get `[]bool{true, false, true}` with a nil error. When I call it with two missing keys, I should get `[]bool{false, false}` with a nil error; a missing key is a per-key false result, not a failed batch call. The returned slice must always be in the same positional order as the input keys, and a nil policy must use the client's default batch policy.\n\nThe method should perform the existence check as a batch operation that requests no bin data. It should honor batch read filtering by leaving filtered-out positions false and returning an error chain that includes `types.FILTERED_OUT`. If the server returns an unexpected per-key error, or returns bin data for an existence-only response, the method should return a non-nil `Error` instead of silently marking that key true."} {"task_id": "format-code-task-000309", "source_id": "format-code-task-000309", "domain": "code", "task_path": "tasks/format-code-task-000309", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:329445341db3c020e3ac20752dfa2b874b40f4cb4a8d8403443cd06bdb31977e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the Go client to expose expression operation builders so an expression can be used as an Aerospike `*Operation` in `Client.Operate`, query execute operations, and other operation lists. The public API should include `ExpReadOp(name string, exp *Expression, flags ExpReadFlags) *Operation` and `ExpWriteOp(binName string, exp *Expression, flags ExpWriteFlags) *Operation`.\n\n`ExpReadOp` should create a read operation whose result bin name is `name` and whose payload is the supplied expression plus the supplied read flags. For example, `ExpReadOp(\"sum\", ExpNumAdd(ExpIntBin(\"a\"), ExpIntVal(5)), ExpReadFlagDefault)` should be sendable as an operation and should return the server-evaluated value under result name `sum`. `ExpReadFlagDefault` should be `0`, and `ExpReadFlagEvalNoFail` should be `1 << 4` so unknown or non-bin expression evaluation failures can be ignored.\n\n`ExpWriteOp` should create a write operation whose target bin is `binName` and whose payload is the supplied expression plus the supplied write flags. For example, `ExpWriteOp(\"total\", ExpNumAdd(ExpIntBin(\"a\"), ExpIntBin(\"b\")), ExpWriteFlagDefault)` should be sendable as an operation and should write the server-evaluated expression result into bin `total`. The write flags should include `ExpWriteFlagDefault == 0`, `ExpWriteFlagCreateOnly == 1 << 0`, `ExpWriteFlagUpdateOnly == 1 << 1`, `ExpWriteFlagAllowDelete == 1 << 2`, `ExpWriteFlagPolicyNoFail == 1 << 3`, and `ExpWriteFlagEvalNoFail == 1 << 4`.\n\nBoth builders should serialize the expression operation payload as an array containing the packed expression and the integer flags, then store that payload as the operation value. Calling either builder with the same expression, name, and flags should produce an equivalent `*Operation` and should not touch the filesystem, network, or global state. If the supplied expression cannot be packed, the builder should panic instead of returning a partial operation."} {"task_id": "format-code-task-000310", "source_id": "format-code-task-000310", "domain": "code", "task_path": "tasks/format-code-task-000310", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6710d2b16e9edd286364dc410eb68851d9ec331a518e5741a81912a0d5e2b88f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nInconsistency in numba mode when passing scalar to function\nIn numba mode:\n```python\nimport aesara\naesara.config.mode = \"NUMBA\"\nimport aesara.tensor as at\n\nx = at.scalar(name=\"x\")\nf = aesara.function([x], 2*x)\nprint(repr(f(10))) # prints 20.0\n```\n\nIn c mode:\n\n```python\nimport aesara\nimport aesara.tensor as at\n\nx = at.scalar(name=\"x\")\nf = aesara.function([x], 2*x)\nprint(repr(f(10))) # prints array(20.)\n```\n\nThis came up in https://github.com/pymc-devs/pymc/issues/5937. Found by [bherwerth](https://github.com/bherwerth)."} {"task_id": "format-code-task-000311", "source_id": "format-code-task-000311", "domain": "code", "task_path": "tasks/format-code-task-000311", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ff7d0408231ca8c1b6b3ff28f1a36f62f138eed09ad420452c3317cf690c5df1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAutofix(): Indexing error\nThe autofix function produces the following error: \n> IndexError: boolean index did not match indexed array along dimension 0; dimension is 138 but corresponding boolean dimension is 2\n\nCode to reproduce the error: \n```python\ndef test_autofix():\n x = load_sample()\n s = bct.autofix(bct.binarize(bct.threshold_proportional(x, .41)))\n assert np.allclose(np.sum(s), 7752)\n```\n\nThis is due to the following line: \nhttps://github.com/aestrivex/bctpy/blob/c8cfdeeca7d2437b93754ea3237f3e44b9c22957/bct/utils/other.py#L272\n\nProposed substituition:\n\n```python\nW[np.where(np.isinf(W))] = 0\nW[np.where(np.isnan(W))] = 0\n```"} {"task_id": "format-code-task-000312", "source_id": "format-code-task-000312", "domain": "code", "task_path": "tasks/format-code-task-000312", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:97faf0f0c337e15f1ad1d6c2e1e44f4952ed7ddb648304235a96db57f0ddcd0b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n我给 dynamic preferences 的 Section 配了 verbose_name（比如 General settings），但在 Django admin 的偏好列表、右侧 section 过滤器和编辑页里看到的还是 general，Python 里 str(section) 打出来也是 general。我一开始还以为是 verbose_name 没生效，或者我配置写错了。\n\nExpected outcomes:\n- `Section.__str__()` should use a configured human-readable section name when one is provided.\n- Sections without a configured human-readable name should continue to display their original section name, and an empty section should still stringify to an empty string.\n- In the Django admin preference changelist, each preference’s section display should prefer the human-readable section name and fall back to the original section name when the section cannot be resolved.\n- In the Django admin section filter, filter option labels should prefer the human-readable section name and fall back to the original section name when the section cannot be resolved.\n- In the Django admin preference edit page, the read-only section display should prefer the human-readable section name and fall back to the original section name when the section cannot be resolved.\n\nImplementation notes:\n- Keep existing preference registration, lookup, filtering, and editing behavior intact; this change is only about the externally visible section label shown to users.\n- The exact data flow, helper structure, and admin customization approach are up to the implementation, as long as the observable API and admin behavior above are satisfied."} {"task_id": "format-code-task-000313", "source_id": "format-code-task-000313", "domain": "code", "task_path": "tasks/format-code-task-000313", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c42294e6d0eba6253f6b3e68b173f870d0d2f668df2511441c41df6d607d8d09", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Torus' `Configuration` class to support executable Python configuration files through a stateful API: `c = Configuration()`, `c.load(fname)`, and later `c.reload()`. `load(fname)` should accept a filesystem path string, execute that Python file as a module, remember the path for future reloads, and expose the file's public configuration globals through the existing configuration accessors. For example, if the file defines `SCHEMAS = {'main': {'match': 'app.*', 'intervals': {}}}`, then after `c.load(path)`, `c.schema('main')` should return that loaded schema and `c.schemas('app.hit')` should include it. If the file defines `AGGREGATES = [('app.total', 'app.*')]`, those aggregate rules should be available through `c.aggregates()` and should participate when `c.process(stat, value, timestamp)` expands matching stats.\n\nThe same load call should register `TRANSFORMS` and `MACROS` dictionaries by name, so `c.transform('name')` returns the callable supplied by the config file and `c.macro('name')` returns the macro mapping supplied by the config file. If the file defines `DEBUG`, `c.debug` should reflect that value after loading. Config-defined schema functions should be able to refer to Torus' `long_or_float` helper from the loaded module namespace.\n\nLifecycle hooks should run when present: a callable `on_load` in a config file is called during `load(fname)`, and a callable `on_reload` is called during `reload()`. `reload()` should clear the currently loaded debug value, schemas, aggregate rules, transforms, and macros, then re-execute every path already loaded on that `Configuration` instance. If a config file changes between `load()` and `reload()`, the accessors should reflect the new file contents after reload, and entries that only existed before reload should no longer be present. Calling `Configuration().reload()` before any files are loaded should succeed and leave the configuration empty. Passing a missing or unreadable path to `load(fname)` should raise the underlying `IOError` to the caller."} {"task_id": "format-code-task-000315", "source_id": "format-code-task-000315", "domain": "code", "task_path": "tasks/format-code-task-000315", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e08accdc49aaebe250e710d92cbc973fa99960e4524c81b0b139331ec5f25549", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the scheduler objects to offer a direct run-and-wait convenience API for one-off jobs, in addition to the lower-level queue/result APIs. For asynchronous use, `AsyncScheduler.run_job(func_or_task_id: str | Callable[..., Any], *, args: Iterable[Any] | None = None, kwargs: Mapping[str, Any] | None = None, job_executor: str | UnsetValue = unset, metadata: MetadataType | UnsetValue = unset) -> Any` should be awaitable after the scheduler has been initialized and started. For synchronous use, `Scheduler.run_job(...) -> Any` should expose the same arguments and block through the scheduler's background event loop.\n\nWhen I call `await scheduler.run_job(add, args=[2, 3])`, it should enqueue exactly one job, wait until that job is released, fetch its result, and return `5`. When I call `await scheduler.run_job(greet, kwargs={\"name\": \"Ada\"})`, it should pass the keyword arguments to the task and return that task's return value. The `job_executor` argument should select the named executor for this one job, so `job_executor=\"threadpool\"` or `job_executor=\"processpool\"` runs the callable through that executor instead of the task default.\n\nIf the task raises an exception, `run_job()` should raise that same exception to the caller instead of returning a `JobResult`. If the recorded job outcome is a missed start deadline, it should raise `JobDeadlineMissed`; if the job is cancelled, it should raise `JobCancelled`. If the scheduler has not been initialized yet, calling `run_job()` should fail the same way as the scheduler's other initialized-only methods."} {"task_id": "format-code-task-000316", "source_id": "format-code-task-000316", "domain": "code", "task_path": "tasks/format-code-task-000316", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0da66b0a6a6ee964c90c4ecb11e4d292407e2c77204dcceb975fda09b3ce77cf", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the Go LINQ `Query` type to support lazy ordering and sorting methods. The public API should include `func (q Query) OrderBy(selector func(any) any) OrderedQuery`, `func (q Query) OrderByT(selectorFn any) OrderedQuery`, `func (q Query) OrderByDescending(selector func(any) any) OrderedQuery`, `func (q Query) OrderByDescendingT(selectorFn any) OrderedQuery`, `func (oq OrderedQuery) ThenBy(selector func(any) any) OrderedQuery`, `func (oq OrderedQuery) ThenByT(selectorFn any) OrderedQuery`, `func (oq OrderedQuery) ThenByDescending(selector func(any) any) OrderedQuery`, `func (oq OrderedQuery) ThenByDescendingT(selectorFn any) OrderedQuery`, `func (q Query) Sort(less func(i, j any) bool) Query`, and `func (q Query) SortT(lessFn any) Query`.\n\nFor `From([]int{3, 1, 2}).OrderBy(func(v any) any { return v.(int) })`, iterating the returned query should yield `1, 2, 3`. For `From([]string{\"aa\", \"b\", \"cc\"}).OrderByDescending(func(v any) any { return len(v.(string)) })`, iterating should yield the length-2 strings before `\"b\"`. Chained ordering should support multiple keys: given records with fields `{group, name}`, `OrderBy(group).ThenByDescending(name)` should sort by `group` ascending and, within equal groups, by `name` descending. `Sort` should let callers provide the direct less-than comparison; for `From([]int{3, 1, 2}).Sort(func(i, j any) bool { return i.(int) < j.(int) })`, iteration should yield `1, 2, 3`.\n\nThe typed variants should accept strongly typed functions through `any` and behave the same as the untyped variants, such as `OrderByT(func(v int) int { return v })` and `SortT(func(i, j int) bool { return i < j })`. Invalid typed function signatures should panic with the existing signature-validation errors, for example `OrderByT(func(i, j int) int { return i })` should panic with `OrderByT: parameter [selectorFn] has a invalid function signature. Expected: 'func(T)T', actual: 'func(int,int)int'`, and `SortT(func(i, j int) string { return \"\" })` should panic with `SortT: parameter [lessFn] has a invalid function signature. Expected: 'func(T,T)bool', actual: 'func(int,int)string'`.\n\nThese methods should be referentially transparent query transformations: creating the ordered query should not mutate the input slice or consume the source immediately, and iterating the same returned query with the same source values should produce the same ordered results each time. Empty inputs should iterate as empty results, and iteration should respect an early `yield` return of `false` without requiring callers to consume the entire sorted output."} {"task_id": "format-code-task-000317", "source_id": "format-code-task-000317", "domain": "code", "task_path": "tasks/format-code-task-000317", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9ccc1d702d444688df9ea4c5ee773c8e2fef58381ec3a3905558965a87ca6997", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n我想在 v2 的 `.ahoy.yml` 里给命令写更详细的 `description`，不然现在 `ahoy --help` / `ahoy --help` 只看得到 usage，长说明写了也像没生效一样。顺手也想把 v2 这些开发命令整理一下：build 相关走 goreleaser，加个 release 命令，旧的 godep 别名就不用留了。\n\nExpected outcomes:\n- Command descriptions: a command-level `description:` in a v2 `.ahoy.yml` command is treated as user-visible help text, including descriptions written as YAML multiline block strings.\n- Per-command help: `ahoy --help` shows the command’s `description:` content when it is configured, while commands without a description continue to work normally.\n- Global help: `ahoy --help` shows configured descriptions in the `COMMANDS:` section in addition to each command’s `usage:` summary; multiline descriptions remain readable as multiple lines, and commands without descriptions still show only their normal summary.\n- v2 development commands: in the repository’s v2 `.ahoy.yml`, `build` uses `goreleaser build --config ../.goreleaser.yml --snapshot --clean --single-target`, and `build-all` uses `goreleaser build --config ../.goreleaser.yml --snapshot --clean`.\n- v2 release command: the repository’s v2 `.ahoy.yml` defines a `release` command that runs `goreleaser release --config ../.goreleaser.yml --clean`.\n- v2 dependency command cleanup: the repository’s v2 `.ahoy.yml` no longer defines the old `godep` alias command.\n\nImplementation notes:\n- Preserve existing v2 command behavior except where the configured help text or the listed development commands are intentionally changed.\n- The data flow, formatting mechanism, and validation location for command descriptions are implementation choices, as long as the observable help output and command configuration behavior match the outcomes above.\n- Keep the help output readable for both single-line and multiline descriptions without requiring callers to use a different command-line entry point."} {"task_id": "format-code-task-000318", "source_id": "format-code-task-000318", "domain": "code", "task_path": "tasks/format-code-task-000318", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:dd9153643ff6ec85317118dbc1bf42d0618c0eb8c35b8d486ccd9558835c8a5e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Gnomix's command-line inference run to support an optional BED segment export controlled by the YAML config key `inference.bed_file_output`. When I run `python3 gnomix.py ` with `bed_file_output: True` in `config.yaml`, or run the training-mode command with a config file that sets that key to true, inference should use the generated `query_results.msp` file to create `/query_results_bed` if that directory does not already exist.\n\nFor each haplotype column in the MSP output, the command should write one tab-separated BED-style file under `query_results_bed`, naming files by replacing dots in the haplotype name with underscores and appending `.bed` (for example `NA001.0` becomes `NA001_0.bed`). Each file should have the columns `chm`, `spos`, `epos`, `ancestry`, `sgpos`, and `egpos` and no row index column. Consecutive MSP windows with the same ancestry call for that haplotype should be collapsed into a single segment: the segment starts at the first window's physical and genetic start positions, ends at the last consecutive window's physical and genetic end positions, and a new row begins only when the ancestry call changes. The `ancestry` column should contain the reference population names in the model's population order, not just numeric labels. On a successful BED export, the command should exit with status 0; if the BED output directory cannot be created or a BED file cannot be written, the command should fail non-zero instead of silently skipping those files."} {"task_id": "format-code-task-000319", "source_id": "format-code-task-000319", "domain": "code", "task_path": "tasks/format-code-task-000319", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:622fcabb343b07455c1325fe12a94448ebe348bf85a53c32d230ca256c325c1b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Allow customizing the `.webp` filename rule\n\nMost WebP converters I've used (the ones that batch-convert `.jpg` / `.png` to `.webp`) keep the original extension and just append `.webp`. So `logo.png` becomes `logo.png.webp` on disk, not `logo.webp`. This is actually the default in a lot of tools.\n\nWith the current plugin I can't make the generated CSS line up with that. For example I have:\n\n```css\n.logo {\n background: url(/logo.png);\n}\n```\n\nand my actual file on disk is `/logo.png.webp`. But the plugin always rewrites the url to `/logo.webp`, which 404s.\n\nLooking at the existing options (`modules`, `webpClass`, `noWebpClass`) there's nothing that controls how the filename is rewritten — the `.png` → `.webp` substitution is hardcoded.\n\nCould we get an option to control how the new filename is derived from the old one? That way users whose webp files sit next to the originals as `name.png.webp` (or any other naming convention) can configure it without forking the plugin.\n\nI'd expect the new option to be something like `rename`.\n\nThanks!"} {"task_id": "format-code-task-000320", "source_id": "format-code-task-000320", "domain": "code", "task_path": "tasks/format-code-task-000320", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:448f013ac8c4c7fbaa2b7c877fd058750e7307cb06f51cb261ed09c47eadafc8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `Ants.get_replay(self) -> dict` available on an initialized or finished `Ants` game object. It should take no arguments and return a JSON-serializable revision 3 replay summary of the game state at the moment it is called.\n\nThe top-level dictionary should include `revision: 3`, `players`, timing settings, radius settings, engine and player seeds, food settings, `map`, `food`, `ants`, `hills`, `scores`, `bonus`, `hive_history`, `winning_turn`, `ranking_turn`, and `cutoff`. The `map` value should contain `rows`, `cols`, and `data`, where `data` is the rendered replay map; for scenario games, the replay map should reflect the original map layout.\n\nEach food entry should be `[row, col, start_turn, end_turn]`, using the current turn plus one when that food is still present, and should append the gathering owner when the food was gathered. Each ant entry should be `[initial_row, initial_col, spawn_turn, end_turn, owner, orders]`, using the current turn plus one for living ants, the death turn for killed ants, and a single concatenated string for the ant's recorded orders. Each hill entry should be `[row, col, owner, end_turn]`, using the current turn plus one for active hills and the hill's end turn for razed hills.\n\nCalling `get_replay()` repeatedly without advancing the game should return equal dictionaries, and it should not advance turns, alter scores, change ant/food/hill state, or write files."} {"task_id": "format-code-task-000322", "source_id": "format-code-task-000322", "domain": "code", "task_path": "tasks/format-code-task-000322", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1c7df96d51e375e20b089bb02a1ec226c9cb933ede7698d886a3ccfe1d61e3f8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### `FileResponse` doesn't implement Range requests properly\n\nI'm serving large files (video) through `web.FileResponse` and a browser/player that does Range requests is misbehaving against the aiohttp server. Digging into what aiohttp actually returns vs. what other static file servers (nginx, apache) return, several things look wrong:\n\n**1. No `Content-Range` on 206 responses**\n\nWhen a client sends e.g. `Range: bytes=0-1023`, aiohttp does respond with `206 Partial Content` and the right body slice, but the response is missing the `Content-Range: bytes 0-1023/` header. RFC 7233 requires it on 206 responses, and some clients refuse to use the partial response without it (they fall back to redownloading from scratch, or just break).\n\n**2. No `Accept-Ranges` advertised**\n\nEven on a normal 200 response for a static file, aiohttp doesn't tell the client that range requests are supported. Clients that probe with a HEAD first (e.g. download managers, video players doing seek) don't know they can issue Range requests.\n\n**3. Out-of-range start range doesn't return 416**\n\nIf the file is 200 bytes and the client sends `Range: bytes=99999-`, aiohttp currently returns the request as if it were satisfiable. Per RFC 7233 the server should return `416 Range Not Satisfiable` here.\n\n**4. Out-of-range tail range breaks the response**\n\nIf the file is 200 bytes and the client sends `Range: bytes=-99999` (give me the last 99999 bytes), the request blows up instead of just returning the whole file (which is what nginx etc. do — clamp to start of file).\n\nMinimal repro for #4:\n\n```python\nfrom aiohttp import web\n\nasync def handler(request):\n return web.FileResponse('./small_file.txt') # ~200 bytes\n\napp = web.Application()\napp.router.add_get('/', handler)\nweb.run_app(app)\n```\n\n```\n$ curl -H 'Range: bytes=-99999' -i http://localhost:8080/\n```\n\n…doesn't give back the file the way I'd expect.\n\n---\n\nWhile we're talking about conditional/range stuff, it would also be nice if `FileResponse` honored `If-Unmodified-Since` and `If-Range` — right now only `If-Modified-Since` is checked, so a client using `If-Range` to revalidate before resuming a download can't actually do conditional ranged GETs against an aiohttp-served file.\n\nCould `FileResponse` be made RFC 7233-compliant for Range requests?"} {"task_id": "format-code-task-000323", "source_id": "format-code-task-000323", "domain": "code", "task_path": "tasks/format-code-task-000323", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1e04d13573d9547702bc8516a42689d840a70e32918202cbad085c415ac8f652", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\naiohttp 3.8.1\n\n`await websocket.receive_json()` Will die with `TypeError` on `WSMsgType.CLOSED`.\n\n#2784 is fixed, yes:\n\n```python\n async def __anext__(self) -> WSMessage:\n msg = await self.receive()\n if msg.type in (WSMsgType.CLOSE, WSMsgType.CLOSING, WSMsgType.CLOSED):\n raise StopAsyncIteration\n return msg\n```\n\nYes, bug is fixed in iterator but NOT fixed for `websocket.receive_*` functions, like `receive_json()`\n\nI would raise RuntimeError or so if corresponding message \"type\" (like bytes, str, or json) can not be received.\n\nA dedicated public exception class (something like `WSMessageTypeError`) exported from `aiohttp` would be ideal so users can catch it specifically."} {"task_id": "format-code-task-000324", "source_id": "format-code-task-000324", "domain": "code", "task_path": "tasks/format-code-task-000324", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e27557b2426b19000d7bff62982c1dd396d0ae60d8d2743b8fccdf64786dd727", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want aiohttp_session to provide a plain JSON cookie storage class named SimpleCookieStorage for simple and test aiohttp.web apps. It should be constructible as SimpleCookieStorage(*, cookie_name: str = \"AIOHTTP_SESSION\", domain: str | None = None, max_age: int | None = None, path: str = \"/\", secure: bool | None = None, httponly: bool = True, samesite: str | None = None, encoder: Callable[[object], str] = json.dumps, decoder: Callable[[str], Any] = json.loads), and it should be usable with aiohttp_session.setup(app, SimpleCookieStorage()) or session_middleware(SimpleCookieStorage()).\n\nWhen load_session(request) runs and the configured cookie is absent, it should return a new empty Session with identity None and the storage max_age. When the configured cookie contains JSON like {\"created\": 1000, \"session\": {\"a\": 1, \"b\": 2}}, load_session(request) should decode it and return a non-new Session whose mapping contains a=1 and b=2. If the default decoder receives malformed JSON from the cookie, the JSON decoding exception should propagate instead of silently creating another session.\n\nWhen save_session(request, response, session) runs for a changed session containing values such as {\"a\": 1, \"c\": 3}, it should serialize the session data as JSON into the configured cookie on the response, including the session's created timestamp. If the session is empty or invalidated, saving should write an empty JSON object payload ({}) through the same cookie-saving path. Custom encoder and decoder callables passed to the constructor should be used for saving and loading instead of the JSON defaults."} {"task_id": "format-code-task-000325", "source_id": "format-code-task-000325", "domain": "code", "task_path": "tasks/format-code-task-000325", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d09d444b7e9d1095d4e7a5b974c4c465cca1f5d3c6b95a61c2184c754fe4de04", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Add an opt-in retry middleware for Bot API requests\n\nIndividual Bot API calls currently have no reusable retry policy, even though aiogram already has a request-middleware pipeline and backoff utilities. Add a public `RetryRequestMiddleware` importable from `aiogram.client.session.middlewares.retry` so applications can register retry behavior on a session without changing `Bot` calls or a custom session implementation.\n\nThe middleware constructor must accept `max_attempts` (default `3`), an optional `BackoffConfig`, and an optional asynchronous `sleep` callable. Omitting the backoff configuration must still produce a usable middleware; a supplied configuration controls the delay sequence. The supplied sleep callable receives the chosen delay in seconds and is awaited. `max_attempts` is the total number of downstream calls, including the initial immediate call, and values less than one must be rejected with `ValueError`.\n\nA successful downstream result must be returned unchanged after one call and without sleeping. Retry only failures represented by `TelegramNetworkError`, `TelegramServerError` (including its subclasses), or `TelegramRetryAfter`. `TelegramEntityTooLarge` is permanent despite inheriting from `TelegramNetworkError` and must not be retried. Every other exception, including other Telegram API errors and application exceptions, must propagate unchanged after the first call.\n\nBetween retryable failures, use successive delays from a fresh `Backoff` based on the configured `BackoffConfig`. For `TelegramRetryAfter`, wait for the greater of that backoff delay and the exception's `retry_after` value. Never sleep after the last permitted attempt. If all attempts fail, re-raise the exact exception from the last attempt. Every attempt must invoke the downstream callable with the same `bot` and `method` objects received by the middleware.\n\nOne middleware instance may be reused for sequential and concurrent requests. Each invocation must start its own backoff sequence, and concurrent invocations must not share counters, delays, results, or errors. Cancellation from either the downstream request or the sleep callable must propagate immediately, with no further attempt.\n\nPreserve normal `RequestMiddlewareManager` composition: retrying re-enters middleware registered downstream of the retry middleware for every attempt, while middleware registered upstream surrounds the whole retry operation and runs once."} {"task_id": "format-code-task-000326", "source_id": "format-code-task-000326", "domain": "code", "task_path": "tasks/format-code-task-000326", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e2ad26277e52e54164755f10153032239f5f0fa4adfa8ff5a21a179a8335f23a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Support formatted poll questions and answer options\n\nTelegram now lets bots send polls whose **question** and **answer options** carry text\nformatting (bold, italic, custom emoji, …). Right now our `send_poll` support is stuck on the\nold shape: answer options are plain strings only, and there's no way to format the question.\nLet's bring the poll API up to date.\n\n## What we need\n\n**A new input type for answer options.** Introduce a public type that represents a single answer\noption to be sent (not the one returned in poll results). It must be importable from the\ntop-level types package and carry:\n\n- `text` — the option text (required).\n- `text_parse_mode` — optional parse mode for the option text.\n- `text_entities` — an optional list of message entities for the option text. When constructed\n from raw data (e.g. a list of dicts), these must be parsed into the usual message-entity\n objects, like every other entities field in the library.\n\nOnly the fields actually provided should appear when the object is serialized (no spurious\n`null`s for the optional fields).\n\n**`send_poll` should accept formatted options.** The `options` argument must accept a list whose\nitems are *either* plain strings *or* the new input-option objects, freely mixed in the same\nlist. Plain strings must stay plain strings, and option objects must be preserved as-is — passing\nan option object where only a string used to be allowed must now work.\n\n**`send_poll` should accept a formatted question.** Add two new parameters:\n\n- `question_parse_mode` — parse mode for the poll question. Like the explanation parse mode, it\n should fall back to the bot's default parse mode when not given explicitly.\n- `question_entities` — an optional list of message entities for the question, usable instead of\n `question_parse_mode`. These must be parsed into message-entity objects.\n\nBoth must be real parameters of the method (and of the corresponding `Bot.send_poll` helper), not\njust arbitrary extra keyword arguments.\n\n**Returned poll data should expose the new entities.** The poll object returned by Telegram now\nincludes formatting entities for the question, and each answer option in poll results now includes\nformatting entities for its text. Add the corresponding optional fields so that:\n\n- the poll object gains a `question_entities` list, and\n- the answer-option result object gains a `text_entities` list,\n\nboth parsed into message-entity objects when present.\n\nExisting call sites that pass `options=[\"A\", \"B\"]` and never touch the new fields must keep working\nexactly as before.\n"} {"task_id": "format-code-task-000327", "source_id": "format-code-task-000327", "domain": "code", "task_path": "tasks/format-code-task-000327", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:48ed0d504d6f9d5dcc80442e8eec9b88aa1ed0a5574f10daff04fd7aa241bc94", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI'm running Airbyte's HubSpot source on an account with a lot of custom contact/deal properties, and those streams keep failing with HubSpot HTTP 414 / `Request-URI Too Long`. Smaller HubSpot objects sync fine, so it looks like the connector is building a request URL that's too large when it includes all the properties.\n\n# Expected outcomes\n\n- HubSpot streams that need to request very large sets of object properties should split those properties across multiple requests conservatively enough to avoid a single request URL becoming too long.\n- Property requests should remain safe for realistic HubSpot property names, including names that are not simple alphanumeric identifiers.\n- Small property sets should continue to be requested normally without unnecessary behavior changes, and all requested properties should still be covered without dropping or duplicating them.\n- Airbyte's HubSpot source definition/spec metadata should point at the fixed connector version: `dockerImageTag: 0.1.69`, `dockerImage: \"airbyte/source-hubspot:0.1.69\"`, and `LABEL io.airbyte.version=0.1.69`.\n\n# Implementation notes\n\nThe exact internal structure, helper functions, and location of the URL-length checks are up to the implementer. The important behavior is that HubSpot property requests remain complete while avoiding overlong request URLs across both large and realistic property sets."} {"task_id": "format-code-task-000328", "source_id": "format-code-task-000328", "domain": "code", "task_path": "tasks/format-code-task-000328", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f25580986fccd86508ecc5129f3a4c7b10e7f1098f89d9f9dfb05934804e7873", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a Zenefits source connector\n\nWe want a new Airbyte source connector for [Zenefits](https://developers.zenefits.com/), the HR /\npeople-management platform, so users can replicate their Zenefits data into Airbyte. Please build it\nas a Python connector using the repository's connector framework (the Airbyte CDK), following the\nsame conventions the other Python sources in this repo use.\n\nThe connector should be importable as the `source_zenefits` package, and its CDK `AbstractSource`\nimplementation should be the class `SourceZenefits` (reachable as `source_zenefits.source.SourceZenefits`).\n\n## Configuration\n\nThe connector takes a single required config field, `token` — the Bearer token a user generates in\nthe Zenefits portal. Every request the connector makes to the Zenefits API must be authenticated by\nsending this token in an `Authorization: Bearer ` HTTP header.\n\n## API shape\n\nAll endpoints live under the base URL `https://api.zenefits.com/`. Requests are plain HTTP `GET`s.\n\nA successful list response is an envelope of the form:\n\n```json\n{\n \"data\": {\n \"data\": [ { ...record... }, { ...record... } ],\n \"next_url\": \"https://api.zenefits.com/core/people?starting_after=\"\n }\n}\n```\n\n- The actual records to emit live under `data.data`.\n- `data.next_url` drives pagination: when it is a non-empty URL there is another page to fetch and\n the connector must follow it to retrieve the remaining records; when it is `null`/absent the\n stream is complete. All records across all pages must be emitted, in order.\n\n## Streams\n\nExpose exactly the following 11 full-refresh streams. Each stream's name must be exactly as listed,\nand each must read from the given path relative to the base URL:\n\n| stream name | path |\n|-----------------------|-----------------------------------|\n| `people` | `core/people` |\n| `employments` | `core/employments` |\n| `departments` | `core/departments` |\n| `locations` | `core/locations` |\n| `labor_groups` | `core/labor_groups` |\n| `labor_group_types` | `core/labor_group_types` |\n| `custom_fields` | `core/custom_fields` |\n| `custom_field_values` | `core/custom_field_values` |\n| `vacation_requests` | `time_off/vacation_requests` |\n| `vacation_types` | `time_off/vacation_types` |\n| `time_durations` | `time_attendance/time_durations` |\n\n`SourceZenefits.streams(config)` must return one stream instance per row above (11 in total).\n\n## Connection check\n\n`SourceZenefits.check_connection(logger, config)` validates the supplied token by making a request\nto the Zenefits API. It returns a `(bool, error)` tuple: `(True, None)` when the API responds\nsuccessfully, and `(False, )` (a falsy/`None` second element only on success — otherwise a\ntruthy error object describing the failure) when the request fails, e.g. the token is rejected with\nan HTTP error status. It must not raise in that failure case.\n"} {"task_id": "format-code-task-000329", "source_id": "format-code-task-000329", "domain": "code", "task_path": "tasks/format-code-task-000329", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f398bde651973ec7f533e96b614e59f36db3bba14432f084950aad97ab70ef39", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\nI'm running into a sync failure with the S3 source when the file format is `jsonl`: rows that include a nested object field like `{\"meta\": {\"foo\": \"bar\"}}` fail before any Airbyte records are produced, while similar JSONL files with only flat scalar fields sync fine. I first thought my S3 config was wrong, but it seems tied to nested objects in the JSONL payload.\n\n# Expected outcomes\n- JSONL files that contain nested object fields should sync successfully instead of failing before any records are produced.\n- Flat scalar JSONL files should continue to sync normally, and nested-object JSONL should still emit records for each input row.\n- The S3 source's published version metadata should be updated consistently to `0.1.20` in its public definition/spec and built image metadata.\n\n# Implementation notes\n具体的解析策略、内部数据结构、以及校验放置位置由实现者自行决定；只要外部可见的同步结果、记录产出、以及公开版本元数据满足上述行为即可。"} {"task_id": "format-code-task-000330", "source_id": "format-code-task-000330", "domain": "code", "task_path": "tasks/format-code-task-000330", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:eb812d1da0cb159098546ff7ecdc7912ae4e0898baff4ff4d632246ca2b4ca71", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Marketo incremental sync fails when records contain null cursor values\n\nI'm using the Marketo source connector to do incremental syncs of a few streams (programs, activities, etc.). For most of the records this works fine, but the sync blows up partway through when it hits records where the cursor field (e.g. `createdAt` / `updatedAt`) comes back as `null` from the Marketo API.\n\nIt seems Marketo can legitimately return records with a null timestamp for the cursor field — not a missing key, but explicitly `null` — and the connector doesn't cope with that. The sync errors out instead of just moving past those records, so I can't get a clean incremental sync to complete on streams where any record happens to have a null cursor value.\n\nI'd expect the connector to tolerate this: if an individual record doesn't have a usable cursor value, the state advancement logic shouldn't crash the whole sync. Falling back to something sensible (e.g. the configured start date) for the purpose of state tracking would be fine — the important thing is the sync should keep running and finish, the way it does for records that do have a cursor value.\n\nReproducing is just \"run an incremental sync against a Marketo account that has any record with a null `createdAt`/`updatedAt` in one of the incremental streams\" — it consistently fails for me."} {"task_id": "format-code-task-000331", "source_id": "format-code-task-000331", "domain": "code", "task_path": "tasks/format-code-task-000331", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5b7a2764dd41b393a8fee385edce9796b091939a60356b6d57f197851c6c03e8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## No way to disable response gzip compression in `HttpServer`\n\nLooking at `HttpServer`, when the servlet context is built it always installs a `GzipHandler` for response compression:\n\n```java\n// -- gzip handler\ncontext.insertHandler(new GzipHandler());\n```\n\nThere's no corresponding switch on `HttpServerConfig` — every service that uses airlift's http-server gets gzipped responses whether it wants them or not.\n\nThis is a problem in a few setups we have:\n\n- Services sit behind a reverse proxy / CDN that already handles content negotiation and compression. Doing gzip a second time at the origin is just wasted CPU.\n- Some downstream consumers want to look at raw response bodies (debugging, capturing traffic, simple clients that don't speak `Accept-Encoding`) and the automatic gzip layer gets in the way.\n- For some endpoints the payloads are already compressed (images, pre-gzipped blobs) and re-running them through `GzipHandler` is pointless.\n\nI'd like to be able to turn the response compression off via configuration, the same way other server features are toggled in `HttpServerConfig` (e.g. the `http-server.log.compression.enabled` flag for request log compression). The default should remain \"compression on\" so existing deployments don't change behavior — this is purely an opt-out for the cases above.\n\nThe new property would be something like `http-server.compression.enabled` on `HttpServerConfig`."} {"task_id": "format-code-task-000332", "source_id": "format-code-task-000332", "domain": "code", "task_path": "tasks/format-code-task-000332", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:bc95ee4920fa1ff9a3c1fb895837bbd4708c18f5c15f9e07d1c5f79d5934d4c6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want PeekingDuck's dabble pipeline nodes to support zone occupancy counting from object detections through two node objects: `dabble.bbox_to_btm_midpoint.Node(config: Dict[str, Any] = None, **kwargs: Any)` with `run(inputs: Dict[str, Any]) -> Dict[str, Any]`, and `dabble.zone_count.Node(config: Dict[str, Any] = None, **kwargs: Any)` with `run(inputs: Dict[str, Any]) -> Dict[str, Any]`.\n\nThe `bbox_to_btm_midpoint` node should read normalized bounding boxes from `inputs[\"bboxes\"]` and the image array from `inputs[\"img\"]`, then return `{\"btm_midpoint\": [...]}` where each point is the integer bottom midpoint in image pixels. For an image with shape `(400, 600, 3)` and one bbox `[0.1, 0.2, 0.3, 0.4]`, `run()` should return `{\"btm_midpoint\": [(120, 160)]}`. For the same image with no boxes, it should return `{\"btm_midpoint\": []}`, and it should not mutate the input image or bbox list.\n\nThe `zone_count` node should accept a config containing `input`, `output`, `resolution`, and `zones`, prepare the configured polygons at construction time, and count incoming `inputs[\"btm_midpoint\"]` points on each call to `run()`. Zone coordinates may be all non-negative integer pixel coordinates, or all fractions between 0 and 1 that are resolved against `resolution`; for `resolution: [1280, 720]`, the fractional zone `[[0.5, 0], [1, 0], [1, 1], [0.5, 1]]` should resolve to `[(640, 0), (1280, 0), (1280, 720), (640, 720)]` in the returned `\"zones\"` list. With zones `[[[0, 0], [640, 0], [640, 720], [0, 720]], [[0.5, 0], [1, 0], [1, 1], [0.5, 1]]]` and points `[(2, 2), (3, 3), (720, 700), (650, 50)]`, `run()` should return `\"zone_count\": [2, 2]`; with no points it should return counts `[0, 0]` and the same resolved zones. If a single configured zone mixes pixel coordinates and fractional coordinates, constructing the node should raise `ValueError` explaining that the zone must be all pixel-wise points or all fractions of the frame between 0 and 1."} {"task_id": "format-code-task-000333", "source_id": "format-code-task-000333", "domain": "code", "task_path": "tasks/format-code-task-000333", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6bbc15db1bbcae636d901e6d484dca8b4f315bd6ef0ab24bdd38369849533ea8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n我现在用 `darker` 的时候想临时按项目外的 Black 设置来跑，比如这次要换个 line length，或者保留字符串引号别被规范化，但好像只能靠 Black 配置文件控制。能不能让 `darker` 命令行里也能直接传这些 Black 格式化选项，而且我手动传的时候就按我这次传的来？\n\nExpected outcomes:\n- CLI line length: `darker` accepts `-l LINE_LENGTH` and `--line-length LINE_LENGTH`, and formatting uses the provided maximum line length for the current run.\n- CLI string normalization: `darker` accepts `-S` and `--skip-string-normalization`, and formatting preserves string quotes and prefixes instead of normalizing them for the current run.\n- CLI precedence over config: when `-c/--config` is used together with an explicit CLI line-length or skip-string-normalization option, the explicit CLI value takes precedence over the corresponding Black config setting for that run.\n- Config compatibility: when a Black configuration read through `-c/--config` requests skipped string normalization, `darker` formats consistently with that configuration.\n- User-facing help and documentation: `darker --help` and the README document the supported Black-related command-line options, including `-l/--line-length` and `-S/--skip-string-normalization`.\n\nImplementation notes:\n- Keep the behavior compatible with existing `darker` workflows, including formatting only the relevant changed portions.\n- The internal representation of option values, where configuration is merged, and how formatting options are passed through are implementation details.\n- Preserve existing command-line behavior for users who do not pass the new options."} {"task_id": "format-code-task-000334", "source_id": "format-code-task-000334", "domain": "code", "task_path": "tasks/format-code-task-000334", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f4697c5861852f80910a01d27bb2503078f1d4bc51be451a763a13e5394d15be", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\npypa/pip#10909 is so far the \"highest profile\" discussion I've found on this topic.\n\nIt seems that [pip](https://github.com/pypa/pip) and [colorama](https://github.com/tartley/colorama) are going to support only `FORCE_COLOR=` and `NO_COLOR=`, but not `PY_COLORS`.\n\nOn the other hand, Pytest does support `PY_COLORS={1|0}` and `NO_COLOR=` (with `PY_COLORS` taking precedence).\n\nSo in the end it may be best to do the same as Pytest:\n- `PY_COLORS={1|0}` (taking precedence over other options)\n- `NO_COLOR=`\n- and possibly `FORCE_COLOR=` (similar to [pip](https://github.com/pypa/pip) and [colorama](https://github.com/tartley/colorama))\n\nAs for the `true`/`false` discussion, I'm going to make the call to only support `PY_COLORS={0|1}` for now. Thanks @MatthijsBurgh for initiating the discussion and motivating me to look further into the topic!\n\n_Originally posted by @akaihola in https://github.com/akaihola/darker/pull/353#discussion_r846752942_"} {"task_id": "format-code-task-000335", "source_id": "format-code-task-000335", "domain": "code", "task_path": "tasks/format-code-task-000335", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:04130077d0645d11223a3bac0c12d9918abd6ae39c9c8eecaa5fe0cedefda9ad", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Make the filter for the homepage graph configurable\n\nThe big graph at the top of the console home page (the one showing Gbps over time) only shows flows where the input interface boundary is `external`. That criterion is baked into the query and I can't find any way to change it from the configuration.\n\nThis is awkward in a couple of deployments I'm running:\n\n- One instance is on a network where I haven't bothered classifying interfaces as external/internal — everything stays `undefined`. The homepage graph is just flat at zero even though the per-source-AS / per-country widgets right below it clearly show traffic going through. New users land on the page and assume Akvorado isn't actually receiving anything.\n- Another instance is used purely for internal observability inside a single AS. There's no \"edge\" in any meaningful sense and I'd much rather have that graph show the sum of *all* captured flows than nothing at all.\n- And in general, \"edge ingress\" isn't always the most useful default — depending on the deployment I might want to filter on a different boundary, a specific exporter group, etc.\n\nIt would be great if this could be exposed in the console configuration alongside the existing keys like `homepage-top-widgets`, `dimensions-limit` and `cache-ttl`, so each operator can pick whatever criterion makes sense for their setup — including no extra filter at all (i.e. just sum everything). The current behaviour should remain the default so existing deployments don't suddenly start showing different numbers.\n\nA natural name for the new console config key would be something like `homepage-graph-filter`."} {"task_id": "format-code-task-000336", "source_id": "format-code-task-000336", "domain": "code", "task_path": "tasks/format-code-task-000336", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0a8257400bf99379bcd332cfee94115a6948b5e4ed94c489e3f0088208754cc5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Front seat \"Auto\" heating triggers the wrong seat\n\nI'm using the tesla_custom integration in Home Assistant. My car has the auto seat climate feature on the front seats, so the heated seat selects for `left` and `right` show an `Auto` option in addition to Off/Low/Medium/High.\n\nWhen I pick any of Off/Low/Medium/High the seat heater behaves correctly — the left select controls the driver's seat, the right select controls the passenger's seat.\n\nBut as soon as I pick `Auto` from either of the front seat selects, the wrong seat reacts (or nothing visible happens on the seat I actually selected). Switching back to Low/Medium/High on that same select then works fine again, so the select itself is wired up to the right seat — it's specifically the `Auto` path that ends up targeting a different seat than the one I clicked.\n\nSame thing happens in reverse: if `Auto` is currently active and I switch the select to one of the manual levels, the integration first tries to turn auto climate off, and that \"off\" call also seems to land on the wrong seat.\n\nIt looks like the regular seat heater command and the auto seat climate command don't agree on which seat index means \"driver\" vs \"passenger\", and the integration is feeding the same index to both. Could the Auto branch be fixed so that picking Auto on the left select actually turns on auto climate for the driver's seat (and right → passenger)?"} {"task_id": "format-code-task-000337", "source_id": "format-code-task-000337", "domain": "code", "task_path": "tasks/format-code-task-000337", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:19add4a7d8aa8b749e02f7b1a7adfdf3ad062efe8992879aab9d0afa14c45cb8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[BUG] Resize with 2D masks crashes\nAdditional target in Resize fails.\n\n```python\nimport numpy as np\nimport albumentations as A\n\nimage = np.zeros((256, 256, 3), dtype=np.uint8)\nmask = np.ones((256, 256), dtype=np.uint8) # 2D mask\n\ntransform = A.Compose(\n [\n A.Resize(height=512, width=512), # Resize image + mask\n ],\n additional_targets={\"semantic_mask\": \"mask\"},\n)\n\naugmented = transform(image=image, semantic_mask=mask)\nprint(augmented[\"semantic_mask\"].shape)\n```\n\ngenerates error:\n```\n File \"C:\\Users\\aselimc\\micromamba\\envs\\dinov3\\Lib\\site-packages\\albumentations\\augmentations\\geometric\\resize.py\", line 833, in apply_to_mask\n return fgeometric.resize(mask, (self.height, self.width), interpolation=interpolation)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\aselimc\\micromamba\\envs\\dinov3\\Lib\\site-packages\\albucore\\decorators.py\", line 42, in wrapped_function\n result = func(img, *args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\aselimc\\micromamba\\envs\\dinov3\\Lib\\site-packages\\albumentations\\augmentations\\geometric\\functional.py\", line 261, in resize\n return resize_cv2(img, target_shape, interpolation)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\aselimc\\micromamba\\envs\\dinov3\\Lib\\site-packages\\albumentations\\augmentations\\geometric\\functional.py\", line 352, in resize_cv2\n return resize_fn(img)\n ^^^^^^^^^^^^^^\n File \"C:\\Users\\aselimc\\micromamba\\envs\\dinov3\\Lib\\site-packages\\albucore\\utils.py\", line 92, in __process_fn\n chunk = img[:, :, index : index + 4]\n ~~~^^^^^^^^^^^^^^^^^^^^^^^^^\nIndexError: too many indices for array: array is 2-dimensional, but 3 were indexed\n```\n\nbut error does not occur if I just pass it as \n```python\naugmented = transform(image=image, mask=mask)\n```\n\nalbumentationsx 2.0.10\nalbucore 0.0.33"} {"task_id": "format-code-task-000338", "source_id": "format-code-task-000338", "domain": "code", "task_path": "tasks/format-code-task-000338", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5993b8ffbef2dd3c37fbf6bda9b06231f7128bc68a3ffd65479f5f994101b387", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `CoarseDropout` ignores the new `*_range` parameters\n\nI've been migrating my pipeline to the new range-style API of `CoarseDropout` since the docs say `min_holes`/`max_holes`/`min_height`/`max_height`/`min_width`/`max_width` are deprecated and we should use `num_holes_range`, `hole_height_range`, `hole_width_range` instead.\n\nHere's roughly what I'm doing:\n\n```python\nimport albumentations as A\nimport numpy as np\n\nimg = np.zeros((256, 256, 3), dtype=np.uint8) + 255\n\naug = A.CoarseDropout(\n num_holes_range=(4, 8),\n hole_height_range=(20, 40),\n hole_width_range=(20, 40),\n p=1.0,\n)\n\nout = aug(image=img)[\"image\"]\n```\n\nI expected to get somewhere between 4 and 8 dropout regions, each roughly 20-40 px in width and height. What I actually see is just a single small hole around 8x8 pixels, the same as if I hadn't passed any range at all. Increasing the values in the `*_range` tuples makes no difference — the output looks identical.\n\nInspecting the transform after construction confirms it:\n\n```python\nprint(aug.num_holes_range, aug.hole_height_range, aug.hole_width_range)\n# (1, 8) (8, 8) (8, 8) <- not what I passed in\n```\n\nSo the new-style parameters seem to be silently overridden by something during init. This basically means there is no way to actually use the non-deprecated API right now — everyone gets the default 8×8 single-hole behavior regardless of what they pass.\n\nPassing the deprecated `max_holes` / `max_height` / `max_width` still works as before, but the whole point of the new `*_range` parameters is to be the path forward, so they should actually take effect when the user provides them."} {"task_id": "format-code-task-000339", "source_id": "format-code-task-000339", "domain": "code", "task_path": "tasks/format-code-task-000339", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:248687bf9e25f8442c502f6a2d58ffd147d77a6e2b9a0a71e108b4947c28d8b8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## RandomToneCurve: float32 images aren't supported, and there's no per-channel mode\n\nI'm building an augmentation pipeline for a model that consumes float32 images (values in `[0, 1]`). Most albumentations transforms (`RandomBrightnessContrast`, `HueSaturationValue`, etc.) handle float32 fine, but `RandomToneCurve` doesn't:\n\n```python\nimport numpy as np\nimport albumentations as A\n\nimg = np.random.rand(256, 256, 3).astype(np.float32)\nA.RandomToneCurve(scale=0.1, p=1.0)(image=img)\n```\n\nThis raises an error complaining about the image dtype. I'd have to convert to uint8 just for this one transform and convert back, which is awkward in a `Compose` pipeline. Could `RandomToneCurve` (and the underlying `move_tone_curve`) support float32 the same way the other photometric transforms do?\n\nWhile we're at it — currently `RandomToneCurve` samples one curve and applies the same remapping to every channel, so it only changes brightness/contrast, never color. For color-augmentation experiments (à la WB augmenter / \"color constancy\" style augmentation) it'd be really useful to have an option that samples an independent tone curve per channel, so the transform can also shift the color balance. The single-curve behavior should stay the default to keep existing pipelines reproducible.\n\nIt would also be nice if `move_tone_curve` didn't assume RGB/grayscale specifically — I sometimes work with 4-channel (RGBA) or multi-spectral images, and it'd be good if the same code path just worked for any number of channels.\n\nI'd expect the new option to be exposed as something like a `per_channel` flag on `RandomToneCurve`."} {"task_id": "format-code-task-000340", "source_id": "format-code-task-000340", "domain": "code", "task_path": "tasks/format-code-task-000340", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:33810a957b2c8d0fd27fba789541a6832f531e483be9a70071ef3ef4fe334e5d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nError when arkade get helmfile mac m1\n\n\nArkade fails when trying to install helmfile:\n\n```bash\narkade get helmfile\nDownloading: helmfile\n2022/09/02 11:34:33 Looking up version for helmfile\n2022/09/02 11:34:34 Found: v0.144.0\nDownloading: https://github.com/roboll/helmfile/releases/download/v0.144.0/helmfile_darwin_386\nError: incorrect status for downloading tool: 404\n```\n\n## Expected Behaviour\n\n\n\narkade successfully installs helmfile\n\n## Current Behaviour\n\n\n\n404 not found\n\n## Are you a GitHub Sponsor yet (Yes/No?)\n\n\n\n\n- [ ] Yes\n- [x] No\n\n## Possible Solution\n\n\n\n1. helmfile migrated to dedicated org, owner field should be updated to `helmfile`\n2. in the new repo binary is replaced by archive with `tar.gz`, so we need to update template\n3. Condition should be added \n```go\n{{- else if or (eq .Arch \"aarch64\") (eq .Arch \"arm64\") -}}\n{{$arch = \"arm64\"}}\n```\n\n## Steps to Reproduce (for bugs)\n\n\n1. run `arkade get helmfile`\n4.\n5.\n6.\n\n## Context\n\n\n\n\n\nCan't install helmfile via arkade\n\n## Your Environment\n\n* What Kubernetes distribution are you using?\n\n```\nClient Version: version.Info{Major:\"1\", Minor:\"25\", GitVersion:\"v1.25.0\", GitCommit:\"a866cbe2e5bbaa01cfd5e969aa3e033f3282a8a2\", GitTreeState:\"clean\", BuildDate:\"2022-08-23T17:36:43Z\", GoVersion:\"go1.19\", Compiler:\"gc\", Platform:\"darwin/arm64\"}\nKustomize Version: v4.5.7\n```\n\n* Operating System and version (e.g. Linux, Windows, MacOS):\n\n```\nDarwin mariakot-osx 21.6.0 Darwin Kernel Version 21.6.0: Wed Aug 10 14:28:35 PDT 2022; root:xnu-8020.141.5~2/RELEASE_ARM64_T8101 arm64\n```\n\n* What arkade version is this?\n\n```\nVersion: 0.8.39\nGit Commit: 2e76c2c3681a3796b03483c3c04fb2ebcba0fa3e\n```"} {"task_id": "format-code-task-000341", "source_id": "format-code-task-000341", "domain": "code", "task_path": "tasks/format-code-task-000341", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4a4ad85368928b165816d559e3be32205b1322156ccbf2e1b0be51017ab9b4a6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI need the stateful `Vyper` configuration object to accept command-line values through `bind_args(parser_or_mapping) -> None`, `bind_arg(key: str, value: Any) -> None`, `bind_arg_values(values: Mapping[str, Any]) -> None`, and `bind_arg_value(key: str, value: Any) -> None`. When `bind_args` receives an `argparse.ArgumentParser`, it should parse the process command line, place every parser default into the configuration's default source, and place only parsed values that differ from their defaults into the command-line source. For example, with `--port` defaulting to `5000`, no flag should make `v.get(\"port\")` return `5000`, while `--port 7000` should make it return `7000`.\n\nWhen `bind_args` receives a mapping such as `{\"port\": 7000, \"debug\": None}`, it should bind `port` and ignore the null `debug` value; direct single-key binding should store the supplied value under a case-insensitive key, while a null value passed to `bind_arg_value` should raise `ValueError`. Command-line values should outrank environment, config-file, remote key/value, and default values, but an explicit `v.set(...)` override should still win. If an argparse option uses `choices` and the command line supplies an invalid choice, `bind_args` should surface argparse's `SystemExit` parser error. Separate `Vyper` instances should keep independently bound command-line values, and binding should update the instance so subsequent `get`, `is_set`, `all_keys`, and `all_settings` calls observe them."} {"task_id": "format-code-task-000342", "source_id": "format-code-task-000342", "domain": "code", "task_path": "tasks/format-code-task-000342", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a8d1d0e2a11969628d56cebca213c1aedd44c54074b91065c51afda2448ea8d5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nCan't reimplement MustParse behaviour when using NewParser\n`MustParse` is really helpful: https://github.com/alexflint/go-arg/blob/74af96c6ccf404613c251bca4814d32c69047c5f/parse.go#L85-L95\n\nHowever, if you want to use a custom `Config`, you have to call `NewParser()` which returns a `*Parser` but hasn't gone through the steps `MustParse` takes. Though I can implement most of `MustParse` myself again by calling `parser.Parse(os.Args)` first and then switching on the resulting `err` like `MustParse` does and calling the exported `WriteHelpForSubcommand()` and `FailWithSubcommand()`.\n\nI can't access `version` and `lastCmd` since those are unexported and there doesn't appear to be an exported equivalent. Version I can still get to using something like `&args.Version()` or having a `func Version()`, but `lastCmd` appears completely unaccessible."} {"task_id": "format-code-task-000343", "source_id": "format-code-task-000343", "domain": "code", "task_path": "tasks/format-code-task-000343", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9b2a375f36336d84fa8b5d337e9b3eefe4f01b598ad1769b541e3af22730f3d0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want configured `Network` instances to provide network-level estimate aggregation methods after a parser or caller has populated `network.partitions` and `network.batch_size`.\n\nThe public methods should be `get_memory_usage_estimate() -> int`, `get_inter_latency(delay, partition_list=None)`, `get_cycle(pipeline, partition_list=None) -> int`, `get_latency(freq, pipeline, delay, partition_list=None) -> float`, `get_throughput(freq, pipeline, delay, partition_list=None) -> float`, and `get_interval(partition_list=None) -> int`. For all methods that accept `partition_list`, passing `None` should mean all partitions in their current order, except `get_interval` should support an explicit list of partition indexes for multi-FPGA estimates.\n\nFor memory usage, the method should inspect each partition's first input and output node, take the maximum input workload times that partition's batch size, take the maximum output workload times that partition's batch size and weight-reload factor, add those maxima, double the result, and return the ceiling as an integer. For example, if the selected partition states imply input sizes `[10, 25]` and output sizes `[7, 60]`, `get_memory_usage_estimate()` should return `170`.\n\nFor cycle aggregation, `get_cycle(False)` should sum `partition.get_cycle()` across the selected partitions. If three partitions report cycles `100`, `200`, and `50`, then `get_cycle(False)` should return `350`, and `get_cycle(False, partition_list=[0, 2])` should return `150`. With pipelining enabled, `get_cycle(True)` should return `int(max(partition.get_interval()) * network.batch_size + sum(partition.get_pipeline_depth()))`; for intervals `20` and `35`, pipeline depths `5` and `7`, and `batch_size == 4`, it should return `152`.\n\n`get_inter_latency(delay, partition_list)` should return `0` when only one partition is selected and otherwise return `len(partition_list) * delay`, so two selected partitions with `delay == 0.0003` add `0.0006` seconds. `get_latency(freq, pipeline, delay, partition_list)` should convert the aggregated cycle count using `freq` in MHz, then add the inter-partition latency; for `get_cycle(...) == 500`, `freq == 250`, one selected partition, and any delay, the latency should be `500 / 250000000`. `get_throughput(...)` should return `network.batch_size / get_latency(...)` in frames per second.\n\nFor multi-FPGA networks, `get_interval(partition_list)` should assert that the network is marked as multi-FPGA and return the maximum `partition.get_interval()` among the requested partitions. If the network is not marked as multi-FPGA, calling `get_interval(...)` should raise `AssertionError` with a message explaining that it only works for multi-FPGA implementations. These methods should be read-only: calling them must not change `network.batch_size`, `network.partitions`, or any partition estimate values."} {"task_id": "format-code-task-000345", "source_id": "format-code-task-000345", "domain": "code", "task_path": "tasks/format-code-task-000345", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:bb01dd20f1a95afc3782a668a607aac4410caa64da0c737e658420a3f6874d6d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a `alicloud_cs_application` resource\n\nWe support managing container (Swarm) clusters through the provider, but there's no way to\ndeclare the *applications* that run on top of a cluster. Please add a new resource type,\n`alicloud_cs_application`, that deploys a Compose-style application onto an existing container\ncluster from a YAML template.\n\nThe resource must be registered with the provider and be a well-formed, importable Terraform\nresource (it supports create/read/update/delete and plain `terraform import`).\n\n### Configurable attributes\n\n- `cluster_name` — (required, force-new) the name of the target cluster the application is\n deployed to.\n- `name` — (required, force-new) the application name. Validate it at plan time: it must be\n 1–64 characters, may contain only ASCII letters, digits and hyphens, and must start with a\n letter or digit (a leading hyphen is not allowed). Anything else — an empty value, a value\n longer than 64 characters, a leading hyphen, spaces, underscores, dots, or non-ASCII\n characters such as Chinese characters — must produce a validation error.\n- `template` — (required) the application definition as a Compose-style YAML document. Reject a\n value that is not well-formed YAML with a validation error at plan time. Because YAML that is\n only cosmetically different (indentation, key ordering, trailing whitespace) describes the same\n application, the value must be stored in a normalized/canonical form so that equivalent\n documents do not show up as spurious diffs.\n- `description` — (optional) free-form description.\n- `version` — (optional) application version, defaulting to `\"1.0\"`.\n- `environment` — (optional) a map of string environment variables.\n- `latest_image` — (optional) boolean, defaults to `false`.\n- `blue_green` — (optional) boolean, defaults to `false`.\n- `blue_green_confirm` — (optional) boolean, defaults to `false`.\n\n### Read-only attributes\n\n- `services` — the list of service names that make up the application.\n- `default_domain` — the application's default domain.\n"} {"task_id": "format-code-task-000346", "source_id": "format-code-task-000346", "domain": "code", "task_path": "tasks/format-code-task-000346", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0f16cc588518f47f75acd9e162150d610ec14ef15147f3a61be81672001cdc3c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Missing ZMSCORE support\n\nI'm using miniredis to unit-test a service that talks to Redis through go-redis. The service has a hot path where it needs to look up scores for a batch of members in a sorted set (a leaderboard), so instead of firing off N round-trips of ZSCORE it uses ZMSCORE to get them all in one call.\n\nWhen I point the same code at miniredis in tests, the ZMSCORE call doesn't go through — miniredis doesn't seem to know about that command. ZSCORE works fine, so it looks like ZMSCORE just hasn't been wired up yet.\n\nCould ZMSCORE be added? It's been a standard Redis sorted-set command for a while now and it'd be nice to be able to cover code paths that use it without having to rewrite them to loop over ZSCORE just for tests.\n\nIt would also be useful to have a matching Go helper on `Miniredis` / `RedisDB` alongside the existing `ZScore(...)`, so test setup/assertions can query multiple members at once directly without going through the protocol."} {"task_id": "format-code-task-000347", "source_id": "format-code-task-000347", "domain": "code", "task_path": "tasks/format-code-task-000347", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:485a7151662318917a22a1d772815db1e29feaef17bf0ef9453b81b4696b482f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI'm trying to pull a list of all the chart repositories under one of my Container Registry instances so I can iterate over them in Terraform, but the alicloud provider doesn't seem to expose a data source for that — I can list image repos but not chart ones. Could you add something on the alicloud provider side so I can query chart repos under a given CR instance? Ideally I'd also be able to filter by name or by a set of ids, and dump the result to a file if I want.\n\nExpected outcomes:\n- Data source availability: Terraform configurations can use `data \"alicloud_cr_chart_repositories\"` to read chart repositories for a Container Registry instance.\n- Required input: `alicloud_cr_chart_repositories.instance_id` is required, so validation or planning reports a missing required argument when it is omitted.\n- Repository results: `alicloud_cr_chart_repositories.repositories` returns the matching chart repositories, with each item exposing `chart_repository_id`, `create_time`, `instance_id`, `id`, `repo_name`, `repo_namespace_name`, `repo_type`, and `summary`.\n- Stable identifiers: each repository `id`, and each value in the top-level `ids` output, identifies a chart repository using its instance, namespace, and repository name; the top-level `names` output contains the repository names.\n- Filtering: `alicloud_cr_chart_repositories.name_regex` limits results to repository names matching the regular expression, and invalid regular expressions are rejected during configuration validation.\n- Filtering by identifiers: `alicloud_cr_chart_repositories.ids` can be supplied to return only repositories whose identifiers are in the supplied set.\n- Optional file output: when `alicloud_cr_chart_repositories.output_file` is set, the data source writes the returned repository list to that path.\n\nImplementation notes:\n- Follow the provider’s existing conventions for Terraform data sources, schema validation, state shape, pagination, retries, and optional file output.\n- The specific helper functions, internal data structures, request construction, and filtering location are implementation details, as long as the public Terraform behavior above is satisfied.\n- Do not require users to manage chart repositories with a new resource in order to use this data source; it should work for existing chart repositories in the target instance."} {"task_id": "format-code-task-000348", "source_id": "format-code-task-000348", "domain": "code", "task_path": "tasks/format-code-task-000348", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:dead70eede3a5be808c77b7184aaea34db67385fb237f4ae4ae5c7202d5d5db6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Bring the EMR (v2) cluster resource up to date with newer service capabilities\n\nThe `alicloud_emrv2_cluster` resource is missing a few configuration options that the\nunderlying EMR service now supports. Please extend the resource so users can express them\nin their Terraform configurations.\n\n## System disk encryption\n\nToday the node-attributes block of a cluster only lets users turn on encryption for *data*\ndisks (`data_disk_encrypted` / `data_disk_kms_key_id`). Add the equivalent options for the\n**system** disk:\n\n- `system_disk_encrypted` — a boolean toggle for whether the node's system disk is encrypted.\n Optional.\n- `system_disk_kms_key_id` — the KMS key id used to encrypt the system disk. Optional, and a\n string. An empty value is not a meaningful key id and must be rejected during validation\n (i.e. configuring it as `\"\"` should produce a validation error); any non-empty value is\n accepted.\n\nBoth options describe how a node is provisioned at creation time, so — like the existing\nnode-attribute fields — they are immutable: changing either of them must force the cluster to\nbe recreated rather than updated in place.\n\n## Newer enumerated values\n\nTwo existing fields reject values that the service now accepts. Widen their accepted sets\n(without dropping any value already accepted today, and while still rejecting anything outside\nthe set):\n\n- The node group type now also supports `MASTER-EXTEND`, in addition to the existing\n `MASTER`, `CORE`, `TASK`, and `GATEWAY`.\n- A bootstrap script's execution moment now also supports `BEFORE_START`, in addition to the\n existing `BEFORE_INSTALL` and `AFTER_STARTED`.\n\nThese should behave like the other validated string fields: a configuration using one of the\naccepted values plans cleanly, and a value outside the accepted set is reported as invalid.\n"} {"task_id": "format-code-task-000349", "source_id": "format-code-task-000349", "domain": "code", "task_path": "tasks/format-code-task-000349", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c4c8a1de5ee7c32269af98da665324349deb312af84e72f9d3a6a2e962c2734d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `image-syncer` root command to load synchronization inputs from user-supplied configuration files before it starts copying registry images.\n\nThe CLI should support `image-syncer --images ` with an optional `--auth `, and it should also support the deprecated combined form `image-syncer --config `. These files may be JSON, YAML, or YML; paths with any other extension should fail during startup before any registry transfer begins. For example, running `image-syncer --images rules.txt` should exit non-zero and report an initialization error that includes `decode image file rules.txt error: only one of yaml/yml/json format is supported`.\n\nIf neither `--config` nor `--images` is provided, startup should fail non-zero with an initialization error that includes `neither config.json nor images.json is provided`. If `--images` is provided without `--auth`, startup should continue and log a warning that no authentication information was found because neither `config.json` nor `auth.json` was provided.\n\nFor separate auth files, each auth entry should load `username`, `password`, and `insecure`, and environment variables in usernames and passwords should be expanded. When building transfer tasks, source and destination repositories should receive the auth entry whose configured repository key matches the repository path on a slash boundary; if no auth entry matches, the command should log that access for that repository will be anonymous.\n\nThe image rules file should map each source image to either one destination string or a list of destination strings. Environment variables in destination strings should be expanded, duplicate destinations in a list should be collapsed, and each source/destination pair should become a sync task. Empty destination strings, empty destination lists, and destination lists containing non-string values should fail during startup; for an empty destination, the error should include `empty destination is not supported for source: `.\n\nThe command should also accept repeatable `--os ` and `--arch ` filters, ignore empty filter values, and pass the remaining filters into the generated sync tasks. `image-syncer --help` should list `--config`, `--auth`, `--images`, `--os`, and `--arch` alongside the other root command flags and exit 0."} {"task_id": "format-code-task-000350", "source_id": "format-code-task-000350", "domain": "code", "task_path": "tasks/format-code-task-000350", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7e9a7d217d968ea334dee9b0fda011d6d133d308364e7ac5f56ababa2386b867", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the Go utility package to expose `GenerateRepoURLs(url string, externalTagsOrDigest func(string, string) ([]string, error)) ([]*RepoURL, error)` as a one-shot parser and expander for Docker image references. When I call it with `\"nginx:v1\"` and a tag provider that would return `[]string{\"latest\"}`, it should return one `RepoURL` whose `String()` is `docker.io/library/nginx:v1`, `GetRegistry()` is `docker.io`, `GetRepo()` is `library/nginx`, and `GetTagOrDigest()` is `v1`; the provider should not be consulted for this concrete tag. When I call it with `\"127.0.0.1:300/library/nginx:v1,v2\"`, it should return two entries for the same registry and repository, preserving the tag order `v1`, then `v2`.\n\nIf the URL has no tag or digest, such as `\"registry.hub.docker.io/library/nginx\"`, the function should call the supplied tag provider with registry `registry.hub.docker.io` and repository `library/nginx`, then return one `RepoURL` per provided tag using those tags in provider order. If the URL uses a tag regex like `\"test-regex/test:/b+/\"`, it should fetch all tags through the provider, compile the regex between the slashes, and return only matching tags; for provider output `[]string{\"aaa\", \"bbb\"}`, the result should contain only tag `bbb`. If the URL uses a digest like `\"registry.cn-beijing.aliyuncs.com/hhyasdf/hybridnet@sha256:df2ef9e979fc063645dcbed51374233c6bcf4ab49308c0478702565e96b9bc9e\"`, the returned `RepoURL` should preserve that digest, use `@` in `GetRepoWithTagOrDigest()`, and report repository `hhyasdf/hybridnet`.\n\nInvalid input should return a non-nil error instead of partial results: a regex tag must have both leading and trailing `/`, an invalid regex should report a regex error, an invalid repository reference should report a parse error, and an error returned by the tag provider should be propagated with context. For the same input and a deterministic tag provider, repeated calls should produce equal ordered results. The function should not write files, perform network access directly, or keep hidden global state; all external tag lookup must happen through the supplied callback."} {"task_id": "format-code-task-000352", "source_id": "format-code-task-000352", "domain": "code", "task_path": "tasks/format-code-task-000352", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e3cca81dac0140ab5000e945d52616f3cceb8ea6c08fa1481754cce344671292", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nWhen I move a chassis to a different rack in Ralph, the blades under it still show the old rack afterward, which makes the location data inconsistent. Could you make the directly attached child assets follow the parent’s rack when I save that change?\n\n## Expected outcomes\n\n- When an existing parent data center asset is saved after its rack changes, each directly attached child data center asset reflects the parent’s new rack.\n- Child assets that are not directly attached to the moved parent are outside the required synchronization scope.\n- Saving a parent without changing its rack should not introduce unrelated location changes.\n\n## Implementation notes\n\n- The specific place where the synchronization is performed is up to the implementation, as long as the externally observable saved asset state is consistent afterward.\n- Preserve existing validation and save behavior for data center assets while adding the rack synchronization behavior."} {"task_id": "format-code-task-000354", "source_id": "format-code-task-000354", "domain": "code", "task_path": "tasks/format-code-task-000354", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:977ef36c6e6bfc51d14a400e091b84befdd64cfef9b40a314f95ae5e51327a45", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### SecurityScan piling up multiple records per host\n\nWe push scan results into ralph from an external vulnerability scanner. The integration runs on a schedule and `POST`s the latest scan for each host to `/api/security-scans/` (using `host_ip` to identify the target).\n\nAfter running this for a while I noticed that ralph keeps **every** scan we've ever submitted for a given host. So for a single machine I end up with dozens of `SecurityScan` rows, all pointing at the same `base_object`, and the API/admin happily lets them accumulate forever.\n\nConceptually that doesn't make sense for our use case — there is only ever one \"current\" scan result for a host. The previous one is stale the moment a new scan finishes; we don't want history, we want the latest state. The Security Info tab in the admin already kind of acknowledges this (it only shows one entry per host anyway), so the underlying data model storing N of them feels wrong.\n\nWhat I'd expect:\n\n- Re-`POST`ing a scan for a host that already has one should replace the previous one rather than add another row.\n- Each host should only ever have one `SecurityScan` associated with it.\n\nIt would also be nice if existing databases (ours has been running for a while and already has the duplicates) got cleaned up to the \"one per host\" state when this is fixed — otherwise we'd have to write our own cleanup script before the new behaviour can be enforced.\n\nIs there a reason scans were modelled as \"many per host\" originally? If not, can we move to a one-scan-per-host model?"} {"task_id": "format-code-task-000357", "source_id": "format-code-task-000357", "domain": "code", "task_path": "tasks/format-code-task-000357", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2491d9b6e6ae3fc60b3075beb22adf79fe1859aa933c7e063f3e3ee927b721eb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAllenNLP accepts local and downloaded archives in several public paths, but their safety rules have drifted: dataset archives opened through `cached_path(..., extract_archive=True)` receive only partial tar checks, ZIP files bypass those checks, and model archives opened through `extracted_archive()` are extracted directly. A malformed archive must never be able to write outside its destination or leave a cache in a half-replaced state. Please apply one consistent archive-safety contract across these surfaces and to additional files selected by `archive_model(..., include_in_archive=...)`.\n\nFor extraction, ordinary nested regular files and directories in both tar and ZIP archives must continue to extract with their names and contents intact. Before writing any member, inspect the complete archive and reject an absolute member name or any name whose normalized destination escapes the extraction root. Symbolic links, hard links, and non-file/non-directory members such as FIFOs or device entries are unsupported and must also be rejected; this includes ZIP entries marked as Unix symbolic links. `cached_path()` and `extracted_archive()` must raise `ValueError` that identifies the offending member. A rejected archive must create nothing outside the destination and must not publish an extraction directory or its metadata.\n\n`force_extract=True` is a transactional refresh: validate and stage the replacement before replacing a successful cached extraction. If refresh fails, the previous extracted directory, its contents, and its metadata remain usable. The `archive!member` form must likewise confine the selected member to the extracted root: normal relative nested selections still work, while absolute or escaping selections raise `ValueError` identifying the requested path even when a file with that name exists beside the extraction directory.\n\nModel archive creation needs the matching outbound boundary. `include_in_archive` must retain support for relative glob patterns and relative directories under the serialization directory, storing matched content under relative archive names. `archive_model()` itself must reject, with `ConfigurationError` identifying the target, absolute selections, selections that escape the serialization directory, reserved built-in archive targets (`config.json`, weights names, or `vocabulary`), and any selected tree containing a symbolic link. Validate all additions before publishing the requested archive, so an invalid selection leaves no partial output archive."} {"task_id": "format-code-task-000359", "source_id": "format-code-task-000359", "domain": "code", "task_path": "tasks/format-code-task-000359", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d1ed7ef6493e85ad209592debc08384a43ad2d768d674c32e8d188ba3d4a963d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI'm running `ensure-sca-scanner` on a repo that uses JFrog Pipelines, and Rule 10 is still reporting that there's no SCA scanner configured. The pipeline has a step that runs `jf xr ...`, so I'm not sure if allero is missing the JFrog pipeline config or just not recognizing the Xray CLI command.\n\nExpected outcomes:\n- Rule 10 / `ensure-sca-scanner` treats SCA scanner configuration found in JFrog Pipelines as satisfying the scanner requirement.\n- A repository that only configures its SCA scan through JFrog Pipelines should not be reported as missing an SCA scanner.\n- Commands using the JFrog Xray CLI form `jf xr ...` should be recognized as SCA scanner invocations when they appear in supported CI/CD pipeline command steps.\n- Existing SCA scanner detection in GitHub Actions and GitLab CI should continue to work.\n\nImplementation notes:\n- The exact parsing approach, data structures, and helper organization are up to the implementer.\n- Keep the behavior focused on Rule 10’s externally observable pass/fail result rather than on any particular internal representation."} {"task_id": "format-code-task-000360", "source_id": "format-code-task-000360", "domain": "code", "task_path": "tasks/format-code-task-000360", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b437ee60bd32f295a5e70cc7a6a6a2838801a24b8a8e601651fc09f24bab4cab", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nWhen I run AllThePlaces spiders for non-US locations, the `phone` field comes out in whatever local formatting the site used, and some perfectly valid international numbers get counted as invalid. Could we make phone output/validation handle real international phone numbers more consistently, so parseable numbers are emitted in a normalized form instead of each spider’s raw formatting?\n\n## Expected outcomes\n\n- Phone normalization:\n - When an item has a parseable valid `phone` value and enough country context is available, the emitted `phone` value should use a consistent international representation rather than preserving arbitrary source formatting.\n - Valid national-format phone numbers for non-US countries should be normalized using their item country context, not treated as US-shaped strings.\n - Existing output continues to use the `phone` field; this change should not require individual spiders to hand-normalize common phone-number formatting variations.\n\n- Phone validation:\n - Valid international or country-local phone numbers should not be counted as invalid merely because they do not match a US-style phone pattern.\n - Phone-like values that are syntactically shaped like a phone number but are not valid real numbers should still be reported through the existing `atp/field/phone/invalid` statistic.\n\n## Implementation notes\n\n- The specific parsing/normalization mechanism, data structures, and where the shared handling is wired into the item-processing flow are implementation choices.\n- Prefer a centralized behavior that applies consistently across spiders rather than adding one-off formatting fixes to individual spiders.\n- Keep the externally observable contract focused on emitted `phone` values and existing validation statistics."} {"task_id": "format-code-task-000361", "source_id": "format-code-task-000361", "domain": "code", "task_path": "tasks/format-code-task-000361", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:bf796b35bfa89d718a97cf04ebdb0fe3d219ea12fa06581167f1efed4ac36444", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `notesium links` CLI command to print markdown note link relationships from my `NOTESIUM_DIR` notes directory. The command should be invoked as `notesium links [filename]` and should accept `--color`, `--outgoing`, `--incoming`, `--dangling`, and `--filename=` in addition to a positional `` filename.\n\nWhen I run `notesium links` on a directory containing timestamp-named markdown notes, it should exit 0 and print one line per outgoing markdown note link in title-sorted source-note order, using `source.md:: → ` for existing targets and the target filename for dangling targets. For example, if `64214a1d.md` is titled `richard feynman` and links on line 3 to `642146c7.md` titled `physicist`, stdout should include `64214a1d.md:3: richard feynman → physicist`; if a note links to missing `12345678.md`, stdout should include that filename as the target.\n\nWhen I run `notesium links 64214a1d.md`, it should exit 0 and print only links related to that note: outgoing links as `target.md:1: outgoing ` and incoming links as `source.md:: incoming `. `notesium links --outgoing 64214a1d.md` should print only existing outgoing targets as `target.md:1: `, and `notesium links --incoming 642146c7.md` should print only existing incoming sources as `source.md:: `. If `--incoming` and `--outgoing` are both supplied with a filename, it should print both groups with the `incoming` and `outgoing` labels.\n\nWhen I run `notesium links --dangling`, it should exit 0 and print only broken note links as `source.md:: → `. `--color` should wrap the label or source-title portion that the command highlights with the standard Notesium ANSI color and reset sequences while preserving the same fields and line order. Invalid option combinations should fail during argument parsing: `notesium links --outgoing` and `notesium links --incoming` without a filename should exit nonzero and report `filename is required`, while `notesium links --dangling 64218087.md` should exit nonzero and report `filename not supported`. The command should also appear in `notesium help` with its filename argument and the supported flags."} {"task_id": "format-code-task-000362", "source_id": "format-code-task-000362", "domain": "code", "task_path": "tasks/format-code-task-000362", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:8a033b245440a0286e6c95e0fa77d706a04afef3e05e4947061981cfd94c6da6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `notesium lines` CLI command to search note contents line by line from the directory selected by `NOTESIUM_DIR`. The invocation should be `notesium lines [--color] [--prefix=title] [--filter=QUERY]`, and the top-level usage text should list the `lines` command with those three flags.\n\nOn the happy path it should exit 0 and print matching non-empty lines from every `.md` note as `filename:line_number: text`, preserving the original line number while skipping blank lines. For example, if `6421460b.md` starts with `# book`, `notesium lines` should include `6421460b.md:1: # book`; if line 3 of `64214930.md` is `a fundamental theory in physics that provides a description of the`, output should include `64214930.md:3: a fundamental theory in physics that provides a description of the`.\n\nWith `--prefix=title`, each emitted line should insert the note title before the line text, where the title is taken from the first `# ` heading encountered in that note. For example, `notesium lines --prefix=title` should print `6421460b.md:1: book # book`, and a later line in a note titled `quantum mechanics` should look like `64214930.md:3: quantum mechanics a fundamental theory in physics that provides a description of the`. With `--color --prefix=title`, only that inserted title prefix should be wrapped in the default cyan ANSI code `\\x1b[0;36m` and reset with `\\x1b[0m`.\n\nWith `--filter=QUERY`, matching should be case-insensitive and applied to the line text. Space-separated tokens are implicit AND terms, `|` inside a token is OR, a token starting with `!` is NOT, and single- or double-quoted phrases should be treated as one token. For example, `--filter='Einstein'` should match lines containing `einstein`, `--filter='einstein albert'` should require both words, `--filter='american|german'` should accept either word, `--filter='einstein !physicist'` should exclude lines containing `physicist`, and `--filter='\"theory of quantum\"'` should require that exact phrase. Combining filtering and `--prefix=title` should keep the same `filename:line_number: title text` output format.\n\nIf the filter matches nothing, the command should still exit 0 with no stdout. If an unknown option is passed, such as `notesium lines --bogus`, it should write an error containing `unrecognized option: --bogus` to stderr and exit non-zero."} {"task_id": "format-code-task-000363", "source_id": "format-code-task-000363", "domain": "code", "task_path": "tasks/format-code-task-000363", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b3abb277933470c8f3e330ea2d52775f3eac689c45ec0b1b69e329a639e08fc7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[Bug]: pydantic ValidationError when calling close_all_positions\n### Is there an existing issue for this?\n\n- [X] I have searched the existing issues\n\n### Current Behavior\n\nWhen I run the code in **Steps To Reproduce** I get the error posted in **Anything else?**.\n\nNot much else to add.\n\n### Expected Behavior\n\n_No response_\n\n### SDK Version I encountered this issue in\n\n0.10\n\n### Steps To Reproduce\n\n```markdown\ntrading_client = TradingClient(alpaca_key, secret, url_override='https://api.alpaca.markets')\ntrading_client.close_all_positions(cancel_orders=False)\n```\n\n\n### Filled out the Steps to Reproduce section?\n\n- [X] I have entered valid steps to reproduce my issue or have attached a minimally reproducible case in code that shows my issue happening; and understand that without this my issue will be flagged as invalid and closed after 30 days.\n\n### Anything else?\n\nTraceback (most recent call last):\n File \"C:\\Users\\Ben\\PycharmProjects\\StockTrader\\main.py\", line 418, in \n print(trading_client.close_all_positions(cancel_orders=True))\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\Ben\\PycharmProjects\\StockTrader\\venv\\Lib\\site-packages\\alpaca\\trading\\client.py\", line 287, in close_all_positions\n return parse_obj_as(List[ClosePositionResponse], response)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"pydantic\\tools.py\", line 38, in pydantic.tools.parse_obj_as\n File \"pydantic\\main.py\", line 341, in pydantic.main.BaseModel.__init__\n account = trading_client.get_account()\npydantic.error_wrappers.ValidationError: 18 validation errors for ParsingModel[List[alpaca.trading.models.ClosePositionResponse]]\n__root__ -> 0 -> body -> available\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> existing_qty\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> held_for_orders\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> symbol\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> id\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> client_order_id\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> created_at\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> updated_at\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> submitted_at\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> asset_id\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> symbol\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> asset_class\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> order_type\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> type\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> side\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> time_in_force\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> status\n field required (type=value_error.missing)\n__root__ -> 0 -> body -> extended_hours\n field required (type=value_error.missing)"} {"task_id": "format-code-task-000364", "source_id": "format-code-task-000364", "domain": "code", "task_path": "tasks/format-code-task-000364", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5178eb2f2ee486812af1f21f9eaabf207cd0defdef4a491b1dac84eb179abe3d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add ACH relationship and bank funding endpoints to BrokerClient\n\nThe Broker API lets you fund accounts. We support two of the funding building\nblocks today only as raw HTTP — please add first-class support to `BrokerClient`\nfor **ACH relationships** and **recipient banks**, including the request/response\nmodels and enums they need. Everything should be importable from the\n`alpaca.broker` package the same way the existing models and enums are.\n\nAll of the new client methods accept `account_id` (and any other resource id) as\neither a `UUID` or a `str`. An id that is not a valid UUID must raise\n`ValueError` before any HTTP request is made.\n\n## ACH relationships\n\nAdd these methods:\n\n- `create_ach_relationship_for_account(account_id, ach_data)` — creates one ACH\n relationship by POSTing to `/accounts/{account_id}/ach_relationships`, and\n returns the created relationship as an `ACHRelationship`. There are two ways to\n create a relationship: by supplying raw bank/ACH details, or by supplying a\n Plaid processor token. `ach_data` must therefore be accepted as either of the\n two corresponding request models. If `ach_data` is neither, raise a\n `ValueError` and do not make any HTTP request.\n\n- `get_ach_relationships_for_account(account_id, statuses=None)` — GETs\n `/accounts/{account_id}/ach_relationships` and returns a `List[ACHRelationship]`.\n `statuses` is an optional list of `ACHRelationshipStatus`; when provided and\n non-empty it must be sent as a single `statuses` query parameter whose value is\n the statuses joined with commas, preserving the given order. When omitted,\n `None`, or empty, no `statuses` query parameter is sent.\n\n- `delete_ach_relationship_for_account(account_id, ach_relationship_id)` — DELETEs\n `/accounts/{account_id}/ach_relationships/{ach_relationship_id}`. It returns\n `None` on success.\n\n### Request models\n\n- `CreateACHRelationshipRequest` — for creating a relationship from raw ACH\n details, with fields: `account_owner_name` (str), `bank_account_type` (a\n `BankAccountType`), `bank_account_number` (str), `bank_routing_number` (str),\n and an optional `nickname` (str).\n- `CreatePlaidRelationshipRequest` — for creating a relationship from a Plaid\n processor token, with a single `processor_token` (str) field.\n\n### `ACHRelationship` model\n\nParses the API response into typed attributes, including at least: `id` (UUID),\n`account_id` (UUID), `status` (an `ACHRelationshipStatus`), `account_owner_name`,\n`bank_account_type` (a `BankAccountType`), `bank_account_number`, and\n`bank_routing_number`.\n\n## Recipient banks\n\nAdd these methods:\n\n- `create_bank_for_account(account_id, bank_data)` — creates one bank by POSTing\n to `/accounts/{account_id}/recipient_banks` and returns the created `Bank`.\n- `get_banks_for_account(account_id)` — GETs `/accounts/{account_id}/recipient_banks`\n and returns a `List[Bank]`.\n- `delete_bank_for_account(account_id, bank_id)` — DELETEs\n `/accounts/{account_id}/recipient_banks/{bank_id}` and returns `None` on success.\n\n### `CreateBankRequest` request model\n\nFields: `name` (str), `bank_code_type` (an `IdentifierType`), `bank_code` (str),\n`account_number` (str), and the optional international-location fields `country`,\n`state_province`, `postal_code`, `city`, and `street_address` (all str).\n\nA bank is either **domestic** (identified by an ABA routing number) or\n**international** (identified by a BIC). The location fields are meaningful only\nfor international banks. Enforce this at construction time:\n\n- When `bank_code_type` is `IdentifierType.ABA`, none of the five location fields\n may be set; setting any of them raises `ValueError`.\n- When `bank_code_type` is `IdentifierType.BIC`, all five location fields are\n required; leaving any of them unset raises `ValueError`.\n\n### `Bank` model\n\nParses the API response into typed attributes, including at least: `id` (UUID),\n`account_id` (UUID), `name`, `status` (a `BankStatus`), `account_number`,\n`bank_code`, and `b"} {"task_id": "format-code-task-000365", "source_id": "format-code-task-000365", "domain": "code", "task_path": "tasks/format-code-task-000365", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:43b7d7e0bbc7eed6550bff037e835b5fbd6433e22e6bdcde5c0705bc78b6222a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[Bug]: potentially unnecessary parameter is required when submitting OCO orders\n### Is there an existing issue for this?\n\n- [X] I have searched the existing issues\n\n### Current Behavior\n\n1) When creating a LImitOrderRequest for an OCO order, an error is returned unless the parameter \"limit_price\" is also included\n2) As a workaround, I use limit_price=1 in all orders. This parameter doesn't seem to be present in the order confirmation details.\n\n### Expected Behavior\n\n1) I assume a LImitOrderRequest is what is needed for an OCO order, since it is used in bracket and OTO orders. \n2) The parameter \"limit_price\" is not necessary here, because \"take_profit\" and \"stop_loss\" are included.\n\n### SDK Version I encountered this issue in\n\n0.14.0\n\n### Steps To Reproduce\n\n```markdown\noco_order_data = LimitOrderRequest(symbol='AAPL', qty=1, side=OrderSide.SELL, time_in_force=TimeInForce.GTC, order_class=OrderClass.OCO, take_profit=TakeProfitRequest(limit_price=300), stop_loss=StopLossRequest(stop_price=50)\n```\n\n\n### Filled out the Steps to Reproduce section?\n\n- [X] I have entered valid steps to reproduce my issue or have attached a minimally reproducible case in code that shows my issue happening; and understand that without this my issue will be flagged as invalid and closed after 30 days.\n\n### Anything else?\n\nWould it be too much to ask for an example of how to submit OCO orders in the sparse documentation?"} {"task_id": "format-code-task-000366", "source_id": "format-code-task-000366", "domain": "code", "task_path": "tasks/format-code-task-000366", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b421a7db46910a29f24875a7375f3b885ac1545b93a537cebbaf12d67828b0da", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Flesh out order management on the Trading API client\n\nRight now `TradingClient` can only submit a new order (`submit_order`). That's not\nenough to actually manage orders against the Alpaca trading API — there's no way to\nlist existing orders, look one up, change one, or cancel them. Please round out the\nclient so it covers the rest of the `/v2/orders` routes.\n\nAdd the following methods to `TradingClient`:\n\n- **`get_orders(filter=None)`** — returns a `list` of `Order` objects by issuing a\n `GET` to `/v2/orders`. The optional `filter` is a `GetOrdersRequest` describing the\n standard query parameters: order `status`, `limit`, `after`/`until` timestamps, sort\n `direction`, `nested`, `side`, and a list of `symbols`. Every field is optional and\n only the ones that are set should be sent. When the filter supplies a list of\n symbols, they must be transmitted as a single comma-separated `symbols` query\n parameter (e.g. `symbols=SPY,AAPL`), not as repeated parameters. Calling it with no\n filter must still work and simply send no query parameters.\n\n- **`get_order_by_id(order_id, filter=None)`** — returns the `Order` retrieved from\n `GET /v2/orders/{order_id}`. The optional `filter` is a `GetOrderByIdRequest`\n carrying the `nested` flag.\n\n- **`replace_order(order_id, order_data=None)`** — updates an existing order via\n `PATCH /v2/orders/{order_id}` and returns the resulting `Order`. The optional\n `order_data` is a `ReplaceOrderRequest` describing the fields that may be changed\n (`qty`, `time_in_force`, `limit_price`, `stop_price`, `trail`, `client_order_id`);\n again, only fields that are set should be sent.\n\n- **`cancel_orders()`** — issues a `DELETE` to `/v2/orders` and returns a `list`\n describing the cancellation outcome of each order. The API responds with an array of\n objects each carrying the order `id` and an integer HTTP `status`; surface each as a\n `CancelOrderResponse` exposing `id` (a `UUID`) and `status` (an `int`).\n\nWherever a method takes an order id, accept either a `UUID` instance or a UUID-formatted\nstring. If the value is neither (for example an arbitrary non-UUID string), raise a\n`ValueError` before any request is made.\n\nThe new request objects (`GetOrdersRequest`, `GetOrderByIdRequest`,\n`ReplaceOrderRequest`) and the `CancelOrderResponse` model should be importable from the\nsame place as the existing order request models (e.g. `MarketOrderRequest`), and the\nrequest objects should follow the existing convention of omitting unset fields when\nserialized.\n\nOne more thing: an `Order`'s `client_order_id` is a user-supplied identifier that is not\nnecessarily a UUID, so make sure an `Order` whose `client_order_id` is an arbitrary\nstring parses correctly.\n"} {"task_id": "format-code-task-000367", "source_id": "format-code-task-000367", "domain": "code", "task_path": "tasks/format-code-task-000367", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ce7bf195dc5bb7f0618b276bfb4f7de076b64ba6070b756e1eaec5e01dea439b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the package entry point to expose a NestJS dynamic module registration API through `UnleashClientModule`. The class should provide `static forRoot(config: { config: UnleashConfig; isGlobal: boolean }): DynamicModule`: when called with `{ config: { url: 'http://unleash.test/api/', appName: 'api' }, isGlobal: true }`, it returns module metadata whose `module` is `UnleashClientModule`, `global` is `true`, `exports` includes `UnleashClientProvider`, `UNLEASH_CLIENT_CONFIG`, and `Reflector`, and `providers` includes a value provider for `UNLEASH_CLIENT_CONFIG` with the same object, a factory provider that creates `UnleashClientProvider` from the nested `config`, `Reflector`, and `UnleashGuard`. Calling `forRoot` with the same config and `isGlobal: false` should return equivalent metadata with `global` set to `false`, and the supplied config object must not be mutated.\n\nI also want `static forRootAsync(options: { imports?: ModuleMetadata['imports']; useFactory?: (...args: any[]) => Promise<{ config: UnleashConfig }> | { config: UnleashConfig }; inject?: any[]; isGlobal?: boolean }): DynamicModule`. When called with an imports array, an inject array, a factory function, and `isGlobal: true`, it should preserve the imports on the returned module metadata, set `global` to `true`, export `UnleashClientProvider`, `UNLEASH_CLIENT_CONFIG`, and `Reflector`, register `UNLEASH_CLIENT_MODULE_OPTIONS` with the supplied factory and inject list, register `UNLEASH_CLIENT_CONFIG` as the resolved module options, and register a factory provider that creates `UnleashClientProvider` from the resolved nested `config`. When called as `forRootAsync({ isGlobal: false })`, it should return `imports: []`, `global: false`, no `UNLEASH_CLIENT_MODULE_OPTIONS` factory provider, and still include the client provider, config token provider, `Reflector`, and `UnleashGuard`. These registration calls should only construct Nest module metadata; repeated calls with the same input should return equivalent results and should not open network connections, write files, or mutate global state."} {"task_id": "format-code-task-000368", "source_id": "format-code-task-000368", "domain": "code", "task_path": "tasks/format-code-task-000368", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:72cf7f9f2a0784f1458e8c5f8e14d6d3f23b5bbf76ea189fba007972ae28e86f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want to import `UnleashClientProvider` from `@alpha018/nestjs-unleash-client` and use it as a Nest injectable, stateful facade over the Node Unleash SDK. The class should be constructible as `new UnleashClientProvider(config: UnleashConfig)`, expose `onModuleInit(): void`, `onModuleDestroy(): void`, a readonly `unleashClient: Unleash` getter, `isEnabled(name: string, context?: Context): boolean`, and `getFeatureToggleDefinition(toggleName: string): FeatureInterface | undefined`.\n\nWhen `onModuleInit()` is called with a config such as `{ url: 'http://localhost:4242/api', instanceId: 'test-instance', appName: 'test-app' }`, it should create one `Unleash` SDK client with that config and make it available through `unleashClient`. During initialization it should register SDK `error` and `warn` listeners; an error event should be logged through Nest's `Logger.error` with the error message and stack, and a warning event should be logged through `Logger.warn` with the warning message.\n\nAfter initialization, `provider.isEnabled('checkout', { userId: '42' })` should return the boolean produced by the SDK client's `isEnabled('checkout', { userId: '42' })`; for example it returns `true` when the SDK returns `true`, and `false` when the SDK returns `false`. `provider.getFeatureToggleDefinition('checkout')` should return the SDK client's feature toggle definition object for that name, and should return `undefined` when the SDK has no definition for that name. When `onModuleDestroy()` is called, the provider should destroy the retained SDK client by calling its `destroy()` method."} {"task_id": "format-code-task-000369", "source_id": "format-code-task-000369", "domain": "code", "task_path": "tasks/format-code-task-000369", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f4d3998ac4e14715f8c93f75d77824c3d26ba3c89f4c500fa818c4e254b5390d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI need the existing SSHClient object to expose `ExecuteCommand(cmd string) (string, error)` for running dashboard metric collection commands over its already-open SSH connection. Each call should trim surrounding whitespace only for allowlist checking, accept only commands whose trimmed text starts with one of these prefixes: `lscpu `, `top -`, `which `, `nvidia-smi `, `amd-smi `, `rocm-smi `, `free -`, or `df -`; any other command, such as `uname -a`, must return an empty string and an error whose message is `command not in allowed list: `, without opening an SSH session or sending the command to the remote host.\n\nFor an allowed command, `ExecuteCommand` should open a fresh SSH session on that client, run the command with combined stdout/stderr capture, close the session before returning, and return the captured text as a Go string. If the remote command exits successfully, the error return is nil; for example, if the remote server responds to `free -m | grep Mem:` with `Mem: 1000 400 600\\n`, the method returns that exact string and nil. If the remote command writes `warn\\n` and exits non-zero for an allowed command such as `df -h | grep -E '^/dev/'`, the method still returns `warn\\n` together with the SSH command error. If creating the session fails, return an empty string and the session creation error."} {"task_id": "format-code-task-000370", "source_id": "format-code-task-000370", "domain": "code", "task_path": "tasks/format-code-task-000370", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:41d30d3b051ab77408e4961edea508e1b22c7f2adf3fa36a7eb3c0278ee87e10", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI'm passing `cutoff_time` into `Entity.head()` / `EntitySet.head()`, but the rows I get back still include records after that timestamp. I also hit weird behavior with `Variable.head(cutoff_time=...)`: sometimes it looks like the cutoff is ignored, and in one case it errored instead of returning the expected variable values.\n\n## Expected Outcomes\n\n- `Entity.head(n=..., cutoff_time=...)` and `EntitySet.head(entity_id, n=..., cutoff_time=...)` should apply the cutoff to the returned data before limiting to the requested number of rows.\n- Timestamp-style cutoffs should exclude rows at or after the cutoff time.\n- Tabular cutoffs supplied as a two-column pandas `DataFrame` should constrain results using the provided instance identifiers and their associated cutoff times.\n- `Variable.head(n=..., cutoff_time=...)` should return the requested variable as a single-column `DataFrame` while honoring the same cutoff behavior.\n- Invalid `cutoff_time` inputs should fail clearly with `ValueError` and the message `cutoff_time must be None, a Datetime, a pd.Timestamp, or a pd.DataFrame`.\n\n## Implementation Notes\n\nThe filtering location, helper structure, and internal data flow are implementation details. Preserve the existing public API shape for `Entity.head`, `EntitySet.head`, and `Variable.head`, and make the externally visible returned data and errors match the behaviors above."} {"task_id": "format-code-task-000371", "source_id": "format-code-task-000371", "domain": "code", "task_path": "tasks/format-code-task-000371", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:651201025c0403d4b1c5516985fc72c953ae2cab7aec1d4ce0146f551101d997", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nCannot initialize a ColumnSchema without initialized Ordinal logical type\nPR #870 made a change which allowed users to initialize a column schema object with an ordinal logical type, without first instantiating the ordinal logical type, as long as the column schema was not tied to data. This seems to have been broken with the 0.4.0 release of woodwork, and it is no longer possible to do this.\n\nThe ability to use a non-instantiated Ordinal logical type is needed/preferred for the Featuretools integration.\n\n#### Code Sample - Woodwork 0.3.1\n\n```python\n>>> from woodwork.column_schema import ColumnSchema\n>>> from woodwork.logical_types import Ordinal\n>>> ColumnSchema(logical_type=Ordinal)\n\n\n```\n\n#### Code Sample - Woodwork 0.4.0\n\n```python\n>>> from woodwork.column_schema import ColumnSchema\n>>> from woodwork.logical_types import Ordinal\n>>> ColumnSchema(logical_type=Ordinal)\nTraceback (most recent call last):\n File \"\", line 1, in \n File \"/Users/nate.parsons/dev/tmp/ordinal-test/env/lib/python3.8/site-packages/woodwork/column_schema.py\", line 35, in __init__\n logical_type = logical_type()\n File \"/Users/nate.parsons/dev/tmp/ordinal-test/env/lib/python3.8/site-packages/woodwork/logical_types.py\", line 369, in __init__\n raise TypeError(\"Must use an Ordinal instance with order values defined\")\nTypeError: Must use an Ordinal instance with order values defined\n\n```"} {"task_id": "format-code-task-000372", "source_id": "format-code-task-000372", "domain": "code", "task_path": "tasks/format-code-task-000372", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:31bba0dbd50e999c11ce9cc2bb49ee73733487de3f521e20b77b7c988ebe6a53", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n```\n>>> a = Const(1) - Const(2) \n>>> a.shape()\nunsigned(3)\n```\n\nThe correct result here (`-1`) is obviously not representable in an unsigned shape."} {"task_id": "format-code-task-000373", "source_id": "format-code-task-000373", "domain": "code", "task_path": "tasks/format-code-task-000373", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b1428d192c47aa8cb11b349ea57fc010b267a420770bf5a1ba912cdd71752da6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the browser SDK to expose a stateful `Identify` builder that I can construct with `new amplitude.Identify()` and use with fluent methods before passing it into identify-related SDK calls. The constructor takes no arguments, starts with no operations, and each method returns the same `Identify` instance so calls can be chained.\n\nThe builder should support `set(property, value)`, `setOnce(property, value)`, `add(property, value)`, `append(property, value)`, `prepend(property, value)`, `preInsert(property, value)`, `postInsert(property, value)`, `remove(property, value)`, `unset(property)`, and `clearAll()`. A session like `new amplitude.Identify().set('plan', 'pro').unset('legacy').add('credits', 3).setOnce('signup_complete', true)` should accumulate `$set: { plan: 'pro' }`, `$unset: { legacy: '-' }`, `$add: { credits: 3 }`, and `$setOnce: { signup_complete: true }` operation buckets. A session like `new amplitude.Identify().append('tests', ['a']).prepend('queue', 'first').preInsert('ids', 10).postInsert('ids2', 20).remove('old_ids', 4)` should accumulate `$append`, `$prepend`, `$preInsert`, `$postInsert`, and `$remove` buckets with those property/value pairs.\n\nOnly the first operation for a given property should be kept on a single `Identify` object: `new amplitude.Identify().setOnce('role', 'admin').add('role', 1).unset('role')` should keep only `$setOnce: { role: 'admin' }`. `add(property, value)` should accept numbers and strings, including numeric strings, but reject arrays and objects by logging an error, leaving the builder unchanged for that call, and still returning the same instance. `unset(property)` should store the property under `$unset` with the value `'-'`.\n\n`clearAll()` should accumulate exactly `$clearAll: '-'` when called on an otherwise empty builder. A builder that already contains `$clearAll` should ignore later property operations, and a builder that already contains any other operation should ignore a later `clearAll()` call. Multiple `clearAll()` calls on the same empty builder should remain idempotent and should not log warnings. Two separate `Identify` instances must keep independent operation state."} {"task_id": "format-code-task-000377", "source_id": "format-code-task-000377", "domain": "code", "task_path": "tasks/format-code-task-000377", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3488f7f3e9fb32e514bc82f7b9dcf26d73407d70aa8e35b4f3ee52c4d1aa1acc", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `WordCloud.recolor(random_state=None, color_func=None, colormap=None) -> WordCloud` so I can take a `WordCloud` instance that has already been generated and change only the colors of its existing words without recomputing placement or font sizes.\n\nA typical session should work like `wc = WordCloud(random_state=0).generate(\"alpha beta beta gamma gamma gamma\")`, then `same = wc.recolor(color_func=lambda **kwargs: \"rgb(255, 0, 0)\")`; the method should return the same `WordCloud` instance, and subsequent image, array, file, or SVG output from that instance should use the new red word colors while preserving the existing word positions, orientations, normalized frequencies, and font sizes. The color function should be called once for each placed word with the word text, font size, position, orientation, `font_path`, and the active random state.\n\nIf `color_func` is omitted and `colormap` is omitted, recoloring should use the instance's current color function. If `colormap` is provided and `color_func` is not, recoloring should draw replacement colors from that matplotlib colormap. Passing an integer `random_state` should make recoloring deterministic: calling `wc.recolor(random_state=10)` twice on the same generated layout should produce equal rendered output. Calling `WordCloud().recolor()` before any `generate` or `generate_from_frequencies` call should raise `ValueError` instead of creating a layout."} {"task_id": "format-code-task-000379", "source_id": "format-code-task-000379", "domain": "code", "task_path": "tasks/format-code-task-000379", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3afe070c45e5dfbf2389a598691d9486facce19c2d509c03215a3a516e44a885", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI am seeing `UserWarning: Now `main` is the default parent branch` even when using the default `unstaged` mode.\n\nFwiw this seems like kind of an annoying warning anyway? Fires once per worker also when using pytest-xdist."} {"task_id": "format-code-task-000380", "source_id": "format-code-task-000380", "domain": "code", "task_path": "tasks/format-code-task-000380", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:731de114537d01c84224de3b8e5e93aeb1d1efe1c18524aede0354b40a09734a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nArchive manifest filtering currently uses `internal/file.GlobMatch(pattern, name)`, but its wildcard behavior is byte-oriented and allows ordinary stars to consume directory separators. That makes archive paths behave differently from path-aware globbing elsewhere in Syft and makes patterns unexpectedly broad. Make this matcher path-aware while keeping its boolean API and existing simple matches compatible.\n\nMatching is case-sensitive and applies to the entire pattern and name. Do not clean or normalize either value: a leading slash, repeated slash, or `.` component remains significant. Ordinary characters match literally.\n\nUse these metacharacter rules:\n\n- `*` matches zero or more Unicode code points other than `/`; `?` matches exactly one Unicode code point other than `/`.\n- Exactly two stars used as a complete path component form a recursive `**` wildcard. It matches zero or more complete components, so `/usr/**/package.json` matches both `/usr/package.json` and deeper paths. A leading recursive component can match a root-level file, a trailing `/**` includes the base path itself and its descendants, and a bare `**` matches any whole path, including the empty path.\n- Star runs that are not exactly `**` as a complete component are non-recursive and behave like one ordinary `*`. Thus `ab**cd` stays within one component, and a complete `***` component matches one component rather than crossing separators.\n- Character classes match one non-separator Unicode code point. Support listed characters, inclusive ranges such as `[0-9]` and `[α-ω]`, and negation when either `!` or `^` is the first class character.\n- Backslash quoting makes `*`, `?`, `[`, `]`, and backslash itself literal rather than metacharacters.\n\nMalformed patterns must return `false` without panicking. This includes an unclosed or empty class, a descending range such as `[z-a]`, and a trailing escape. Preserve the existing behavior covered by the current `TestGlobMatch` table, including empty/exact matches and ordinary wildcard backtracking."} {"task_id": "format-code-task-000381", "source_id": "format-code-task-000381", "domain": "code", "task_path": "tasks/format-code-task-000381", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6bb76405ca1bd32110b205f66449dbf251ed328cb4b51a0652f1195cad74a12a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `trash-list` to support its normal listing mode for trashed entries. When I run `trash-list` with no action flag, or with one or more `--trash-dir DIR` options, it should read each selected trash directory's `info/*.trashinfo` records, decode the stored `Path=` value, combine it with the trash directory's volume, and print one line per entry to stdout as ` `. For example, a trashinfo record with `Path=/file1` and `DeletionDate=2000-01-01T00:00:01` in a trash directory on `/` should produce `2000-01-01 00:00:01 /file1` and exit 0.\n\nI also need the hidden listing modes to keep working: `trash-list --size` should print the backup copy size in bytes instead of the deletion date, and `trash-list --files` should append the backing trash file path as ` -> `. If a trashinfo file cannot be read, the IOError text should be written to stderr and listing should continue. If a trashinfo file cannot parse its `Path=` field, stderr should contain `Parse Error: : Unable to parse Path.` while the command continues and exits 0. When a top-level trash directory is skipped because its parent is not sticky or because its parent is a symlink, stderr should contain `TrashDir skipped because parent not sticky: ` or `TrashDir skipped because parent is symlink: ` respectively."} {"task_id": "format-code-task-000382", "source_id": "format-code-task-000382", "domain": "code", "task_path": "tasks/format-code-task-000382", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7399361eabb5d7d19d765e541c52eb40efd7a043e01d95c71bc05b4b392b466f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the installed command-line tools to include a `trash-rm PATTERN` command that permanently removes only trash entries whose original location matches the given shell-style pattern. When I run `trash-rm '*.o'`, it should scan the current user's home trash and volume trash directories, read each entry's `.trashinfo` metadata, match the pattern against the basename of the original path, and remove both the matching backup copy under `files/` and its matching `.trashinfo` file under `info/`; entries such as `/tmp/foo.o` and `/home/me/bar.o` should disappear, while `/tmp/foo.c` should remain. When I run `trash-rm /home/me/project/*.log`, the leading slash should make matching happen against the full reconstructed original location, so `/home/me/project/app.log` matches but `app.log` in another original directory does not.\n\nThe command should produce no stdout on a successful removal pass and should exit with status code 0, including the case where no trash entry matches the pattern. If no pattern is provided, `trash-rm` should print this usage message to stderr and exit with status code 8: `Usage:\\n trash-rm PATTERN\\n\\nPlease specify PATTERN.\\ntrash-rm uses fnmatch.fnmatchcase to match patterns, see https://docs.python.org/3/library/fnmatch.html for more details.` If a `.trashinfo` file cannot provide a parseable `Path` value while scanning, the command should report `trash-rm: : unable to parse 'Path'` on stderr, skip that malformed entry, and continue processing the rest."} {"task_id": "format-code-task-000383", "source_id": "format-code-task-000383", "domain": "code", "task_path": "tasks/format-code-task-000383", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1750d6e290c921ad85a7a469871f44151cb6760a802df0c05ad16000271aa1ea", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a configured FIX client instance to expose `subscribe_to_data(**kargs) -> None` and `unsubscribe_to_data(**kargs) -> None` for sending FIX 4.4 MarketDataRequest messages through the quote session. When I call `subscribe_to_data(**{'55': 'EUR/USD'})`, it should send exactly one message to the quote session with message type `MarketDataRequest` (`35=V`), a generated market-data request ID (`262`), subscription type snapshot plus updates (`263=1`), market depth defaulting to top of book (`264=1`), two market-data entry types (`267=2`) for bid and offer, full-refresh update mode by default (`265=0`), and one related symbol group for `55=EUR/USD`.\n\nWhen I call `unsubscribe_to_data(**{'55': 'EUR/USD'})`, it should send the same kind of MarketDataRequest to the quote session, but with subscription type disable previous snapshot plus updates (`263=2`). If I omit `55`, the symbol should default to `EUR/USD`; if I provide `264`, `267`, or `265`, those FIX tag options should be reflected in the outbound message. Passing a non-integer value for `264` or `267` should raise `ValueError` and should not send a message."} {"task_id": "format-code-task-000384", "source_id": "format-code-task-000384", "domain": "code", "task_path": "tasks/format-code-task-000384", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:49371a9b01742ddf2126d5e33e4cabc355db4bb9f4436e05b3f3f4dbf70b383d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Refactor websocket dial into its own package with extensible options\n\nRight now the websocket client lives in `pkg/conn` alongside the generic `Conn` interface and `RetryableError`, and the dial entrypoint looks like:\n\n```go\nconn.DialWebsocket(ctx, url, token)\n```\n\nA couple of things make this awkward:\n\n1. The websocket-specific code (dialer, retryable status codes, message-type checks, the `WebsocketConn` wrapper) is mixed into the same package as the transport-agnostic `Conn` interface. They're really two different concerns.\n\n2. `DialWebsocket` takes its configuration as positional arguments. Today it's just `token`, but as soon as we want to expose anything else for the dialer (TLS config, custom headers, timeouts, etc.) we'd have to either keep growing the positional parameter list or break every caller.\n\nI'd like to pull the websocket client out into its own subpackage under `pkg/conn`, and make the dial function take an extensible set of options instead of hard-coded positional args, so the bearer token is just the first such option and additional ones can be added later without churn.\n\nCallers are in `agent/endpoint.go` (client side, currently passes `e.conf.Auth.APIKey`) and `server/server/upstream/server.go` (server side, just wraps an already-upgraded `*websocket.Conn`) — both should be updated to the new package.\n\nBehaviour of the connection itself (binary-only messages, retryable status code handling, wrapping the underlying `*websocket.Conn`) should stay the same; this is purely a packaging + API-shape change."} {"task_id": "format-code-task-000385", "source_id": "format-code-task-000385", "domain": "code", "task_path": "tasks/format-code-task-000385", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2684dd35d2427cc1a8553cfe2ccd1a5b55c0eef74213c7eb5b3aa5556f4249f0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Rendering some Jira issue descriptions crashes or produces broken markdown\n\nI'm using jira-cli to view tickets in my terminal (`jira issue view ...`). For most issues this works fine, but a handful of tickets either crash the CLI or render with the entire bottom half of the description swallowed into one giant code block.\n\nAfter narrowing it down, both cases come from Jira wiki code blocks in the description. I can reproduce by feeding the relevant snippets through `jirawiki.Parse`.\n\n**Case 1 — crash on a code block with a stray colon in the opening tag**\n\nSome tickets in our project have descriptions that look roughly like:\n\n```\n{code:}\ndo_the_thing();\n{code}\n```\n\n(I've also seen variants where the part after the colon is something the original author probably didn't intend to be parsed as an attribute.)\n\nRunning `jira issue view` on these tickets exits immediately with a runtime error instead of printing the ticket — so I can't read the description at all from the CLI.\n\n**Case 2 — closing tag on the same line as content eats the rest of the description**\n\nThis one doesn't crash, it just renders wrong. Input:\n\n```\n{code}\nsome line of code\nfinal line of code{code}\n\nThen some more description text below.\nAnother paragraph.\n```\n\nThe `{code}` closer is glued to the end of the last code line (this happens a lot when people paste code into Jira's editor and the cursor ends up right before the closer). Expected output is a fenced code block containing the two code lines, followed by the prose below it. What I actually get is a fenced code block that never closes — every line after `{code}` (including the prose paragraphs) becomes part of the code block, and the trailing ``` is misplaced.\n\nBoth feel like they should just work — the descriptions render fine in Jira's web UI, and the CLI shouldn't be picky about whether the closing tag has its own line. And it definitely shouldn't crash on an input that Jira itself accepts."} {"task_id": "format-code-task-000386", "source_id": "format-code-task-000386", "domain": "code", "task_path": "tasks/format-code-task-000386", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b981d909fe3ab852c250d671c3db66ac90cff5d1be7498274f52c5a541d6ebec", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\n我们用 fxhttpclient 发了不少外部请求，现在 `/metrics` 上能看到 server 端的指标，但 client 这边啥都没有，没法做请求量、耗时、错误率的监控和告警。能不能也给 client 加一套 Prometheus 指标，至少有请求数和耗时分布，带上 method/status/url 这些 label？最好能在 config 里开关，status 也能按 2xx/3xx 这种聚合一下，不然 url 一多基数就爆了。\n\n# Expected outcomes\n\n- Configuration:\n - The HTTP client module accepts a `modules.http.client.metrics` configuration block.\n - Metrics collection can be enabled or disabled through `modules.http.client.metrics.collect.enabled`.\n - The metrics namespace and subsystem can be configured through `modules.http.client.metrics.collect.namespace` and `modules.http.client.metrics.collect.subsystem`, with sensible defaults based on the application name and the HTTP client subsystem.\n - Request-duration histogram buckets can be overridden through `modules.http.client.metrics.buckets`.\n - Status label normalization can be enabled through `modules.http.client.metrics.normalize`.\n\n- Emitted metrics:\n - When client metrics collection is enabled, each outgoing HTTP client request increments a Prometheus counter named with the configured namespace/subsystem prefix and the `client_requests_total` suffix.\n - The request counter exposes `method`, `status`, and `url` labels.\n - When status normalization is enabled, the counter’s `status` label uses HTTP status classes such as `2xx`; when it is disabled, the label remains specific to the response status.\n - When client metrics collection is enabled, outgoing HTTP client requests are observed in a Prometheus histogram named with the configured namespace/subsystem prefix and the `client_request_duration_seconds` suffix.\n - The request-duration histogram exposes `method` and `url` labels and includes bucket, sum, and count series.\n\n- Documentation and integration:\n - The fxhttpclient documentation describes the client metrics configuration and shows the expected metrics behavior.\n - The fxhttpclient module documentation and examples mention the metrics module integration needed to expose these metrics.\n\n# Implementation notes\n\nThe concrete instrumentation mechanism, data structures, registration location, and helper functions are implementation details. Preserve the existing HTTP client behavior for logging, tracing, transport configuration, decoration, and request execution while adding the client-side metrics behavior described above."} {"task_id": "format-code-task-000388", "source_id": "format-code-task-000388", "domain": "code", "task_path": "tasks/format-code-task-000388", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e1508fca61ceac08ad79417cea3fcf2a9b94e7437aa44bfed8cb6b4480401742", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Feature request: allow the `community.general.sudoers` module to produce rules that preserve environment variables\n\nI'm using `community.general.sudoers` to manage `/etc/sudoers.d` entries on my hosts. One of my users (let's call her `alice`) needs to run a small upload script via sudo:\n\n```yaml\n- name: Allow alice to sudo /usr/local/bin/upload\n community.general.sudoers:\n name: allow-alice-upload\n user: alice\n commands: /usr/local/bin/upload\n```\n\nThe script she's running depends on a couple of environment variables that are configured in her shell (things like `AWS_PROFILE`, `HTTP_PROXY`, and a custom `PATH` addition). When she runs the command through sudo, those variables get stripped, so the script fails because the config it expects isn't there.\n\nIf I were hand-writing the sudoers file I would just slap the `SETENV` tag on the rule, e.g.:\n\n```\nalice ALL=(ALL) SETENV: /usr/local/bin/upload\n```\n\nThat way she can invoke it with `sudo -E /usr/local/bin/upload` (or `sudo FOO=bar /usr/local/bin/upload`) and the environment is preserved. This is a pretty standard sudoers feature.\n\nAs far as I can tell, the `sudoers` module doesn't currently expose any way to emit this tag in the generated rule. There's already an option that controls whether `NOPASSWD:` is included, but nothing equivalent for `SETENV:`, so the only workaround is to stop using the module for these rules and drop a hand-written file into `/etc/sudoers.d`, which defeats the purpose.\n\nCould the module be extended to optionally include `SETENV:` in the rule it writes? Existing playbooks should keep behaving the same by default — I'd only want this turned on for the rules where I explicitly ask for it."} {"task_id": "format-code-task-000389", "source_id": "format-code-task-000389", "domain": "code", "task_path": "tasks/format-code-task-000389", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2adc782e024e62b23114c74f38f354e85e9ec09f8e4cba6f49a90eb56fd8e05c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nRule can't contain an address and netmask if the connection-type is 'local' with 3.9.0\n\n\n\n\n##### SUMMARY\n\n\nWe got the following snippet in our playbooks:\n\n```\n- name: \"Enable datata access to its PostgreSQL DB\"\n become: true\n become_user: postgres\n community.postgresql.postgresql_pg_hba:\n dest: \"/var/lib/pgsql/data/pg_hba.conf\"\n contype: local\n databases: datadata\n users: datata\n notify: Restart PostgreSQL\n```\n\nThat is working flawless since quite some time. With the release of 3.9.0 things started failing with the following:\n\n```\n{\n \"msg\": \"Error modifying rules:\\nRule can't contain an address and netmask if the connection-type is 'local'\",\n \"invocation\": {\n \"module_args\": {\n \"dest\": \"/var/lib/pgsql/data/pg_hba.conf\",\n \"contype\": \"local\",\n \"databases\": \"datadata\",\n \"users\": \"datadata\",\n \"method\": \"md5\",\n \"address\": \"samehost\",\n \"backup\": false,\n \"create\": false,\n \"keep_comments_at_rules\": false,\n \"state\": \"present\",\n \"rules_behavior\": \"conflict\",\n \"overwrite\": false,\n \"unsafe_writes\": false,\n \"backup_file\": null,\n \"comment\": null,\n \"netmask\": null,\n \"options\": null,\n \"rules\": null,\n \"mode\": null,\n \"owner\": null,\n \"group\": null,\n \"seuser\": null,\n \"serole\": null,\n \"selevel\": null,\n \"setype\": null,\n \"attributes\": null\n }\n },\n \"_ansible_no_log\": false,\n \"changed\": false\n}\n```\n:top: this is a json from an AWX job.\nPinning the release before that, 3.8.0, gives the expected result.\n\nTrying to pinpoint the issue I've stumbled across https://github.com/ansible-collections/community.postgresql/pull/772 that did some larger refactoring to the handling of parameters, including the `address` that should be omitted if `contype` is set to `local`.\nWith 3.9.0 the `address` parameter is set to `samehost` if nothing is passed no matter if contype is set to local.\nSetting the address to `null`, `~` or leaving it empty results in a `argument 'address' is of type and we were unable to convert to str: 'None' is not a string and conversion is not allowed`.\n\n\n##### ISSUE TYPE\n- Bug Report\n\n##### COMPONENT NAME\n\npostgresql.postgresql_pg_hba in Version 3.9.0\n\n##### ANSIBLE VERSION\n\n\nAWX 21.12.0\nquay.io/ansible/awx-ee:latest, sha256:7dc75b8723ee9f63ce716fa46f3427895dce756834deeef1d9986af084978c4b\n\n```paste below\n\n```\n\n##### COLLECTION VERSION\n\nAWX output for collection install:\n```paste below\nStarting galaxy collection install process\nProcess install dependency map\nStarting collection install process\nDownloading https://galaxy.ansible.com/api/v3/plugin/ansible/content/published/collections/artifacts/community-rabbitmq-1.3.0.tar.gz to /var/lib/awx/projects/.__awx_cache/_23__extrusionos_via_ansible/stage/tmp/ansible-local-5128215awgqocb/tmponshxnzu/community-rabbitmq-1.3.0-b6rwufqd\nInstalling 'community.rabbitmq:1.3.0' to '/var/lib/awx/projects/.__awx_cache/_23__extrusionos_via_ansible/stage/requirements_collections/ansible_collections/community/rabbitmq'\ncommunity.rabbitmq:1.3.0 was installed successfully\nDownloading https://galaxy.ansible.com/api/v3/plugin/ansible/content/published/collections/artifacts/lvrfrc87-git_acp-2.2.0.tar.gz to /var/lib/awx/projects/.__awx_cache/_23__extrusionos_via_ansible/stage/tmp/ansible-local-5128215awgqocb/tm"} {"task_id": "format-code-task-000390", "source_id": "format-code-task-000390", "domain": "code", "task_path": "tasks/format-code-task-000390", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7f1fdcd1ee21e4912248d85058a72aef449ab406186e9aa9bbcb34bea10ca674", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nrun_job_template removes relative path from controller URL\n### Please confirm the following\n\n- [X] I agree to follow this project's [code of conduct](https://docs.ansible.com/ansible/latest/community/code_of_conduct.html).\n- [X] I have checked the [current issues](https://github.com/ansible/ansible-rulebook/issues) for duplicates.\n- [X] I understand that ansible-rulebook is open source software provided for free and that I might not receive a timely response.\n\n### Bug Summary\n\nWhen specifying a URL with a relative path to the controller (AWX), the API call is sent to the absolute path /api/v2.\n\nConfigured eda.yml:\n\n```\n$ cat eda.yaml \n# eda.yaml\napiVersion: eda.ansible.com/v1alpha1\nkind: EDA\nmetadata:\n name: my-eda\nspec:\n automation_server_url: http://awx-demo-service.awx/awx/\n service_type: ClusterIP\n ingress_type: ingress\n```\n\nHere's the output of the rulebook activation:\n\n```\n2024-01-19 23:32:13,505 - ansible_rulebook.websocket - INFO - websocket ws://my-eda-daphne:8001/api/eda/ws/ansible-rulebook connected\n2024-01-19 23:32:13,552 - ansible_rulebook.job_template_runner - INFO - Attempting to connect to Controller http://awx-demo-service.awx/awx\n2024-01-19 23:32:13,553 - ansible_rulebook.job_template_runner - ERROR - Error connecting to controller 404, message='Not Found', url=URL('http://awx-demo-service.awx/api/v2/config/')\n2024-01-19 23:32:13,554 - ansible_rulebook.cli - ERROR - Terminating 404, message='Not Found', url=URL('http://awx-demo-service.awx/api/v2/config/') \n```\n\nAfter removing the leading \"/\" from UNIFIED_TEMPLATE_SLUG and CONFIG_SLUG. It started to work locally for me.\n\n\n\n### Environment\n\neda-server-operator 2.10.0 with quay.io/ansible/ansible-rulebook:v1.0.4\n\n### Steps to reproduce\n\nYou'll need AWX installed on a different path than '/`.\n\nI installed ansible-rulebook from brew (OSX).\n\nExport the following variable (CLI options or documentation didn't work).\n\n```\nexport EDA_CONTROLLER_URL=\"http://localhost:8081/awx/\"\nexport EDA_CONTROLLER_TOKEN=\"ZesMMFepb2WYfHNEfvv1pMz0PMbUkE\"\n```\n\nStart a simple rulebook using job_run_template.\nVerify it doesn't connect.\n\nMy output looked like this on ansible-rulebook ran locally:\n\n```\n$ ansible-rulebook -r rulebooks/hello.yaml -vv \n2024-01-19 20:02:55,303 - asyncio - DEBUG - Using selector: KqueueSelector\n2024-01-19 20:02:55,304 - ansible_rulebook.app - DEBUG - Loading rules from the file system rulebooks/hello.yaml\n2024-01-19 20:02:55,309 - ansible_rulebook.condition_parser - DEBUG - [Identifier(value='event.i'), '==', Integer(value=1)]\n2024-01-19 20:02:55,309 - ansible_rulebook.job_template_runner - INFO - Attempting to connect to Controller http://localhost:8081/awx/\n2024-01-19 20:02:55,317 - ansible_rulebook.cli - ERROR - Terminating Expecting value: line 1 column 1 (char 0)\n```\n\nOnce I remove the trailing slash (there is a different error):\n\n```\n$ ansible-rulebook -r rulebooks/hello.yaml -vv \n2024-01-19 20:16:46,145 - asyncio - DEBUG - Using selector: KqueueSelector\n2024-01-19 20:16:46,145 - ansible_rulebook.app - DEBUG - Loading rules from the file system rulebooks/hello.yaml\n2024-01-19 20:16:46,150 - ansible_rulebook.condition_parser - DEBUG - [Identifier(value='event.i'), '==', Integer(value=1)]\n2024-01-19 20:16:46,151 - ansible_rulebook.job_template_runner - INFO - Attempting to connect to Controller http://localhost:8081/awx/\n2024-01-19 20:16:46,259 - ansible_rulebook.app - INFO - AAP Version 23.6.0\n2024-01-19 20:16:46,259 - ansible_rulebook.app - INFO - Starting sources\n2024-01-19 20:16:46,259 - ansible_rulebook.app - INFO - Starting rules\n2024-01-19 20:16:46,259 - ansible_rulebook.engine - INFO - run_ruleset\n2024-01-19 20:16:46,260 - drools.ruleset - INFO - Using jar: /opt/homebrew/lib/python3.9/site-packages/drools/jars/drools-ansible-rulebook-integration-runtime-1.0.5-SNAPSHOT.jar\n2024-01-19 20:16:46,508 - drools.ruleset - DEBUG - Creating Drools Ruleset\n2024-01-19 20:16:46 801 [main] INFO org.drools.ansible.rulebook.integration."} {"task_id": "format-code-task-000391", "source_id": "format-code-task-000391", "domain": "code", "task_path": "tasks/format-code-task-000391", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:091d283ace840c30358b919341ce73c47105543d52f4dda7862a9a7fc004df30", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `credential` module gives a confusing error when `credential_type` is omitted\n\nI'm using the `ansible.controller.credential` module (from the awx collection) to manage credentials against an AWX 23.4.x controller. I had a typo in my playbook and forgot to specify `credential_type`. Instead of telling me the parameter was missing, the module failed with an unrelated-looking error from somewhere deeper in the module.\n\nRoughly what I had:\n\n```yaml\n- name: Add machine credential\n ansible.controller.credential:\n name: my-cred\n organization: Default\n # credential_type accidentally omitted\n state: present\n inputs:\n username: joe\n password: secret\n```\n\nWhen I run this, the task fails, but the error message doesn't make it obvious that the problem is just a missing parameter — it looks like something went wrong internally. It took me a while to realize I'd simply forgotten `credential_type`.\n\nLooking at the module docs, `credential_type` doesn't seem to be marked as required, even though in practice you basically always need it to create/look up a credential. It would be much friendlier if the module treated it as a required parameter and failed up front with a clear \"missing required argument\" message, the same way `name` already does."} {"task_id": "format-code-task-000392", "source_id": "format-code-task-000392", "domain": "code", "task_path": "tasks/format-code-task-000392", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:13a2e0145e46d9a12b4705062ee0a71bf664609ce1832209a59ef1c530741252", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## OpenStack credential is missing a region field\n\nI'm using AWX to run Ansible playbooks against an OpenStack cloud (OVH in my case). With OVH — and a few other providers — you have to tell the OpenStack SDK which region to talk to, otherwise authentication / API calls just don't go anywhere useful.\n\nWhen I configure an \"OpenStack\" credential in AWX, the form lets me fill in:\n\n- Username\n- Password (API Key)\n- Host (Authentication URL)\n- Project (Tenant Name)\n- Project (Domain Name)\n- Domain Name\n- Verify SSL\n\n…but there's no field for the region. The `clouds.yaml` that AWX generates from this credential ends up looking like:\n\n```yaml\nclouds:\n devstack:\n auth:\n auth_url: ...\n username: ...\n password: ...\n project_name: ...\n verify: true\n```\n\ni.e. no `region_name` anywhere. For providers like OVH this isn't enough to actually use the cloud — the upstream `ansible-collections-openstack` inventory config (https://github.com/openstack/ansible-collections-openstack/blob/8255ec4c80f186aa7851f023b85007a593b6f42f/scripts/inventory/openstack.yml#L14) clearly expects region to be settable, and a plain `openstack.yaml` written by hand can include it just fine. AWX just doesn't expose it.\n\nIt would be great if the built-in OpenStack credential type had a region field so that:\n\n1. Users can fill it in from the credential form in the UI.\n2. When AWX renders the cloud config file used by the OpenStack modules / inventory, the region the user entered actually shows up there, so playbooks targeting OVH-style providers work without manual workarounds.\n\nLeaving it empty should keep behaving like today (region simply not set), so existing OpenStack credentials that don't need a region aren't affected.\n\n(For consistency with how other project-scoped fields are named on this credential, the new input id would be something like `project_region_name`.)"} {"task_id": "format-code-task-000393", "source_id": "format-code-task-000393", "domain": "code", "task_path": "tasks/format-code-task-000393", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7e113ad677f35c3a45ac5a3b4f21cd2b505c54509dbd87b37d5cad484ce0550d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nDouble-nested ansible_collections folders cause errors\n# Issue Type\n\n- Bug report\n\n# Molecule and Ansible details\n\nansible - 3.1.0\nansible-base - 2.10.6\nmolecule - 3.2.3\n\nMolecule installation method (one of):\n\n- pip\n\nAnsible installation method (one of):\n\n- pip\n\n# Desired Behavior\n\nBecause of some weirdness I'm running my tests under a double folder structure that goes something like\n\n/home/user/src/ansible_collections/some/more/paths/ansible_collections/namespace/collection/roles\n\nWhen living there, I should still have no issues with running Molecule\n\n# Actual Behaviour\n\nErrors arise from line 424 in molecule/provisioner/ansible.py which assumes that collection_indicator occurs exactly one time in the path structure.\n\nResulting output:\n\n```console\n$ tox -e roles-test_role-docker\nroles-test_role-docker installed: ansible==3.1.0,ansible-base==2.10.5,ansible-lint==5.0.3,appdirs==1.4.4,arrow==1.0.3,attrs==20.3.0,bcrypt==3.2.0,binaryornot==0.4.4,boto==2.49.0,boto3==1.17.25,botocore==1.20.25,bracex==2.1.1,Cerberus==1.3.2,certifi==2020.12.5,cffi==1.14.5,chardet==4.0.0,click==7.1.2,click-completion==0.5.2,click-help-colors==0.9,colorama==0.4.4,commonmark==0.9.1,cookiecutter==1.7.2,cryptography==3.4.6,decorator==4.4.2,distro==1.5.0,docker==4.4.4,dogpile.cache==1.1.2,enrich==1.2.6,flake8==3.8.4,idna==2.10,iniconfig==1.1.1,iso8601==0.1.14,Jinja2==2.11.3,jinja2-time==0.2.0,jmespath==0.10.0,jsonpatch==1.31,jsonpointer==2.0,keystoneauth1==4.3.1,MarkupSafe==1.1.1,mccabe==0.6.1,molecule==3.2.2,molecule-containers==0.2.1,molecule-docker==0.2.4,molecule-ec2==0.3,molecule-openstack==0.3,molecule-podman==0.3.0,molecule-vagrant==0.6.1,munch==2.5.0,netifaces==0.10.9,openstacksdk==0.54.0,os-client-config==2.1.0,os-service-types==1.7.0,packaging==20.9,paramiko==2.7.2,pathspec==0.8.1,pbr==5.5.1,pluggy==0.13.1,poyo==0.5.0,py==1.10.0,pycodestyle==2.6.0,pycparser==2.20,pyflakes==2.2.0,Pygments==2.8.1,PyNaCl==1.4.0,pyparsing==2.4.7,pytest==6.2.2,pytest-testinfra==6.1.0,python-dateutil==2.8.1,python-slugify==4.0.1,PyYAML==5.4.1,requests==2.25.1,requestsexceptions==1.4.0,rich==9.13.0,ruamel.yaml==0.16.13,ruamel.yaml.clib==0.2.2,s3transfer==0.3.4,selinux==0.2.1,shellingham==1.4.0,six==1.15.0,stevedore==3.3.0,subprocess-tee==0.2.0,testinfra==6.0.0,text-unidecode==1.3,toml==0.10.2,typing-extensions==3.7.4.3,urllib3==1.26.3,wcmatch==8.1.2,websocket-client==0.58.0,yamllint==1.26.0\nroles-test_role-docker run-test-pre: PYTHONHASHSEED='4259097906'\nroles-test_role-docker run-test: commands[0] | molecule --debug -c /var/home/ghelling/src/ansible_collections/meta_ansible_templates/ansible_collections/test_ns/test_collection/tests/molecule.yml test -s docker\nDEBUG Validating schema /var/home/ghelling/src/ansible_collections/meta_ansible_templates/ansible_collections/test_ns/test_collection/roles/test_role/molecule/docker/molecule.yml.\nINFO docker scenario test matrix: dependency, lint, cleanup, destroy, syntax, create, prepare, converge, idempotence, side_effect, verify, cleanup, destroy\nINFO Running docker > dependency\nDEBUG: ANSIBLE ENVIRONMENT:\nANSIBLE_COLLECTIONS_PATH: /var/home/ghelling/src/:/var/home/ghelling/.ansible\nANSIBLE_COLLECTIONS_PATHS: /var/home/ghelling/src/ansible_collections/meta_ansible_templates/ansible_collections/test_ns/test_collection/../../../\nANSIBLE_FORCE_COLOR: '1'\n\nDEBUG: MOLECULE ENVIRONMENT:\nMOLECULE_DEBUG: 'True'\nMOLECULE_DEPENDENCY_NAME: galaxy\nMOLECULE_DRIVER_NAME: docker\nMOLECULE_ENV_FILE: /var/home/ghelling/src/ansible_collections/meta_ansible_templates/ansible_collections/test_ns/test_collection/roles/test_role/.env.yml\nMOLECULE_EPHEMERAL_DIRECTORY: /var/home/ghelling/.cache/molecule/test_role/docker\nMOLECULE_FILE: /var/home/ghelling/.cache/molecule/test_role/docker/molecule.yml\nMOLECULE_INSTANCE_CONFIG: /var/home/ghelling/.cache/molecule/test_role/docker/instance_config.yml\nMOLECULE_INVENTORY_FILE: /var/home/ghelling/.cache/molecule/test_role/docker/inventory/ansible_inventory.yml\nMOLECULE_PRO"} {"task_id": "format-code-task-000394", "source_id": "format-code-task-000394", "domain": "code", "task_path": "tasks/format-code-task-000394", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1373b6549fce44690ae7a691da85908718f8361a630f11b1c1067ec1d735b93d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Molecule workflow commands to let me exclude one or more scenarios from a run. For commands such as `molecule test`, `molecule create`, `molecule converge`, `molecule check`, `molecule destroy`, `molecule dependency`, `molecule cleanup`, `molecule prepare`, `molecule idempotence`, `molecule side-effect`, `molecule syntax`, `molecule verify`, and `molecule reset`, users should be able to pass `--exclude NAME` or `-e NAME` multiple times alongside the existing scenario targeting options.\n\nWhen I run a workflow command with `--all --exclude skipped`, Molecule should discover all matching scenarios but not execute the scenario named `skipped`; if every other scenario succeeds, the command should exit 0 and the excluded scenario should produce no action output or report entry. When I target explicit scenarios, for example `molecule test -s default -s experimental --exclude experimental`, only `default` should run and the command should exit 0 if that remaining scenario succeeds. Exclude values should also accept glob-style patterns, so `molecule test --all --exclude 'db_*'` skips scenarios whose names match `db_*`, and nested scenario names such as `appliance_vlans/merged` can be skipped by exact name or by patterns like `appliance_*/*`.\n\nIf an excluded scenario name does not exist, Molecule should not fail just because of that exclusion; it should simply run the non-excluded selected scenarios. If all explicitly targeted scenarios are excluded, the command should not execute those scenarios, and it should behave consistently with an empty resulting selection rather than reporting the excluded names as missing. The option should appear in the workflow command help as a repeatable scenario exclusion flag using both `--exclude` and `-e`."} {"task_id": "format-code-task-000395", "source_id": "format-code-task-000395", "domain": "code", "task_path": "tasks/format-code-task-000395", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:046212cf0eb2f32988e0bfb3fe09e83c2c35eac8ae2d8d641ad67496e4058e89", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the wxPython app bundle to include a `RunScript.py` command that can be placed in an Xcode Run Script build phase and run as `python RunScript.py`. The command should read Xcode's build environment variables: `EFFECTIVE_PLATFORM_NAME`, `DWARF_DSYM_FOLDER_PATH`, `BUILD_STYLE`, `DWARF_DSYM_FILE_NAME`, `EXECUTABLE_NAME`, and `PROJECT_DIR`. On every run it should print the effective platform name as the first stdout line and the dSYM folder path as the second stdout line.\n\nFor `BUILD_STYLE=Debug`, it should print `跳过debug模式`, exit with code 0, and not move or register any dSYM. For `EFFECTIVE_PLATFORM_NAME=-iphonesimulator`, it should print `跳过模拟器编译情况`, exit with code 0, and not move or register any dSYM. For an eligible device release build, it should move the generated dSYM bundle from `${DWARF_DSYM_FOLDER_PATH}/${DWARF_DSYM_FILE_NAME}` into `${PROJECT_DIR}/dSYM/${EXECUTABLE_NAME}.${YYYYmmddHHMMSS}.app.dSYM`, where the timestamp uses the current local time with year, month, day, hour, minute, and second. Before moving, stdout should include `moving to `.\n\nAfter the move succeeds and the destination exists, the command should open `/Applications/DSYMTools.app/Contents/Resources/dsym.db`, create `archives (file_path varchar(10) unique)` if needed, insert the moved destination path into that table, commit the transaction, and exit with code 0. If a required environment variable is missing or the source dSYM cannot be moved, the command should fail non-zero and should not silently insert a registry row. Packaging with `setup.py py2app` should include `RunScript.py` in the app resources so users can install or copy that build-phase command from the bundled app."} {"task_id": "format-code-task-000396", "source_id": "format-code-task-000396", "domain": "code", "task_path": "tasks/format-code-task-000396", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5059752083776c2a141064acb863020fd40cba4265af02344547d63dd018c513", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nPlacement props not working in DatePicker and TimePicker\n### Reproduction link\n\n[https://ant.design/components/date-picker#components-date-picker-demo-placement](https://ant.design/components/date-picker#components-date-picker-demo-placement)\n\n### Steps to reproduce\n\njust click on placement and see if it affecting the datepicker\n\n### What is expected?\n\nif we select topRight then datePicker should be reflected as \"topRight\" side rather than bottomLeft \n\n### What is actually happening?\n\nit is always placed at bottomLeft\n\n| Environment | Info |\n| --- | --- |\n| antd | 5.15.3 |\n| React | 18 |\n| System | Ubuntu 20 |\n| Browser | Chrome |\n\n"} {"task_id": "format-code-task-000398", "source_id": "format-code-task-000398", "domain": "code", "task_path": "tasks/format-code-task-000398", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:94058d164512e5736f9d2f660f95fe96caa5055f408b308ff7e2202b2cd2da07", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nHey, I'm running Antrea with the Flow Aggregator and I'd really like a stable way to tag the flow records with a unique cluster ID so I can tell flows from different clusters apart downstream. Right now the shipped manifests do not provide a persistent cluster identity that the Flow Aggregator and other Antrea components can discover and use out of the box. Would be great if this worked across all supported install variants without me having to hand-roll a ConfigMap and ClusterRole myself.\n\n## Expected Outcomes\n\n- Cluster identity storage: all shipped Antrea install manifests expose an initially empty cluster identity `ConfigMap` named `antrea-cluster-identity` in namespace `kube-system`, labeled consistently with Antrea resources.\n- Shared read access: the manifests provide a reusable `ClusterRole` named `antrea-cluster-identity-reader` whose permissions are limited to reading the cluster identity `ConfigMap`.\n- Controller persistence access: shipped RBAC allows the Antrea Controller to read and update the cluster identity `ConfigMap` while preserving its existing access to related Antrea `ConfigMap` resources.\n- Flow Aggregator consumption access: shipped Flow Aggregator RBAC allows the Flow Aggregator ServiceAccount to read the persisted cluster identity through the shared reader role.\n- Kustomize base propagation: the base manifest resources include the shared cluster identity reader role so downstream generated install manifests can inherit it.\n\n## Implementation Notes\n\n- Keep the solution focused on externally observable Kubernetes resources and RBAC behavior; the exact manifest organization and generation workflow may follow existing repository conventions.\n- RBAC should remain scoped to the cluster identity resource needed by consumers rather than granting broad ConfigMap permissions.\n- The cluster identity storage should be reusable by components beyond the Flow Aggregator without requiring users to create custom manifests manually."} {"task_id": "format-code-task-000399", "source_id": "format-code-task-000399", "domain": "code", "task_path": "tasks/format-code-task-000399", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:035116aa9b0cb53148646519cbdaed519b68f5e1aecc63468987486fd8f62072", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nPanic is found in log after agent successful initialization [feature/externalnode branch]\n**Describe the bug**\nFound panic in log after agent initialization for feature/externalnode branch. It doesn't impact the agent running, but we will see the panic in log.\n`I0610 07:12:23.519375 14587 agent.go:447] Agent initialized NodeConfig=NodeName: mengdie-k8s0-0, OVSBridge: br-int, PodIPv4CIDR: , PodIPv6CIDR: , NodeIPv4: , NodeIPv6: , TransportIPv4: , TransportIPv6: , Gateway: , NetworkConfig=&{noEncap %!v(PANIC=String method: runtime error: index out of range [-1]) {%!v(PANIC=String method: runtime error: index out of range [-1]) } [] true false}`\n\nIt seems that the issue is caused by empty value for trafficEncryptionMode.\nIn this case, it will return TrafficEncryptionModeInvalid which is -1 since we do not specify default value for it.\n`const (\n\tTrafficEncryptionModeNone TrafficEncryptionModeType = iota\n\tTrafficEncryptionModeIPSec\n\tTrafficEncryptionModeWireGuard\n\tTrafficEncryptionModeInvalid = -1\n)`\n\n**To Reproduce**\nRun antrea-agent in ExternalNode nodeType\n\n**Expected**\nThere should be no PANIC in logs since agent initialized successfully."} {"task_id": "format-code-task-000400", "source_id": "format-code-task-000400", "domain": "code", "task_path": "tasks/format-code-task-000400", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a1c1ca3e8374e4ae8505994025b058cf9f55a4fa4e9fe12659e74e54b91167d5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Share the resize-hotspot construction logic across header cells\n\nEvery header cell (row, column, corner, frozen column…) draws an invisible \"resize\nhotspot\" — a thin rectangle the user can grab to drag-resize a row height or a column\nwidth. Today each cell type builds that rectangle's canvas attributes by hand, copying\nthe same block of literals (fill, opacity, cursor, the `appendInfo` metadata the\ninteraction layer reads back when a drag starts). The copies have already drifted apart,\nwhich makes resize bugs easy to introduce and hard to fix in one place.\n\nPlease extract this into a small, reusable interaction utility so that building a resize\nhotspot — and getting the foreground group the hotspots live in — is done the same way\neverywhere.\n\nExpose it as a resize interaction utility module importable at\n`@/utils/interaction/resize`, providing two named functions: `getResizeAreaAttrs` (the\nhotspot attribute builder) and `getResizeAreaGroupById` (the group lookup). How you\nstructure the internals is up to you.\n\n## What to provide\n\n**1. Build the hotspot rectangle attributes from a single config.**\n\nGiven a config describing one hotspot, produce the attribute object that gets handed to\nthe canvas `addShape('rect', { attrs })` call. The config carries:\n\n- `type` — `'row'` for a row-height hotspot, `'col'` for a column-width hotspot.\n- `effect` — what the drag changes: `'field'`, `'cell'` or `'tree'`.\n- `theme` — the resize-area style, with at least `size`, `background` and\n `backgroundOpacity`.\n- `offsetX`, `offsetY`, `width`, `height` — the geometry of the *cell* the hotspot\n belongs to (used later to compute the drag delta and guide-line position).\n- optional `id` (the field id) and `caption` (the dimension value).\n\nThe returned attributes must be:\n\n- `fill` set to the theme background, and `fillOpacity` to the theme background opacity.\n- `cursor` set to `\"-resize\"` (so `'col-resize'` / `'row-resize'`).\n- `width` and `height` describing only the *thickness* of the hotspot, leaving the other\n dimension for the caller to fill in: a `'col'` hotspot is a vertical bar, so its\n `width` is the theme `size` and its `height` is `null`; a `'row'` hotspot is a\n horizontal bar, so its `height` is the theme `size` and its `width` is `null`.\n- an `appendInfo` object that the interaction layer can later read off the shape to drive\n the resize. It must flag the shape as a resize hotspot, and carry the `type`, `effect`,\n `id`, `caption` (when given), `offsetX`, `offsetY`, and the cell `width`/`height` from\n the config. The `theme` must not be copied into `appendInfo`.\n\nNote the deliberate split: the top-level `width`/`height` are the rendered bar's\nthickness, while the `width`/`height` inside `appendInfo` are the original cell\ndimensions.\n\n**2. Get (or lazily create) a named resize-area group.**\n\nResize hotspots of a kind share one group under the sheet's foreground group, looked up\nby a fixed id. Provide a helper that, given the sheet and a group id, returns that group:\nthe already-existing child group with that id if there is one, otherwise a freshly added\ngroup registered under that id. Calling it repeatedly with the same id must never create\nduplicate groups. If the sheet has no foreground group yet, it should return nothing\nrather than throw.\n"} {"task_id": "format-code-task-000401", "source_id": "format-code-task-000401", "domain": "code", "task_path": "tasks/format-code-task-000401", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a0663ff1ba4e215f9e4c727e89589258d8963ffd623d2f36d13b45cbbefea3c8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want calendar charts created with `anychart.calendar(opt_data, opt_csvSettings)` to expose a calendar-specific data view through `chart.data()` that keeps only valid calendar points. A row is valid when its `x` field can be parsed as a date/time by AnyChart formatting rules and its `value` field converts to a number; rows with an invalid date, null/unparseable date, or `NaN` value should be excluded from the chart data view. The retained rows should be ordered by ascending parsed timestamp, regardless of the input order. For example, if I create a chart with rows for `2020-03-02` value `5`, `not-a-date` value `7`, `2020-01-01` value `10`, and `2020-02-01` value `bad-number`, iterating `chart.data().getIterator()` should visit only `2020-01-01` then `2020-03-02`. If I later call `chart.data(newRows)`, the same filtering and chronological ordering should be applied to the new data before `chart.data().getIterator()` is used. This behavior should be deterministic for the same input data and should not mutate the input rows."} {"task_id": "format-code-task-000402", "source_id": "format-code-task-000402", "domain": "code", "task_path": "tasks/format-code-task-000402", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e235a0eab722de7b6ad8602f790e067118f97285c9d7870d4d8b78c1bbf7653e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAmazon SES: support extra_headers, metadata, tags for template sends\nOriginally, AWS's `ses::SendBulkEmail` API didn't allow specifying email headers. Since Anymail's Amazon SES backend also uses custom headers for `metadata` and `tags`, this meant you [couldn't use](https://anymail.dev/en/stable/esps/amazon_ses/#batch-sending-merge-and-esp-templates) any of the following message options together with a `template_id`:\n\n- `headers` (a.k.a. `extra_headers`)\n- `metadata` or `merge_metadata`\n- `tags` (except a single tag when using `AMAZON_SES_MESSAGE_TAG_NAME`)\n\nIn March, [AWS added a new `ReplacementHeaders` parameter](https://aws.amazon.com/about-aws/whats-new/2024/03/amazon-ses-headers-sending-email/) for `ses::SendBulkEmail`, which allows per-recipient custom headers. In early May, the [new parameter was made available in boto3](https://github.com/boto/boto3/blob/1.34.98/CHANGELOG.rst#L14).\n\nAnymail should use SES's new ReplacementHeaders to support extra_headers, metadata, merge_metadata, and tags in template sends.\n\n(Related: #371)"} {"task_id": "format-code-task-000403", "source_id": "format-code-task-000403", "domain": "code", "task_path": "tasks/format-code-task-000403", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:8e29de2a577e753d1e24442fecd48e0735b0852c837ef149860bcee31714b5e2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAdd bulk \"clear\" to Browse -> DAG Runs UI\n**Description**\n\nAirflow DAG Run UI (Browse -> DAG Runs) is really useful for managing many DAG Runs, creating specific DAG Runs and so on.\nIt would be great to have an additional option to \"With Selected DAG Runs\" -> \"Clear\" to reset all task instances under those DAG Runs.\n\n**Use case / motivation**\n\nWhen rerunning DAGs especially during development cycles it is tedious to go into the Tree view and clear many DAG Runs manually. It would be great to have some way to do this in bulk / over a range of DAG Runs. \n\nThe DAG Run UI seems like a good place to add this since it is already a kind of administrative area over all the DAG Runs and has \"with selected\" options like \"delete\", \"mark failed\", etc. Discussion welcome :)\n\n**Related Issues**\nNone"} {"task_id": "format-code-task-000404", "source_id": "format-code-task-000404", "domain": "code", "task_path": "tasks/format-code-task-000404", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6606374226ada796ea3413eef0b5d2d1f6d3779ad34431ec00f116beaa3140b6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Description\n\nCurrently the oracle hook it not setting a CURRENT_SCHEMA after connecting with the Database.\nIn a lot of use cases we have production and test databases with seperate connections and database schemas e.g. TEST.Table1, PROD.Table1\n\"Hard-coding\" the database schema in SQL Scripts is not elegant due to having different Airflow Instances for developing and Production.\n\nAn Option would be to store the database schema in a airflow Variable and getting it into the sql script with JINJA.\nIn Large SQL files with several tables it is not elegant either, because for every table a query to the metadatabase is made.\n\nWhy not using the Schema parameter in the Airflow Connections and executing \n`ALTER SESSION SET CURRENT_SCHEMA='SCHEMA'`\nright after successfully connecting to the database?\n\nAn alternative would be to use option `Connection.current_schema ` of Library cx_Oracle.\nhttps://cx-oracle.readthedocs.io/en/6.4.1/connection.html#Connection.current_schema\n\n\n\n### Use case/motivation\n\nIt makes Query development much easier by storing environment attributes directly in the Airflow Connection.\nYou have full flexibility without touching your SQL Script.\nIt makes separation of Test and Production environments and connections possible.\n\n\n\n### Related issues\n\n_No response_\n\n### Are you willing to submit a PR?\n\n- [ ] Yes I am willing to submit a PR!\n\n### Code of Conduct\n\n- [X] I agree to follow this project's [Code of Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)"} {"task_id": "format-code-task-000405", "source_id": "format-code-task-000405", "domain": "code", "task_path": "tasks/format-code-task-000405", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c96b9342eaac24a1a9a843209cf1d08ebd5562b85364eb64bbab17c5fc9ef2a1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nWebserver doesn't mask rendered fields for pending tasks\n### Apache Airflow version\n\n2.2.5 (latest released)\n\n### What happened\n\nWhen triggering a new dagrun the webserver will not mask secrets in the rendered fields for that dagrun's tasks which didn't start yet.\n\nTasks which have completed or are in state running are not affected by this.\n\n### What you think should happen instead\n\nThe webserver should mask all secrets for tasks which have started or not started.\n\n\"Screenshot\n.\n\n### How to reproduce\n\nCreate a variable `my_secret` and run this DAG\n\n```python\nfrom datetime import timedelta\n\nfrom airflow import DAG\nfrom airflow.operators.bash import BashOperator\nfrom airflow.sensors.time_delta import TimeDeltaSensor\nfrom airflow.utils.dates import days_ago\n\nwith DAG(\n \"secrets\",\n start_date=days_ago(1),\n schedule_interval=None,\n) as dag:\n wait = TimeDeltaSensor(\n task_id=\"wait\",\n delta=timedelta(minutes=1),\n )\n\n task = wait >> BashOperator(\n task_id=\"secret_task\",\n bash_command=\"echo '{{ var.value.my_secret }}'\",\n )\n```\n\nWhile the first task `wait` is running, displaying rendered fields for the second task `secret_task` will show the unmasked secret variable.\n\n\"Screenshot\n\n\n\n### Operating System\n\nDebian (Astronomer Airflow Docker image)\n\n### Versions of Apache Airflow Providers\n\n```\napache-airflow-providers-amazon==1!3.2.0\napache-airflow-providers-cncf-kubernetes==1!3.0.0\napache-airflow-providers-elasticsearch==1!3.0.2\napache-airflow-providers-ftp==1!2.1.2\napache-airflow-providers-google==1!6.7.0\napache-airflow-providers-http==1!2.1.2\napache-airflow-providers-imap==1!2.2.3\napache-airflow-providers-microsoft-azure==1!3.7.2\napache-airflow-providers-mysql==1!2.2.3\napache-airflow-providers-postgres==1!4.1.0\napache-airflow-providers-redis==1!2.0.4\napache-airflow-providers-slack==1!4.2.3\napache-airflow-providers-sqlite==1!2.1.3\napache-airflow-providers-ssh==1!2.4.3\n```\n\n### Deployment\n\nAstronomer\n\n### Deployment details\n\n_No response_\n\n### Anything else\n\nWe have seen this issue also in Airflow 2.2.3.\n\n### Are you willing to submit PR?\n\n- [ ] Yes I am willing to submit a PR!\n\n### Code of Conduct\n\n- [X] I agree to follow this project's [Code of Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)"} {"task_id": "format-code-task-000406", "source_id": "format-code-task-000406", "domain": "code", "task_path": "tasks/format-code-task-000406", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b40b062aed81c1579a7b63fd40d9b3e568f510b859005907489d8a6f9c1b59fb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAttributeError: 'CeleryKubernetesExecutor' object has no attribute 'send_callback'\n### Apache Airflow version\n\n2.3.0 (latest released)\n\n### What happened\n\nThe issue started to occur after upgrading airflow from v2.2.5 to v2.3.0. The schedulers are crashing when DAG's SLA is configured. Only occurred when I used `CeleryKubernetesExecutor`. Tested on `CeleryExecutor` and it works as expected.\n\n```\nTraceback (most recent call last): \n File \"/home/airflow/.local/bin/airflow\", line 8, in \n sys.exit(main()) \n File \"/home/airflow/.local/lib/python3.9/site-packages/airflow/__main__.py\", line 38, in main \n args.func(args) \n File \"/home/airflow/.local/lib/python3.9/site-packages/airflow/cli/cli_parser.py\", line 51, in command \n return func(*args, **kwargs) \n File \"/home/airflow/.local/lib/python3.9/site-packages/airflow/utils/cli.py\", line 99, in wrapper \n return f(*args, **kwargs) \n File \"/home/airflow/.local/lib/python3.9/site-packages/airflow/cli/commands/scheduler_command.py\", line 75, in scheduler \n _run_scheduler_job(args=args) \n File \"/home/airflow/.local/lib/python3.9/site-packages/airflow/cli/commands/scheduler_command.py\", line 46, in _run_scheduler_job \n job.run() \n File \"/home/airflow/.local/lib/python3.9/site-packages/airflow/jobs/base_job.py\", line 244, in run \n self._execute() \n File \"/home/airflow/.local/lib/python3.9/site-packages/airflow/jobs/scheduler_job.py\", line 736, in _execute \n self._run_scheduler_loop() \n File \"/home/airflow/.local/lib/python3.9/site-packages/airflow/jobs/scheduler_job.py\", line 824, in _run_scheduler_loop \n num_queued_tis = self._do_scheduling(session) \n File \"/home/airflow/.local/lib/python3.9/site-packages/airflow/jobs/scheduler_job.py\", line 919, in _do_scheduling \n self._send_dag_callbacks_to_processor(dag, callback_to_run) \n File \"/home/airflow/.local/lib/python3.9/site-packages/airflow/jobs/scheduler_job.py\", line 1179, in _send_dag_callbacks_to_processor\n self._send_sla_callbacks_to_processor(dag) \n File \"/home/airflow/.local/lib/python3.9/site-packages/airflow/jobs/scheduler_job.py\", line 1195, in _send_sla_callbacks_to_processor\n self.executor.send_callback(request) \nAttributeError: 'CeleryKubernetesExecutor' object has no attribute 'send_callback'\n```\n\n### What you think should happen instead\n\nWork like previous version\n\n### How to reprod"} {"task_id": "format-code-task-000407", "source_id": "format-code-task-000407", "domain": "code", "task_path": "tasks/format-code-task-000407", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:06a66955763cd2ea79f469718187781d7f86741041f83c0b4823b168a408f3e6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAdd one_done trigger rule\n### Body\n\nAction: trigger as soon as 1 upstream task is in success or failuire\nThis has been requested in https://stackoverflow.com/questions/73501232/how-to-implement-the-one-done-trigger-rule-for-airflow\nI think this can be useful for the community.\n\n**The Task:**\nAdd support for new trigger rule `one_done`\nYou can use as reference previous PRs that added other trigger rules for example: https://github.com/apache/airflow/pull/21662\n\n### Committer\n\n- [X] I acknowledge that I am a maintainer/committer of the Apache Airflow project."} {"task_id": "format-code-task-000408", "source_id": "format-code-task-000408", "domain": "code", "task_path": "tasks/format-code-task-000408", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:dc2e29f078b0f308f9f88bb646eb0573fb1c02ea6715a28acf2c8e331105acbe", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nGCSToBigQueryOperator fails when schema_object is specified without schema_fields\n### Apache Airflow Provider(s)\n\ngoogle\n\n### Versions of Apache Airflow Providers\n\napache-airflow 2.5.0\napache-airflow-providers-apache-beam 4.1.0\napache-airflow-providers-cncf-kubernetes 5.0.0\napache-airflow-providers-google 8.6.0\napache-airflow-providers-grpc 3.1.0\n\n### Apache Airflow version\n\n2.5.0\n\n### Operating System\n\nDebian 11\n\n### Deployment\n\nOfficial Apache Airflow Helm Chart\n\n### Deployment details\n\nKubernetesExecutor\n\n### What happened\n\nGCSToBigQueryOperator allows multiple ways to specify schema of the BigQuery table:\n\n1. Setting autodetect == True\n1. Setting schema_fields directly with autodetect == False\n1. Setting a schema_object and optionally a schema_object_bucket with autodetect == False\n\nThis third method seems to be broken in the latest provider version (8.6.0) and will always result in this error:\n\n```\n[2022-12-16, 21:06:18 UTC] {taskinstance.py:1772} ERROR - Task failed with exception\nTraceback (most recent call last):\n File \"/home/airflow/.local/lib/python3.9/site-packages/airflow/providers/google/cloud/transfers/gcs_to_bigquery.py\", line 395, in execute\n self.configuration = self._check_schema_fields(self.configuration)\n File \"/home/airflow/.local/lib/python3.9/site-packages/airflow/providers/google/cloud/transfers/gcs_to_bigquery.py\", line 524, in _check_schema_fields\n raise RuntimeError(\nRuntimeError: Table schema was not found. Set autodetect=True to automatically set schema fields from source objects or pass schema_fields explicitly\n```\n\nThe reason for this is because [this block](https://github.com/apache/airflow/blob/25bdbc8e6768712bad6043618242eec9c6632618/airflow/providers/google/cloud/transfers/gcs_to_bigquery.py#L318-L320) where `if self.schema_object and self.source_format != \"DATASTORE_BACKUP\":`. fails to set self.schema_fields. It only sets the local variable, schema_fields. When self._check_schema_fields is subsequently called [here](https://github.com/apache/airflow/blob/25bdbc8e6768712bad6043618242eec9c6632618/airflow/providers/google/cloud/transfers/gcs_to_bigquery.py#L395), we enter the [first block](https://github.com/apache/airflow/blob/25bdbc8e6768712bad6043618242eec9c6632618/airflow/providers/google/cloud/transfers/gcs_to_bigquery.py#L523-L528) because autodetect is false and schema_fields is not set.\n\n### What you think should happen instead\n\nNo error should be raised if autodetect is set to False and a valid schema_object is provided\n\n### How to reproduce\n\n1. Create a simple BigQuery table with a single column col1:\n\n```sql\nCREATE TABLE `my-project.my_dataset.test_gcs_to_bigquery` (col1 INT);\n```\n\n2. Upload a json blob for this object to a bucket (e.g., data/schemas/table.json)\n3. Upload a simple CSV for the source file to load to a bucket (e.g., data/source/file.csv)\n4. Run the following command:\n\n```py\n gcs_to_biquery = GCSToBigQueryOperator(\n task_id=\"gcs_to_bigquery\",\n destination_project_dataset_table=\"my-project.my_dataset.test_gcs_to_bigquery\",\n bucket=\"my_bucket_name\",\n create_disposition=\"CREATE_IF_NEEDED\",\n write_disposition=\"WRITE_TRUNCATE\",\n source_objects=[\"data/source/file.csv\"],\n source_format=\"CSV\",\n autodetect=False,\n schema_object=\"data/schemas/table.json\",\n )\n```\n\n### Anything else\n\n_No response_\n\n### Are you willing to submit PR?\n\n- [X] Yes I am willing to submit a PR!\n\n### Code of Conduct\n\n- [X] I agree to follow this project's [Code of Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)"} {"task_id": "format-code-task-000409", "source_id": "format-code-task-000409", "domain": "code", "task_path": "tasks/format-code-task-000409", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c526b23ea053191e6cb115c8b9b9fb47560281a5ee1ff69f5f86509f72f7e4b2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nBigQueryCreateDataTransferOperator will log AWS credentials when transferring from S3\n### Apache Airflow Provider(s)\n\ngoogle\n\n### Versions of Apache Airflow Providers\n\n[apache-airflow-providers-google 8.6.0](https://airflow.apache.org/docs/apache-airflow-providers-google/8.6.0/)\n\n\n\n### Apache Airflow version\n\n2.5.0\n\n### Operating System\n\nDebian GNU/Linux 11 (bullseye)\n\n### Deployment\n\nOfficial Apache Airflow Helm Chart\n\n### Deployment details\n\n_No response_\n\n### What happened\n\nWhen creating a transfer config that will move data from AWS S3, an access_key_id and secret_access_key are provided (see: https://cloud.google.com/bigquery/docs/s3-transfer).\n\nThese parameters are logged and exposed as XCom return_value.\n\n### What you think should happen instead\n\nAt least the secret_access_key should be hidden or removed from the XCom return value\n\n### How to reproduce\n\n```\nPROJECT_ID=123\nTRANSFER_CONFIG={\n \"destination_dataset_id\": destination_dataset,\n \"display_name\": display_name,\n \"data_source_id\": \"amazon_s3\",\n \"schedule_options\": {\"disable_auto_scheduling\": True},\n \"params\": {\n \"destination_table_name_template\": destination_table,\n \"file_format\": \"PARQUET\",\n \"data_path\": data_path,\n \"access_key_id\": access_key_id,\n \"secret_access_key\": secret_access_key\n }\n },\n\ngcp_bigquery_create_transfer = BigQueryCreateDataTransferOperator(\n transfer_config=TRANSFER_CONFIG,\n project_id=PROJECT_ID,\n task_id=\"gcp_bigquery_create_transfer\",\n)\n```\n\n### Anything else\n\n_No response_\n\n### Are you willing to submit PR?\n\n- [ ] Yes I am willing to submit a PR!\n\n### Code of Conduct\n\n- [X] I agree to follow this project's [Code of Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)"} {"task_id": "format-code-task-000410", "source_id": "format-code-task-000410", "domain": "code", "task_path": "tasks/format-code-task-000410", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:deacb9c852ceee25d491d3e50a2514ace8b5c90035dd7615f89fd0e668f134dd", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nBigQueryToGCSOperator does not wait for completion\n### Apache Airflow Provider(s)\n\ngoogle\n\n### Versions of Apache Airflow Providers\n\napache-airflow-providers-google==7.0.0\n\n### Apache Airflow version\n\n2.3.2\n\n### Operating System\n\nDebian GNU/Linux\n\n### Deployment\n\nOfficial Apache Airflow Helm Chart\n\n### Deployment details\n\n_No response_\n\n### What happened\n\n[Deferrable mode for BigQueryToGCSOperator #27683](https://github.com/apache/airflow/pull/27683) changed the functionality of the `BigQueryToGCSOperator` so that it no longer waits for the completion of the operation. This is because the `nowait=True` parameter is now [being set](https://github.com/apache/airflow/pull/27683/files#diff-23c5b2e773487f9c28b75b511dbf7269eda1366f16dec84a349d95fa033ffb3eR191).\n\n### What you think should happen instead\n\nThis is unexpected behavior. Any downstream tasks of the `BigQueryToGCSOperator` that expect the CSVs to have been written by the time they are called may result in errors (and have done so in our own operations).\n\nThe property should at least be configurable.\n\n### How to reproduce\n\n1. Leverage the `BigQueryToGcsOperator` in your DAG.\n2. Have it write a large table to a CSV somewhere in GCS\n3. Notice that the task completes almost immediately but the CSVs may not exist in GCS until later.\n\n### Anything else\n\n_No response_\n\n### Are you willing to submit PR?\n\n- [X] Yes I am willing to submit a PR!\n\n### Code of Conduct\n\n- [X] I agree to follow this project's [Code of Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)"} {"task_id": "format-code-task-000411", "source_id": "format-code-task-000411", "domain": "code", "task_path": "tasks/format-code-task-000411", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e884ac54c8364445483cfd8d1ad1ec5fbda03286a160d235449dc82beeaf68af", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nMake `webserver_config.py` location customizable\n**Description**\n`webserver_config.py` location is [hard-coded](https://github.com/apache/airflow/blob/c2db0dfeb13ee679bf4d7b57874f0fcb39c0f0ed/airflow/configuration.py#L769) as follows:\n```\nWEBSERVER_CONFIG = AIRFLOW_HOME + '/webserver_config.py'\n```\nIt would be great if this path is customizable.\n\n**Use case / motivation**\nSince `AIRFLOW_HOME/config` is already in `PYTHONPATH` (though that's also hard-coded), it's common for user to mount all config files under `AIRFLOW_HOME/config`. However, this `webserver_config.py` needs extra handling, either by mounting using a `subPath` explicitly, or in my case, creating a symbolic link from `AIRFLOW_HOME/webserver_config.py` to `AIRFLOW_HOME/config/webserver_config.py` in the image."} {"task_id": "format-code-task-000412", "source_id": "format-code-task-000412", "domain": "code", "task_path": "tasks/format-code-task-000412", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f466a789909cbf6443104446911dad14d69ac6814e23ef111c36e09e644adf60", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nDeserialization of old xcom data fails after upgrade to 2.6.1 from 2.5.2 when calling /xcom/list/ [GET]\n### Apache Airflow version\n\n2.6.1\n\n### What happened\n\nAfter upgrading from airflow 2.5.2 to 2.6.1 calling the endpoint `xcom/list/` we get the following exception:\n```\n[2023-06-07T12:16:50.050+0000] {app.py:1744} ERROR - Exception on /xcom/list/ [GET]\nTraceback (most recent call last):\n File \"/home/airflow/.local/lib/python3.10/site-packages/flask/app.py\", line 2529, in wsgi_app\n response = self.full_dispatch_request()\n File \"/home/airflow/.local/lib/python3.10/site-packages/flask/app.py\", line 1825, in full_dispatch_request\n rv = self.handle_user_exception(e)\n File \"/home/airflow/.local/lib/python3.10/site-packages/flask/app.py\", line 1823, in full_dispatch_request\n rv = self.dispatch_request()\n File \"/home/airflow/.local/lib/python3.10/site-packages/flask/app.py\", line 1799, in dispatch_request\n return self.ensure_sync(self.view_functions[rule.endpoint])(**view_args)\n File \"/home/airflow/.local/lib/python3.10/site-packages/flask_appbuilder/security/decorators.py\", line 139, in wraps\n return f(self, *args, **kwargs)\n File \"/home/airflow/.local/lib/python3.10/site-packages/flask_appbuilder/views.py\", line 554, in list\n widgets = self._list()\n File \"/home/airflow/.local/lib/python3.10/site-packages/flask_appbuilder/baseviews.py\", line 1177, in _list\n widgets = self._get_list_widget(\n File \"/home/airflow/.local/lib/python3.10/site-packages/flask_appbuilder/baseviews.py\", line 1076, in _get_list_widget\n count, lst = self.datamodel.query(\n File \"/home/airflow/.local/lib/python3.10/site-packages/flask_appbuilder/models/sqla/interface.py\", line 500, in query\n query_results = query.all()\n File \"/home/airflow/.local/lib/python3.10/site-packages/sqlalchemy/orm/query.py\", line 2773, in all\n return self._iter().all()\n File \"/home/airflow/.local/lib/python3.10/site-packages/sqlalchemy/engine/result.py\", line 1476, in all\n return self._allrows()\n File \"/home/airflow/.local/lib/python3.10/site-packages/sqlalchemy/engine/result.py\", line 401, in _allrows\n rows = self._fetchall_impl()\n File \"/home/airflow/.local/lib/python3.10/site-packages/sqlalchemy/engine/result.py\", line 1389, in _fetchall_impl\n return self._real_result._fetchall_impl()\n File \"/home/airflow/.local/lib/python3.10/site-packages/sqlalchemy/engine/result.py\", line 1813, in _fetchall_impl\n return list(self.iterator)\n File \"/home/airflow/.local/lib/python3.10/site-packages/sqlalchemy/orm/loading.py\", line 151, in chunks\n rows = [proc(row) for row in fetch]\n File \"/home/airflow/.local/lib/python3.10/site-packages/sqlalchemy/orm/loading.py\", line 151, in \n rows = [proc(row) for row in fetch]\n File \"/home/airflow/.local/lib/python3.10/site-packages/sqlalchemy/orm/loading.py\", line 984, in _instance\n state.manager.dispatch.load(state, context)\n File \"/home/airflow/.local/lib/python3.10/site-packages/sqlalchemy/event/attr.py\", line 334, in __call__\n fn(*args, **kw)\n File \"/home/airflow/.local/lib/python3.10/site-packages/sqlalchemy/orm/mapper.py\", line 3702, in _event_on_load\n instrumenting_mapper._reconstructor(state.obj())\n File \"/home/airflow/.local/lib/python3.10/site-packages/airflow/models/xcom.py\", line 128, in init_on_load\n self.value = self.orm_deserialize_value()\n File \"/home/airflow/.local/lib/python3.10/site-packages/airflow/models/xcom.py\", line 677, in orm_deserialize_value\n return BaseXCom._deserialize_value(self, True)\n File \"/home/airflow/.local/lib/python3.10/site-packages/airflow/models/xcom.py\", line 659, in _deserialize_value\n return json.loads(result.value.decode(\"UTF-8\"), cls=XComDecoder, object_hook=object_hook)\n File \"/usr/local/lib/python3.10/json/__init__.py\", line 359, in loads\n return cls(**kw).decode(s)\n File \"/usr/local/lib/python3.10/json/decoder.py\", line 337, in decode\n obj, end = self.raw_decode(s, idx=_w(s, 0).end()"} {"task_id": "format-code-task-000413", "source_id": "format-code-task-000413", "domain": "code", "task_path": "tasks/format-code-task-000413", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d7211a7e1f5c85252d85edb45fb8d11f9f6f7b5da0ce26668d21995bf5f1d680", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nScheduler is spending most of its time in clear_not_launched_queued_tasks function\n### Apache Airflow version\n\n2.7.1\n\n### What happened\n\nAirflow running the clear_not_launched_queued_tasks function on a certain frequency (default 30 seconds). When we run the airflow on a large Kube cluster (pods more than > 5K). Internally the clear_not_launched_queued_tasks function loops through each queued task and checks the corresponding worker pod existence in the Kube cluster. Right this existence check using list pods Kube API. The API is taking more than 1s. if there are 120 queued tasks, then it will take ~ 120 seconds (1s * 120). So, this leads the scheduler to spend most of its time in this function rather than scheduling the tasks. It leads to none of the jobs being scheduled or degraded scheduler performance.\n\n### What you think should happen instead\n\nIt would be nice to get all the airflow worker pods in a one/few batch calls rather than for each task. These batch calls helps to speed the processing of clear_not_launched_queued_tasks function call. \n\n### How to reproduce\n\nRun the airflow on large Kube clusters (> 5K pods). Simulate the airflow to run the 100 parallel DAG runs for every minute. \n\n### Operating System\n\nCent OS 7\n\n### Versions of Apache Airflow Providers\n\n2.3.3, 2.7.1\n\n### Deployment\n\nOther Docker-based deployment\n\n### Deployment details\n\nTerraform based airflow deployment\n\n### Anything else\n\n_No response_\n\n### Are you willing to submit PR?\n\n- [x] Yes I am willing to submit a PR!\n\n### Code of Conduct\n\n- [X] I agree to follow this project's [Code of Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)"} {"task_id": "format-code-task-000414", "source_id": "format-code-task-000414", "domain": "code", "task_path": "tasks/format-code-task-000414", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9348f452ec8298fa3ff2704e1e4a1574705c451cbb9a678ffd36c338b702db1a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAllow better filtering of event logs (audit log)\n### What do you see as an issue?\n\nThe REST API get event logs endpoint only allows filtering logs by a single event type. This only includes that one event type and all others are discarded. One cannot filter out even a single event type.\n\n### Solving the problem\n\nIn order to use this fully, I think we need to be able to pass multiple event names and also specify if we are including or excluding them.\n\nIn the webserver we use `audit_view_excluded_events` and `audit_view_included_events`. We should have something similar for the rest API for a user to specify what they want to include and exclude.\n\n### Anything else\n\n_No response_\n\n### Are you willing to submit PR?\n\n- [ ] Yes I am willing to submit a PR!\n\n### Code of Conduct\n\n- [X] I agree to follow this project's [Code of Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)"} {"task_id": "format-code-task-000415", "source_id": "format-code-task-000415", "domain": "code", "task_path": "tasks/format-code-task-000415", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:030e80c283526efdafeb0ae577b6450ac457ea74b94d2c6e64abd4d4e951a496", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nFTPHook doesn't not allow to change port\n### Apache Airflow Provider(s)\n\nftp\n\n### Versions of Apache Airflow Providers\n\n3.8.0\n\n### Apache Airflow version\n\n2.6.3\n\n### Operating System\n\nUbuntu 20.04\n\n### Deployment\n\nOfficial Apache Airflow Helm Chart\n\n### Deployment details\n\nIrrelevant as it is in source code\n\n### What happened\n\nWhen setting up an FTP connection with a custom port (different from 21), the FTP Hook does not read it at all and simply passes the hostname and uses the default port set by FTP_PORT in ftplib.FTP_PORT (used by ftplib.FTP.port), thus preventing any modification of default port 21. \nPS: The 'connect' function of the FTP class reads port from the self.port attribute;\n\n### What you think should happen instead\n\nThe FTP Hook should be able to read the port (if set) and pass it to the FTP object created \n\n### How to reproduce\n\nSet up any FTP connection with and test it on Airflow UI. \n\n### Anything else\n\n_No response_\n\n### Are you willing to submit PR?\n\n- [ ] Yes I am willing to submit a PR!\n\n### Code of Conduct\n\n- [X] I agree to follow this project's [Code of Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)"} {"task_id": "format-code-task-000416", "source_id": "format-code-task-000416", "domain": "code", "task_path": "tasks/format-code-task-000416", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6079a16e2fc5b25d3dcedd6ca78b0ce510a9642e7b65bebfb66c415bdda45fbe", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Apache Airflow Provider(s)\n\ncommon-sql\n\n### What happened\n\nTwo unrelated annoyances around `DbApiHook` that I hit on the same DAG.\n\n**1. `hook.placeholder` re-queries the metadata DB on every access**\n\nI have a Postgres-backed custom hook subclassing `DbApiHook`, and a task that inserts a lot of rows via `insert_rows`. While poking at why the task is slower than I expected and why I see so much load on my Airflow metadata DB, I noticed the `placeholder` property goes through `get_connection(...)` every single time it's read. Since the insert path reads it per row to build the SQL template, every inserted row triggers a metadata DB roundtrip just to read the connection back. For a task pushing tens of thousands of rows this is very noticeable.\n\nThe connection itself isn't changing during a task — reading it once per hook instance should be enough.\n\n**2. Warning about an invalid placeholder in `extra` is unhelpful**\n\nIn a different Connection I tried to override the placeholder via the `extra` field, e.g.\n\n```json\n{ \"placeholder\": \":1\" }\n```\n\nwhich isn't in the supported set, so it's (correctly) ignored. But the warning I get in the task log looks roughly like:\n\n```\nPlaceholder defined in Connection 'postgres_conn_id' is not listed in 'DEFAULT_SQL_PLACEHOLDERS' and got ignored. Falling back to the default placeholder '%s'.\n```\n\nTwo problems with this message:\n\n- `'postgres_conn_id'` here is not my actual connection id — it's the literal name of the attribute on the hook class. So when I went to look up \"which of my connections is misconfigured?\" I couldn't find one with that id at all and got pretty confused.\n- The message never tells me which placeholder value was rejected. If I had several connections / had recently edited extras, I'd want to see that `':1'` was the offending value so I can go fix the right one.\n\n### What you think should happen instead\n\n- Reading `placeholder` on a hook instance shouldn't keep hammering the metadata DB; the connection lookup should only happen once per hook.\n- The \"invalid placeholder\" warning should identify (a) the actual connection id the bad value came from and (b) the bad placeholder value itself, so it's actually actionable.\n\n### Versions of Apache Airflow Providers\n\napache-airflow-providers-common-sql (current main)\n\n### Deployment\n\nOther\n\n### Anything else\n\nBoth are reproducible on any subclass of `DbApiHook` — the slow path just needs a bulk insert, the warning just needs an `extra` with a non-standard `placeholder`."} {"task_id": "format-code-task-000417", "source_id": "format-code-task-000417", "domain": "code", "task_path": "tasks/format-code-task-000417", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4de1ddaa99ca52d76bcf1bc747d57298d9874fb3339d0d2012ac6036303d6217", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## DAGs are being deactivated after upgrade even though they are still parsed and present\n\nAfter upgrading Airflow, several of my DAGs unexpectedly turned **inactive** in the UI. They no longer show up in the active list, can't be triggered, and look like they were removed — but the DAG files are still right where they always were and the DAG processor is still happily parsing them on every loop.\n\n### Setup\n\nIn my deployment, not all DAG files live directly under the `dags_folder` configured in `airflow.cfg`. Some of them are loaded from paths outside that directory — this is a legitimate setup for me (multiple sources of DAGs / a custom layout where the scheduler's DAG processor sees DAGs whose `fileloc` is not nested under the main `dags_folder` path). This worked fine on previous versions.\n\n### What I see\n\nAfter the upgrade:\n\n- The DAGs whose file path is **not** under the configured `dags_folder` get marked stale and deactivated, basically right after the scheduler starts.\n- In the scheduler logs I see lines like:\n\n ```\n DAG is missing and will be deactivated.\n ```\n\n for every one of these DAGs, even though the file is clearly still there and is being successfully parsed (no import errors, `last_parsed_time` keeps advancing).\n\n- DAGs whose file path **is** under the main `dags_folder` are unaffected — only the \"outside\" ones get killed.\n\n### What I expected\n\nA DAG should only be considered stale / be deactivated when it has actually disappeared (the file is gone, or it's no longer being parsed within the stale threshold). As long as the DAG processor is still parsing the file successfully, the DAG should stay active — regardless of whether `fileloc` happens to live under the main `dags_folder` path or somewhere else.\n\nThe previous behavior (only deactivating DAGs that are genuinely missing / no longer parsed in time) is what I need back. Right now this effectively breaks any setup where DAGs come from more than one root directory."} {"task_id": "format-code-task-000418", "source_id": "format-code-task-000418", "domain": "code", "task_path": "tasks/format-code-task-000418", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:df6da86374e3f9cf19e78ffb02f8a03c585fa4683232234e2ea69966495f5e72", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Apache Airflow version\n\n2.10.5\n\n### If \"Other Airflow 2 version\" selected, which one?\n\n_No response_\n\n### What happened?\n\nWhen configuring Vault as secrets backend with AWS IAM authentication, for example, when installing Airflow on AWS EKS, and using the IRSA (IAM Role for Service Accounts) of the Airflow Pods to authenticate to Vault, the STS authentication call will fail if Vault is configured with a STS region other than `us-east-1`. This is a pretty common scenario whenever one is not using that region and wants to minimize the latency of STS API calls.\nThe `region` parameter is never passed to `hvac.api.auth_methods.Aws.iam_login()` by the `airflow.providers.hashicorp._internal_client.vault_client._VaultClient._auth_aws_iam()` method, which [defaults to `us-east-1`](https://github.com/hvac/hvac/blob/ea3a6520cc08f69470494cce0ac26a2ab025f91d/hvac/api/auth_methods/aws.py#L747). \n\nExample error message:\n```\nhvac.exceptions.InvalidRequest: error making upstream request: received error code 403 from STS: \n \n Sender\n SignatureDoesNotMatch\n Credential should be scoped to a valid region. \n \n 84b47cd7-fc7b-442f-b75b-309b93d2b2ee\n\n, on post http://vault-active.vault.svc.cluster.local:8200/v1/auth/aws/login\n```\n\n### What you think should happen instead?\n\nThe AWS region should be configurable as an input parameter of VaultBackend (and _VaultClient), and should default to the boto3 configuration if not provided (from env vars, instance metadata, etc.).\n\n### How to reproduce\n\nJust configure Vault with a regional STS endpoint different than `us-east-1`.\n\n### Operating System\n\nDebian 12 Bookworm\n\n### Versions of Apache Airflow Providers\n\n`apache-airflow-providers-hashicorp==4.0.0`\n\n### Deployment\n\nOfficial Apache Airflow Helm Chart\n\n### Deployment details\n\n_No response_\n\n### Anything else?\n\n_No response_\n\n### Are you willing to submit PR?\n\n- [x] Yes I am willing to submit a PR!\n\n### Code of Conduct\n\n- [x] I agree to follow this project's [Code of Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)"} {"task_id": "format-code-task-000419", "source_id": "format-code-task-000419", "domain": "code", "task_path": "tasks/format-code-task-000419", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9b7f677bef3a12de42de0e07b3f9209bd9b9931406bf9f0220b34f1b56eff25d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## GKEClusterCreateOperator and GKEClusterDeleteOperator fail at execute time\n\nI'm using Airflow to manage GKE clusters from a DAG. My tasks look roughly like:\n\n```python\ncreate = GKEClusterCreateOperator(\n task_id='create_cluster',\n project_id='my-gcp-project',\n location='us-central1-a',\n body={'name': 'analytics-cluster', 'initial_node_count': 3},\n gcp_conn_id='my_gcp_conn',\n)\n\ndelete = GKEClusterDeleteOperator(\n task_id='delete_cluster',\n project_id='my-gcp-project',\n location='us-central1-a',\n name='analytics-cluster',\n gcp_conn_id='my_gcp_conn',\n)\n```\n\nBoth operators fail as soon as the task starts executing. The `_check_input` validation passes fine (all of `project_id`, `location`, `name`/`body` are populated), so the failure is happening once the operator actually tries to talk to GCP.\n\nIt also looks like the `gcp_conn_id` I configure on the operator isn't being used the way I'd expect — even when I leave it at the default `google_cloud_default`, the behavior is the same broken one, so something about how the operator hands off to the underlying hook seems off.\n\nThe operators are documented as the standard way to create/delete a GKE cluster from a DAG, and the parameters I'm passing match what the docstrings show, so I'd expect this end-to-end flow to just work. Right now neither operator is usable."} {"task_id": "format-code-task-000420", "source_id": "format-code-task-000420", "domain": "code", "task_path": "tasks/format-code-task-000420", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b6aa396275b7aae15c7861d89d772c1ccfd5e62378480b4b49081d490e0a4136", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Add support for the `decimal` logical type in Python AVRO\n\nThe Avro spec defines a `decimal` logical type that annotates either a `bytes` or `fixed` schema with `precision` and `scale` properties — typically used for representing exact-precision values like monetary amounts. The Java implementation handles this, but the Python library currently has no real support for it.\n\nFor example, I have a schema like:\n\n```json\n{\n \"type\": \"record\",\n \"name\": \"Transaction\",\n \"fields\": [\n {\n \"name\": \"amount\",\n \"type\": {\n \"type\": \"bytes\",\n \"logicalType\": \"decimal\",\n \"precision\": 4,\n \"scale\": 2\n }\n }\n ]\n}\n```\n\nParsing this with `avro.schema.parse(...)` succeeds, but the `logicalType` / `precision` / `scale` keys just end up sitting in the \"other properties\" bucket of the parsed schema — there's no decimal-aware behaviour anywhere downstream. Concretely:\n\n- if I try to write `{\"amount\": Decimal(\"12.34\")}` the writer fails validation, because a `bytes` field is expected to be a `str`;\n- the same problem happens for a `fixed`-typed decimal;\n- and on the read side I just get raw bytes back instead of a `Decimal`.\n\nI'd like the Python implementation to actually support the `decimal` logical type on both `bytes`- and `fixed`-backed schemas, so a Python `decimal.Decimal` value can be round-tripped through serialization and deserialization, with the bytes encoded per the Avro spec (two's-complement big-endian of the unscaled integer).\n\nIt would also be very useful if schema parsing validated `precision` / `scale` at parse time — misconfigured schemas (e.g. non-positive precision, negative scale, scale greater than precision, or a precision that wouldn't even fit in the declared fixed size) should fail loudly rather than silently producing garbage at write time.\n\nTracked upstream as AVRO-1816."} {"task_id": "format-code-task-000421", "source_id": "format-code-task-000421", "domain": "code", "task_path": "tasks/format-code-task-000421", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:032d9905c4b43ccf37ea4eb9404d41b8913c1657bef048a6232c0b69d96ece74", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `CarbonStore` to support semantic long-term memory recall with a public method `recall(query: str | vector, *, top_k: int = 5, session_id: str | None = None, actor: str | None = None, namespace: str | None = None, min_salience: float | None = None, include_expired: bool = False, model: str | None = None, embedder: Embedder | None = None, metric: \"cosine\" | \"dot\" | \"l2\" = \"cosine\") -> list[MemoryHit]`.\n\nIn a session where memories have already been recorded with embeddings, `store.recall(\"vim editor\", top_k=3, embedder=emb)` should return ranked `MemoryHit` objects sorted from best score to worst; if the stored memories are \"loves vim\", \"prefers emacs\", and \"likes nano\" with vectors closest to the query in that order, the first hit's `memory.content` should be \"loves vim\". Each `MemoryHit` should expose the recalled `MemoryItem` as `.memory` and the numeric similarity as `.score`.\n\nFilters should compose with AND. For example, after recording one embedded memory in session `s1` and one in session `s2`, `store.recall(\"a\", session_id=\"s1\", top_k=5, embedder=emb)` should return only memories whose `memory.session_id == \"s1\"`; similarly `actor`, `namespace`, and `min_salience` should restrict results to the matching actor, namespace, and salience threshold. Recall should only return entities recorded as memories, not ordinary document entities that also have chunks and embeddings.\n\nExpired memories should be hidden by default: if \"old\" has an `expires_at` timestamp in the past and \"fresh\" has no expiry, `store.recall(\"old\", embedder=emb)` should return only \"fresh\". Passing `include_expired=True` should include both expired and unexpired matching memories.\n\nIf no memories match, recall should return an empty list. A string query should require an embedder so it can be encoded; a vector query should work with `model=` and no embedder. The `metric` argument should choose cosine, dot, or negative L2 scoring for ranking."} {"task_id": "format-code-task-000422", "source_id": "format-code-task-000422", "domain": "code", "task_path": "tasks/format-code-task-000422", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b22bfd6a7e8459af869b7d2a2862b4dee55f26de4e2777f18fa492e358c7956c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\ncordova plugin rm fails when multiple asset elements add content to plugins/ folder\n# Bug Report\n\n## Problem\nIf a plugin's plugin.xml file contains multiple `asset `elements which add content to the `plugins/` folder the operation `cordova plugin rm ` fails with message:\n\n```\nError during processing of action! Attempting to revert...\nTypeError [ERR_INVALID_ARG_TYPE]: The \"code\" argument must be of type number. Received type string ('ENOENT')\n at process.set [as exitCode] (node:internal/bootstrap/node:123:9)\n at /Users/CDM/desb1224/cordova/cordova-12.0.0/node_modules/cordova/bin/cordova:32:22\n```\n\n### What is expected to happen?\nThe operation should be removed completely.\n\n\n### What does actually happen?\nThe plugin is not completely removed.\n\n\n\n## Information\n\nThe problem is caused by code in the method `asset.uninstall` in `\\cordova-ios\\lib\\plugman\\pluginHandlers.js`. After removing the asset file another call is made to `removeFileF `to remove the plugin folder:\n```\n removeFileF(path.resolve(project.www, 'plugins', plugin.id));\n ...\n\nfunction removeFileF (file) {\n fs.rmSync(file, { recursive: true, force: true });\n}\n```\nThe recursive call in `removeFileF` deletes any other assets sharing the same path, and on attempting to delete those assets the operation fails with the error given above\n\nThe plugin can be removed if ios platform is removed first.\n\nThis error is related to changes in made to resolve issue [https://github.com/apache/cordova-ios/issues/1443](https://github.com/apache/cordova-ios/issues/1443):\n[https://github.com/apache/cordova-ios/pull/1446/files](https://github.com/apache/cordova-ios/pull/1446/files)\n\n\n### Command or Code\n\nAdd a plugin containing multiple assets in the same folder, for example with something along the following lines in plugin.xml:\n```\n\t\n\t\n```\nAttempting to remove the plugin (when ios platform is installed) shoudl trigger the error.\n\n\n### Environment, Platform, Device\n\nError occurs on development machine.\n\n### Version information\n\n- Cordova 12.0.0\n- cordova-ios 7.1.1\n\n\n\n## Checklist\n\n\n- [X] I searched for existing GitHub issues\n- [X] I updated all Cordova tooling to most recent version\n- [X] I included all the necessary information above"} {"task_id": "format-code-task-000426", "source_id": "format-code-task-000426", "domain": "code", "task_path": "tasks/format-code-task-000426", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a0b524514effdd02bd7b92f204be874b8fdd2ea921d48308d22530291a464379", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want local PouchDB map/reduce queries to support the built-in `_stats` reducer through `db.query()`. In a session like `const db = new PouchDB(name); await db.bulkDocs({docs: [{val: 'bar'}, {val: 'bar'}, {val: 'baz'}]});`, calling `await db.query({map: function (doc) { emit(doc.val, 1); }, reduce: '_stats'}, {reduce: true, group_level: 999})` should return grouped rows where the `bar` row has `value` equal to `{sum: 2, count: 2, min: 1, max: 1, sumsqr: 2}`. If the map function emits number arrays, `_stats` should calculate one stats object per array position; for emitted values `[1,2,3]`, `[4,5,6]`, and `[7,8,9]`, the reduced value should be `[{sum: 12, count: 3, min: 1, max: 7, sumsqr: 66}, {sum: 15, count: 3, min: 2, max: 8, sumsqr: 93}, {sum: 18, count: 3, min: 3, max: 9, sumsqr: 126}]`. If there are no values to reduce, `_stats` should produce `{sum: 0, min: null, max: null, count: 0, sumsqr: 0}`. Invalid emitted values should reject the `db.query()` promise: strings are invalid, mixing scalar numbers with arrays is invalid, and arrays with inconsistent lengths are invalid."} {"task_id": "format-code-task-000427", "source_id": "format-code-task-000427", "domain": "code", "task_path": "tasks/format-code-task-000427", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6219ee480cdd0dc87d2eeeb62f96733d66a1d909428abac77a74f25ca6824252", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n1. Please describe the issue you observed:\n 1. start local RocketMQ cluster \n 2. use example/consumer/simple/main.go start a consumer.\n 3. use ./mqadmin resetOffset tool to reset offset.\n 4. consumer process exit with.\n \n INFO[0005] receive reset consumer offset request... broker=\"127.0.0.1:30931\" consumerGroup=groupB timestamp=631123200000 topic=topicB\npanic: runtime error: index out of range [1] with length 1\ngoroutine 113 [running]:\ngithub.com/apache/rocketmq-client-go/v2/internal.(*ResetOffsetBody).Decode(0xc000461d20, {0xc00040c100, 0xfd, 0xfd})\n /Users/bytedance/github/rocketmq-client-go/internal/model.go:309 +0xa45\ngithub.com/apache/rocketmq-client-go/v2/internal.GetOrNewRocketMQClient.func5(0xc000416a80, {0x14ed5c8, 0xc0003260c0})\n /Users/bytedance/github/rocketmq-client-go/internal/client.go:303 +0x3f8\ngithub.com/apache/rocketmq-client-go/v2/internal/remote.(*remotingClient).processCMD.func2()\n /Users/bytedance/github/rocketmq-client-go/internal/remote/remote_client.go:207 +0x94\ngithub.com/apache/rocketmq-client-go/v2/primitive.WithRecover(0xc000421ef0)\n /Users/bytedance/github/rocketmq-client-go/primitive/base.go:100 +0x47\ncreated by github.com/apache/rocketmq-client-go/v2/internal/remote.(*remotingClient).processCMD\n /Users/bytedance/github/rocketmq-client-go/internal/remote/remote_client.go:206 +0x378\nExiting.\n ```\n\n2. Please tell us about your environment:\n mac local \n master client-go code.\n rocketmq broker with fastjson-1.2.76.jar\n\n3. Other information (e.g. detailed explanation, logs, related issues, suggestions on how to fix, etc):\nthe test code in decode ResetOffsetBody json format is different from mine\n\nmine from rocketmq broker response rpc\n```json\n{\"offsetTable\":{{\"brokerName\":\"RaftNode00\",\"queueId\":0,\"topic\":\"topicB\"}:0,{\"brokerName\":\"RaftNode00\",\"queueId\":1,\"topic\":\"topicB\"}:0,{\"brokerName\":\"RaftNode00\",\"queueId\":2,\"topic\":\"topicB\"}:0,{\"brokerName\":\"RaftNode00\",\"queueId\":3,\"topic\":\"topicB\"}:0}}\n```\n\nany different from fastjson config ?"} {"task_id": "format-code-task-000428", "source_id": "format-code-task-000428", "domain": "code", "task_path": "tasks/format-code-task-000428", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c32a5965bbace00a6bae9a1f85a92c4f3aec47ba6a3ea3cd213e5cc4672fd43f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a runtime factory for optimizers\n\nSINGA's C++ model library provides several gradient-descent optimizers (vanilla\nSGD with momentum, Nesterov, Adagrad and RMSProp). Today the only way to obtain\none is to name the concrete class at compile time. Code that drives training\nfrom a configuration or from a binding layer needs to pick the optimizer at\nruntime from a plain type name, so we want a small factory in the optimizer\npublic API.\n\nAdd a function\n\n```cpp\nstd::shared_ptr singa::CreateOptimizer(const std::string& type);\n```\n\nthat constructs the optimizer selected by `type` and returns it through the\ncommon `Optimizer` base class, so the caller can drive it polymorphically\n(`Setup`, `Apply`, etc.) without knowing the concrete type. It must be reachable\nfrom the optimizer's public header, i.e. any translation unit that includes the\noptimizer header can call it.\n\nRequirements:\n\n- The accepted type names are exactly `\"SGD\"`, `\"Nesterov\"`, `\"Adagrad\"` and\n `\"RMSProp\"`. Each must yield an optimizer that performs that algorithm's\n update: an object created with a given name, once set up and applied, must\n produce the same parameter update as constructing that optimizer class\n directly and applying it the same way.\n- The returned handle must be usable purely through the `Optimizer` base\n interface.\n- Passing any other (unrecognized) string must fail fast — abort with a fatal\n error rather than returning a null or otherwise unusable optimizer.\n\nNo change to the optimizers' update math is expected; this is about being able\nto obtain them by name.\n"} {"task_id": "format-code-task-000430", "source_id": "format-code-task-000430", "domain": "code", "task_path": "tasks/format-code-task-000430", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f1436edc487a7a34496f2bf32f68e6304d5cb2bd30dc0eb8e2ce900b5031b872", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Replica-aware node selection for the round-robin selector\n\nOur cluster places the shards of a group across the available data nodes using the\nround-robin node selector. Today the selector can only tell you which node owns a\ngiven shard. We're adding replication, so a shard now has a primary copy plus zero\nor more replicas, and we need the selector to tell us which node owns each copy.\n\nPlease extend the round-robin selector so a caller can ask for the node that owns a\nspecific replica of a shard, identified by a zero-based replica index. Replica `0`\nis the primary copy.\n\nRequirements for the observable behavior:\n\n- Asking for the node of a shard **without** specifying a replica must keep working\n exactly as it does now, and must return the same node as asking for replica `0`\n of that shard. Existing callers must not have to change.\n- For the same shard, distinct replica indices must map to **distinct** nodes, as\n long as there are at least that many data nodes available. With `N` data nodes,\n the copies for a shard (the primary together with its replicas, replica indices\n `0` through `N-1`) must be spread across `N` different nodes — no two copies of\n the same shard may land on the same node.\n- Selection must be deterministic and stable: repeated lookups for the same\n (group, shard, replica) always return the same node, regardless of how many other\n lookups happen in between or in what order.\n- The existing error behavior must be preserved for every replica index: if there\n are no data nodes, or the group/shard is unknown, picking any replica of that\n shard must return an error rather than a node.\n"} {"task_id": "format-code-task-000431", "source_id": "format-code-task-000431", "domain": "code", "task_path": "tasks/format-code-task-000431", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:48f547fd3bf484a70b0526f7b7343551e1c5093442093d9d38d9c51faaef6ae5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nSolr has a feature that gives an HTTP endpoint to ask Solr Nodes to gracefully terminate.\n\nKuberentes has container lifecycle hooks that allow for post-start and pre-stop actions: https://kubernetes.io/docs/tasks/configure-pod-container/attach-handler-lifecycle-event/#define-poststart-and-prestop-handlers\n\nWe should explicitly use the ./solr stop -p option in Solr and call this as a pre-stop action. That way Solr can determine how to best stop itself, until a timeout is reached and kubernetes stops the process itself.\n\nThere are some niceties given for stopping in docker-solr, but I'm not sure how much we want to dedicate ourselves to this image of Solr. Something to think about."} {"task_id": "format-code-task-000433", "source_id": "format-code-task-000433", "domain": "code", "task_path": "tasks/format-code-task-000433", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:32392d60c76b33974fa5fff815eb657ec130ca2757938a86226225617484f153", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n`{\"controller\": \"component\", \"controllerGroup\": \"apps.kubeblocks.io\", \"controllerKind\": \"Component\", \"Component\": {\"name\":\"mongo-mongodb\",\"namespace\":\"default\"}, \"namespace\": \"default\", \"name\": \"mongo-mongodb\", \"reconcileID\": \"1e07239f-4193-481b-8a68-9782c1817228\", \"error\": \"role selector for service is not defined, service: default, role: leader\"`"} {"task_id": "format-code-task-000434", "source_id": "format-code-task-000434", "domain": "code", "task_path": "tasks/format-code-task-000434", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7c64b722773830308ac7e54d4db3cfad1d7c4a245e95a46840841011975f1a6f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\ndredd init generates invalid file\n**Describe the bug**\nI ran `dredd init` which generates a `dredd.yml` config file.\nWhen I run `dredd --dry-run` I get the following error:\n\n```\nerror: unknown tag ! at line 4, column 47:\n ... g:yaml.org,2002:js/undefined> ''\n```\n\nLooking in the generated config there is this line:\n```yaml\nlanguage: ! ''\n```\n\nThis seems to be what is broken.\n\nDeleting that line fixes the problem.\n\n**What's your `dredd --version` output?**\n\n```\ndredd v8.0.0 (Darwin 18.2.0; x64)\n```"} {"task_id": "format-code-task-000435", "source_id": "format-code-task-000435", "domain": "code", "task_path": "tasks/format-code-task-000435", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e1f9bcb68ecaf2bf1569d43a4e2ae50465a00fa714a5c5f2286da761eba7871f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nbug: `SitemapRequestList.persistState()` throws when sitemap loading has finished\nIf `SitemapRequestList` finishes parsing the remote sitemaps and `persistState()` is called, it throws an exception.\n\nThis is because after parsing the sitemaps fully, the `SitemapRequestList` [closes the internal stream with `.push(null)`](https://github.com/apify/crawlee/blob/f3eb99d9fa9a7aa0ec1dcb9773e666a9ac14fb76/packages/core/src/storages/sitemap_request_list.ts#L364).\n\nWhen `persistState()` is called afterwards, the contents of the stream are read into a (persisted) list. To not mutate the internal state with `persistState()`, we return the stream contents back to the original stream - if the stream had been closed, this causes an exception (push after EOF)."} {"task_id": "format-code-task-000436", "source_id": "format-code-task-000436", "domain": "code", "task_path": "tasks/format-code-task-000436", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ee733b5930de350ec090b11c2c85dc0f9245d1e9176102961828cd0ddd011d89", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Support graphql-java 14\n\n`federation-jvm` is currently pinned to `graphql-java` 13.0 (see `graphql-java.version` in the parent `pom.xml`). graphql-java 14 has been out for a while now and we'd like to be able to use it together with this library.\n\nWhen I tried to override the `graphql-java` version to 14 in my own project (which depends on `graphql-java-support`), the build no longer compiles — there are a number of source-incompatible API changes between graphql-java 13 and 14 that this module hits directly. So just bumping the dependency in a downstream consumer doesn't work, the library itself needs to be updated.\n\nCould the `graphql-java` dependency be bumped to 14 and the code adapted accordingly?\n\nI understand this would be a backwards-incompatible change for consumers that are still on 13, but at this point I think it's worth it — staying on 13 is becoming a real problem for projects that want to take advantage of fixes/features in newer graphql-java releases.\n\nFor reference, my use case is a federated subgraph built with `Federation.transform(...)`, exposed via `_service { sdl }` to an Apollo gateway, so anything around schema generation, the runtime wiring, and the SDL printing path needs to keep working end-to-end after the upgrade (the SDL has to be something the gateway will accept for composition)."} {"task_id": "format-code-task-000437", "source_id": "format-code-task-000437", "domain": "code", "task_path": "tasks/format-code-task-000437", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c215cb637646d1da302a8620330180959e42b0fee77e2f393eb34d178f9121c7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nComplete the MIL implementation of reverse_sequence so sequence models can be evaluated and validated consistently with TensorFlow's reverse_sequence contract. The current public builder can lower the operation, but materialized inputs do not produce a compile-time value and malformed lengths or axes are allowed to reach later stages.\n\nFor mb.reverse_sequence, lengths is a rank-1 tensor with one entry for every position of the batch axis. Normalize negative seq_axis and batch_axis values against the input rank; both axes must be in range and must be different. When either axis is omitted, use TensorFlow-compatible defaults seq_axis=0 and batch_axis=1. For each batch index, reverse exactly the first lengths[index] elements along seq_axis while leaving the remaining elements and all other dimensions unchanged. A length of zero leaves that batch slice unchanged, and a length equal to the sequence dimension reverses the complete slice. Materialized lengths must be within [0, size of the sequence axis], and invalid rank, length count, bounds, or axis combinations must be rejected during graph construction with a clear ValueError or IndexError.\n\nWhen x and lengths are materialized constants, the builder must expose the exact reversed tensor through value inference. The result must retain the input tensor's shape and dtype, including for non-leading batch/sequence axes and negative-axis spellings. Symbolic graph construction must continue to infer the unchanged output shape and dtype without constant folding."} {"task_id": "format-code-task-000438", "source_id": "format-code-task-000438", "domain": "code", "task_path": "tasks/format-code-task-000438", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5181ba874e8d7c2edb6c677b7bcbe669adfae35645ba23bc916c48126e2d740a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a `generate(model: torch.nn.Module, tokenizer: Any, input_text: str, max_length: int, temperature: float = 1.2, top_p: float = 0.95, top_k: int = 0, scaler_state: Optional[Dict[str, Any]] = None, device: torch.device = torch.device(\"mps\")) -> str` function for autoregressive text generation.\n\nThe function should encode `input_text` with `tokenizer.encode(input_text, return_tensors=\"pt\")`, move the token tensor to `device`, then append exactly `max_length` sampled tokens before returning `tokenizer.decode(..., clean_up_tokenization_spaces=True)` for the full prompt plus continuation. If `max_length=0`, it should return the decoded prompt without running the model; for example, with a tokenizer whose `encode(\"hi\")` returns `[[10, 11]]` and whose decode maps `[10, 11]` to `\"hi\"`, `generate(model, tokenizer, \"hi\", max_length=0, device=torch.device(\"cpu\"))` returns `\"hi\"`.\n\nFor each generated token, it should run the model functionally on the whole growing sequence with a causal mask, read `outputs[\"output_representation\"][:, -1, :]`, divide those logits by `temperature`, apply softmax, optionally filter the probabilities, sample one token with `torch.multinomial(..., num_samples=1)`, and concatenate it to the sequence. With a toy tokenizer where `encode(\"A\")` returns `[[0]]`, decode maps `[0, 1, 1]` to `\"ABB\"`, and a model whose final-token logits make token `1` the only remaining token under `top_k=1`, `generate(model, tokenizer, \"A\", max_length=2, top_k=1, device=torch.device(\"cpu\"))` should return `\"ABB\"`.\n\n`top_p` should implement nucleus filtering when `0.0 < top_p < 1.0`: sort probabilities descending, keep the smallest prefix whose cumulative mass reaches the threshold, zero the rest, and renormalize before sampling. `top_k` should apply when `top_k > 0`: cap `k` at the vocabulary size, zero probabilities below the kth-highest probability threshold, and renormalize. If both controls are enabled, nucleus filtering should run before top-k filtering.\n\nWhen `scaler_state` is provided, generation should run the model under `torch.autocast` using `scaler_state[\"precision\"]`; otherwise it should run without autocast. The function should perform generation under `torch.no_grad()` and return a Python string."} {"task_id": "format-code-task-000439", "source_id": "format-code-task-000439", "domain": "code", "task_path": "tasks/format-code-task-000439", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:eea977b9671f600a5b7fac63d635d6845f89bd1c5501bd3a77d3a2066173bf5d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nRuntime selection is used by checksum and package-processing paths, but broad selectors currently cannot express exceptions, and slash-form selectors accept unsupported platform names. Extend the existing `runtime.GetRuntimes` and `runtime.GetRuntimesFromEnvs` APIs into a validated, ordered selector contract without changing their signatures.\n\nA selector body is exactly one of `all`, an OS (`darwin`, `linux`, or `windows`), an architecture (`amd64` or `arm64`), or a valid `OS/architecture` pair from those values. Trim surrounding whitespace before parsing; empty bodies, internal whitespace, extra slash components, or unknown OS/architecture values are invalid. `GetRuntimes` accepts positive selector bodies only. It must retain the established expansions: an OS expands to `amd64` then `arm64`; an architecture expands across `darwin`, `windows`, then `linux`; `all` expands in canonical order (`darwin/amd64`, `darwin/arm64`, `linux/amd64`, `linux/arm64`, `windows/amd64`, `windows/arm64`); and an exact pair yields only that pair.\n\n`GetRuntimesFromEnvs` additionally accepts an exclusion written as one leading `!` immediately followed by a selector body. Positive selectors form an ordered union: process them in input order, use each selector's expansion order, and keep only the first occurrence of each runtime. Exclusions subtract every matching runtime regardless of whether they appear before or after a positive selector, so a later positive selector must not re-add an excluded runtime and `all` must not bypass exclusions. When a non-nil list has at least one positive selector, its starting set is empty. When it contains exclusions but no positive selector, start from the canonical `all` set and subtract. Preserve the existing distinction that a nil list means canonical `all`, while an explicitly empty non-nil list means no runtimes.\n\nValidation is atomic. If either API receives an invalid selector, return a nil runtime slice and a non-nil error rather than any partial selection. For every non-empty invalid token, the error text must include that token after surrounding whitespace has been trimmed so callers can identify the bad entry. Trim outer whitespace before recognizing `!`; a bare or repeated `!` is invalid. Results are caller-owned: every call must return fresh slice storage and fresh `Runtime` values, so mutating one result cannot change a later result from either API."} {"task_id": "format-code-task-000440", "source_id": "format-code-task-000440", "domain": "code", "task_path": "tasks/format-code-task-000440", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c9b654acbb0d4502433f558f15176d3cf8f7766d3901abda40776676c33dec4c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## CloudFormation `Condition:` references don't seem to be evaluated\n\nI'm scanning a CloudFormation template that uses the top-level `Conditions:` block, and several places in the template reference those conditions by name. The references don't appear to take effect during scanning — it's as if the condition is never actually evaluated.\n\nMinimal template that shows the problem:\n\n```yaml\nParameters:\n Env:\n Type: String\n Default: prod\n\nConditions:\n IsProd: !Equals [!Ref Env, \"prod\"]\n\nResources:\n MyBucket:\n Type: AWS::S3::Bucket\n Properties:\n BucketName: !If\n - IsProd\n - prod-bucket\n - dev-bucket\n\nOutputs:\n MaybeArn:\n Condition: IsProd\n Value: !GetAtt MyBucket.Arn\n```\n\nWith `Env=prod` I'd expect `IsProd` to evaluate to true, so:\n\n- the `!If` should pick the `prod-bucket` branch, and\n- the `MaybeArn` output (gated by `Condition: IsProd`) should be present.\n\nWhen I scan this, the behaviour I see downstream doesn't match either expectation — it looks like the `Condition: IsProd` reference in the output, and condition references in general, are not being resolved against the `Conditions:` block at all. Plain things like `!Equals`, `!Ref` on parameters, etc. seem fine on their own; it's specifically the by-name reference to a declared condition that doesn't seem to do anything.\n\nBoth forms of the reference should work — the long form:\n\n```yaml\nOutputs:\n MaybeArn:\n Condition: IsProd\n Value: ...\n```\n\nand the YAML short-tag form:\n\n```yaml\nSomeValue: !Condition IsProd\n```\n\nIt would be great if these references were actually resolved to the value of the named condition from the `Conditions:` block, so that anything depending on them (If branches, conditional outputs, etc.) behaves the way the template intends."} {"task_id": "format-code-task-000441", "source_id": "format-code-task-000441", "domain": "code", "task_path": "tasks/format-code-task-000441", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:efd4dc50b9670fc2384ead9c41ab013ecac07aa47bb66938ee5600034b393c53", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nHey, I'm noticing something weird with starboard. When I run `starboard scan vulnerabilityreports deploy/my-app` from the CLI, the VulnerabilityReport that gets created has the Deployment itself as its owner. But when the starboard operator running in my cluster scans the same Deployment, the report it produces is owned by the current ReplicaSet instead. So I end up with two reports for the same workload that never line up, and anything that looks them up by owner ref gets confused. Same kind of thing happens for Pods controlled by Jobs/CronJobs — the CLI just attributes the report to whatever I passed in, while the operator walks up the chain. Can you take a look?\n\n## Expected outcomes\n\n- CLI-created reports should use the same externally visible report owner that the operator would use for the same built-in Kubernetes workload, instead of always using the object named on the command line.\n- Deployment scans should be reported against the Deployment’s active child workload object, so the generated report labels and owner references line up with operator-created reports for that Deployment.\n- Workloads reached through normal Kubernetes controller ownership chains, including Pods managed through Job/CronJob-style controllers, should be attributed consistently with the operator’s owner-selection behavior; unmanaged workloads or workloads controlled by unrelated/custom controllers should not be incorrectly reassigned.\n- The same owner-resolution behavior should be applied to both vulnerability reports and config-audit reports created by the CLI.\n- When a Deployment’s expected active child workload cannot be found, callers should receive a distinguishable not-found condition rather than silently falling back to the Deployment itself.\n- Unsupported workload objects should fail with a clear error that identifies the unsupported kind/type.\n\n## Implementation notes\n\n- Keep the fix focused on externally visible report ownership: report labels, owner references, and any public owner-resolution behavior should agree with the operator’s semantics.\n- The internal structure, helper names, lookup strategy, and call sites used to perform owner resolution are up to the implementation.\n- Use Kubernetes object metadata and controller relationships consistently with existing repository conventions."} {"task_id": "format-code-task-000442", "source_id": "format-code-task-000442", "domain": "code", "task_path": "tasks/format-code-task-000442", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:68c509f7bfd2de77acefbcef47e56cfa2b5bb71920f0f9df68532f4432f1f0b0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a public Go library function `vm.Eval(ctx expr.EvalContext, arg expr.Node) (value.Value, bool)` that evaluates an already-parsed SQL-style expression node against a row/message context. Callers can build the node with the existing parser, for example `expr.MustParse(\"int5 + 5\")`, and pass a context such as `datasource.NewContextSimpleNative(map[string]interface{}{\"int5\": 5, \"str5\": \"5\", \"email\": \"bob@bob.com\", \"urls\": []string{\"abc\", \"123\"}, \"hits\": map[string]int64{\"google.com\": 5}})`.\n\nFor `vm.Eval(ctx, expr.MustParse(\"int5 + 5\"))`, it should return a `value.IntValue` whose underlying value is `int64(10)` and `ok == true`. For `vm.Eval(ctx, expr.MustParse(\"toint(str5) + 6\"))` after `builtins.LoadAllBuiltins()`, it should return `int64(11)` with `ok == true`, meaning function nodes evaluate their arguments and call the registered evaluator. For boolean and comparison expressions, `vm.Eval(ctx, expr.MustParse(\"email == \\\"bob@bob.com\\\"\"))` should return boolean true, while `vm.Eval(ctx, expr.MustParse(\"email != \\\"bob@bob.com\\\"\"))` should return boolean false.\n\nThe evaluator should support SQL-style operators over typed values: arithmetic `+`, `-`, `*`, `/`, `%`; boolean `AND`, `OR`, `&&`, `||`, `!`; comparisons `=`, `==`, `!=`, `>`, `>=`, `<`, `<=`; `LIKE` patterns using `*` and `%`; `contains`; `IN`; `INTERSECTS`; and `BETWEEN`. For example, `vm.Eval(ctx, expr.MustParse(\"email LIKE \\\"bob*\\\"\"))` should return true, `vm.Eval(ctx, expr.MustParse(\"[1,2,3,5] contains int5\"))` should return true, and `vm.Eval(ctx, expr.MustParse(\"\\\"google.com\\\" IN hits\"))` should return true. Missing fields should not panic: `vm.Eval(ctx, expr.MustParse(\"missing > 21\"))` should return boolean false with `ok == true`, while unsupported type combinations such as `vm.Eval(ctx, expr.MustParse(\"\\\"hello\\\" == split(\\\"hell-no\\\", \\\",\\\")\"))` should report failure with `ok == false`. Repeating the same deterministic expression with the same context should produce the same value, and evaluating an expression should not mutate the caller's context or perform filesystem or network I/O."} {"task_id": "format-code-task-000444", "source_id": "format-code-task-000444", "domain": "code", "task_path": "tasks/format-code-task-000444", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ec78a7603e31c2bf6ea67ce2a1a269dadbe68143ba3af09af29b19c3941f752f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Time to refresh the CI setup\n\nThe `.github/workflows/pre-commit.yml` workflow and the pre-commit config are starting to feel out of date and the test suite is needlessly slow / network-dependent. Filing this so we don't lose track. Concretely:\n\n**Stale CI dependencies.** The workflow is still pinned to old major versions of the standard GitHub Actions (checkout etc.), and the pre-commit hook revs (black, isort, the built-in hooks repo) have all gone through several releases since the versions we have committed. We should bump them and re-run the formatters, which will probably cause a small amount of churn in the codebase as the newer black makes slightly different formatting choices.\n\n**Python version matrix.** We're still testing against 3.9 in the matrix even though it's basically at end of life, and we're not testing against 3.12 at all. Time to drop the old one and add the new one. The package metadata should also be updated to reflect the minimum supported Python.\n\n**Tests reach out to the network.** Several tests load `gpt2` from HuggingFace as a small \"real\" model to exercise the merging machinery. This has two annoying consequences:\n- CI runs spend non-trivial time downloading the weights on every fresh runner.\n- On a developer machine without network access (planes, locked-down corp networks, etc.) the test suite just fails to start.\n\nSince these tests don't actually care about the *values* in the weights — they only care that the architecture is exercised end-to-end — we don't need the real pretrained checkpoint. They should be reworked to use a small in-process model with the same architecture so no network is required to run `pytest`.\n\nWhile doing the above we may also want to clean up a couple of harmless warnings that show up in the test output and add a couple of additional sanity-check pre-commit hooks for the non-Python config files we have in the repo."} {"task_id": "format-code-task-000445", "source_id": "format-code-task-000445", "domain": "code", "task_path": "tasks/format-code-task-000445", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7c8c9f1374985175ecfb5cbafdadc0b9a47389fc84a51aa916f5d08e37b5b9a5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## websocket example1 is more verbose than it needs to be, and client/server disagree on the wire format\n\nWhile going through `topics/web/sockets/example1`, the server-side handler in `main.go` looks more complicated than it should be for a teaching example. It manually allocates a 512-byte buffer, calls `ws.Read` into it, checks `n > 0`, slices the bytes back into a string, builds a `Message`, and then constructs a `json.Encoder` around the connection to write the response. For an introductory websocket sample this is a lot of plumbing, and most of it is exactly the kind of boilerplate the `golang.org/x/net/websocket` package is supposed to spare you from.\n\nWhile reading through it I also noticed the two sides don't actually agree on the wire format:\n\n- The browser side in `static/app.js` sends the raw input string:\n ```js\n ws.send(val);\n ```\n- The server, however, never decodes that as JSON on the way in (it just reads bytes), but then turns around and JSON-encodes the response, and the client side does `JSON.parse(evt.data)` on what comes back.\n\nSo the inbound direction is \"raw text\" and the outbound direction is \"JSON\", even though everything around it is dressed up like it's a JSON conversation. It's confusing to read as an example — a learner can't tell what the intended convention is.\n\nTwo things I'd like to see:\n\n1. Simplify the server handler so it isn't doing manual buffer management and manual JSON encoding when the websocket library can handle that for you. The current shape buries the actual lesson (read a message, transform it, write a message back) under buffer-size and byte-slice noise.\n2. Make the client and the server agree: if the response is JSON, the request should be JSON too, so the example is internally consistent and a reader doesn't have to wonder whether the asymmetry is intentional.\n\nThe externally visible behavior should stay the same — the browser types a string, the server echoes back a `Message` with `original`, `formatted` (uppercased), and `received`, and the page renders it. This is purely about making the example shorter and self-consistent."} {"task_id": "format-code-task-000448", "source_id": "format-code-task-000448", "domain": "code", "task_path": "tasks/format-code-task-000448", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ebfdf272414497eb1eb65109541720927b440781812e4f3b61337660482a0bb2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nargocd-tls-certs-cm is overwritten on any change as of v0.2.0\n**Describe the bug**\nPrior to v0.2.0, certificates for TLS validation could be managed via the ArgoCD GUI or by controlling the argocd-tls-certs-cm ConfigMap directly. The value of argocd-tls-certs-cm was set initially when ArgoCD was deployed but not altered thereafter.\n\nAfter upgrading tov0.2.0, this ConfigMap is reset to the value of tls.intitialCerts in the ArgoCD CR after any change. This renders the certificate management via the GUI or previous direct ConfigMap control useless and deletes any certificates that were added to the ConfigMap prior to the upgrade.\n\nWorkaround:  Add the needed cert to ArgoCD CR spec  tls.initialCerts.  \nSee https://argocd-operator.readthedocs.io/en/latest/reference/argocd/#tls-options.\n\nWorkaround example: https://github.com/iam-veeramalla/openshift-gitops-examples/blob/master/argocd/GITOPS-1725/argocd-initialTLScerts.yaml\n\n**Expected behavior**\nRevert back to the original behavior\n\n**Screenshots**\nIf applicable, add screenshots to help explain your problem.\n\n**Additional context**\nAdd any other context about the problem here."} {"task_id": "format-code-task-000449", "source_id": "format-code-task-000449", "domain": "code", "task_path": "tasks/format-code-task-000449", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:69cba9906273558ed449bac242faeff2ec1f3e6302a50f1f654c43feb27dffe2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n**Feature request: allow extra command line arguments on the repo-server via the ArgoCD CR**\n\nWe're running Argo CD through the operator and recently wanted to tune a flag on the argocd-repo-server (for example, something like `--reposerver.max.combined.directory.manifests.size 10M`) that doesn't have a dedicated knob in the ArgoCD spec yet. The repo-server Deployment is fully managed by the operator, so when we just `kubectl edit` the Deployment to add the flag to the container command, the next reconcile reverts it.\n\nI noticed that `spec.server.extraCommandArgs` already gives us an escape hatch for the argocd-server component — we can use it to pass extra flags that the operator doesn't surface as first-class fields, and they show up on the server Pod's command. There doesn't seem to be an equivalent for the repo-server side of `spec`, though.\n\nThe practical problem this creates: every time Argo CD upstream adds a new repo-server flag we want to use, we either have to wait for an operator release that exposes it as a typed field, or we have to fork/patch the operator. For flags we only need temporarily, or that are niche enough that they may never get a dedicated field, this is a lot of friction.\n\nWould it be possible to expose the same kind of escape hatch on the repo section of the ArgoCD CR, so users can declare a list of extra command line arguments that get added on top of whatever the operator already generates for the repo-server (without replacing any of those defaults)? That way the operator stays in charge of the baseline command, but we don't get blocked on operator releases whenever upstream introduces a new repo-server flag.\n\nThe new field on `spec.repo` could be named something like `extraRepoCommandArgs`."} {"task_id": "format-code-task-000450", "source_id": "format-code-task-000450", "domain": "code", "task_path": "tasks/format-code-task-000450", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f069a9b20f6cc9b0acbbb28e19545d51001a1a84917de01b4b08b1720693544f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n我在同一个 asyncio event loop 里用 `asyncio.gather` 并发构建多个 Hera `Workflow`，每个里面都有自己的 `with DAG(...)` 和 tasks，但跑出来的 `to_yaml()` 有时会把 A workflow 的 task 混到 B 里，有时又少几个 task，重复跑结果还不一样。\n\n**Expected outcomes**\n- 在同一个 asyncio event loop 里并发构建多个 `Workflow` 时，各自 `with` 块内声明的 `DAG` 和 task 只能进入对应的 `Workflow`，不能互相串扰。\n- 同样的并发输入重复运行时，生成结果应保持稳定；不应出现 task 混入、遗漏，或因交错时序不同而变化的 YAML 内容。\n- 顺序创建多个 `Workflow` 也应保持彼此隔离，后一个构建不能继承前一个构建残留的节点。\n\n**Implementation notes**\n具体的状态保存方式、上下文切换方式和内部组织由实现者自行决定；只要满足并发和顺序场景下的隔离与稳定输出即可。"} {"task_id": "format-code-task-000451", "source_id": "format-code-task-000451", "domain": "code", "task_path": "tasks/format-code-task-000451", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1ade211abf06173614bebd12ab3b4de733de7cfa5d8fd15b457c8852a9e24395", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nArgo CD's dex integration overwrites a custom dex.config.oauth2 yaml block\n**Describe the bug**\n\nWhen\n\n**To Reproduce**\n\nDeploy Argo CD with the following data for the `argo-cm` ConfigMap\n\n```yaml\napiVersion: v1\ndata:\n application.instanceLabelKey: argocd.argoproj.io/instance\n dex.config: |\n oauth2:\n passwordConnector: ldap # I want this enabled in dex to leverage the password grant type against LDAP\n connectors:\n - type: ldap\n name: LDAP\n id: ldap\n config:\n host: myldap.company.com\n usernamePrompt: Username\n userSearch:\n baseDN: \"cn=users,dc=example,dc=com\"\n username: uid\n idAttr: uid\n emailAttr: uid\n nameAttr: uid\n groupSearch:\n baseDN: \"cn=groups,dc=freeipa,dc=example,dc=com\"\n nameAttr: cn\n url: https://argocd.company.com/\nkind: ConfigMap\n```\n\nWhen I examine generated dex config in the argocd-dex-server pod I find this\n\n```yaml\n# /shared/dex.yaml \nconnectors:\n- config:\n groupSearch:\n baseDN: cn=groups,dc=freeipa,dc=example,dc=com\n nameAttr: cn\n host: myldap.company.com\n userSearch:\n baseDN: cn=users,dc=example,dc=com\n emailAttr: uid\n idAttr: uid\n nameAttr: uid\n username: uid\n usernamePrompt: Username\n id: ldap\n name: LDAP\n type: ldap\ngrpc:\n addr: 0.0.0.0:5557\nissuer: https://argocd.company.com//api/dex\noauth2:\n skipApprovalScreen: true # passwordConnector was removed\nstaticClients:\n- id: argo-cd\n name: Argo CD\n redirectURIs:\n - https://argocd.company.com/auth/callback\n secret: \"\"\n- id: argo-cd-cli\n name: Argo CD CLI\n public: true\n redirectURIs:\n - http://localhost\n - http://localhost:8085/auth/callback\nstorage:\n type: memory\ntelemetry:\n http: 0.0.0.0:5558\nweb:\n http: 0.0.0.0:5556\n\n```\n\nI expected to see \n\n```yaml\noauth2:\n passwordConnector: ldap\n skipApprovalScreen: true\n```\nbut instead I got:\n\n```yaml\noauth2:\n skipApprovalScreen: true\n```\n\n**Expected behavior**\n\nI want to be able to generate an oauth token with dex using the password grant-type when integrating with LDAP. Therefore, when I pass in the following yaml to the argocd-cm ConfMap\n\n```yaml\napiVersion: v1\ndata:\n application.instanceLabelKey: argocd.argoproj.io/instance\n dex.config: |\n oauth2:\n passwordConnector: ldap\n # Rest of the config....\n```\nargocd-dex will take that configuration into account when generating the dex configuration file.\n\n**Version**\n\n```shell\nargocd: v2.1.2+7af9dfb\n BuildDate: 2021-09-02T18:05:23Z\n GitCommit: 7af9dfb3524c13e941ab604e36e49a617fe47d2e\n GitTreeState: clean\n GoVersion: go1.16.5\n Compiler: gc\n Platform: linux/amd64```"} {"task_id": "format-code-task-000452", "source_id": "format-code-task-000452", "domain": "code", "task_path": "tasks/format-code-task-000452", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:06bbee3c31e3c218a96f92571322db6a1728016459e3dfa36bf47ddf8639a525", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\npublic Api could create invalid workflow which has message Pod \"\" is invalid: spec.containers[1].image: Required value\nChecklist:\n\n * [X] I've included the version.\n * [X] I've included reproduction steps.\n * [ ] I've included the workflow YAML.\n * [ ] I've included the logs.\n \n**What happened**:\n\n**What you expected to happen**:\n\n**How to reproduce it (as minimally and precisely as possible)**:\nkubectl -n argo port-forward deployment/argo-server 2746:2746\n\nPOST http://localhost:2746/api/v1/workflows/argo-performance\n\nPOST Body: with wrong container properties imagex\n```json\n{\n \"namespace\": \"argo-performance\",\n \"workflow\": {\n \"apiVersion\": \"argoproj.io/v1alpha1\",\n \"kind\": \"Workflow\",\n \"metadata\": {\n \"name\": \"hello-world-right-env-12\"\n },\n \"spec\": {\n \"entrypoint\": \"whalesay\",\n \"templates\": [\n {\n \"name\": \"whalesay\",\n \"container\": {\n \"imagex\": \"docker/whalesay:latest\",\n \"command\": [\n \"cowsay\"\n ],\n \"args\": [\n \"hello world\"\n ],\n \"env\": []\n }\n }\n ]\n }\n }\n}\n```\n\nget response with status 200\n\n```json\n{\n \"metadata\": {\n \"name\": \"hello-world-right-env-13\",\n \"namespace\": \"argo-performance\",\n \"selfLink\": \"/apis/argoproj.io/v1alpha1/namespaces/argo-performance/workflows/hello-world-right-env-13\",\n \"uid\": \"6466d400-0440-46e5-9a21-9f141ef17431\",\n \"resourceVersion\": \"2441046\",\n \"generation\": 1,\n \"creationTimestamp\": \"2020-06-11T13:21:28Z\"\n },\n \"spec\": {\n \"templates\": [\n {\n \"name\": \"whalesay\",\n \"arguments\": {},\n \"inputs\": {},\n \"outputs\": {},\n \"metadata\": {},\n \"container\": {\n \"name\": \"\",\n \"command\": [\n \"cowsay\"\n ],\n \"args\": [\n \"hello world\"\n ],\n \"resources\": {}\n }\n }\n ],\n \"entrypoint\": \"whalesay\",\n \"arguments\": {}\n },\n \"status\": {\n \"startedAt\": null,\n \"finishedAt\": null\n }\n}\n```\n\nget error from argo CLI\n\n```bash\n➜ ~ argo get hello-world-right-env-12 -n argo-performance\nName: hello-world-right-env-12\nNamespace: argo-performance\nServiceAccount: default\nStatus: Error\nMessage: Pod \"hello-world-right-env-12\" is invalid: spec.containers[1].image: Required value\nConditions:\n Completed True\nCreated: Thu Jun 11 21:08:49 +0800 (17 minutes ago)\nStarted: Thu Jun 11 21:08:49 +0800 (17 minutes ago)\nFinished: Thu Jun 11 21:08:50 +0800 (17 minutes ago)\nDuration: 1 second\n\nSTEP TEMPLATE PODNAME DURATION MESSAGE\n ⚠ hello-world-right-env-12 whalesay hello-world-right-env-12 0s Pod \"hello-world-right-env-12\" is invalid: spec.containers[1].image: Required value\n```\n\n**Anything else we need to know?**:\n\nI cannot create this workflow via argo CLI\n```bash\n➜ argo git:(master) ✗ argo submit examples/hello-world.yaml -n argo-performance\n2020/06/11 21:11:50 Failed to parse workflow: error unmarshaling JSON: while decoding JSON: json: unknown field \"imagex\"\n```\n```yaml\napiVersion: argoproj.io/v1alpha1\nkind: Workflow\nmetadata:\n name: hello-world-001\nspec:\n entrypoint: whalesay\n templates:\n - name: whalesay\n container:\n imagex: docker/whalesay:latest\n command: [cowsay]\n args: [\"hello world\"]\n env: []\n```\n\n**Environment**:\n\n- Argo version:\n```\n$ argo version\n➜ ~ argo version\nargo: v2.8.1\n BuildDate: 2020"} {"task_id": "format-code-task-000453", "source_id": "format-code-task-000453", "domain": "code", "task_path": "tasks/format-code-task-000453", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:373460084a9fab9a08a17b4805821f13588c3069b8ab293e4262c7b639ec9391", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nWorkflow can't be terminated because of the containerSet template that was referenced in the DAG task\n## Summary\n\nThe problem occurs when running workflow with DAG task that is referencing `containerSet` template.\n\nIf you will try to terminate the workflow at the moment when DAG nodes have been initialized (in `Pending`) but not started yet, then the workflow will stuck at the running phase.\n\nI pressed \"Terminate\" button when nodes `step-1` and `step-2` were created and the result displayed below:\n\n![image](https://user-images.githubusercontent.com/58072595/151021410-28cfaa78-4f3e-4299-88eb-ee9614ff31ee.png)\n\n## Diagnostics\n\nThe behavior is consistent and can be reproduced with the following workflow:\n\n```yaml\napiVersion: argoproj.io/v1alpha1\nkind: Workflow\nmetadata:\n generateName: container-set-termination-demo-\nspec:\n entrypoint: main\n templates:\n - name: main\n dag:\n tasks:\n - name: using-container-set-template\n template: problematic-container-set\n - name: problematic-container-set\n containerSet:\n containers:\n - name: step-1\n image: alpine\n command:\n - sh\n - -c\n - \"sleep 10\"\n - name: step-2\n image: alpine\n command:\n - sh\n - -c\n - \"sleep 10\"\n```\nI have tried to change entrypoint from `main` to `problematic-container-set` and then the bug is not reproducible. So the issue is specific to DAG templates.\n\nAfter the failed termination the workflow has following state:\n```yaml\napiVersion: argoproj.io/v1alpha1\nkind: Workflow\nmetadata:\n annotations:\n workflows.argoproj.io/pod-name-format: v1\n creationTimestamp: \"2022-01-25T16:46:10Z\"\n generateName: container-set-termination-demo-\n generation: 6\n labels:\n workflows.argoproj.io/phase: Running\n workflows.argoproj.io/resubmitted-from-workflow: container-set-termination-demo-rjx8r\n name: container-set-termination-demo-449lw\n namespace: argo\n resourceVersion: \"45572\"\n uid: cbbc9bb8-63c5-4ce1-abce-c8fe7192cdea\nspec:\n activeDeadlineSeconds: 300\n arguments: {}\n entrypoint: main\n podSpecPatch: |\n terminationGracePeriodSeconds: 3\n shutdown: Terminate\n templates:\n - dag:\n tasks:\n - arguments: {}\n name: using-container-set-template\n template: problematic-container-set\n inputs: {}\n metadata: {}\n name: main\n outputs: {}\n - containerSet:\n containers:\n - command:\n - sh\n - -c\n - sleep 10\n image: alpine\n name: step-1\n resources: {}\n - command:\n - sh\n - -c\n - sleep 10\n image: alpine\n name: step-2\n resources: {}\n inputs: {}\n metadata: {}\n name: problematic-container-set\n outputs: {}\n ttlStrategy:\n secondsAfterCompletion: 600\nstatus:\n artifactRepositoryRef:\n artifactRepository:\n archiveLogs: true\n s3:\n accessKeySecret:\n key: accesskey\n name: my-minio-cred\n bucket: my-bucket\n endpoint: minio:9000\n insecure: true\n secretKeySecret:\n key: secretkey\n name: my-minio-cred\n configMap: artifact-repositories\n key: default-v1\n namespace: argo\n conditions:\n - status: \"False\"\n type: PodRunning\n finishedAt: null\n nodes:\n container-set-termination-demo-449lw:\n children:\n - container-set-termination-demo-449lw-773192577\n displayName: container-set-termination-demo-449lw\n finishedAt: null\n id: container-set-termination-demo-449lw\n name: container-set-termination-demo-449lw\n phase: Running\n progress: 1/1\n startedAt: \"2022-01-25T16:46:10Z\"\n templateName: main\n templateScope: local/container-set-termination-demo-449lw\n type: DAG\n container-set-termination-demo-449lw-773192577:\n boundaryID: container-set-termination-demo-449lw\n children:\n - container-set-termination-d"} {"task_id": "format-code-task-000454", "source_id": "format-code-task-000454", "domain": "code", "task_path": "tasks/format-code-task-000454", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3d0a8e20567d0f9a97c05266381cf08303c1b25cab2a24e1347497e1946d960d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nInternal Server Error When Fetching Logs From Inline Workflows\n### Pre-requisites\n\n- [X] I have double-checked my configuration\n- [X] I can confirm the issues exists when I tested with `:latest`\n- [ ] I'd like to contribute the fix myself (see [contributing guide](https://github.com/argoproj/argo-workflows/blob/master/docs/CONTRIBUTING.md))\n\n### What happened/what you expected to happen?\n\n### What happened?\nWhen I run a workflow that contains an inline component, I cannot view the logs for that container. The Argo UI shows an \"Internal Server Error\". However, when I check the minio browser, I can find the main.log file and retrieve it's contents without any issues.\n\nThis issue only occurs on inline components. When running other types of templates, I don't see this issue. \n\nFor example, these workflows run fine and I can see the logs in the Argo UI:\n[hello-world](https://github.com/argoproj/argo-workflows/blob/master/examples/hello-world.yaml)\n[steps](https://github.com/argoproj/argo-workflows/blob/master/examples/steps.yaml)\n\nBut, these workflows recreate the issue:\n[steps-inline-workflow](https://github.com/argoproj/argo-workflows/blob/master/examples/steps-inline-workflow.yaml)\n[dag-inline-workflow](https://github.com/argoproj/argo-workflows/blob/master/examples/dag-inline-workflow.yaml)\n\nThis issue also occurs with actual artifacts coming from containers. For example I can run a simple argosay container and generate a .txt file that shows up in the Argo UI. However if I move that container into an inline definition, the logs, and the artifact itself throw errors in the UI (while showing up just fine in minio). \n\n### What you expected to happen?\nI expected the logs and the artifacts to be accessible from the Argo UI despite the container being defined as inline.\n\n\n### Version\n\n3.4.1 and latest\n\n### Paste a small workflow that reproduces the issue. We must be able to run the workflow; don't enter a workflows that uses private images.\n\n```YAML\napiVersion: argoproj.io/v1alpha1\nkind: Workflow\nmetadata:\n generateName: hello-world-\n labels:\n workflows.argoproj.io/archive-strategy: \"false\"\nspec:\n entrypoint: main\n templates:\n - name: main\n steps:\n - - name: a\n inline:\n container:\n image: argoproj/argosay:v2\n args: [ echo, hello, /mnt/file.txt ]\n outputs:\n artifacts:\n - name: file\n path: /mnt/file.txt\n archive:\n none: { }\n```\n\n\n### Logs from the workflow controller\n\n```\ntime=\"2022-10-14T18:50:19.077Z\" level=info msg=\"Update leases 200\"\ntime=\"2022-10-14T18:50:24.082Z\" level=info msg=\"Get leases 200\"\ntime=\"2022-10-14T18:50:24.087Z\" level=info msg=\"Update leases 200\"\ntime=\"2022-10-14T18:50:29.093Z\" level=info msg=\"Get leases 200\"\ntime=\"2022-10-14T18:50:29.097Z\" level=info msg=\"Update leases 200\"\ntime=\"2022-10-14T18:50:34.101Z\" level=info msg=\"Get leases 200\"\ntime=\"2022-10-14T18:50:34.106Z\" level=info msg=\"Update leases 200\"\ntime=\"2022-10-14T18:50:36.217Z\" level=info msg=\"Processing workflow\" namespace=quilter-dev-sergiy workflow=hello-world-62qpl\ntime=\"2022-10-14T18:50:36.221Z\" level=info msg=\"Get configmaps 200\"\ntime=\"2022-10-14T18:50:36.221Z\" level=info msg=\"resolved artifact repository\" artifactRepositoryRef=\"quilter-dev-sergiy/#\"\ntime=\"2022-10-14T18:50:36.221Z\" level=info msg=\"Updated phase -> Running\" namespace=quilter-dev-sergiy workflow=hello-world-62qpl\ntime=\"2022-10-14T18:50:36.221Z\" level=info msg=\"Steps node hello-world-62qpl initialized Running\" namespace=quilter-dev-sergiy workflow=hello-world-62qpl\ntime=\"2022-10-14T18:50:36.221Z\" level=info msg=\"StepGroup node hello-world-62qpl-2899326954 initialized Running\" namespace=quilter-dev-sergiy workflow=hello-world-62qpl\ntime=\"2022-10-14T18:50:36.222Z\" level=info msg=\"Pod node hello-world-62qpl-75685911 initialized Pending\" namespace=quilter-dev-sergiy workflow=hello-world-62qpl\ntime=\"2022-10-14T18:5"} {"task_id": "format-code-task-000455", "source_id": "format-code-task-000455", "domain": "code", "task_path": "tasks/format-code-task-000455", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:80f81b5defc61ce17e39414d96c91d5987c0e60d65842848307712c7cfa4ec6d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI have a database schema that has an enum and some tables within a postgresql schema. When I run `atlas schema diff` or `atlas schema apply` it fails to emit the `CREATE TYPE` enum definition. Everything works fine if the enum is in the default schema, this only happens if the enum is part of another schema.\n\nSteps to reproduce:\n\nFirst, create `repro.sql` with the contents below\n```sql\nCREATE SCHEMA test;\nCREATE TYPE test.state AS ENUM ('a', 'b', 'c' );\nCREATE TABLE test.the_table (\n id int4 NOT NULL,\n value test.state NOT NULL\n);\n```\nThen run the following\n```sh\ntouch empty.sql\natlas schema diff --dev-url docker://postgres/15/test --from file://empty.sql --to file://repro.sql --format '{{ sql . \" \" }}'\n```\nThis will print out\n```sql\n-- Add new schema named \"test\"\nCREATE SCHEMA \"test\";\n-- Create \"the_table\" table\nCREATE TABLE \"test\".\"the_table\" (\n \"id\" integer NOT NULL,\n \"value\" \"test\".\"state\" NOT NULL\n);\n```\n... which does not have the definition for the `test.state` enum even though `test.the_table` refers to it.\n\nExtra notes:\n- `atlas schema inspect` does emit HCL that includes the `test.state` enum."} {"task_id": "format-code-task-000456", "source_id": "format-code-task-000456", "domain": "code", "task_path": "tasks/format-code-task-000456", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a2deaf252ad4c9d823dca49374570663b243d61816587b0c8e21c3da8b137290", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI'm using `schemaspec.Resource.As` to decode into a struct that has an interface field and a `[]MyInterface` field. The child blocks are registered extension types that implement that interface, but after `As` runs those fields are still nil/empty and the same blocks are still sitting in the remainder. I thought my registration might be wrong, but decoding the concrete extension types directly works.\n\nExpected outcomes:\n- Interface slice fields: when `(*Resource).As(target interface{}) error` decodes a struct containing a field of type `[]MyInterface`, registered child extension blocks whose concrete types implement `MyInterface` should be decoded into their concrete values and appended to that slice.\n- Interface slice fields: child blocks successfully decoded into a matching interface slice field should be treated as consumed and should not remain in the decoded resource remainder/extras.\n- Single interface fields: when `(*Resource).As(target interface{}) error` decodes a struct containing a field of type `MyInterface`, and exactly one registered child extension block implements that interface, the field should be set to the decoded concrete extension value.\n- Ambiguous single interface fields: when a single interface field could be filled by more than one matching registered child block, `(*Resource).As(target interface{}) error` should return an error using the format `more than one blocks implement %q`.\n\nImplementation notes:\n- The concrete mechanism for discovering registered extension types, selecting matching child blocks, and consuming decoded blocks is up to the implementation.\n- Preserve existing decoding behavior for concrete extension fields, resource fields, attributes, and non-matching remainder content.\n- Matching should be based on externally observable registration and Go interface assignability semantics, not on hard-coded block names used by a particular test."} {"task_id": "format-code-task-000457", "source_id": "format-code-task-000457", "domain": "code", "task_path": "tasks/format-code-task-000457", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:54de913ee8978d28da7a79e009f0fc5c1732479cf08d6035dab8375b7ecab5a9", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nFailed due to duplicate Message-ID header\nI've been using this project for a while and it's awesome, thanks a lot! I stumbled upon an error for the first time today. Error log from CloudWatch is below:\n\n```\n2017-05-03T11:38:51.802Z\tf2ab5592-2ff4-11e7-97ca-a1db4277fedb\t{ level: 'error',\nmessage: 'sendRawEmail() returned error.',\nerror: \n{ [InvalidParameterValue: Duplicate header 'Message-ID'.]\nmessage: 'Duplicate header \\'Message-ID\\'.',\ncode: 'InvalidParameterValue',\ntime: Wed May 03 2017 11:38:51 GMT+0000 (UTC),\nrequestId: '14956f7e-2ff5-11e7-8290-6fdc1cb44eda',\nstatusCode: 400,\nretryable: false,\nretryDelay: 16.828257404267788 },\nstack: 'InvalidParameterValue: Duplicate header \\'Message-ID\\'.\\n at Request.extractError (/var/runtime/node_modules/aws-sdk/lib/protocol/query.js:47:29)\\n at Request.callListeners (/var/runtime/node_modules/aws-sdk/lib/sequential_executor.js:105:20)\\n at Request.emit (/var/runtime/node_modules/aws-sdk/lib/sequential_executor.js:77:10)\\n at Request.emit (/var/runtime/node_modules/aws-sdk/lib/request.js:673:14)\\n at Request.transition (/var/runtime/node_modules/aws-sdk/lib/request.js:22:10)\\n at AcceptorStateMachine.runTo (/var/runtime/node_modules/aws-sdk/lib/state_machine.js:14:12)\\n at /var/runtime/node_modules/aws-sdk/lib/state_machine.js:26:10\\n at Request. (/var/runtime/node_modules/aws-sdk/lib/request.js:38:9)\\n at Request. (/var/runtime/node_modules/aws-sdk/lib/request.js:675:12)\\n at Request.callListeners (/var/runtime/node_modules/aws-sdk/lib/sequential_executor.js:115:18)' }\n```\n\nRepro step: create a Skype account using foobar@mydomain.com. The email containing the verification code fails to be forwarded."} {"task_id": "format-code-task-000458", "source_id": "format-code-task-000458", "domain": "code", "task_path": "tasks/format-code-task-000458", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3c89b0f9b2e8f30b40fbec55cb1b717a6969a99731ebff91e6f96fc69500e464", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n`convert_mfile` fails on recent IPython versions\n============================================\n\nI'm trying to convert one of my Matlab `.m` files into an IPython notebook using `pymatbridge.publish.convert_mfile`, so I can re-run it interactively with the `%%matlab` magic. Something like:\n\n```python\nfrom pymatbridge.publish import convert_mfile\nconvert_mfile('my_script.m')\n```\n\nOn older IPython this used to work, but with a recent IPython install it blows up partway through and no `.ipynb` is written. The notebook construction code in `publish.py` seems to be calling into `IPython.nbformat` in a way that doesn't match what current IPython exposes anymore, so the function never gets to actually serialize anything.\n\nCould `convert_mfile` be updated so it works against current `IPython.nbformat` and produces a notebook file that opens cleanly in the notebook UI? The end result I want is the same as before — a `.ipynb` with the matlab-magic loader at the top, then alternating markdown / `%%matlab` code cells corresponding to the `%`-comments and code blocks in the m-file."} {"task_id": "format-code-task-000459", "source_id": "format-code-task-000459", "domain": "code", "task_path": "tasks/format-code-task-000459", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:482ead1966473dd108f7934684a9b2345a558827eebb480ee2c30a8651c3bc0a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI need the public page lifecycle methods for an existing `Jbig2Ctx`: `int jbig2_complete_page(Jbig2Ctx *ctx)`, `Jbig2Image *jbig2_page_out(Jbig2Ctx *ctx)`, and `void jbig2_release_page(Jbig2Ctx *ctx, Jbig2Image *image)`. After I create a decoder context and feed enough JBIG2 data for a page, `jbig2_complete_page(ctx)` should mark the current decoded page as available and return 0. If the current page has no image buffer, `jbig2_complete_page(ctx)` should report a fatal error through the configured error callback and return -1. For streams whose final pending segment declares an unknown data length of `0xffffffff`, `jbig2_complete_page(ctx)` should warn, use the bytes already buffered as that segment body, parse it, advance the segment state, and return -1 with a warning if that parse fails.\n\nOnce a page is complete, `jbig2_page_out(ctx)` should return a referenced `Jbig2Image *` for the first available completed page, with the page pixel data, width, height, and stride preserved for the caller. If no completed page is available, `jbig2_page_out(ctx)` should return `NULL`. Calling `jbig2_page_out(ctx)` twice in a row should not return the same page twice unless another page has completed in between, because the first call hands that page to the caller. When I am done with a returned image, `jbig2_release_page(ctx, image)` should release the library's reference and mark that page as no longer held by the caller. Calling `jbig2_release_page(ctx, NULL)` should be a no-op, and passing an image pointer that did not come from that context's `jbig2_page_out` should report a warning through the error callback without crashing."} {"task_id": "format-code-task-000460", "source_id": "format-code-task-000460", "domain": "code", "task_path": "tasks/format-code-task-000460", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:24fd0fed60519d6f151c03f9b62f2b9f762fd42b227fe8134a4127dcdde78fc7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Works tab is empty for artists that only have collections (no artist series)\n\nOn the artist page, the **Works** tab currently renders the artist series rail beneath the top works rail. For artists who don't have any artist series associated with them but who *do* belong to one or more collections, this section just shows up empty — the collections never appear on the Works tab at all.\n\nFor comparison, the **Overview** tab shows the collections rail correctly for these same artists. So the data is clearly available; it's just that the Works tab seems to assume every artist will have at least one artist series.\n\n#### Steps to reproduce\n1. Find an artist who is featured in at least one collection but has no artist series (we have a few of these in production).\n2. Navigate to `/artist//works`.\n3. Notice that nothing related to series or collections renders in that area — just the top works rail and the artwork filter below.\n4. Now navigate to `/artist/` (Overview). The collections show up fine.\n\n#### Expected\nOn the Works tab, when an artist has no artist series, we should fall back to showing their collections so the slot isn't wasted (and so users can still discover the artist via collections from this tab). When the artist *does* have artist series, behavior should stay as it is today — show the artist series rail.\n\nThe Overview tab should continue to behave as it currently does (collections only — no change there)."} {"task_id": "format-code-task-000461", "source_id": "format-code-task-000461", "domain": "code", "task_path": "tasks/format-code-task-000461", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5c0dbefd04a2ca2dad732942c3b259e363b967cb342f7b2f475ea2bb7bb66e2b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want server-side `Meteor.SmartCollection` live queries to honor the `fields` option when a cursor is observed or published. The API shape is `var coll = new Meteor.SmartCollection(name, options); var cursor = coll.find(selector, {fields: projection}); cursor.observeChanges(callbacks);`, and the same projected output should flow through a cursor returned from `Meteor.publish`.\n\nFor an inclusion projection such as `{aa: 1, \"bb.c\": 1}`, when a matching document is `{_id: \"aa\", aa: 10, bb: {c: 10, k: 12}, cc: 20}`, the `added` callback should receive id `\"aa\"` and document `{_id: \"aa\", aa: 10, bb: {c: 10}}`. `_id` should be included by default for inclusion projections, but `{aa: 1, _id: 0}` should omit `_id` from the delivered document fields.\n\nFor an exclusion projection such as `{aa: 0, \"bb.c\": 0}`, when a matching document is `{_id: \"aa\", aa: 50, bb: {c: 45, k: 12}, cc: 20}`, the delivered document should be `{_id: \"aa\", bb: {k: 12}, cc: 20}`. The same projection rules should apply to later `changed` field maps, so with `{aa: 1}` a change map `{aa: 50, bb: 30}` should notify observers only with `{aa: 50}` and update the live-query cache only for that projected field.\n\nThe projection validator should reject unsupported field specs by throwing `Error`: field names containing `$` or values that are not numeric `0` or `1` are invalid, and mixing inclusion and exclusion in the same projection is invalid except for excluding `_id` alongside included fields."} {"task_id": "format-code-task-000462", "source_id": "format-code-task-000462", "domain": "code", "task_path": "tasks/format-code-task-000462", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:81d00cf1b9eb1eb16b3b59c5f7b05514cb6cc62b1ac9a8635ca7431dbd9c9369", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `asdf uninstall ` command to remove an installed runtime version for a plugin. When I run `asdf uninstall dummy 1.0.0` and that version is installed, the command should run any configured `pre_asdf_uninstall_dummy` hook with `1.0.0`, run the plugin's optional `uninstall` callback, delete that version's install directory, run any configured `post_asdf_uninstall_dummy` hook with `1.0.0`, then refresh shims so executables from the removed version are no longer advertised. The uninstall callback should receive `ASDF_INSTALL_TYPE`, `ASDF_INSTALL_VERSION`, and `ASDF_INSTALL_PATH`, and hook or callback stdout/stderr should pass through to the command output. For example, with uninstall hooks and a callback that prints `custom uninstall`, stdout should contain `pre_asdf_uninstall_dummy 1.0.0`, then `custom uninstall`, then `post_asdf_uninstall_dummy 1.0.0`, and the command should exit 0.\n\nIf either `` or `` is missing, `asdf uninstall` should print `No plugin given` on stderr and exit non-zero. If the requested version is `latest`, it should print `'latest' is a special version value that cannot be used for uninstall command` on stderr and exit non-zero. If the version is not installed, it should print `No such version` on stderr and exit non-zero. If config loading, a hook, the plugin callback, install-directory removal, or shim refresh fails, the command should print the underlying error to stderr and exit non-zero."} {"task_id": "format-code-task-000463", "source_id": "format-code-task-000463", "domain": "code", "task_path": "tasks/format-code-task-000463", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:332021f84f2f999c3acd182ac371153e50cc38fe4df83d3b6a5e4a351f50598c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `asdf shimversions ` to show which installed plugin versions provide a shimmed command. When I install versions such as `dummy 3.0` and `dummy 1.0`, run `asdf reshim dummy`, and then run `asdf shimversions dummy`, stdout should contain one provider line per recorded version in the format ` `, for example `dummy 3.0` and `dummy 1.0`, and the command should exit 0.\n\nIf I run `asdf shimversions` without a command name, it should print `usage: asdf shimversions ` to stderr and exit nonzero. If the asdf config cannot be loaded, it should report `error loading config: ` to stderr and fail nonzero. If the shim metadata for the requested command cannot be read, the command should fail nonzero without printing made-up provider lines."} {"task_id": "format-code-task-000464", "source_id": "format-code-task-000464", "domain": "code", "task_path": "tasks/format-code-task-000464", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ebcbde0d4f0c8dc030eb41d89c6981e55364dc163859e0c5de9af82e626d987b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `asdf set` command to update the version selection file used by future asdf resolution. The command should be invoked as `asdf set [ ...]`, should print nothing on stdout on success, and should exit with status 0 after writing the requested tool and versions.\n\nBy default, running `asdf set lua 5.2.3` in a directory with no version file should create `.tool-versions` in the current directory containing `lua 5.2.3` followed by a newline. If that file already contains `lua 1.1.1`, the command should replace that line with `lua 5.2.3`; if the file contains other tools, it should preserve those lines and append `lua 5.2.3` at the end. Inline comments on the replaced tool line and comment-only lines for unrelated content should be preserved, and multiple requested versions such as `asdf set python 3.12.1 3.11.8` should be written on one line after the tool name.\n\nThe command should support `--home`/`-u` to write the user home `.tool-versions` file instead of the current directory file, and `--parent`/`-p` to find and update the closest existing `.tool-versions` file in a parent directory. If both `--home` and `--parent` are supplied, it should fail with a non-zero exit and print `home and parent flags cannot both be specified; must be one location or the other` to stderr. If no arguments are supplied, it should fail non-zero and print `tool and version must be provided as arguments`; if only a tool is supplied, it should fail non-zero and print `version must be provided as an argument`. If `--parent` is supplied and no parent version file exists, it should fail non-zero and print `No .tool-versions version file found in parent directory` to stderr.\n\nVersion arguments may be literal versions, `ref:...`, `path:...`, `system`, `latest`, or `latest:`. For `latest` selectors, `asdf set` should resolve the selector for the named plugin first and write the resolved concrete version string to the version file; if the selector cannot be resolved, the command should fail non-zero instead of writing a `latest` token."} {"task_id": "format-code-task-000466", "source_id": "format-code-task-000466", "domain": "code", "task_path": "tasks/format-code-task-000466", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c93e43db97e0b036710d2d71f23d287d4e4847ac56feb8c5eaafbcf8a279e03c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nRewriting TypedDict can lead to SyntaxError\npyupgrade checks if the keys of a TypedDict are strings, but not if they are valid for usage in class syntax.\n\nHere's a minimal input:\n\n```python\nfrom typing_extensions import TypedDict\n\nMyDict = TypedDict(\"MyDict\", {\"my-key\": str, \"another key\": int})\n```\n\n\nThis is the rewrite produced by `pyupgrade --py36-plus minimal.py`:\n\n```python\nfrom typing_extensions import TypedDict\n\nclass MyDict(TypedDict):\n my-key: str\n another key: int\n```\n\nThe `-` and the space in the keys are invalid syntax."} {"task_id": "format-code-task-000467", "source_id": "format-code-task-000467", "domain": "code", "task_path": "tasks/format-code-task-000467", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c4a83dfb0b55a9c7f6d3ceb973fb39e22a54120cc8773e11777fc3243f882f06", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nFailing test case using pyupgrade 2.25.0:\n\n```\n$ echo 'a=b=c=None; print(\"Query:%8s %s %s\" % (a, b, c))' > test.py ; pyupgrade --py36-plus test.py ; cat test.py\nRewriting test.py\na=b=c=None; print(f\"Query:{a:>8} {b} {c}\")\n```\n\n```pycon\n>>> a=b=c=None; print(\"Query:%8s %s %s\" % (a, b, c))\nQuery: None None None\n>>> a=b=c=None; print(f\"Query:{a:>8} {b} {c}\")\nTraceback (most recent call last):\n File \"\", line 1, in \nTypeError: unsupported format string passed to NoneType.__format__\n```\n\nThe problem is while ``f\"{None}\"`` works, advanced string format arguments do not:\n\n```pycon\n>>> \"%8s\" % None\n' None'\n```\n\n```pycon\n>>> f\"{None:>8}\"\nTraceback (most recent call last):\n File \"\", line 1, in \nTypeError: unsupported format string passed to NoneType.__format__\n```\n\nPossible workaround:\n\n```python\n>>> a=b=c=None; print(f\"Query:{str(a):>8} {b} {c}\")\nQuery: None None None\n```\n\nThis issue was found in https://github.com/biopython/biopython/pull/3724"} {"task_id": "format-code-task-000468", "source_id": "format-code-task-000468", "domain": "code", "task_path": "tasks/format-code-task-000468", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f53e06d5a8a1e89a47775583aac08b42b7de6365d4ff988c39376d772410a6ce", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `git-deps` to report dependency analysis results directly in the terminal when I run it with commit-ish arguments and no `--serve` or `--json` mode. For a single target commit such as `git-deps 4f27a1e^!`, stdout should contain one dependency SHA per line, without repeating the same dependency more than once for that target; in the git-deps repository this includes lines like `3374b8419a45d91d3c0631be11c8cf893b272217`, `3a1dd42fd6114a634ba7cf037ce61e2aee76db73`, and `b1967573e81a8100a4cc778936de0ba0a8a8f5cb`. If the target is the root commit, for example `git-deps b196757^!`, the command should exit successfully and produce no dependency lines. When I ask for multiple targets, either by passing multiple commit-ish arguments or by using a revision range that expands to more than one revision, each dependency line should be an edge formatted as ` `. The same edge format should be used with recursive reporting, so `git-deps -r 4f27a1e^!` should include lines such as `4f27a1ee2b5fd63a58311a20e2aed0a24eda8da2 3374b8419a45d91d3c0631be11c8cf893b272217` and recursive downstream edges like `3374b8419a45d91d3c0631be11c8cf893b272217 b1967573e81a8100a4cc778936de0ba0a8a8f5cb`. With `--log`, terminal reporting should print each newly discovered dependency as a `git log -n1` block using color-enabled git output; in multi-commit mode it should prefix each dependency with ` depends on:` and, if a dependency commit has already been shown, print `commit (already shown above)` instead of showing the full log block again. The happy path should exit 0 after all selected revisions are analyzed and their terminal output is written as discoveries are emitted."} {"task_id": "format-code-task-000469", "source_id": "format-code-task-000469", "domain": "code", "task_path": "tasks/format-code-task-000469", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6399d237c31c5d0a0b27f676a3cf57c2e9edc80ab1138c452dfc17683ecdc7eb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a \"consecutive irrelevant\" stopping rule\n\nThe active learning loop decides when to stop screening through a *stopping\nmechanism*: a small object with a `stop(results, data)` method that returns\n`True` when the review should halt. The project already ships a handful of these\n(for example `StoppingDefault`, `StoppingN`, `StoppingQuantile`,\n`StoppingIsFittable`), and an `ActiveLearningCycle` delegates to whatever object\nis passed as its `stopping` criterion.\n\nA very common heuristic in screening is missing: stop once the reviewer has seen\na run of irrelevant records in a row. Please add a new stopping mechanism for\nthis, exposed from the same place as the existing stopping rules and named\n`StoppingNConsecutiveIrrelevant`.\n\nBehaviour:\n\n- It is constructed with a single threshold `n` — the number of consecutive\n irrelevant records that should trigger stopping.\n- `stop(results, data)` follows the same calling convention as the other\n stopping rules. `results` is a `pandas.DataFrame` holding the records labeled\n so far in labeling order, with a `label` column where `1` means relevant and\n `0` means irrelevant. `data` is the full collection of records being screened;\n only its length (the total number of records) is relevant.\n- It returns `True` when the most recently labeled `n` records are *all*\n irrelevant. The run must be contiguous and at the end of `results`: scattered\n irrelevant records that do not form a trailing run of length `n` must not\n trigger stopping.\n- It also returns `True` when there is nothing left to screen — i.e. every\n record has been labeled (`len(results) >= len(data)`) or `data` is empty.\n- Otherwise it returns `False`. In particular, when fewer than `n` records have\n been labeled and the pool is not yet exhausted, it must return `False`.\n\nThe new mechanism must work both on its own and when handed to an\n`ActiveLearningCycle` as its stopping criterion (so that `cycle.stop(results,\ndata)` reflects the rule above).\n"} {"task_id": "format-code-task-000471", "source_id": "format-code-task-000471", "domain": "code", "task_path": "tasks/format-code-task-000471", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7c8bf790c44dfd79435f00500a1164e3253d3570a0ba17951d83a541fa676077", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAllow regex matching in S3KeySensorAsync\n**Is your feature request related to a problem? Please describe.**\nThe wildcard matching uses `fnmatch.fnmatch` that uses unix wildcard pattern to filter bucket keys but it has a limitation. It cannot distinguish between `archived_uploads/dataset.csv` and `dataset.csv` when your wildcard pattern is `*.csv`. As S3 does not use directories, that wildcard pattern cannot filter \"top level\" bucket keys that do not have the delimeter `/`. Because of this, S3KeySensorAsync always goes in success state even if you have not uploaded anything new.\n\nHowever, there is a workaround, which is to use some prefix in your bucket keys for new file uploads and to not upload at the top level. E.g. If you are uploading new files, upload them as `new/dataset.csv` and the wildcard pattern for this case will be `new/*.csv`.\n\n**Describe the solution you'd like**\nA flag variable `use_regex` when set to `True` should use regex pattern.\n\nhttps://github.com/astronomer/astronomer-providers/blob/6b29b0428ca3a2f26c31dc6040f3a6f9b5d01d67/astronomer/providers/amazon/aws/hooks/s3.py#L151-L157\n\n```python\n elif use_regex:\n keys = await self.get_file_metadata(client, bucket_name, key)\n key_matches = [k for k in keys if re.match(pattern=key, string=k[\"Key\"])]\n if not key_matches:\n return False\n```\n\nE.g. If you have bucket keys `archived_uploads/dataset1.csv`, `archived_uploads/dataset2.csv`, `archived_uploads/dataset3.csv` and `dataset4.csv`, this is the pattern you will need to only poke for keys that do not have the delimeter '/', in this case `dataset4.csv` does not have this delimeter: \n\n```python\nS3KeySensorAsync(\n task_id='sense_datasets',\n bucket_name='default',\n # Negative Lookbehind regex pattern which tells regex to not match names that are being preceded by the delimeter \"/\".\n bucket_key=r'(? bool:\n if any((file_metadata for file_metadata in files if re.match(pattern=r'(? None` class that analyzes the entities and weighted synapses already stored in a `KnowledgeGraph`. The analyzer should expose `build_graph() -> networkx.DiGraph`, `pagerank() -> dict[str, float]`, `betweenness_centrality() -> dict[str, float]`, `detect_communities() -> dict[int, set[str]]`, `find_bridges() -> list[tuple[str, str]]`, `find_keystones() -> list[str]`, `connected_components() -> list[set[str]]`, `shortest_path(source: str, target: str) -> list[str] | None`, and `predict_links(top_n: int = 10) -> list[tuple[str, str, float]]`.\n\nFor a graph containing entities `a`, `b`, `c`, `d`, and `e`, with synapses `a -> b` at strength `0.8`, `b -> c` at `0.6`, `c -> d` at `0.4`, `a -> c` at `0.3`, and `d -> e` at `0.2`, `build_graph()` should return a directed graph with 5 nodes, 5 edges, and the `a` to `b` edge weight set to `0.8`. Calling `pagerank()` on that analyzer should return one score for each entity, all between `0.0` and `1.0`, and in this sample `a` should rank lower than `d`. Calling `betweenness_centrality()` should return one score for each entity, and bridge-like middle nodes such as `b` or `c` should score higher than the leaf `e` in the sample network. `detect_communities()` should return numbered communities whose member sets together cover all graph entities, and on an empty graph it should return `{}`.\n\n`find_bridges()` should return bridge edges from the undirected view of the graph, `find_keystones()` should return articulation-point entity IDs from that same undirected view, and `connected_components()` should return connected component sets covering every entity. `shortest_path(\"a\", \"e\")` in the sample should prefer stronger synapses by using inverse strength as cost, so it should return `[\"a\", \"b\", \"c\", \"d\", \"e\"]`; if either endpoint is missing or no path connects the endpoints, it should return `None` rather than raising. On an empty graph, `pagerank()` and `betweenness_centrality()` should return `{}`. `predict_links()` should use Adamic-Adar style scores on the undirected graph, omit existing edges, sort suggestions by descending score, honor `top_n`, and return `[]` when there are fewer than two nodes or no non-edge suggestions."} {"task_id": "format-code-task-000475", "source_id": "format-code-task-000475", "domain": "code", "task_path": "tasks/format-code-task-000475", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2482503ddd57c835e7a33514f4cd87533cbcb577f8779c43f41b174ac1bb63a3", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nWay to reset data-react-beautiful-dnd-draggable for server side rendering\n## Bug or feature request?\nBug\n\n### Expected behavior\n`data-react-beautiful-dnd-draggable` shouldn't increment based on how many times it was run on the server but start fresh for each request.\n\n### Actual behavior\nEach time the server render is hit `data-react-beautiful-dnd-draggable` is incremented by 1.\ni.e. after 6 reloads console throws:\n```\nWarning: Prop `data-react-beautiful-dnd-draggable` did not match. Server: \"6\" Client: \"0\"\n```\n\n### Steps to reproduce\nServer side render and reload couple times, each reload the `data-react-beautiful-dnd-draggable` gets incremented by 1.\n\n### Browser version\nChrome\n\nhttps://github.com/atlassian/react-beautiful-dnd/blob/master/src/view/style-marshal/style-marshal.js#L11\n\nPossible solution to this would be to expose a function that resets the counter that one could call during a server side render like so:\n```\nexport function resetCounter() {\n count = 0;\n}\n```"} {"task_id": "format-code-task-000476", "source_id": "format-code-task-000476", "domain": "code", "task_path": "tasks/format-code-task-000476", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6d39f1fe89a85d643b54a728855561ca9051e2397d10f798633cb3ec1bd33d06", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `gosaic index [PATHS...]` with no `--clean`, `--list`, or `--rm` flag to populate the local image index from one or more file or directory paths. When I run `gosaic --dsn sqlite3:///tmp/gosaic.sqlite3 index ./tiles`, it should recursively discover regular `.jpg`, `.jpeg`, and `.png` files under `./tiles`, convert stored paths to absolute paths, print a line containing `Indexing N images...` where `N` is the number of discovered usable image paths, and exit with code 0 after inserting the records.\n\nEach indexed image should store the absolute path, MD5 content hash, width, height, EXIF orientation, and a reusable aspect-ratio record. JPEG EXIF orientation values 5 through 8 should swap the stored width and height, and missing or zero orientation should be treated as orientation 1. If two discovered files have the same MD5 hash, only the first should be inserted so later `gosaic index --list` output contains one path for that image content.\n\nRunning `gosaic index` without any paths and without stdin paths should fail with exit code 1 and print `Must specify paths to index`. If a supplied path cannot be walked or an individual image cannot be hashed, opened, or inspected, the command should print an error message naming that path and continue indexing the remaining discovered images when possible."} {"task_id": "format-code-task-000477", "source_id": "format-code-task-000477", "domain": "code", "task_path": "tasks/format-code-task-000477", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d4dc8778bcf1751f540cf5187f2f5c26a9db77cc970895731c3d547135f12aaf", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Add support for `proof_of_possession` on resource servers\n\nI'm using auth0-deploy-cli to manage my tenant configuration as code. One of my APIs needs proof-of-possession (DPoP) enabled on the access tokens it issues — Auth0's Management API exposes a `proof_of_possession` setting on resource servers for exactly this.\n\nWhen I add `proof_of_possession` to the resource server entry in my YAML config and run an import, the deploy CLI doesn't recognize the field — the resource server schema only covers things like `enforce_policies`, `token_dialect`, scopes, etc.\n\nCould the resource server schema be extended so that `proof_of_possession` is a recognized, validated field? I'd like to manage this setting through the CLI alongside the other resource server properties instead of having to configure it out-of-band."} {"task_id": "format-code-task-000478", "source_id": "format-code-task-000478", "domain": "code", "task_path": "tasks/format-code-task-000478", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6a8fdfa3bc606efb72eed8c5619dd37faaaefeef046bf27979b4d136aabab630", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI’m using the Java SDK’s ManagementAPI and I need to trigger Auth0’s verification email job from there, but right now I don’t see any Jobs API support so I have to hand-roll the HTTP call. It would be great if the SDK let me send a verification email for a user and get back a typed job response I can inspect, like the job id/status/type and creation time.\n\nExpected outcomes:\n- Management API callers can access Jobs support through `ManagementAPI.jobs()`, receiving a `JobsEntity` that exposes Jobs-related operations.\n- `JobsEntity.sendVerificationEmail(String userId, String clientId)` returns a `Request` for sending a verification email job for the given user.\n- Executing the verification email request sends a POST request to Auth0’s verification email job endpoint, includes the required `user_id` in the JSON request body, and includes `client_id` only when a non-empty client id is provided.\n- Calling the verification email operation without a user id is rejected by the SDK before sending a request.\n- The SDK provides `com.auth0.json.mgmt.jobs.Job` for the Jobs API response, with readable `getId()`, `getStatus()`, `getType()`, and `getCreatedAt()` values from the response JSON.\n- `Job` JSON handling ignores unknown response fields and does not serialize null fields.\n\nImplementation notes:\n- Follow the existing Management API entity and request patterns in the SDK.\n- The internal data structures, validation location, and JSON binding details are up to the implementation as long as the public SDK behavior above is satisfied.\n\n## Required output literals (exact-match contract)\n\nWhen executing Jobs Management API requests, the implementation MUST authenticate the HTTP request using the configured API token:\n\n- HTTP header name: `Authorization` - exact wire-level header required by the existing Management API request pattern.\n- HTTP header value: `Bearer ` followed by the configured API token - exact bearer-token prefix required for Auth0 Management API authentication.\n"} {"task_id": "format-code-task-000479", "source_id": "format-code-task-000479", "domain": "code", "task_path": "tasks/format-code-task-000479", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:716f718183f2a374d5d1f08c2214e5044e97c11385d62cfcb9e12c3669227cb4", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nTerms and Privacy checkbox not available with Passwordless mode of Lock\n**Issue:**\n\nCurrently when using passwordless signup we aren't able to get a checkbox with the `signUpTerms` text and the `mustAcceptTerms` keys set.\n\nThis is limited only to the `Auth0LockPasswordless` where as it seems to be working fine when using the database with username-password version of the lock.\n\nWe would like to be able to ensure this checkbox/notice display to be in compliance with GDPR and other privacy standards.\n\n- [x] Code snippet or sample project that reproduces the bug\n\n```js\n var clientID = \"XXXXXXXXXXXXXXXXXXXX\";\n var domain = \"freecodecamp.auth0.com\";\n\n var options = {\n languageDictionary: {\n title : \"freeCodeCamp.org\",\n signUpTerms: \"By choosing to continue you confirm that you have read and agreed to...\"\n },\n mustAcceptTerms: true,\n theme: {\n ...\n },\n ...\n socialBigButtons : true,\n passwordlessMethod: 'link',\n auth: {\n redirectUrl: 'http://localhost:3000/',\n responseType: 'token id_token',\n params: {\n scope: 'openid profile email'\n }\n },\n allowedConnections: [\n \"email\", \"google-oauth2\", \"facebook\", \"github\"\n ],\n };\n\n var lock = new Auth0LockPasswordless(clientID, domain, options);\n lock.show();\n```\n\n- [x] Screenshots when appropriate\n\n![image](https://user-images.githubusercontent.com/1884376/41974178-d7c3b1ce-7a34-11e8-8904-6b4496721a0a.png)\n\n\n- [x] Lock version\n\n```\n11.7.2\n```\n\n- [x] Browser & OS\n\n```\nNot relevant\n```\n\nMake sure to include **as much information as possible** for us to understand and reproduce the bug, that way we can fix it as quickly as possible."} {"task_id": "format-code-task-000480", "source_id": "format-code-task-000480", "domain": "code", "task_path": "tasks/format-code-task-000480", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:694a39cb53997d9662d52de6132423ecdee75b71e946975f0f5376c2ab2cd054", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Version\n\n\nv4.38.2\n\n### Deployment Method\n\n\nDocker\n\n### Reverse Proxy\n\n\nTraefik\n\n### Reverse Proxy Version\n\n\n2.11.0\n\n### Description\n\n\nIn upgrading to 4.38 I changed my Traefik middleware to use the (default provided) `/api/authz/forward-auth` route. Then, I noticed requests to Home Assistant's `/api` were getting a 401 response, despite having valid (for Home Assistant) `Authorization` headers. The error comes from Authelia, even though the request should be bypassed.\n\n### Reproduction\n\n\nThis can be reproduced using something simple like `traefik/whoami`\n\nAn example docker-compose.yml:\n\n```yml\nservices:\n whoami:\n image: traefik/whoami\n labels:\n - traefik.enable=true\n - traefik.http.routers.${COMPOSE_PROJECT_NAME}-whoami.rule=Host(`whoami.${HOMELAB_BASE_DOMAIN}`)\n - traefik.http.routers.${COMPOSE_PROJECT_NAME}-whoami.entrypoints=web,websecure\n - traefik.http.routers.${COMPOSE_PROJECT_NAME}-whoami.middlewares=auth@file\n```\n\nwhere `auth@file` is this Traefik middleware:\n\n```yml\nhttp:\n middlewares:\n auth:\n forwardAuth:\n # Using the internal container + port\n address: http://authelia:9091/api/authz/forward-auth\n trustForwardHeader: true\n authResponseHeaders:\n - Authorization\n - Proxy-Authorization\n - Remote-User\n - Remote-Groups\n - Remote-Email\n - Remote-Name\n```\n\nand the `access_control` which should bypass Authelia:\n\n```yml\naccess_control:\n default_policy: one_factor\n\n rules:\n - domain: whoami.${HOMELAB_BASE_DOMAIN}\n policy: bypass\n```\n\nWith Authelia + Traefik + whoami running, now you can try curl:\n\n```sh\n# This works:\ncurl -L https://whoami.${HOMELAB_BASE_DOMAIN}/\n\n# This gets a 401 from Authelia:\ncurl -L -H 'Authorization: Bearer test' https://whoami.${HOMELAB_BASE_DOMAIN}/\n```\n\n### Expectations\n\n\nIdeally Authelia wouldn't be doing much when the requested route is bypassed.\n\n### Configuration (Authelia)\n\n\n```yaml\ntheme: auto\n\ntotp:\n issuer: auth.${HOMELAB_BASE_DOMAIN}\n\nauthentication_backend:\n password_reset:\n disable: true\n\n file:\n path: /config/users.yml\n\naccess_control:\n default_policy: one_factor\n #default_policy: two_factor\n\n rules:\n - domain: whoami.${HOMELAB_BASE_DOMAIN}\n policy: bypass\n\nsession:\n inactivity: '5m'\n expiration: '1h'\n remember_me: '1M'\n\n redis:\n host: redis\n\n cookies:\n - domain: ${HOMELAB_BASE_DOMAIN}\n authelia_url: https://auth.${HOMELAB_BASE_DOMAIN}\n default_redirection_url: https://dashboard.${HOMELAB_BASE_DOMAIN}\n\nstorage:\n local:\n path: /output/db.sqlite3\n\nnotifier:\n filesystem:\n filename: /output/notifications\n```\n\n\n### Build Information\n\n\n```shell\nLast Tag: v4.38.2\nState: tagged clean\nBranch: v4.38.2\nCommit: 573e79c8d34e118195731424df15d3e2c989495e\nBuild Number: 27687\nBuild OS: linux\nBuild Arch: amd64\nBuild Compiler: gc\nBuild Date: Sat, 16 Mar 2024 02:33:12 +1100\nExtra:\n\nGo:\n Version: go1.22.1\n Module Path: github.com/authelia/authelia/v4\n Executable Path: github.com/authelia/authelia/v4/cmd/authelia\n Settings:\n -buildmode: pie\n -compiler: gc\n -trimpath: true\n DefaultGODEBUG: httplaxcontentlength=1,httpmuxgo121=1,tls10server=1,tlsrsakex=1,tlsunsafeekm=1\n CGO_ENABLED: 1\n GOARCH: amd64\n GOOS: linux\n GOAMD64: v1\n vcs: git\n vcs.revision: 573e79c8d34e118195731424df15d3e2c989495e\n vcs.time: 2024-03-15T15:24:32Z\n vcs.modified: true\n Dependencies:\n authelia.com/provider/oauth2@v0.1.1 (h1:JHMWB8aieYW++7a+t3RIB0fkxcLMHyT7NUKPEAl6cls=)\n filippo.io/edwards25519@v1.1.0 (h1:FNf4tywRC1HmFuKW5xopWpigGjJKiJSV0Cqo0cJWDaA=)\n github.com/Azure/go-ntlmssp@v0.0.0-20221128193559-754e69321358 (h1:mFRzDkZVAjdal+s7s0MwaRv9igoPqLRdzOLzw/8Xvq8=)\n github.com/Gurpartap/logrus-stack@v0.0.0-20170710170904-89c00d8a28f4 (h1:vdT7QwBhJJEVNFMBNhRSFDRCB6O16T28VhvqRgqFyn8=)\n github.com/andybalholm/brotl"} {"task_id": "format-code-task-000481", "source_id": "format-code-task-000481", "domain": "code", "task_path": "tasks/format-code-task-000481", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e84d20a1afa3eeb959345fd7777ea49e2af00155aaaec11bff0857a525ec3d85", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `` element instance to expose an async `setAuthenticationConfiguration(securitySchemeId, { token, clientId, clientSecret, redirectUri })` method so host applications can configure credentials after a spec is loaded. The method should look up the loaded OpenAPI security scheme by `securitySchemeId` and reject by throwing `Error('SecuritySchemeNotFound')` when no matching scheme exists.\n\nWhen I call `await explorer.setAuthenticationConfiguration('bearerAuth', { token: 'abc123' })` for an HTTP bearer scheme, the matching scheme should store `Bearer abc123` as the applied credential, and a token that already starts with `Bearer ` should not get a duplicate prefix. When I call `await explorer.setAuthenticationConfiguration('basicAuth', { token: 'user:password' })` for an HTTP basic scheme, it should apply an `Authorization` value of `Basic ${btoa('user:password')}`.\n\nThe method should trim and save OAuth client configuration from the options object: `clientId`, `clientSecret`, and `redirectUri`. `redirectUri` should be resolved with `new URL(..., window.location.href).toString()`, so relative redirect paths become absolute URLs. After updating the scheme, the element should check the current page URL for an OAuth response, request a render update, and later executable requests for operations that require that security scheme should receive the applied credential through the existing request flow."} {"task_id": "format-code-task-000482", "source_id": "format-code-task-000482", "domain": "code", "task_path": "tasks/format-code-task-000482", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:14af191f5a6693c1287eb2fe30ec012743dd6577cd0eac46722fa780b9742277", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Watch API on Postgres occasionally reports namespace modifications as deletions, corrupting the schema cache\n\nWe're running SpiceDB in production with Postgres as the datastore, and we have the experimental schema cache enabled (the one that's driven by the Watch API). After running for a while, we start seeing requests fail with errors like:\n\n```\nunknown namespace \"document\"\n```\n\n…even though the namespace is clearly still defined — `zed schema read` against the same cluster shows it's there. Restarting SpiceDB clears the cache and the errors go away, but they come back after some more schema activity.\n\nWhile digging into this we started subscribing to the Watch API directly to see what events were actually being emitted when our team edits the schema. What we observed is that **a plain modification to a namespace** (e.g. adding a new relation to an existing `document` namespace and writing the updated schema) sometimes shows up on the Watch stream as a **deletion event for that namespace**, instead of as a \"definition changed\" event. It's not consistent — the same kind of edit will sometimes correctly come through as \"changed\" and sometimes incorrectly come through as \"deleted\". We haven't been able to find a reliable repro on demand; it feels like a race / ordering thing inside SpiceDB.\n\nThis obviously breaks the schema cache: it sees the spurious deletion event, drops the namespace from its cache, and then subsequent permission checks against that namespace fail with \"unknown namespace\" until something else forces a refresh.\n\nA few extra data points:\n\n- We **only** see this on clusters using Postgres as the datastore. We have another cluster on CockroachDB and have never reproduced it there — Watch events for schema edits on CRDB always look correct.\n- The user never actually deleted the namespace. There is no `DeleteNamespace` / `WriteSchema`-with-namespace-removed in the audit log around the time of the bad event. The schema modification was a pure edit.\n- Because the cache is wrong, end users hit the failure even though the schema in the datastore is fine, which makes this pretty painful to operate around.\n\nExpected behavior: when a namespace (or caveat) is modified in a single transaction, the Watch API should consistently emit it as a definition change, regardless of which datastore backend is in use. It shouldn't sometimes look like a delete just because of how the underlying datastore happens to surface the update."} {"task_id": "format-code-task-000483", "source_id": "format-code-task-000483", "domain": "code", "task_path": "tasks/format-code-task-000483", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d242eba46ffbd61096594045d753c6eba2f9430384e22ee45fb0444691abc262", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Remove the required column list from `Automerge.Table`\n\nWhen you create a table, you currently have to pass in a list of column names:\n\n```js\ndoc.authors = new Automerge.Table(['surname', 'forename'])\ndoc.publications = new Automerge.Table(['type', 'authors', 'title', 'publisher', 'edition', 'year'])\n```\n\nBut as far as I can tell, this list of columns isn't actually used for anything meaningful anymore. Since #236, you can no longer add a row by passing an array of values that gets mapped to column names — rows must already be objects with named properties. So the column list doesn't enforce a schema, doesn't validate row shapes, doesn't affect storage, and rows are free to omit some of the \"columns\" or include extra properties not in the list.\n\nIn other words, requiring this argument forces every user to write something that looks meaningful but isn't. It's confusing — new users reasonably assume the column list does something (enforces required fields, fixes a row shape, etc.) and then are surprised when it doesn't.\n\nI think we should just drop the column list from the `Table` API. Creating a table should look like:\n\n```js\ndoc.authors = new Automerge.Table()\ndoc.publications = new Automerge.Table()\n```\n\nRows continue to be plain objects, and by convention users should give rows in the same table the same set of properties, but Automerge doesn't (and currently can't) enforce that. The docs should be updated to reflect this — and to mention that not enforcing a schema is actually somewhat intentional, since different clients in a collaborative app may be running different versions and using slightly different property sets.\n\nIf/when proper schema support is added later, that can be designed properly. The current half-feature isn't getting us closer to that and just adds noise to the API.\n\nThoughts?"} {"task_id": "format-code-task-000484", "source_id": "format-code-task-000484", "domain": "code", "task_path": "tasks/format-code-task-000484", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fb821ab827128b73c0a1d9eca3bdba5d4a97a8e8e626b5e0ffb3ef7433ac1487", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## PI and LogEI acquisition functions crash when called\n\nI'm using RoBO for some Bayesian optimization experiments. The default `EI`\nacquisition function works fine, so I wanted to compare against `PI` and\n`LogEI` on the same problem.\n\nI set them up the same way I set up `EI` (same model, same `X_lower` /\n`X_upper`, same `compute_incumbent` for `LogEI`). The moment I evaluate\neither of them on a candidate point, it blows up somewhere inside the call —\nit never gets to return an acquisition value.\n\nRough sketch of what I'm doing:\n\n```python\n# this works fine\nei = EI(model, X_lower, X_upper, compute_incumbent)\nval = ei(X_test)\n\n# this dies\npi = PI(model, X_lower, X_upper)\nval_p = pi(X_test, incumbent)\n\n# this also dies\nlogei = LogEI(model, X_lower, X_upper, compute_incumbent)\nval_l = logei(X_test)\n```\n\nSince `EI` works on exactly the same model/bounds, I'd expect `PI` and\n`LogEI` to be usable as drop-in alternatives. Right now they aren't — only\n`EI` is actually callable.\n\nCould `PI` and `LogEI` be fixed so they also return a value the way `EI`\ndoes?"} {"task_id": "format-code-task-000485", "source_id": "format-code-task-000485", "domain": "code", "task_path": "tasks/format-code-task-000485", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5b6c1a5701c39ba90376af5da1d3f5016c90e8c0875629b3fae38d77c0b7adff", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `autopkg run` and `autopkg install` to support a check-only mode with `-c` or `--check`, as in `autopkg run --check Firefox.download.recipe` or `autopkg install -c Firefox`. In this mode, after the recipe is found and composed, AutoPkg should shorten that recipe's processor list so execution stops at the final `EndOfCheckPhase` processor and does not run processors that appear after it. A successful check run should then continue through the normal recipe processing path for the remaining steps, print the usual `Processing ...` progress line, and complete with the normal successful exit status when no recipe fails.\n\nIf a recipe passed to `autopkg run --check` or `autopkg install --check` does not contain an `EndOfCheckPhase` processor, AutoPkg should not process that recipe. It should write `Recipe at is missing EndOfCheckPhase Processor, not possible to perform check.` to stderr, count that recipe as failed, continue with any remaining recipes, and exit with the recipe-failure status when the run finishes. The short `-c` flag and long `--check` flag should be accepted anywhere the run/install option parser accepts other options, and command help for run/install should list the option as only checking for new or changed downloads."} {"task_id": "format-code-task-000486", "source_id": "format-code-task-000486", "domain": "code", "task_path": "tasks/format-code-task-000486", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3997e554aacca7de43f57dfcd0dc79c7f7f61736ac3b1cb715084cc13561d100", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `autopkg` CLI to provide a recipe repository listing command. When I run `autopkg repo-list`, it should read the configured `RECIPE_REPOS` preference and print each installed repository as one line in the format ` ()`, sorted by repository path, followed by a blank line. The alias `autopkg list-repos` should behave the same way.\n\nFor example, if `RECIPE_REPOS` contains `/Users/me/Library/AutoPkg/RecipeRepos/com.github.autopkg.recipes` with URL `https://github.com/autopkg/recipes` and `/tmp/custom-recipes` with URL `https://github.com/example/custom-recipes`, stdout should list `/tmp/custom-recipes (https://github.com/example/custom-recipes)` before `/Users/me/Library/AutoPkg/RecipeRepos/com.github.autopkg.recipes (https://github.com/autopkg/recipes)` and then emit the trailing blank line. If there are no configured recipe repositories, stdout should be exactly `No recipe repos.` followed by a newline. The command should accept the common `--prefs ` option so I can point it at a specific preferences file before listing repositories. `autopkg repo-list --help` should show `Usage: autopkg repo-list` and describe that it lists all installed recipe repos, exiting successfully. Top-level `autopkg help` should include both `repo-list` with help text `List installed recipe repos` and `list-repos` with help text `see repo-list`."} {"task_id": "format-code-task-000487", "source_id": "format-code-task-000487", "domain": "code", "task_path": "tasks/format-code-task-000487", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:679370c28f3b07cf3943f177574f2511d87355c3deed9ff342b3154444e8acb5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `ToIPDesc` can't parse IPv6 addresses\n\nI'm trying to use `utils.ToIPDesc` with IPv6 peer addresses and it doesn't work. For example:\n\n```go\ndesc, err := utils.ToIPDesc(\"::1:9651\")\n// err != nil -> \"bad ip format\"\n```\n\nAny IPv6 address goes through this path because `ToIPDesc` looks like it just splits the input on `:` and expects exactly two parts, which is fine for `127.0.0.1:9651` but obviously falls apart the moment the host itself contains colons.\n\nThe reverse direction has the same problem. If I build an `IPDesc` with an IPv6 `net.IP` and call `String()` on it, I get something like\n\n```\n::1:9651\n```\n\nwhich is ambiguous — you can't tell where the address ends and the port begins, and feeding it back into `ToIPDesc` won't round-trip.\n\nIPv4 still needs to keep working as before (`127.0.0.1:9651` etc.), but `ToIPDesc` / `IPDesc.String()` should handle IPv6 in an unambiguous way so nodes can actually be addressed over v6."} {"task_id": "format-code-task-000488", "source_id": "format-code-task-000488", "domain": "code", "task_path": "tasks/format-code-task-000488", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:73160ba415256bbc61a274e55d98cb036434da8ca1d74aff2aa89642bc5c2cb6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `taipy.create_scenario(config: ScenarioConfig, creation_date: Optional[datetime] = None, name: Optional[str] = None) -> Scenario` to create a runtime `Scenario` from an existing scenario configuration. When I pass a configuration with one task that reads `input_cfg` and writes `output_cfg`, the returned scenario should contain that task plus concrete data nodes for the configured input and output, and the scenario should be retrievable through the normal scenario APIs after creation. When I pass `name=\"Forecast run\"`, the returned scenario should expose that display name through its properties. When the configuration has additional data node configs, those data nodes should be included in the returned scenario even when they are not connected to any task.\n\nIf the configuration has a frequency and I pass `creation_date=datetime(2026, 7, 22, 10, 30)`, the scenario should be attached to the cycle matching that frequency bucket, creating the cycle if needed. The first scenario created in a cycle should be marked primary, while later scenarios in the same cycle should not be primary. If the configuration has no frequency, the returned scenario should not belong to a cycle and should not be primary.\n\nIf the configuration defines sequences, each sequence should be created on the returned scenario with the concrete tasks named by that sequence definition. For example, a config with tasks `extract` and `train` plus `sequences={\"pipeline\": [extract_task_cfg, train_task_cfg]}` should return a scenario whose `sequences[\"pipeline\"]` contains those two tasks in that order. If a sequence references a task config that is not part of the same scenario config, `create_scenario()` should raise `SequenceTaskConfigDoesNotExistInSameScenarioConfig` instead of creating a partial scenario. The function should also perform the normal configuration/version lock before creating entities, reject an inconsistent scenario with `InvalidScenario`, persist the new scenario, and publish the scenario creation event."} {"task_id": "format-code-task-000489", "source_id": "format-code-task-000489", "domain": "code", "task_path": "tasks/format-code-task-000489", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6f101c26c643663eb22bf07e5110ad2fd97a0930d853796d5908a2a27580d87a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Taipy's process-wide `Config` singleton to honor a `TAIPY_CONFIG_PATH` environment variable that points to an additional TOML configuration file. When my app sets `os.environ[\"TAIPY_CONFIG_PATH\"]` before a configuration compilation and then calls existing stateful APIs such as `Config.load(filename: str) -> None`, `Config.override(filename: str) -> None`, or `Config.configure_global_app(**properties)`, the TOML file referenced by `TAIPY_CONFIG_PATH` should be loaded into a highest-precedence configuration layer.\n\nFor example, if `base.toml` contains `[TAIPY] custom_property_not_overwritten = true` and `custom_property_overwritten = 10`, and the file referenced by `TAIPY_CONFIG_PATH` contains `[TAIPY] custom_property_overwritten = 11`, then after `Config.load(\"base.toml\")`, `Config.global_config.custom_property_not_overwritten` should be `True` and `Config.global_config.custom_property_overwritten` should be `11`. The same precedence should apply over Python configuration: if code sets `Config.configure_global_app(foo=1)` while `TAIPY_CONFIG_PATH` points to a TOML file containing `[TAIPY] foo = 2`, then the applied `Config.global_config.foo` should be `2` after compilation.\n\nIf `TAIPY_CONFIG_PATH` points to a file that does not exist, configuration compilation should not raise because of that missing environment file; it should log the problem and continue using the normal default, Python, and explicit file configuration layers. `Config.backup(filename: str) -> None` should write the applied configuration after this environment-file overlay has been merged, so backed-up TOML reflects values overridden by the environment-specified file."} {"task_id": "format-code-task-000490", "source_id": "format-code-task-000490", "domain": "code", "task_path": "tasks/format-code-task-000490", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:59add9cef4c6148f00c698ba54ff77840b915f739d5b3cd88bbb133029a3449e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nt.like not handling arrays as expected\nI am trying to use the `t.like` assertion on deep objects with arrays of objects in them.\n\nThe following simplified example fails:\n```ts\nimport test from 'ava';\n\ntest('deep t.like', (t) => {\n t.like({ a: [{ a: 1, b: 2 }] }, { a: [{ a: 1 }] });\n});\n```\n```\n> tsc -b\n\n$ ava dist/ava.spec.js\n\n deep t.like\n\n Difference:\n\n {\n a: [\n {\n a: 1,\n - b: 2,\n },\n ],\n }\n\n › src/ava.spec.ts:4:5\n\n ─\n\n 1 test failed\nerror Command failed with exit code 1.\n```\n\nI expected the above assertion to pass. \nIt seems that elements within arrays need to be deep equal rather than \"like\" for the assertion to pass.\n\n\nAVA version:\n```\n$ ava --version\n3.14.0\n```"} {"task_id": "format-code-task-000491", "source_id": "format-code-task-000491", "domain": "code", "task_path": "tasks/format-code-task-000491", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fdbd3d6becc7286bc4e84643acf2f7443c15811eb066dddb0a3a5639eca5cc99", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `av stack sync` leaves me on a detached HEAD after pruning merged branches\n\nI noticed this on my normal stacked-branches workflow with `av`. Steps:\n\n1. I'm on a feature branch (let's call it `feature-foo`) — verified with\n `git status` before I start.\n2. Run `av stack sync`.\n3. The sync rebases on top of `master`, pushes to GitHub, and asks me whether\n to delete branches that have been merged. I say yes (some parents earlier\n in the stack have already been merged into `master`).\n4. Sync finishes successfully.\n\nAfter it finishes, I'm no longer on `feature-foo`. `git status` shows\n`HEAD detached at ` instead. I have to manually `git checkout\nfeature-foo` to get back to where I was.\n\nIt happens whether or not `feature-foo` itself was one of the branches\ndeleted in the prune step — even when only an ancestor branch in the stack\ngets pruned, I still end up detached.\n\nWhat I'd expect:\n\n- After sync (including the prune step) finishes, I should be back on the\n same branch I was on when I ran `av stack sync`.\n- If the branch I was on actually got deleted during prune (because it was\n merged), then falling back to the default branch (`master`/`main`) would\n be reasonable.\n- Either way, I shouldn't be left on a detached HEAD — that's not a state\n I was in before running the command, and it's easy to miss and then\n accidentally commit onto.\n\nHappy to provide more details if useful."} {"task_id": "format-code-task-000492", "source_id": "format-code-task-000492", "domain": "code", "task_path": "tasks/format-code-task-000492", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:acf3745e983f2e5ce1c0a853b3d59b3e29d1fa282aa98557714388c142397af0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the Panel Highcharts pane classes to support dynamic series insertion on an already-rendered chart. After I create a chart object such as `chart = panel_highcharts.HighChart(object=config)`, `chart.add_series(options: dict, redraw: bool = True, animation: bool = True)` should send that one series options dictionary to the live browser chart without replacing the whole chart configuration.\n\nFor example, if the initial config has one series and I call `chart.add_series(options={\"name\": \"Forecast\", \"data\": [3, 4, 5]}, redraw=False, animation=False)`, the rendered chart should gain exactly that additional series while preserving the existing series, and the redraw and animation choices should be passed through for that insertion. If I call `chart.add_series(options={\"name\": \"Actual\", \"data\": [1, 2, 3]})` again later, that second call should add another series as a distinct update rather than being ignored because a prior add used the same method.\n\nThe same method should be available on the stateful pane objects for `HighChart`, `HighStock`, `HighMap`, and `HighGantt`. Series options supplied through `add_series` should be processed like normal chart configuration before insertion, so callback strings and event channel references inside the added series options are converted before the browser chart receives them."} {"task_id": "format-code-task-000493", "source_id": "format-code-task-000493", "domain": "code", "task_path": "tasks/format-code-task-000493", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:492084ac5c2e5a6607f941f33ea6173a7fa20681e6a3cf6c270fd9b3a3f7620f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Build fails on JRuby for files that contain only front matter\n\nI'm building a site with Awestruct on JRuby. A few of my pages are placeholders — they have a YAML front matter header (title, layout, some metadata) but no body content underneath yet. Something like a `foo.textile` that looks like:\n\n```\n---\ntitle: Coming soon\nlayout: base\n---\n```\n\n(nothing after the closing `---`).\n\nWhen I run the build, it crashes on these files. If I add even a single line of real content below the front matter, the same file builds fine. So it seems specifically tied to \"front matter only, empty body\".\n\nThis is annoying because it's a perfectly reasonable thing to have during development — a stub page that's wired into the navigation but doesn't have its prose written yet — and one such file shouldn't take down the whole site generation.\n\nI'd expect a file like this to just render to empty output and let the build move on."} {"task_id": "format-code-task-000494", "source_id": "format-code-task-000494", "domain": "code", "task_path": "tasks/format-code-task-000494", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:47c9843d0bc8b84dc042aa8f9bb74c4e0e8dd8e740dcee81909ea3b4ab0ee7d4", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## cfn-lint passes invalid AWS::EC2::Subnet templates with IPAM pools\n\nI've been migrating some of my CloudFormation VPC/Subnet code to use IPAM pools, and I noticed that cfn-lint is letting through Subnet templates that CloudFormation actually rejects at deploy time.\n\nTwo patterns I've hit:\n\n**1. No CIDR source specified.** A Subnet needs either `CidrBlock` or `Ipv4IpamPoolId`. If neither is set (e.g. I refactored to switch from a hard-coded CIDR to IPAM and forgot to wire in the IPAM properties), cfn-lint reports no errors, but the stack fails to create:\n\n```yaml\nResources:\n MySubnet:\n Type: AWS::EC2::Subnet\n Properties:\n VpcId: !Ref MyVpc\n AvailabilityZone: us-east-1a\n # no CidrBlock, no Ipv4IpamPoolId\n```\n\n**2. `Ipv4NetmaskLength` without `Ipv4IpamPoolId`.** I had a template where I'd set the netmask length but forgot to also set the IPAM pool id. cfn-lint accepts it, but the netmask length only makes sense paired with the pool id and the deploy fails:\n\n```yaml\nResources:\n MySubnet:\n Type: AWS::EC2::Subnet\n Properties:\n VpcId: !Ref MyVpc\n Ipv4NetmaskLength: 24\n # missing Ipv4IpamPoolId\n```\n\nThe second problem also shows up on `AWS::EC2::VPC` — cfn-lint will happily accept a VPC that sets `Ipv4NetmaskLength` on its own without `Ipv4IpamPoolId`.\n\nWould be great if the schema for these resources caught these configurations at lint time so I get the errors before pushing to CloudFormation."} {"task_id": "format-code-task-000495", "source_id": "format-code-task-000495", "domain": "code", "task_path": "tasks/format-code-task-000495", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:dccd0139ca2cd584c639839fa0c7fe80506282387a924a87045d92f28b252116", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### CloudFormation Lint Version\n\ngit (v1.17.2-2-ga5672f074)\n\n### What operating system are you using?\n\nMac\n\n### Describe the bug\n\n\nPR #3768 causes problems for `AWS::StepFunctions::StateMachine` resources in `json` templates that define `DefinitionString` using intrinsic functions to construct the embedded json string.\n\n[Sample template](https://docs.aws.amazon.com/AWSCloudFormation/latest/UserGuide/aws-resource-stepfunctions-statemachine.html#aws-resource-stepfunctions-statemachine--examples)\n\n```\n$ cfn-lint statemachine.json \nE1022 {'Fn::Join': ['\\n', ['{', ' \"StartAt\": \"HelloWorld\",', ' \"States\" : {', ' \"HelloWorld\" : {', ' \"Type\" : \"Task\", ', ' \"Resource\" : \"arn:aws:lambda:us-east-1:111122223333:function:HelloFunction\",', ' \"End\" : true', ' }', ' }', '}']]} is not of type 'object'\nstatemachine.json:10:17\n\nE3601 'StartAt' is a required property\nstatemachine.json:10:17\n\nE3601 'States' is a required property\nstatemachine.json:10:17\n```\n\n### Expected behavior\n\ncfn-lint passes\n\n### Reproduction template\n\n```json\n{\n \"AWSTemplateFormatVersion\" : \"2010-09-09\",\n \"Description\" : \"An example template for a Step Functions state machine.\",\n \"Resources\": {\n \"MyStateMachine\": {\n \"Type\": \"AWS::StepFunctions::StateMachine\",\n \"Properties\": {\n \"StateMachineName\" : \"HelloWorld-StateMachine\",\n \"StateMachineType\":\"STANDARD\",\n \"DefinitionString\" : {\n \"Fn::Join\": [\n \"\\n\",\n [\n \"{\",\n \" \\\"StartAt\\\": \\\"HelloWorld\\\",\",\n \" \\\"States\\\" : {\",\n \" \\\"HelloWorld\\\" : {\",\n \" \\\"Type\\\" : \\\"Task\\\", \",\n \" \\\"Resource\\\" : \\\"arn:aws:lambda:us-east-1:111122223333:function:HelloFunction\\\",\",\n \" \\\"End\\\" : true\",\n \" }\",\n \" }\",\n \"}\"\n ]\n ]\n },\n \"RoleArn\" : \"arn:aws:iam::111122223333:role/service-role/StatesExecutionRole-us-east-1\",\n \"Tags\": [\n {\n \"Key\": \"keyname1\",\n \"Value\": \"value1\"\n },\n {\n \"Key\": \"keyname2\",\n \"Value\": \"value2\"\n }\n ]\n }\n }\n }\n}\n```"} {"task_id": "format-code-task-000496", "source_id": "format-code-task-000496", "domain": "code", "task_path": "tasks/format-code-task-000496", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e98e8f25d443adb41a721c8e31076583d647af158ecf02f73118c421076daa09", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nTransformation Error 'Fn::ForEach could not be resolved' when using a Mapping Default in 1.21.0\n### CloudFormation Lint Version\n\ncfn-lint 1.21.0\n\n### What operating system are you using?\n\nFedora\n\n### Describe the bug\n\nI've got a template that uses an `Fn::ForEach` with `!FindInMap`, it uses a `DefaultValue`. Most the Mappings don't override the Default.\n\nIn `1.21.0`, when a Mapping wants to rely on the Default cfn-lint fails to lint the template.\n\nFrom my testing it's only affecting 1.21.0\n\n\n### Expected behavior\n\nTo succesfully Transform Fn::ForEach when FindInMap with a Default is used, and to Pass like in versions <1.21.0\n\n### Reproduction template\n\n### Test Code\n\nI wanted to test against multiple versions of `cfn-lint` to narrow down the scope. \n\n```bash\nfor version in \"1.21.0\" \"1.20.2\"; do \n echo \"cfn-lint $version\"\n pipx run --quiet --spec \"git+https://github.com/aws-cloudformation/cfn-lint@v${version}\" cfn-lint test/test.yaml\n echo \"exit code: $?\"\ndone\n```\n\n### Red\n#### Given a Template that Wants to use Default\nWe can make cfn-lint fail when the `Source` key is commented out of the Mapping\n\n```yaml\nAWSTemplateFormatVersion: \"2010-09-09\"\n\nMappings:\n IdentityProviders:\n Name:\n Github:\n - token.actions.githubusercontent.com/someorg\n Jenkins:\n - somehost.somedomain.sometld\n\n Roles:\n A:\n Subjects:\n - repo:organization/repository:ref:*\n # Source: Github\n\nTransform: AWS::LanguageExtensions\n\nResources:\n Fn::ForEach::DeploymentRole:\n - Role\n - - A\n - ${Role}:\n Type: AWS::IAM::Role\n Properties:\n AssumeRolePolicyDocument:\n Version: \"2012-10-17\"\n Statement:\n - Sid: AllowExternalIdP\n Principal:\n Fn::ForEach::PrincipalLoop:\n - IdP\n - !FindInMap\n - IdentityProviders\n - Name\n - !FindInMap\n - Roles\n - !Ref Role\n - Source\n - DefaultValue: Github\n - Federated: !Sub arn:${AWS::Partition}:iam::${AWS::AccountId}:oidc-provider/${IdP}\n Effect: Allow\n Action: sts:AssumeRoleWithWebIdentity\n```\n\n\n```sh\ncfn-lint 1.21.0\nE0001 Error transforming template: Fn::ForEach could not be resolved\ntest/test.yaml:33:23\n\nexit code: 2\ncfn-lint 1.20.2\nexit code: 0\n```\n\n## Green\n### Given a Mapping that supplies Source Value\n\nWe can make cfn-lint pass when the `Source` is included in the Mapping\n\n```yaml\nAWSTemplateFormatVersion: \"2010-09-09\"\n\nMappings:\n IdentityProviders:\n Name:\n Github:\n - token.actions.githubusercontent.com/someorg\n Jenkins:\n - somehost.somedomain.sometld\n\n Roles:\n A:\n Subjects:\n - repo:organization/repository:ref:*\n Source: Github\n\nTransform: AWS::LanguageExtensions\n\nResources:\n Fn::ForEach::DeploymentRole:\n - Role\n - - A\n - ${Role}:\n Type: AWS::IAM::Role\n Properties:\n AssumeRolePolicyDocument:\n Version: \"2012-10-17\"\n Statement:\n - Sid: AllowExternalIdP\n Principal:\n Fn::ForEach::PrincipalLoop:\n - IdP\n - !FindInMap\n - IdentityProviders\n - Name\n - !FindInMap\n - Roles\n - !Ref Role\n - Source\n - DefaultValue: Github\n - Federated: !Sub arn:${AWS::Partition}:iam::${AWS::AccountId}:oidc-provider/${IdP}\n Effect: Allow\n Action: sts:AssumeRoleWithWebIdentity\n\n```\n\nWhen we test cfn-lint passes \n\n```sh\ncfn-lint 1.21.0\nexit code: 0\ncfn-lint 1.20.2\nexit code: 0\n```\n\n### Passes without the Fn::ForEach::PrincipalLoop\n\nwanted to factor this a little further "} {"task_id": "format-code-task-000497", "source_id": "format-code-task-000497", "domain": "code", "task_path": "tasks/format-code-task-000497", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:423b2510ad176a48c1b7933a8e4fc0c4f85c64dac4552a797e64b4d4687390e3", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nFeature request: Add the ability to access the LambdaContext when using batch processing\n### Use case\n\nI'm looking to use the get_remaining_time_in_millis function specifically. In this case, the reason is that my batch size sometimes includes very large items that might time out if they are at the end of the batch. I want to add something to the top of the record_handler that checks the remaining time, and if it's below a threshold it just raises an exception for that record. Then the partial failure mechanism can put the failed items back on the queue.\n\nAs it is right now, if the function just times out, everything is put back, not just the failed items.\n\n### Solution/User Experience\n\nPassing the LambdaContext object into the record_handler would be great.\n\n### Alternative solutions\n\n_No response_\n\n### Acknowledgment\n\n- [X] This feature request meets [Lambda Powertools Tenets](https://awslabs.github.io/aws-lambda-powertools-python/latest/#tenets)\n- [ ] Should this be considered in other Lambda Powertools languages? i.e. [Java](https://github.com/awslabs/aws-lambda-powertools-java/), [TypeScript](https://github.com/awslabs/aws-lambda-powertools-typescript/)"} {"task_id": "format-code-task-000498", "source_id": "format-code-task-000498", "domain": "code", "task_path": "tasks/format-code-task-000498", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:01f392e293b52b8d0b0d10e04702b1584de6c4518b4c9e5d816419843051d7ac", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nFeature request: Additional HTTP Error Code Exceptions\n### Use case\n\nref: https://github.com/aws-powertools/powertools-lambda-python/blob/develop/aws_lambda_powertools/event_handler/exceptions.py\n\nI need to handle additional HTTP error codes, specifically 413 (Payload Too Large). I notice that the current exceptions.py only implements a subset of HTTP error codes (400, 401, 404, 500).\nWhile I understand I can extend ServiceError to create my own custom exceptions, this creates inconsistency in my codebase where some errors come from PowerTools and others from my custom definitions.\n\nAre there plans to expand the list of predefined exceptions to cover more HTTP status codes?\nWhat is the rationale behind which HTTP errors are currently implemented vs. those that aren't?\n\n### Solution/User Experience\n\nAdd status code\n\n### Alternative solutions\n\n```markdown\n\n```\n\n### Acknowledgment\n\n- [x] This feature request meets [Powertools for AWS Lambda (Python) Tenets](https://docs.powertools.aws.dev/lambda/python/latest/#tenets)\n- [x] Should this be considered in other Powertools for AWS Lambda languages? i.e. [Java](https://github.com/aws-powertools/powertools-lambda-java/), [TypeScript](https://github.com/aws-powertools/powertools-lambda-typescript/), and [.NET](https://github.com/aws-powertools/powertools-lambda-dotnet/)"} {"task_id": "format-code-task-000499", "source_id": "format-code-task-000499", "domain": "code", "task_path": "tasks/format-code-task-000499", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f2916e88c2637edcb29767fc08f17c9bf59e54b8897d46bc62400090654e7bb1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Container health status stops updating once container reaches RUNNING\n\nI'm running an ECS task whose container definition has a Docker `HEALTHCHECK` configured (one that flips between `starting` → `healthy` → `unhealthy` depending on a probe). I'd expect the ECS agent to keep my container's health status in sync with what Docker reports, so that `DescribeTasks` and the console reflect the current health.\n\nWhat I actually see: the health value the agent reports is essentially \"whatever Docker happened to say at the moment the container first transitioned to RUNNING\", and after that it never moves again, even when the container clearly becomes unhealthy and `docker inspect` shows the new state.\n\nIf I tail the agent logs while this is happening, I see a steady stream of lines like:\n\n```\nManaged task [...]: redundant container state change. to RUNNING, but already RUNNING\n```\n\n…and nothing else. So Docker is delivering events for my container after it reached RUNNING (which makes sense — health transitions are reported as container events), but the agent treats every one of those events as a no-op because the container's known status is already RUNNING. The health info / other metadata riding along with those events appears to just get dropped.\n\nOther metadata that I'd expect to be refreshed for a running container (e.g. what comes back from Docker about the container after start) seems to be affected by the same thing — once the container has settled into RUNNING, the agent stops folding new information into its in-memory container state.\n\nExpected behavior: while a container is sitting in steady-state RUNNING, the agent should still pick up updated info (especially health status) from subsequent Docker container events for that container, instead of treating \"RUNNING → RUNNING\" events as completely uninteresting. The \"redundant state change\" log is fine, but the side data on those events shouldn't be thrown away.\n\nRepro is basically:\n1. Run a task with a container that has a `HEALTHCHECK` that will eventually flip its health state.\n2. Wait until the container is RUNNING.\n3. Force the healthcheck to go unhealthy (e.g. break whatever the probe depends on).\n4. Observe via the agent introspection / `DescribeTasks` that the agent never reports the new health value."} {"task_id": "format-code-task-000500", "source_id": "format-code-task-000500", "domain": "code", "task_path": "tasks/format-code-task-000500", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:76d6b8e0a76203532af21ba6f0b75bbe1c956d87c61efc4620bb0f86e50200f2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\n现在跑 ECS task 的时候，我只能在 container 级别设 CPU 和内存，没法直接给整个 task 设一个总的上限，资源规划起来挺别扭的。最好能在 task 定义里直接写个 task 级别的 CPULimit 和 Memory，让整个 task 共享这个配额。另外这功能能不能搞个开关控制下，默认开着就行，但像 Windows 这种支持不了的环境，或者 Docker 版本太老不支持的时候，能自动识别然后禁掉、别硬上报错。\n\n# Expected outcomes\n\n- Task definition resource limits\n - The exported task model should accept and emit a task-level CPU limit through the JSON field `CPULimit`.\n - The exported task model should accept and emit a task-level memory limit through the JSON field `Memory`.\n - When these task-level limit fields are unset, normal JSON serialization should omit them rather than adding zero-valued fields.\n\n- Configuration control\n - The feature should be controlled by `ECS_ENABLE_TASK_CPU_MEM_LIMIT`.\n - If `ECS_ENABLE_TASK_CPU_MEM_LIMIT` is not set, the feature should be enabled by default.\n - If `ECS_ENABLE_TASK_CPU_MEM_LIMIT` is set to a false value, the feature should be disabled.\n - On Windows, the feature should remain disabled even if configuration requests enabling it.\n\n- Capability reporting and automatic disablement\n - When the feature is enabled and the local Docker API support is new enough for task-level CPU and memory limits, `DockerTaskEngine.Capabilities()` should advertise task CPU/memory limit support.\n - When the feature is disabled, `DockerTaskEngine.Capabilities()` should not advertise task CPU/memory limit support.\n - When the feature is enabled but the local Docker API support is too old, `DockerTaskEngine.Capabilities()` should not advertise task CPU/memory limit support and should disable the feature for that agent instance rather than returning an unusable capability.\n\n# Implementation notes\n\n- The concrete internal representation, validation location, and capability-detection plumbing are up to the implementer.\n- Preserve existing task, configuration, and capability behavior that is unrelated to task-level CPU/memory limits.\n- Keep platform-specific behavior consistent with existing platform configuration patterns in the repository.\n\n## Required output literals (exact-match contract)\n\nWhen `DockerTaskEngine.Capabilities()` advertises task-level CPU and memory limit support, it MUST include the capability attribute `ecs.capability.task-cpu-mem-limit` exactly. Downstream ECS capability negotiation consumes this attribute string.\n\nThe Docker API support threshold for task-level CPU and memory limits is Docker API version `1.22`: versions at or above this support level may advertise the capability when the feature is enabled, while older versions must not and must disable the feature for that agent instance."} {"task_id": "format-code-task-000501", "source_id": "format-code-task-000501", "domain": "code", "task_path": "tasks/format-code-task-000501", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:63702ff1bccbcc6fc35484b908a7d055a24382cf8b357c5aab7ddf3a8d99a3f2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nTag AWS resources with the CLI version\n[//]: # (Before raising the feature request, please check to see if an existing feature request already exists.)\n\n\n**Description**\n\n[//]: # (A short description of the feature you are proposing.)\n\nAGC creates resources in AWS as part of `account activate` and `context deploy` commands. To aid in debugging and reporting, we should tag these resources with the version of AGC used to create them.\n\n**Use Case**\n\n[//]: # (Why do you need this feature?)\n\nWhen debugging, it may be important to know which version a resource was deployed under. For example, if a user upgrades and are using an older version of the resource, they may encounter issues.\n\nWhen reporting on users, account managers may want to know which version of AGC their users are using.\n\n**Proposed Solution**\n\n[//]: # (Please include prototype/workaround/sketch/reference implementation.)\n\nAdd a `agc-version` tag with the CLI version used to deploy that resource."} {"task_id": "format-code-task-000502", "source_id": "format-code-task-000502", "domain": "code", "task_path": "tasks/format-code-task-000502", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b6feb2d2fda3a1d67a8d4479ddc82c4c4a4248503bd5b244f1fc7f9308c21b0e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nadd WES API endpoint information to output from `agc context describe`\n[//]: # (Before raising the feature request, please check to see if an existing feature request already exists.)\n\n\n**Description**\n\n[//]: # (A short description of the feature you are proposing.)\nAdd the WES API endpoint url for a context as part of the output from `agc context describe `\n\n\n**Use Case**\n\n[//]: # (Why do you need this feature?)\nUsers that write applications that speak GA4GH WES and leverage AGC to manage contexts need a way to retrieve the endpoint to talk to after the context has already been deployed.\n\n**Proposed Solution**\n\n[//]: # (Please include prototype/workaround/sketch/reference implementation.)\nAdd a `WESENDPOINT` to the output from `agc context describe `\n\n```\n$ agc context describe spotCtx\n2021-12-10T23:42:58Z � New version of agc available. Current version is '1.1.1'. The latest version is '1.1.2'\n2021-12-10T23:42:58Z � Describing context 'spotCtx'\nCONTEXT 256 spotCtx true STARTED \nOUTPUTLOCATION s3://agc-733263974272-us-east-2/project/Demo/userid/pwymingJKP3z/context/spotCtx\nWESENDPOINT https://d1jhugndd0.execute-api.us-east-2.amazonaws.com/prod/ga4gh/wes/v1\n```\n\n**Other information**\n\n[//]: # (detailed explanation, stacktraces, related issues, suggestions how to fix, links for us to have context, eg. associated pull-request, stackoverflow, slack, etc)"} {"task_id": "format-code-task-000504", "source_id": "format-code-task-000504", "domain": "code", "task_path": "tasks/format-code-task-000504", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:69ca5fd0dd89f0e62a441d59f203e3f985cca9e09ed7b82dd6e146558dae5098", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Add support for HTTP API Gateway events\n\nThe `events` package in `aws-lambda-go` provides strongly-typed structs for unmarshalling Lambda event payloads, e.g. `APIGatewayProxyRequest` for the REST API Gateway integration. This works great when my Lambda is triggered by a REST API.\n\nHowever, AWS also offers HTTP API Gateway, which is a separate product with a different (and incompatible) event payload format from the classic REST API Gateway proxy integration. I've configured an HTTP API Gateway in front of my Go Lambda, and when I try to reuse `APIGatewayProxyRequest` to decode the incoming event, the fields don't line up — the payload my Lambda actually receives doesn't match this struct, so a bunch of values come back empty / zero and I lose information that's present in the raw JSON.\n\nRight now my only options are:\n\n1. Write my own struct that mirrors the HTTP API payload shape, or\n2. Decode into a `map[string]interface{}` and pull fields out by string keys.\n\nBoth feel wrong for a library whose whole job is to give us typed event structs for every supported Lambda trigger.\n\nCould we add a first-class type in `events/apigw.go` (alongside the existing `APIGatewayProxyRequest`) for the HTTP API Gateway event payload, so it can be used the same way:\n\n```go\nfunc handler(ctx context.Context, req events./* HTTP API Gateway type */) (..., error) {\n // ...\n}\n```\n\nThe shape should follow the official AWS docs for the HTTP API Gateway payload format so users can rely on the JSON tags matching what API Gateway actually sends.\n\nThe new top-level type I'd expect is something like `APIGatewayV2HTTPRequest` (with a corresponding nested request-context type)."} {"task_id": "format-code-task-000505", "source_id": "format-code-task-000505", "domain": "code", "task_path": "tasks/format-code-task-000505", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:57e6109bb0919efb1148e11ff6e0a9ccbc35ecbb20feead6c36a535439dbdaf2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## LoginNodes should fall back to HeadNode's SSH key when no Ssh section is provided\n\nI'm setting up a Slurm cluster with both a HeadNode and a LoginNodes pool. I want to access everything (head node and login nodes) with the same SSH key, so my config looks roughly like:\n\n```yaml\nHeadNode:\n InstanceType: t2.micro\n Networking:\n SubnetId: subnet-xxxxxxxx\n Ssh:\n KeyName: my-key\nLoginNodes:\n Pools:\n - Name: pool1\n InstanceType: t2.micro\n Count: 1\n Networking:\n SubnetIds:\n - subnet-xxxxxxxx\nScheduling:\n Scheduler: slurm\n ...\n```\n\nWhen I run `pcluster create-cluster` with this, validation rejects the config because the `Ssh` block under each LoginNodes pool is treated as required — I have to repeat `Ssh: KeyName: my-key` on every pool even though I already declared the same key for the HeadNode.\n\nThis feels redundant. In the common case where the operator just wants one key for the whole cluster, the HeadNode's `KeyName` is already enough information. Could the LoginNodes `Ssh` (or at least its `KeyName`) be made optional, so that omitting it just reuses what's configured for the HeadNode? Explicitly setting a different `KeyName` on a LoginNodes pool should of course still work and override that fallback."} {"task_id": "format-code-task-000506", "source_id": "format-code-task-000506", "domain": "code", "task_path": "tasks/format-code-task-000506", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a4cb9ea41f24d96c3ed98c412115e8f4c58b792a584cf110a80a41cc7c81721f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### s3manager.Upload API shape — align with upcoming Download manager\n\nI'm working on adding a download manager to `s3manager` (counterpart to the existing upload manager). While sketching the API I realized the two are going to look inconsistent if I leave Upload as-is.\n\nToday, uploading looks like this:\n\n```go\nsvc := s3.New(nil)\nresp, err := s3manager.Upload(svc, &s3manager.UploadInput{...}, nil)\n```\n\ni.e. `Upload` is a package-level function that takes the `*s3.S3` client, the input, and an options struct as positional arguments. The download manager I'm building really wants to be a reusable object — you configure it once (concurrency, part size, which S3 client to use) and then call it many times against different inputs. Forcing the same shape on Upload would mean every call site has to re-pass the client and the opts, and the two managers would read very differently side by side even though they're conceptually symmetric.\n\nI'd like to reshape Upload along the same lines: configure once, then call `Upload` on the configured object as many times as you want, ideally safe to share across goroutines. The S3 client should be part of that one-time configuration rather than a required positional argument on every call — and it'd be nice if it were optional, so simple use cases (`just upload this file with defaults`) don't have to construct an `s3.S3` themselves.\n\nI'm fine with this being a breaking change for Upload callers — better to do it now, before the download manager ships and locks in the asymmetry, than to introduce two differently-shaped APIs and try to reconcile them later.\n\nThe existing knobs in `UploadOptions` (PartSize, Concurrency, LeavePartsOnError) and the zero-value-means-default behavior should keep working the way they do now; this is purely about the call shape, not about changing what the upload itself does.\n\nNaming-wise I'd expect something like `s3manager.NewUploader(opts)` returning the reusable object, with the S3 client living on `UploadOptions` (e.g. an `S3` field)."} {"task_id": "format-code-task-000507", "source_id": "format-code-task-000507", "domain": "code", "task_path": "tasks/format-code-task-000507", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:30d0445cdb227abf79f4517e459395b8ffa0840009959e419bd6700098872e62", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nCloudFront Sign creates invalid signature if URL contains &\nConfirm by changing [ ] to [x] below to ensure that it's a bug:\n- [x] I've gone though the [API reference](https://docs.aws.amazon.com/sdk-for-go/v2/api/)\n- [x] I've checked [AWS Forums](https://forums.aws.amazon.com) and [StackOverflow](https://stackoverflow.com/questions/tagged/aws-sdk-go) for answers\n- [x] I've searched for [previous similar issues](https://github.com/aws/aws-sdk-go-v2/issues) and didn't find any solution\n \n**Describe the bug**\nAfter switching from the V1 SDK to the V2 SDK, signing a URL with more than 2 query parameters results in an invalid signature and cloudfront will return a 403. For a url with 1 or less query parameters, the V1 SDK and V2 SDK signatures are identical. But if you have a URL with more than 2 query parameters, the V2 SDK will return a different signature than the V1 SDK. This looks like a regression from the v1 SDK - an issue was raised and fixed in V1 but it looks like the fix is not included in the v2 SDK https://github.com/aws/aws-sdk-go/issues/2163\n\n**Version of AWS SDK for Go?**\n1.40.18\n\n**Version of Go (`go version`)?**\ngo1.16.4 darwin/amd64\n\n**To Reproduce (observed behavior)**\nCode example copied from this issue: https://github.com/aws/aws-sdk-go/issues/2163\n```\nsigner := sign.NewURLSigner(keyID, privKey)\nsignedURL, _ := signer.Sign(\"https://myuploaddomain.com/\", time.Now().Add(1*time.Hour))\n// signedURL VALID\nsignedURL, _ := signer.Sign(\"https://myuploaddomain.com?key=one\", time.Now().Add(1*time.Hour))\n// signedURL VALID\nsignedURL, _ := signer.Sign(\"https://myuploaddomain.com?key=one&value=two\", time.Now().Add(1*time.Hour))\n// signedURL INVALID returns AccessDenied when uploading\n\n```\n\n**Expected behavior**\nSigning a URL containing two query string parameters should result in the same signature as the V1 SDK."} {"task_id": "format-code-task-000508", "source_id": "format-code-task-000508", "domain": "code", "task_path": "tasks/format-code-task-000508", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5f36c050850e54f023d53a926dc1c3285037c0a45485f59dce94a4b48c0caf95", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `pipeline init` defaults the CodeCommit branch to `master`, but new CodeCommit repos use `main`\n\nI'm trying to set up a Copilot pipeline that's triggered from an AWS CodeCommit repository. My repo is fairly new — when I created it in CodeCommit, the default branch was `main` (which I believe is the current default for new CodeCommit repos).\n\nI ran `copilot pipeline init` from my workspace, picked my CodeCommit repo URL when prompted, and accepted the default branch (i.e. I didn't pass `--git-branch`). When I looked at the generated pipeline manifest, the source branch was set to `master`. After deploying the pipeline, the Source stage points at a branch that doesn't exist in my repo, so the pipeline can't actually pull anything or get triggered by pushes.\n\nIf I explicitly pass the branch name it works fine, so this is just about what the default should be when CodeCommit is the source.\n\nLooking at the docs page for the pipeline manifest, it also says:\n\n> The name of the branch in your repository that triggers the pipeline. The default for GitHub is `main`; the default for Bitbucket and CodeCommit is `master`.\n\nSo at least for the CodeCommit half of that sentence, the documented default doesn't match what new CodeCommit repos actually use. It would be great if the CodeCommit default in `pipeline init` (and the corresponding docs) were updated to match what users actually get out of CodeCommit today, so things \"just work\" without having to override the branch every time."} {"task_id": "format-code-task-000509", "source_id": "format-code-task-000509", "domain": "code", "task_path": "tasks/format-code-task-000509", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:52260dbf6f30f2882c284d45ce43b058dce076a8981232866c0347d04306bcaa", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a manifest interpolation API in the Go `manifest` package that can be used as `manifest.NewInterpolator(appName string, envName string).Interpolate(input string) (string, error)`. The interpolator should parse the input as YAML, substitute `${VAR}` placeholders only in YAML string values, and return the resulting YAML text with normal YAML formatting.\n\n`COPILOT_APPLICATION_NAME` and `COPILOT_ENVIRONMENT_NAME` should be predefined from the constructor arguments. For example, with `manifest.NewInterpolator(\"myApp\", \"test\")`, interpolating `image: repo/${COPILOT_ENVIRONMENT_NAME}:${TAG}` with OS environment variable `TAG=latest` should return YAML containing `image: repo/test:latest`. With `SECURITY_GROUPS=[\"sg-1\",\"sg-2\"]`, interpolating `network:\\n vpc:\\n security_groups: ${SECURITY_GROUPS}` should turn that string value into a YAML sequence containing `sg-1` and `sg-2` rather than leaving it as a JSON-looking string.\n\nEscaped placeholders should remain literal after removing the escape slash: interpolating `command: echo \\\\${NAME}` should produce YAML containing `command: echo ${NAME}` and should not require `NAME` to be set. If a placeholder refers to an unset OS environment variable, `Interpolate` should return an error like `environment variable \"NAME\" is not defined`. If the OS environment tries to override a predefined Copilot variable with a different value, such as setting `COPILOT_ENVIRONMENT_NAME=prod` while the interpolator was created with env name `test`, `Interpolate` should return an error saying that predefined variable cannot be overridden.\n\nCalling `Interpolate` multiple times with the same constructor arguments, input YAML, and OS environment should produce the same output and should not mutate the input string or write to the filesystem or network."} {"task_id": "format-code-task-000510", "source_id": "format-code-task-000510", "domain": "code", "task_path": "tasks/format-code-task-000510", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6ee8b0d8b0f89e11e859284bbfa5504b9ba63cad3321fe6700f71865125b747b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nOption to apply user classnames\nIt would be a very nice feature to allow us to set our own class names on the resulting iframe. Right now it is hardcoded to a static value:\nhttps://github.com/awslabs/amazon-quicksight-embedding-sdk/blob/79bdce13d916ed906d22c198720ca13a42fdb0bd/src/EmbeddableDashboard.js#L139\n\nMaybe something like this?\n```js\n QuickSightEmbedding.embedDashboard({\n url,\n container,\n scrolling: 'no',\n height: 'AutoFit',\n loadingHeight: '400px',\n className: 'my-apps-classname', // <-----\n }).on('error', this.handleError);\n```\n\nFor example, this could be used to apply bootstrap class names to the iframe itself to change the border or shadow, without having to duplicate framework features in custom css."} {"task_id": "format-code-task-000511", "source_id": "format-code-task-000511", "domain": "code", "task_path": "tasks/format-code-task-000511", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d4e57a04bd7d544df911b90f54d01364a426dd045d1d59a98ea24b078feb59f7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI'm looking for an ability to disable the metric extraction and ingestion into CloudWatch Metrics, while still logging them to CloudWatch Logs.\n\nThat ability is important for me as I want to be able to disable ingestion into CloudWatch Metrics on demand for cost saving purposes while still receiving the data in CloudWatch Logs enabling me to analytics with CloudWatch Insights, even though they aren't ingested into CloudWatch Metrics.\n\nFrom reading through the EMF specification (https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/CloudWatch_Embedded_Metric_Format_Specification.html) I didn't notice any flag to indicate if metrics should be ingested into CloudWatch Metrics or not. \n\nIs there an option to choose whether such metrics should get ingested by CloudWatch Metrics or not? If not it'd be great to have such an option.\n\nI'd expect this to be exposed as a config option, something like `Config.disable_metric_extraction` (with a matching `AWS_EMF_DISABLE_METRIC_EXTRACTION` environment variable)."} {"task_id": "format-code-task-000512", "source_id": "format-code-task-000512", "domain": "code", "task_path": "tasks/format-code-task-000512", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:35b47e188a778d973ebdfd45376f72c7ebb4822e7cb26032c9e5eedae36a8cd8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want an `ExpectedNumInstanceSampler` available from `gluonts.transform` for training pipelines that need roughly a target number of split points per time series. I should be able to construct it with keyword fields `num_instances: float`, `min_instances: int = 0`, `axis: int = -1`, `min_past: int = 0`, and `min_future: int = 0`, then call it as `sampler(ts: np.ndarray) -> np.ndarray` to get integer split indices.\n\nFor each successful call, it should compute the inclusive valid split interval from `min_past` through `ts.shape[axis] - min_future`; if that interval is empty, it returns an empty integer array and does not advance its running counters. Otherwise, the sampler should keep state across calls by incrementing `n` and adding the current valid window size to `total_length`, then use `num_instances / (total_length / n)` as the per-index sampling probability.\n\nFor a concrete deterministic case, with `sampler = ExpectedNumInstanceSampler(num_instances=2, min_past=1, min_future=1)`, `ts = np.zeros(5)`, and `np.random.random_sample(4)` producing `[0.9, 0.2, 0.8, 0.1]`, calling `sampler(ts)` should return `np.array([2, 4])`, because the valid split indices are `[1, 2, 3, 4]` and the probability is `2 / 4`. With `sampler = ExpectedNumInstanceSampler(num_instances=1, min_instances=2)`, `ts = np.zeros(3)`, `np.random.random_sample(4)` producing values above the sampling probability for every position, and `np.random.randint(0, 4, size=2)` producing `[3, 1]`, the call should return `np.array([3, 1])` so that the requested minimum number of instances is still met. With `ExpectedNumInstanceSampler(num_instances=4, min_past=3, min_future=2)(np.zeros(4))`, the result should be an empty integer array because there is no valid split interval."} {"task_id": "format-code-task-000513", "source_id": "format-code-task-000513", "domain": "code", "task_path": "tasks/format-code-task-000513", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:879a8228679d87491ab95298e39d66252aecaa585e67cbfb1110d26d100e1a0d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `QuantileForecast.quantile(inference_quantile: Union[float, str]) -> np.ndarray` to let users query a quantile-backed forecast by level and get a NumPy array aligned to the forecast horizon. Given a `QuantileForecast` with `forecast_arrays=np.array([[1.0, 2.0], [10.0, 20.0], [100.0, 200.0]])`, `forecast_keys=[\"0.1\", \"0.5\", \"0.9\"]`, and any valid start period, `forecast.quantile(0.5)` and `forecast.quantile(\"p50\")` should both return `np.array([10.0, 20.0])`. For a stored-decile request such as `forecast.quantile(\"0.1\")`, it should return `np.array([1.0, 2.0])`.\n\nFor a missing quantile level inside the stored range, it should linearly interpolate between the surrounding stored forecast arrays; with the same forecast, `forecast.quantile(0.75)` should return `np.array([66.25, 132.5])`. For requests in the effective left or right tails, it should extrapolate from the two nearest stored tail quantiles using an exponential tail approximation instead of returning NaN. If the forecast has fewer than two stored quantile arrays, or only a stored `\"mean\"` array, a quantile request that is not present should return an all-NaN NumPy array with one value per prediction step.\n\nCalling the method repeatedly with the same forecast and same quantile should return equal arrays, and querying a quantile should not mutate the forecast arrays or write to filesystem, network, or global state."} {"task_id": "format-code-task-000514", "source_id": "format-code-task-000514", "domain": "code", "task_path": "tasks/format-code-task-000514", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1ed2d613daf206f13e3526e19ff5ab06e26a1dd5da583c9542946a4544458e84", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI’ve really been enjoying using Goformation! I found a bug recently though that I’m not sure how to fix (I'm still learning Golang), but I tracked down the reason, and figured you might immediately know how to fix.\n\n**Summary**:\nThe Fn::Join is not joining values together. I've traced this down to find that it is not reading the array inside the array. I believe the \"goformation/intrinsics/fnjoin.go\" file needs to be updated to loop the inner array. Let me know if you’d like me to put this in Github Issues\n \n**# TROUBLESHOOTING WALKTHROUGH**\n \nExample CF snippet from the example template in the repo:\n \n```\nRole:\n Fn::ImportValue:\n !Join ['-', [!Ref 'ProjectId', !Ref 'AWS::Region', 'LambdaTrustRole']]\n```\n \nIf you try to unmarshall, and output the function.Role, it gives you nothing:\n \n```\n > 2018/05/12 08:40:09 Found a AWS::Serverless::Function named GetHelloWorld (Role: )\n```\n\nLet's remove the ImportValue function:\n \n```\n Role: !Join ['-', [!Ref 'ProjectId', !Ref 'AWS::Region', 'LambdaTrustRole']]\n```\n \nIf you try to unmarshall this again, and output the function.Role as before, it gives you just \"-\":\n \n```\n > 2018/05/12 08:40:34 Found a AWS::Serverless::Function named GetHelloWorld (Role: -)\n```\nSo then for fun, I removed the inner array, and this worked:\n \n```\n Role: !Join ['-', !Ref 'AWS::Region']\n```\n \nThe output is:\n \n```\n > 2018/05/12 08:42:03 Found a AWS::Serverless::Function named GetHelloWorld (Role: -us-east-1)\n``` \nHowever, this syntax is in invalid.\n\nI'm pretty sure the \"goformation/intrinsics/fnjoin.go\" file needs to loop the inner array.\n\nThanks!"} {"task_id": "format-code-task-000515", "source_id": "format-code-task-000515", "domain": "code", "task_path": "tasks/format-code-task-000515", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a7e65838b11dd755863e27960a79ac2303c232c262a3f774276f00a58f1b0ea5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want cached music library search exposed consistently through the library API stack: `LibraryAPI.searchLibrary(query: string): Promise`, `LibraryService.searchLibrary(query: string): Promise`, `LibraryDataService.searchLibrary(query: string): Promise`, `libraryGateway.search(query: string): Promise`, and `electronAPI.library.search(query: string): Promise`, backed by the `library:search` IPC channel.\n\nThe search should operate on the already loaded library cache and return matching tracks without scanning files, parsing metadata, or writing cache data. Blank or whitespace-only queries should return all cached tracks in their current cache order. Non-empty queries should be case-insensitive substring searches over each track's title, artist, album, and file name.\n\nFor example, if the cache contains `Blue Monday` by `New Order` on album `Substance` with file name `blue-monday.flac`, and `Around the World` by `Daft Punk` with file name `around.mp3`, searching for `order`, `SUBSTANCE`, or `flac` should return the `Blue Monday` track. Searching for `around` should return the `Around the World` track, and searching for `missing-term` should return an empty array. A non-string query should fail input validation before an IPC request is sent. If the cache-side search or IPC call fails, the public async search methods should resolve to an empty array rather than crashing the caller."} {"task_id": "format-code-task-000516", "source_id": "format-code-task-000516", "domain": "code", "task_path": "tasks/format-code-task-000516", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:161c27c7c4bd672e5578b21ee7934018b981abed5d2313ebe8087a0e6dc4317e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAdd generic type to response headers\n### Is your feature request related to a problem? Please describe.\n\nSometimes APIs use response headers to communicate important metadata, such as rate limit information. At the moment there is no way to type what response headers we expect for a given request, which makes it hard to express/document this.\n\nMy use-case is primarily around code generation from openapi specifications, where I'd like to be able to include the declared response headers in the generated code, eg:\n```yaml\n headers:\n X-RateLimit-Limit:\n \"$ref\": \"#/components/headers/x-rate-limit-limit\"\n X-RateLimit-Remaining:\n \"$ref\": \"#/components/headers/x-rate-limit-remaining\"\n X-RateLimit-Reset:\n \"$ref\": \"#/components/headers/x-rate-limit-reset\"\n```\nShould be expressed in output client functions like: https://github.com/mnahkies/openapi-code-generator/blob/main/integration-tests/typescript-axios/src/generated/api.github.com.yaml/client.ts#L1898-L1946\n\n### Describe the solution you'd like\n\nA new generic parameter added to the `AxiosResponse` and associated header types that could be used similar to this:\n```typescript\ntype ResponseBody = {...}\n\ntype ResponseHeaders = {\n \"x-rate-limit-remaining\": string\n \"x-rate-limit-used\": string\n}\n\nfunction getSomeResource(): Promise> {\n...\n}\n```\n\nUnfortunately there is already a second generic parameter on the `AxiosResponse` so care would be needed to do this in a way that was non-breaking.\n\n### Describe alternatives you've considered\n\nI can probably workaround this on my side, either by moving away from the `AxiosResponse` type, or through an intersection type or similar, but it would be nice if `axios` made it easier to express this kind of information out of the box.\n\n### Additional context/Screenshots\n\nThere are a couple of semi-related issues:\n- https://github.com/axios/axios/issues/5967\n- https://github.com/axios/axios/issues/6677\n\nThat would also be nice to see resolved."} {"task_id": "format-code-task-000517", "source_id": "format-code-task-000517", "domain": "code", "task_path": "tasks/format-code-task-000517", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:df353a0ef57b30dad087f5275bd48f5bbeb63361a5a4eff12b01a08250ef668d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI get ValueError when trying to run Axoltlt on pretraining dataset\nI am having the same issue except my dataset is not large (144 records). I want to finetune a model and my yaml is like this: \n\n```yaml\nbase_model: meta-llama/Meta-Llama-3-8B\nmodel_type: LlamaForCausalLM\ntokenizer_type: AutoTokenizer\n\nload_in_8bit: false\nload_in_4bit: true\nstrict: false\n\nmax_steps: 100\npretraining_dataset:\n - path: dummy_dataset\n type: completion # (I also tried `pretrain`)\ndataset_prepared_path:\nval_set_size: 0.3\noutput_dir: ./outputs/finetuned_model\n\nadapter: qlora\nlora_model_dir:\n\nsequence_len: 8192\nsample_packing: false\neval_sample_packing: false\npad_to_sequence_len: false\n\nlora_r: 16\nlora_alpha: 16\nlora_dropout: 0\nlora_target_modules: [\"q_proj\", \"k_proj\", \"v_proj\", \"o_proj\", \"gate_proj\", \"up_proj\", \"down_proj\"]\nlora_target_linear: true\nlora_fan_in_fan_out:\n\nwandb_project:\nwandb_entity:\nwandb_watch:\nwandb_name:\nwandb_log_model:\n\ngradient_accumulation_steps: 1\nmicro_batch_size: 8\nnum_epochs: 2\noptimizer: paged_adamw_32bit\nlr_scheduler: cosine\nlearning_rate: 0.0002\n\ntrain_on_inputs: false\ngroup_by_length: false\nbf16: auto\nfp16: true\ntf32: false\n\ngradient_checkpointing: true\nearly_stopping_patience:\nresume_from_checkpoint:\nlocal_rank:\nlogging_steps: 4\nxformers_attention:\nflash_attention: true\n\nwarmup_steps: 2\nevals_per_epoch: 1\nsaves_per_epoch: 1\ndebug:\ndeepspeed:\nweight_decay: 0.0\nfsdp:\nfsdp_config:\nspecial_tokens:\n```\n\nAnd my dataset is in JSONL format following this format: \n\n```jsonl\n{\"text\": \"\"}\n{\"text\": \"\"}\n...\n```\n\nAnd I keep getting: \n\n```\n[rank0]: Traceback (most recent call last):\n[rank0]: File \"\", line 198, in _run_module_as_main\n[rank0]: File \"\", line 88, in _run_code\n[rank0]: File \"/home/axolotl/src/axolotl/cli/train.py\", line 70, in \n[rank0]: fire.Fire(do_cli)\n[rank0]: File \"/home/bsci/venv/lib/python3.11/site-packages/fire/core.py\", line 143, in Fire\n[rank0]: component_trace = _Fire(component, args, parsed_flag_args, context, name)\n[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n[rank0]: File \"/home/bsci/venv/lib/python3.11/site-packages/fire/core.py\", line 477, in _Fire\n[rank0]: component, remaining_args = _CallAndUpdateTrace(\n[rank0]: ^^^^^^^^^^^^^^^^^^^^\n[rank0]: File \"/home/bsci/venv/lib/python3.11/site-packages/fire/core.py\", line 693, in _CallAndUpdateTrace\n[rank0]: component = fn(*varargs, **kwargs)\n[rank0]: ^^^^^^^^^^^^^^^^^^^^^^\n[rank0]: File \"/home/axolotl/src/axolotl/cli/train.py\", line 38, in do_cli\n[rank0]: return do_train(parsed_cfg, parsed_cli_args)\n[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n[rank0]: File \"/home/axolotl/src/axolotl/cli/train.py\", line 66, in do_train\n[rank0]: return train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)\n[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n[rank0]: File \"/home/axolotl/src/axolotl/train.py\", line 170, in train\n[rank0]: trainer.train(resume_from_checkpoint=resume_from_checkpoint)\n[rank0]: File \"/home/bsci/venv/lib/python3.11/site-packages/transformers/trainer.py\", line 1539, in train\n[rank0]: return inner_training_loop(\n[rank0]: ^^^^^^^^^^^^^^^^^^^^\n[rank0]: File \"/home/bsci/venv/lib/python3.11/site-packages/transformers/trainer.py\", line 1836, in _inner_training_loop\n[rank0]: for step, inputs in enumerate(epoch_iterator):\n[rank0]: File \"/home/bsci/venv/lib/python3.11/site-packages/accelerate/data_loader.py\", line 677, in __iter__\n[rank0]: next_batch, next_batch_info = self._fetch_batches(main_iterator)\n[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n[rank0]: File \"/home/bsci/venv/lib/python3.11/site-packages/accelerate/data_loader.py\", line 631, in _fetch_batches\n[rank0]: batches.append(next(iterator))\n[rank0]: ^^^^^^^^^^^^^^\n[rank0]: F"} {"task_id": "format-code-task-000519", "source_id": "format-code-task-000519", "domain": "code", "task_path": "tasks/format-code-task-000519", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:8075c3c8e471b069495bf5bba1ef9a4160cfc3a1a4012ee13297ac9149c43a7b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `rastervision run local CFG_MODULE [COMMANDS...]` to be a built-in way to execute a configured pipeline on the current machine using local subprocesses. After the shared `run` command has written the serialized pipeline config, choosing the `local` runner should generate a `Makefile` in the same directory as that config and invoke `make -j -f `.\n\nThe generated Makefile should contain phony numeric targets and an `all` target over every command job. Each non-split command should run once with a recipe shaped like `python -m rastervision.pipeline.cli run_command CFG_JSON_URI COMMAND`. If a selected command is listed in the pipeline's split-capable commands and `--splits N` is greater than 1, it should create N separate targets for that command, each calling `run_command` with `--num-splits N` and its own `--split-ind` value from `0` through `N - 1`.\n\nCommand ordering should be preserved through Makefile dependencies: jobs for the first selected stage have no dependencies, jobs for the next selected stage depend on all jobs produced by the previous selected stage, and so on. For example, with selected commands `chip train predict`, split-capable commands `chip` and `predict`, and `--splits 2`, the Makefile should run two `chip` targets in parallel, make `train` depend on both `chip` targets, and make both `predict` targets depend on `train`. The CLI should exit with code 0 when the `make` process exits 0, and should exit with the same non-zero code when `make` fails."} {"task_id": "format-code-task-000520", "source_id": "format-code-task-000520", "domain": "code", "task_path": "tasks/format-code-task-000520", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7347126ba72ff94faa50eb29651bd5136dba5b0730308c134edd2736bb13ef13", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Raster Vision to provide a raster source API for turning vector label geometries into semantic-segmentation mask chips. The direct API should be `RasterizedSource(vector_source, background_class_id: int, bbox: Box | None = None, all_touched: bool = False, raster_transformers: list = [])`, where `vector_source` supplies a GeoDataFrame with a `class_id` column and a CRS transformer.\n\nAfter construction, `source.dtype` should be `np.uint8`, `source.crs_transformer` should be the same transformer exposed by the vector source, and `source.get_chip(window)` should return a uint8 array shaped `(height, width, 1)`. For a 10 by 10 window with a polygon covering pixel coordinates `[0, 0]` through `[5, 5]`, class id `0`, and `background_class_id=1`, `get_chip(Box.make_square(0, 0, 10))` should return background value `1` everywhere except the covered 5 by 5 region, which should be `0`. For an empty vector source, the same call should return a full chip filled with `background_class_id`. When `all_touched=False`, a tiny polygon that touches but does not contain a pixel center should not burn that pixel; when `all_touched=True`, the touched pixel should receive the geometry class id.\n\nThe source should validate labels during construction. If any label geometry is a `Point` or `LineString`, construction should raise `ValueError` explaining that those geometry types are unsupported unless they are buffered into polygons. If the vector data is non-empty and lacks a `class_id` column, construction should raise `ValueError` saying all label polygons must have a class id.\n\nI also want configuration support through `RasterizerConfig(background_class_id: int, all_touched: bool = False)` and `RasterizedSourceConfig(vector_source: VectorSourceConfig, rasterizer_config: RasterizerConfig)`. `RasterizedSourceConfig` should register as `rasterized_source`, `RasterizerConfig` should register as `rasterizer`, and building the config with `build(class_config, crs_transformer, bbox=None)` should build the configured vector source and return a `RasterizedSource` using the configured background class and `all_touched` values. If the vector source config does not already include a class-inference transformer or point/line buffer transformers, the config should add defaults so GeoJSON labels can be assigned class ids and point or line labels can be buffered into polygons before rasterization."} {"task_id": "format-code-task-000521", "source_id": "format-code-task-000521", "domain": "code", "task_path": "tasks/format-code-task-000521", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:255748b322c2a66f1828bddd436dee8a4ad459b3eb275bf09cc17a6063333c48", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want an initialized `rc_params_management` object to provide `write_RC_file(outputFilename, docstring=None) -> None` so a model run can save its active rc parameters as a reusable rc file. When I call `manager.write_RC_file('/tmp/model.pyabmrc')` without a custom docstring, it should create or overwrite that file, write the built-in default rc-file header, then a blank line, then emit the parameters from the loaded default definitions in their original order. The generated body should preserve comment-only lines and blank lines from the default definitions, and each parameter line should be formatted as `key : value` with any inline comment kept after a single separating space.\n\nThe values written for keys should come from the manager's current rc parameter dictionary, not from the textual defaults, so if the active `random_seed` is `12345` the output should contain `random_seed : 12345`. If a caller passes `docstring='# My model configuration'`, that text should replace the built-in header and be followed by exactly one blank line before the parameter body. The method should ignore active parameter entries that are not present in the loaded default definitions, because generated rc files should contain only keys that can be read back later. It should return `None`, should not mutate the manager's current parameter values, and if the target file cannot be opened for writing it should let the underlying file-open exception propagate instead of silently succeeding."} {"task_id": "format-code-task-000522", "source_id": "format-code-task-000522", "domain": "code", "task_path": "tasks/format-code-task-000522", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6560ae5e40dd9b1eb63edbafbe47887aa62354f754b8ac316261e2fb3dcd072f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a stateful `EosDetector` C++ helper for streaming chat generation stop detection. It should be constructed as `EosDetector(size_t nTokens, const int *tokens, const char **pieces, int paddingLeft, int paddingRight)`, where `tokens` are EOS token ids and `pieces` are the corresponding textual stop markers. It should expose `EosDetectorType append(int tokenId, const char *piece)`, `bool isEos(int tokenId)`, `char *getDelta()`, and `void reset()`, with `EosDetectorType` values `MAYBE_EOS`, `EOS`, and `NOT_EOS`.\n\nFor a detector created with token ids `{10000, 10001}`, stop pieces `{ \"\", \"\" }`, and padding `1, 1`, appending pieces `\"<\"`, `\"eo\"`, and `\"s>\"` should return `MAYBE_EOS`, `MAYBE_EOS`, then `EOS`, and `getDelta()` should return null so the stop marker is not shown to the user. After `reset()`, appending `\"!<\"`, `\"eos\"`, and `\"> \"` should end with `EOS` and `getDelta()` should return only `\"!\"`. If a possible stop prefix turns out not to be a stop, such as appending `\"XY\"`, it should return `NOT_EOS` and `getDelta()` should return the full buffered text `\"XY\"`.\n\nIf `append()` receives an EOS token id with a null piece, it should return `EOS`; any buffered non-stop text before that token should still be returned by `getDelta()`, while a bare EOS token with no buffered text should produce a null delta. `reset()` should clear buffered text so the next append starts a new detection window. `isEos()` should report true for configured EOS token ids and false otherwise, and the detector should release its internal buffers when destroyed."} {"task_id": "format-code-task-000523", "source_id": "format-code-task-000523", "domain": "code", "task_path": "tasks/format-code-task-000523", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:078bd80f630cc117d1d66a2b06318be5002a8c174062ab7b3524c5b824849ab4", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `RawSimulator` to support bucket notification rules as in-memory state for tests. A test should be able to use a `RawSimulator` instance, create and authorize an account, create a bucket, then call `set_bucket_notification_rules(api_url: str, account_auth_token: str, bucket_id: str, rules: Iterable[NotificationRule]) -> list[NotificationRuleResponse]` and `get_bucket_notification_rules(api_url: str, account_auth_token: str, bucket_id: str) -> list[NotificationRuleResponse]` repeatedly against that bucket.\n\nFor a new bucket, `get_bucket_notification_rules(...)` should return `[]`. If I set a rule such as `{'eventTypes': ['b2:ObjectCreated:*'], 'isEnabled': True, 'name': 'test-rule', 'objectNamePrefix': '', 'targetConfiguration': {'customHeaders': [], 'targetType': 'webhook', 'url': 'https://example.com/webhook'}}`, the set call should return and persist a response rule with the same data plus `isSuspended: False` and `suspensionReason: ''`, and a later get call should return the same persisted list. If the input rule includes `isSuspended` or `suspensionReason`, the simulator should ignore those input fields when setting rules. If a rule with the same name already exists, setting it again should merge the existing `targetConfiguration` with the new one so omitted nested fields are preserved and provided nested fields are updated. Calling `set_bucket_notification_rules(..., [])` should clear the bucket's rules and later get calls should return `[]`.\n\nThe raw simulator methods should enforce the same capability checks as the rest of the simulator: setting rules requires `writeBucketNotifications`, reading rules requires `readBucketNotifications`, and an unknown bucket id should fail through the simulator's normal nonexistent-bucket error. For tests that need to simulate provider-side suspension, the bucket simulator object should also provide `simulate_notification_rule_suspension(rule_name: str, reason: str, is_suspended: bool | None = None) -> None`; it should update the matching rule's `isSuspended` and `suspensionReason`, default `isSuspended` to `bool(reason)` when omitted, and raise `ResourceNotFound` when the named rule does not exist."} {"task_id": "format-code-task-000524", "source_id": "format-code-task-000524", "domain": "code", "task_path": "tasks/format-code-task-000524", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:540820b6ce7f81b8a6f0fa6e53dfed0afd6f19ed413fac661b8500e46b2322a8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nAppveyor badge shows internal error if there are no builds\nWhen trying to get an Appveyor badge for a project that doesn't have any builds yet, shields generates an internal error:\n![ss1536772059-5026](https://user-images.githubusercontent.com/2406499/45441442-07a99b80-b68d-11e8-8e34-d847a3a61ac7.png)\n\nGlancing at the service file, it queries their JSON API and extracts the build status from the `build.status` property - however for projects that have no builds the `build` property does not exist."} {"task_id": "format-code-task-000525", "source_id": "format-code-task-000525", "domain": "code", "task_path": "tasks/format-code-task-000525", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:74bc091f9a3adeca37223e1da3eccc51777a6b8676a8fd0e542f90886ca9fb19", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want every PyBoy instance to expose a stateful memory scanner through `pyboy.memory_scanner`, and I also want `MemoryScanner(pyboy)` to be importable from `pyboy.api.memory_scanner` for direct construction. The scanner should provide `scan_memory(target_value=None, start_addr=0x0000, end_addr=0xFFFF, standard_comparison_type=StandardComparisonType.EXACT, value_type=ScanMode.INT, byte_width=1, byteorder=\"little\") -> list[int]` and `rescan_memory(new_value=None, dynamic_comparison_type=DynamicComparisonType.UNCHANGED, byteorder=\"little\") -> list[int]`.\n\n`scan_memory` should read the inclusive address range, compare each integer value to the target, return matching addresses, and remember those addresses plus their values for later rescans. For example, if memory at `0xC000..0xC003` is `[0, 7, 0, 9]`, `scan_memory(0, start_addr=0xC000, end_addr=0xC003)` should return `[0xC000, 0xC002]`; if `target_value` is `None` over that same range, it should return all four addresses. When `byte_width=2`, it should combine adjacent bytes using `byteorder` and avoid starting at the final address if there is not enough room to read a full value.\n\nI need `StandardComparisonType` to support `EXACT`, `LESS_THAN`, `GREATER_THAN`, `LESS_THAN_OR_EQUAL`, and `GREATER_THAN_OR_EQUAL`, and I need `ScanMode.INT` and `ScanMode.BCD` so BCD-encoded game values can be scanned as decimal values. For example, if address `0xC000` contains `0x42`, scanning for `42` with `value_type=ScanMode.BCD` should include `0xC000`.\n\n`rescan_memory` should only consider the addresses saved by the most recent `scan_memory` or `rescan_memory`, update the remembered values for kept addresses, and return the narrowed address list. `DynamicComparisonType.UNCHANGED`, `CHANGED`, `INCREASED`, and `DECREASED` should compare the current value at each candidate address against the remembered value, while `DynamicComparisonType.MATCH` should keep only addresses whose current value equals `new_value`. Calling `rescan_memory(None, DynamicComparisonType.MATCH)` should raise `ValueError`, and passing an unsupported comparison enum to either scan path should raise `ValueError`."} {"task_id": "format-code-task-000526", "source_id": "format-code-task-000526", "domain": "code", "task_path": "tasks/format-code-task-000526", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b8e168fa93ff62a629b33979ffdb403622144acc058cd7963fa692b21103b776", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want each `PyBoy` instance to expose a stateful GameShark cheat handler at `pyboy.gameshark`, and I want the constructor to accept `PyBoy(gamerom, ..., gameshark: str | None = None, ...)` for comma-separated startup cheats. The handler should support `pyboy.gameshark.add(code: str)`, `pyboy.gameshark.remove(code: str, restore_value: bool = True)`, and `pyboy.gameshark.clear_all(restore_value: bool = True)`.\n\nGameShark codes are 8-character hexadecimal strings in `ttvvaaaa` form. For type `01`, `vv` is the byte to force and `aaaa` is a little-endian address, so adding `\"010138CD\"` should force `pyboy.memory[0xCD38]` to `0x01` on every emulator tick until that cheat is removed. A session should work like this: create `pyboy = PyBoy(rom, window=\"null\")`, observe that `pyboy.memory[0xCD38]` is not `0x01`, call `pyboy.gameshark.add(\"010138CD\")`, then after `pyboy.tick()` the memory byte at `0xCD38` is `0x01`. Calling `pyboy.gameshark.remove(\"010138CD\")` should restore the original byte by default, while calling `pyboy.gameshark.remove(\"010138CD\", False)` after another add and tick should leave the forced byte in memory.\n\nMultiple cheats should be independent: after adding `\"010138CD\"` and `\"010138CE\"`, one tick should force both decoded addresses, and `pyboy.gameshark.clear_all()` should remove every active code and restore all original bytes by default. Passing `gameshark=\"010138CD, 010138CE\"` to the `PyBoy` constructor should register both startup cheats, trimming whitespace around each comma-separated code.\n\nInvalid usage should fail clearly. `add()` should raise `ValueError` for codes that are not exactly 8 characters, for decoded addresses below `0x8000`, and for unsupported GameShark types such as `02`; adding the same code twice should raise `PyBoyInvalidInputException`. Removing a code that has not been applied should raise `ValueError`."} {"task_id": "format-code-task-000527", "source_id": "format-code-task-000527", "domain": "code", "task_path": "tasks/format-code-task-000527", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6a07b1ae0e22d36e27d714258907571247d3e45f5a735776182be12d8b2a21b8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nThe training workflow already constructs linear, gradient-boosted-tree, and random-forest contextual bandits, but the serving layer assumes every model is a PyTorch network. Make the public BanditPredictor workflow genuinely backend-agnostic so users can train, serve, persist, and reload any of the documented model families.\n\nA predictor must preprocess one request exactly as today: expand it into the experiment's ordered decision candidates, apply the configured numeric/categorical/dense-product transformations, and preserve candidate IDs in experiment order. For neural models, retain the existing behavior. For sklearn-style estimators used by linear_bandit, gbdt_bandit, and random_forest_bandit, score the resulting float feature matrix through the estimator's public prediction API: regression returns one score per candidate, while binary classification returns the positive-class probability (not a hard label and not both class columns). The returned `scores` shape and `ids` contract must match neural serving.\n\nArtifact persistence must work for both backend families. Saving a predictor must emit configuration plus model data that can be loaded in a fresh process without the original Python object. Reloading must reconstruct the appropriate estimator or neural network from the saved metadata, restore preprocessing and feature-order information, and produce the same scores and IDs as the in-memory predictor for both regression and binary requests. Existing neural save/reload compatibility must remain intact. Unsupported or malformed model artifacts should fail clearly rather than silently producing neural-only behavior."} {"task_id": "format-code-task-000528", "source_id": "format-code-task-000528", "domain": "code", "task_path": "tasks/format-code-task-000528", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d49b5e8c492856926c7c50d516c4d27999cec71c9deeaf006ef9153885653875", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Polymorphic children are not loaded correctly when hydrating a parent\n\nI'm setting up a polymorphic relation where one parent owns many children of different concrete types — pretty much the canonical example from the README (something like an `Owner` that has many `Pet` records, where `Pet` can be a `Cat` or a `Dog`).\n\nMy setup looks roughly like this:\n\n```ts\n@Entity()\nexport class Owner extends BaseEntity {\n @PrimaryGeneratedColumn()\n id: number;\n\n @PolymorphicChildren(() => Pet, { eager: true })\n pets: Pet[];\n}\n\n@Entity()\nexport class Pet extends BaseEntity implements PolymorphicChildInterface {\n @PrimaryGeneratedColumn()\n id: number;\n\n @Column()\n name: string;\n\n @PolymorphicParent(() => Owner)\n owner: Owner;\n\n @Column()\n entityId: number;\n\n @Column()\n entityType: string;\n}\n```\n\nThen in my repository (extending `AbstractPolymorphicRepository`) I do something like:\n\n```ts\nconst owner = await ownerRepo.findOne({ where: { id: 1 } });\n// or explicitly: await ownerRepo.hydrateOne(owner);\nconsole.log(owner.pets);\n```\n\nIn the DB I've manually inserted a few `pet` rows with `entityId = 1` and `entityType = 'Owner'`, so I'm expecting `owner.pets` to contain exactly those rows.\n\nWhat I get instead is wrong — the `pets` array I end up with on the owner does not match what's actually in the database for that owner. It comes back with the wrong length, and the contents aren't what I'd expect for that parent either. Plays out the same whether I call `find`, `findOne`, or `hydrateOne` directly.\n\nI'd expect the hydrated children array to be exactly the set of child rows belonging to that parent — same count, same entities, no duplicates, nothing missing.\n\nIs the parent → many children case actually working? Happy to share a minimal repro if useful."} {"task_id": "format-code-task-000529", "source_id": "format-code-task-000529", "domain": "code", "task_path": "tasks/format-code-task-000529", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:db5aef05968e7c72c636c30384e721f320b00016c7f4130a512485b60b1b2cdb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nError when trying to fix a model\nHi, Bashtage:\n\nI have a problem when I'm trying to `fix` a model. The use case I'm going after is to fit a model, store the parameters on disk and load it again to build a model using `fix`. But when I try that I get the error:\n\n```\n File \"/usr/local/lib/python3.6/site-packages/arch/univariate/base.py\", line 322, in fix\n resids = self.resids(self.starting_values())\n File \"/usr/local/lib/python3.6/site-packages/arch/univariate/base.py\", line 567, in starting_values\n params = np.asarray(self._fit_no_arch_normal_errors().params)\n File \"/usr/local/lib/python3.6/site-packages/arch/univariate/mean.py\", line 538, in _fit_no_arch_normal_errors\n nobs = self._fit_y.shape[0]\nAttributeError: 'NoneType' object has no attribute 'shape'\n```\n\nSimple code to reproduce is this:\n\n```python\nimport datetime as dt\nimport pandas_datareader.data as web\n\nimport arch\n\nst = dt.datetime(1990, 1, 1)\nen = dt.datetime(2016, 1, 1)\n\ndata = web.get_data_yahoo('^GSPC', start=st, end=en)\nreturns = 100 * data['Adj Close'].pct_change().dropna()\n\nresult = arch.arch_model(returns).fit()\nfixed_result = arch.arch_model(returns).fix(result.params)\n```\n\nIt this a supported usecase?"} {"task_id": "format-code-task-000530", "source_id": "format-code-task-000530", "domain": "code", "task_path": "tasks/format-code-task-000530", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6d094ed832b5a283276072dfa03e2aa59040f2c338b41848cd9f8f76790a3cf2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nsort-* function fails on empty seq\nHi,\n\nsort-* functions return an error on empty seq\n```TypeError: 'NoneType' object is not iterable```\n\nTo reproduce.\n1. Open up a REPL and try to sort an empty seq\n``` clojure\nbasilisp.user=> (sort (seq []))\nTraceback (most recent call last):\n File \"C:\\src\\basilisp\\src\\basilisp\\cli.py\", line 306, in repl\n result = eval_str(lsrc, ctx, ns, eof)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\src\\basilisp\\src\\basilisp\\cli.py\", line 52, in eval_str\n last = compiler.compile_and_exec_form(form, ctx, ns)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\src\\basilisp\\src\\basilisp\\lang\\compiler\\__init__.py\", line 165, in compile_and_exec_form\n return getattr(ns.module, final_wrapped_name)()\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"\", line 1, in __lisp_expr___65\n File \"C:\\src\\basilisp\\src\\basilisp\\core.lpy\", line 1160, in sort\n (defn sort\n File \"C:\\src\\basilisp\\src\\basilisp\\core.lpy\", line 1165, in sort__arity1\n (basilisp.lang.runtime/sort coll))\n File \"C:\\src\\basilisp\\src\\basilisp\\lang\\runtime.py\", line 1432, in sort\n return lseq.sequence(sorted(coll, key=key))\n ^^^^^^^^^^^^^^^^^^^^^\nTypeError: 'NoneType' object is not iterable\n```\n\nClojure returns an empty seq instead\n``` clojure\nuser> (sort (seq []))\n()\n```\n\nPR to follow."} {"task_id": "format-code-task-000531", "source_id": "format-code-task-000531", "domain": "code", "task_path": "tasks/format-code-task-000531", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b480e11927fa27bd57fad447aae9e9cb37c8a3e72b894805460e7378c3660ffb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want library users to be able to construct supported scene-text recognizer models by public model id without manually reading Hydra YAML files or wiring model constructors. The core API should be `create_model(experiment: str, pretrained: bool = False, **kwargs) -> torch.nn.Module`, where supported experiment ids include `parseq`, `parseq-tiny`, `parseq-patch16-224`, `abinet`, `trba`, `vitstr`, and `crnn`.\n\nFor `create_model('parseq', pretrained=False)`, return a PARSeq model initialized from the shared defaults plus the `parseq` experiment/model config, with runtime defaults such as `decode_ar=True` and `refine_iters=1`. For `create_model('parseq-tiny', pretrained=False)`, return a PARSeq model whose resolved config has the tiny public variant values, including model name `parseq-tiny`, `embed_dim=192`, `enc_num_heads=3`, and `dec_num_heads=6`. Passing keyword overrides must win over config values; for example, `create_model('parseq', pretrained=False, decode_ar=False, refine_iters=2)` should return a PARSeq model configured with non-autoregressive decoding and two refinement iterations.\n\nWhen `pretrained=True`, the factory should load the released weights for that same model id before returning the model. If the model id has no config, `create_model` should raise `InvalidModelError` rather than a raw file error; if the id is configured but has no supported model class, it should also raise `InvalidModelError`. Repeating the same call with `pretrained=False` should produce a fresh model with the same resolved configuration and should not mutate caller-provided inputs.\n\nI also want Torch Hub entry functions so users can call `torch.hub.load(..., 'parseq', pretrained=True)`, `parseq_tiny(pretrained=False, decode_ar=True, refine_iters=1, **kwargs)`, `parseq(pretrained=False, decode_ar=True, refine_iters=1, **kwargs)`, `parseq_patch16_224(pretrained=False, decode_ar=True, refine_iters=1, **kwargs)`, `abinet(pretrained=False, iter_size=3, **kwargs)`, `trba(pretrained=False, **kwargs)`, `vitstr(pretrained=False, **kwargs)`, and `crnn(pretrained=False, **kwargs)`. Each Torch Hub function should delegate to the same model-id factory and apply its documented default runtime overrides.\n\nFinally, `load_from_checkpoint(checkpoint_path: str, **kwargs) -> torch.nn.Module` should accept checkpoint strings of the form `pretrained=` as a shortcut for constructing the named released model with pretrained weights and any supplied overrides, such as `load_from_checkpoint('pretrained=parseq', decode_ar=False, refine_iters=2)`."} {"task_id": "format-code-task-000532", "source_id": "format-code-task-000532", "domain": "code", "task_path": "tasks/format-code-task-000532", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:dd70c87597ef7a51cf7771899ac4c6a7b723c1c7ce842b14a8f9bf5adae09630", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Confirmable wallet transactions\n\nToday a deposit or withdrawal can be recorded as *unconfirmed* by passing `false` as the\nthird argument to `deposit(...)` / `withdraw(...)` (and their `force*` variants). An unconfirmed\ntransaction is persisted, exposes `confirmed === false`, but it intentionally does **not** affect\nthe wallet balance. What's missing is a way to *later* confirm such a pending transaction.\n\nPlease add the ability to confirm a previously-recorded transaction on the wallet it belongs to.\n\nExpected behaviour:\n\n- A wallet exposes a `confirm($transaction)` method. Confirming a still-unconfirmed transaction\n applies its amount to that wallet's balance (a deposit increases it, a withdrawal decreases it),\n flips the transaction's `confirmed` flag to `true`, and returns `true`.\n\n- A transaction may only be confirmed through the wallet it was recorded against. Trying to\n confirm it through any other wallet must be rejected, and no balances may change.\n\n- A transaction that is already confirmed cannot be confirmed again. The attempt must be rejected\n and the balance left untouched (the transaction stays confirmed).\n\n- Confirming a withdrawal must respect the wallet's available funds: if applying it would make the\n balance go negative, the confirmation must be rejected and the balance left unchanged (the\n transaction stays unconfirmed).\n\n- A rejected `confirm($transaction)` raises an exception. Alongside it, provide a\n `safeConfirm($transaction)` method that performs the exact same confirmation but returns `false`\n instead of raising when the transaction cannot be applied, again without changing any balance.\n\nThe existing deposit/withdraw behaviour, including how unconfirmed transactions are created, must\nstay exactly as it is.\n"} {"task_id": "format-code-task-000533", "source_id": "format-code-task-000533", "domain": "code", "task_path": "tasks/format-code-task-000533", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:80e7447b0d402727d343320b34e94f50abb3e2e4a0c13b8815ef220bb46516c8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nbazel.Runfile fails to parse a MANIFEST file generated for rules_python\nFull minimal reproduction repro with a running github action at: https://github.com/samhowes/rules_go_runfiles_repro\nFailing github action run: https://github.com/samhowes/rules_go_runfiles_repro/runs/2370880759?check_suite_focus=true\nOutput of interest: `Error: panic: error parsing runfiles manifest: C:/users/runneradmin/_bazel_runneradmin/t32zp6hb/execroot/__main__/bazel-out/x64_windows-opt-exec-2B5CBBC6/bin/parent.exe.runfiles/MANIFEST:1: no space`\n\nSample manifest that reproduces this issue: https://github.com/samhowes/rules_go_runfiles_repro/blob/main/sample_MANIFEST\n\nTarget producing the output: `//:parent_output_test`\n\n### What version of rules_go are you using?\n1.16\n\n### What version of gazelle are you using?\n\nv0.23.0\n\n### What version of Bazel are you using?\n\nBuild label: 4.0.0\nBuild target: bazel-out/x64_windows-opt/bin/src/main/java/com/google/devtools/build/lib/bazel/BazelServer_deploy.jar\nBuild time: Thu Jan 21 07:39:16 2021 (1611214756)\nBuild timestamp: 1611214756\nBuild timestamp as int: 1611214756\n\n### Does this issue reproduce with the latest releases of all the above?\n\nThe above are the latest releases\n\n### What operating system and processor architecture are you using?\n\nwindows x64\n\n### Any other potentially useful information about your toolchain?\nits all standard\n\n### What did you do?\n\nI made a `py_binary` target that has a data dependency on a `go_binary` target then tried to execute the `go_binary` from within the `py_binary`. Inside of the `go_binary` I tried to execute ```path, err := bazel.Runfile(\"important.txt\")``` to access a runfile that the `go_binary` has as a data dependency. \n\nOn windows.\n\n### What did you expect to see?\n\nNo err, and a successful location of the runfile.\n\n### What did you see instead?\n\n```error parsing runfiles manifest: C:/users/runneradmin/_bazel_runneradmin/t32zp6hb/execroot/__main__/bazel-out/x64_windows-opt-exec-2B5CBBC6/bin/parent.exe.runfiles/MANIFEST:1: no space```"} {"task_id": "format-code-task-000534", "source_id": "format-code-task-000534", "domain": "code", "task_path": "tasks/format-code-task-000534", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e0d7608b6e66fb1e29ba15ff866405c9e110be317727427503aa46185147f4e6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nconstant-glob fails to lint when the first argument is named\nconstant-glob does not correctly lint constant globs when using a named `include` parameter.\n\nThis raises a warning:\n```\n❯ echo 'glob([\"foo.py\"])' | buildifier -lint=warn -warnings=+constant-glob -type build -\nglob([\"foo.py\"])\n:1: constant-glob: Glob pattern \"foo.py\" has no wildcard ('*'). Constant patterns can be error-prone, move the file outside the glob. (https://github.com/bazelbuild/buildtools/blob/master/WARNINGS.md#constant-glob)\n```\n\nThis fails to raise a warning, even though I'd expect it to:\n```\n❯ echo 'glob(include = [\"foo.py\"])' | buildifier -lint=warn -warnings=+constant-glob -type build -\nglob(include = [\"foo.py\"])\n```"} {"task_id": "format-code-task-000535", "source_id": "format-code-task-000535", "domain": "code", "task_path": "tasks/format-code-task-000535", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7f64205a01c9ae65874b8c2dc0c2af0e405ac7085e68e52e8edea8eb1f6c2b13", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `audiowaveform` command to generate waveform data files directly from audio inputs. When I run `audiowaveform -i test_file_stereo.wav -o out.dat -z 64 -b 8`, it should read the WAV audio, group decoded PCM frames into 64-sample windows, combine stereo channels into a single mono min/max waveform by default, write an 8-bit `.dat` waveform data file, print the normal progress messages to stderr, and exit with code 0.\n\nI also need the same audio-input path to write JSON: `audiowaveform -i test_file_stereo.mp3 -o out.json -z 64 -b 8` should produce a JSON waveform data file with the decoded MP3 represented as min/max points and exit 0. The route should apply to every supported audio input format, including WAV, MP3, FLAC, Ogg/Vorbis, Opus when the build supports it, and raw PCM when the existing raw input flags are supplied, whenever the requested output format is `.dat` or JSON.\n\nThe data-generation path should honor `--split-channels` by preserving separate channel waveforms instead of mixing down to mono. It should honor `--amplitude-scale auto` for waveform data output by scaling the generated points before saving. If the audio file cannot be opened or decoded, the command should not create a successful waveform data output and should exit non-zero."} {"task_id": "format-code-task-000537", "source_id": "format-code-task-000537", "domain": "code", "task_path": "tasks/format-code-task-000537", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2978381e7b6e75b4a01059da25ac7814f67c2794c56edcda2660367b82758cb7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want an installed command named `ntsynt_merge_collinear.py` for post-processing an existing ntSynt synteny block TSV. It should be invoked as `ntsynt_merge_collinear.py --tsv blocks.tsv [--indel N] [--merge N]`, with `--indel` defaulting to 50000 bp and `--merge` defaulting to 1000000 bp. The input is tab-separated rows with at least these columns: block ID, assembly, contig, start, end, strand; extra columns such as minimizer count and prior break reason may be present but should not control the merge.\n\nThe command should group rows by block ID, use the first assembly encountered in the file as the reference assembly, sort blocks by that reference assembly's contig and start coordinate, and then compare each adjacent block pair across every assembly. Adjacent blocks should merge only when every assembly stays on the same contig, every assembly keeps the same strand, all orientation-aware gaps are non-negative, the spread between the largest and smallest gap is at most `--indel`, and the largest gap is less than `--merge`. For `+` strand blocks the gap is `next.start - current.end` and merging extends the current end to the next end; for `-` strand blocks the gap is `current.start - next.end` and merging extends the current start to the next start.\n\nFor example, given `--indel 10 --merge 100` and an input file containing `0\tasmA\tchr1\t100\t200\t+\t7\tNone`, `0\tasmB\tchr7\t500\t600\t+\t7\tNone`, `1\tasmA\tchr1\t220\t300\t+\t4\tNone`, and `1\tasmB\tchr7\t625\t705\t+\t4\tNone`, stdout should be exactly two rows: `0\tasmA\tchr1\t100\t300\t+\t0\tNone` and `0\tasmB\tchr7\t500\t705\t+\t0\tNone`, with exit code 0. Output block IDs should be renumbered from 0, the minimizer-count column should be `0`, and the final column should be `None` for the first output block or the recomputed discontinuity reason for later blocks. When a pair does not merge, the next block's reason should be `id_change`, `ori_change`, `inconsistent_order`, `indel`, or `merge`, chosen in that priority order.\n\nIf an input row has fewer than six tab-separated columns, the command should fail with a non-zero exit and report an error mentioning the expected minimum column count. If `--tsv` is omitted, argparse-style usage should be written to stderr and the exit code should be 2. `--help` should print usage to stdout, include `--tsv`, `--indel`, and `--merge`, and exit 0."} {"task_id": "format-code-task-000538", "source_id": "format-code-task-000538", "domain": "code", "task_path": "tasks/format-code-task-000538", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a77f382aabd2c5d96ed53ebebf2a8215202a9a7d01ed860988c6cd3da08da586", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nIs a template is must in v1.6?\n\nI use c.Ctx.Output.Body() direct write reponse , it works in beego v1.5 . \n\n```\nc.Ctx.Output.Header(\"Content-Type\", \"application/xml; charset=utf-8\")\nc.Ctx.Output.Body(response)\n```\n\nBut wrong in beego v1.6\n\n```\n2016/02/09 21:50:55 [router.go:854][C] Handler crashed with error can't find templatefile in the path:maincontroller/get.tpl \n2016/02/09 21:50:55 [router.go:860][C] /usr/local/go/src/runtime/asm_amd64.s:437 \n2016/02/09 21:50:55 [router.go:860][C] /usr/local/go/src/runtime/panic.go:423 \n2016/02/09 21:50:55 [router.go:860][C] /Users/brian/dev/go/src/github.com/astaxie/beego/controller.go:262 \n2016/02/09 21:50:55 [router.go:860][C] /Users/brian/dev/go/src/github.com/astaxie/beego/controller.go:183 \n2016/02/09 21:50:55 [router.go:860][C] /Users/brian/dev/go/src/github.com/astaxie/beego/router.go:784 \n2016/02/09 21:50:55 [router.go:860][C] /usr/local/go/src/net/http/server.go:1862 \n2016/02/09 21:50:55 [router.go:860][C] /usr/local/go/src/net/http/server.go:1361 \n2016/02/09 21:50:55 [router.go:860][C] /usr/local/go/src/runtime/asm_amd64.s:1721 \n```"} {"task_id": "format-code-task-000539", "source_id": "format-code-task-000539", "domain": "code", "task_path": "tasks/format-code-task-000539", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0d8da1653242dfecf50ff869c85b83a85d3eaef78e3af8ae6b510857b283fdb2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Beego sessions to support reading the session ID from the request URL query string when I explicitly enable that behavior. The public option should be `session.CfgEnableSidInURLQuery(enable bool) session.ManagerConfigOpt`, and `session.ManagerConfig` should also expose an `EnableSidInURLQuery bool` JSON-configurable field so callers can configure it directly.\n\nFor a concrete flow, I should be able to create a manager with `cfg := session.NewManagerConfig(session.CfgCookieName(\"sid\"), session.CfgEnableSidInURLQuery(true))`, then `mgr, err := session.NewManager(\"memory\", cfg)`. If I already have a memory session with ID `abc123` containing `\"user\" = \"alice\"`, then calling `mgr.SessionStart(w, r)` for a request with no session cookie and URL `/?sid=abc123` should return that same store, so `store.SessionID(ctx)` is `\"abc123\"` and `store.Get(ctx, \"user\")` is `\"alice\"`. If there is no matching existing session for the query value, `SessionStart` should create a fresh session ID through the configured provider instead of treating the query value as valid.\n\nWhen URL-query lookup is enabled and parsing the request form fails, `SessionStart` should return that parse error and no session store. When URL-query lookup is disabled, a `?sid=...` value should be ignored and the normal cookie/header/new-session behavior should apply. Beego's web session configuration should also have `SessionEnableSidInURLQuery bool`; when automatic session registration builds the session manager from `BConfig.WebConfig.Session`, that setting should populate the manager's URL-query lookup option."} {"task_id": "format-code-task-000540", "source_id": "format-code-task-000540", "domain": "code", "task_path": "tasks/format-code-task-000540", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3ab03632ea65de70a9a92e5f6f39960236b13618214d320664244b7d833d87cf", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## NoneType binary/comparison ops don't match CPython behavior\n\nI've been running some Python snippets through Batavia and noticed a bunch of operations involving `None` behave differently from CPython 3.\n\n### Comparisons\n\nIn CPython 3, ordering comparisons against `None` raise `TypeError`:\n\n```python\n>>> None < 1\nTypeError: unorderable types: NoneType() < int()\n>>> None <= 1\nTypeError: ...\n>>> None > 1\nTypeError: ...\n>>> None >= 1\nTypeError: ...\n```\n\nIn Batavia, these just return `False` silently, so code that's supposed to blow up keeps running and produces wrong results downstream. Only `==` / `!=` should work against `None` (and they do — that part's fine).\n\n### Bitwise operators\n\n```python\n>>> None & 1\n>>> None ^ 1\n>>> None | 1\n```\n\nIn CPython these raise `TypeError` (\"unsupported operand type(s) for ...\") like the other unsupported binary ops on `None`. In Batavia I get a `NotImplementedError` saying the dunder hasn't been implemented yet, which makes it look like Batavia is incomplete rather than the user's code being wrong.\n\n### Subscripting\n\n```python\n>>> None[0]\n```\n\nCPython says `'NoneType' object is not subscriptable`. Batavia phrases this as if it were a binary operand mismatch (\"unsupported operand type(s) for []: ...\"), which doesn't match and is misleading — subscripting isn't a binary op.\n\n### `pow` / `**`\n\n```python\n>>> None ** 2\n```\n\nCPython's message mentions both `**` and `pow()` in the unsupported-operand text. Batavia's message only mentions `pow`, so anyone using the `**` operator and grepping for it in the error won't find anything.\n\nCould the `None` dunder methods be aligned with CPython here? The other binary ops on `None` (`+`, `-`, `*`, `/`, `//`, `%`, `<<`, `>>`) already raise `TypeError` with CPython-shaped messages, so this is just filling in the remaining cases consistently."} {"task_id": "format-code-task-000541", "source_id": "format-code-task-000541", "domain": "code", "task_path": "tasks/format-code-task-000541", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:36a7df43964ad33edd99da7772e4b60e9e42ae09d2393be4f860cdc43aa0068b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add lifecycle event handlers to `toga.Window`\n\nRight now a `Window` only lets you respond to one lifecycle event, `on_close`.\nApplications frequently need to react to a window changing its focus or\nvisibility — for example, pausing background work when a window is hidden, or\nrefreshing data when it comes back to the foreground.\n\nAdd four new optional event handlers to `toga.Window`:\n\n- `on_gain_focus` — the window became the application's current (focused) window.\n- `on_lose_focus` — the window stopped being the application's current window.\n- `on_show` — the window became visible.\n- `on_hide` — the window stopped being visible.\n\n## Expected behavior\n\n- Each handler must be available **both** as a keyword argument to the `Window`\n constructor **and** as a readable/writable property, in exactly the same way\n the existing `on_close` handler works. Assigning a new callable to the\n property replaces the handler.\n- When a handler has not been provided, reading the property must still return a\n callable (a no-op), consistent with how an unset `on_close` behaves.\n- Whenever a handler fires, it is invoked with the window instance it belongs\n to.\n\nThe handlers fire in response to the following observable state changes:\n\n- **Visibility.** When a window transitions from not-visible to visible, its\n `on_show` handler fires; when it transitions from visible to not-visible, its\n `on_hide` handler fires. This applies whether visibility is changed through\n `show()`/`hide()` or through the `visible` property. A request that does not\n actually change the visibility (e.g. showing an already-visible window) must\n not fire either handler.\n- **Minimize / restore.** Moving a visible window into the minimized state fires\n its `on_hide` handler; restoring it from the minimized state back to a visible\n state fires its `on_show` handler. Switching between other window states does\n not fire these handlers, and a state request that leaves the state unchanged\n fires nothing.\n- **Focus.** When the application's current window changes from one window to a\n different window, the window that was current has its `on_lose_focus` handler\n fired, and the window that becomes current has its `on_gain_focus` handler\n fired. Setting the current window to the window that is already current fires\n neither handler.\n\nAdding these handlers should not disturb the existing `on_close` behavior.\n"} {"task_id": "format-code-task-000542", "source_id": "format-code-task-000542", "domain": "code", "task_path": "tasks/format-code-task-000542", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1990390aefe634ba1db5a7ce5520a9f7d0772a56575d28dcba6d28762be8ea9d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Make `tuple` support the standard operators\n\nRight now Python tuples that get transpiled to the JVM can be created, indexed,\nsliced and iterated, but almost every operator on them is dead — using one\ncurrently blows up with a \"not implemented\" error instead of behaving like\nCPython. Please make tuples support the usual operator protocol so that\ntranspiled code behaves exactly the way it does under CPython 3.5.\n\nConcretely, a tuple should support:\n\n* **Rich comparisons** (`<`, `<=`, `>`, `>=`, `==`, `!=`) against another\n tuple, using Python's lexicographic ordering: compare element by element, and\n at the first position where the two tuples differ the result is decided by\n comparing those two elements; if one tuple is a prefix of the other, the\n shorter one is the smaller. Equality is element-wise and length-sensitive.\n * `==` and `!=` must also work when the other operand is **not** a tuple:\n a tuple is never equal to a non-tuple, so `==` is `False` and `!=` is\n `True` (no error).\n * The ordering comparisons (`<`, `<=`, `>`, `>=`) against a non-tuple operand\n must raise `TypeError`, matching CPython's message for unorderable types.\n\n* **Concatenation** with `+`: `tuple + tuple` produces a new tuple with the\n elements of the left followed by the elements of the right. Adding anything\n that is not a tuple raises `TypeError` with CPython's message.\n\n* **Repetition** with `*`: `tuple * n` for an integer `n` produces a new tuple\n with the elements repeated `n` times (`n` of zero or less yields an empty\n tuple); a boolean counts as `0`/`1`. Multiplying by something that is not an\n integer raises `TypeError` with CPython's message.\n\n* **Truthiness**: an empty tuple is falsy and a non-empty tuple is truthy, so it\n behaves correctly in boolean contexts (`if`, `while`, `not ...`).\n\n* **Unary operators**: `+`, `-` and `~` are not valid on tuples and must raise\n `TypeError` with CPython's message for a bad unary operand.\n\nThe observable contract is simply that each of these operations produces output\n(including raised exceptions and their messages) identical to CPython 3.5 for\nthe same source. Existing tuple behaviour (creation, `repr`, indexing, slicing,\niteration) must keep working.\n"} {"task_id": "format-code-task-000543", "source_id": "format-code-task-000543", "domain": "code", "task_path": "tasks/format-code-task-000543", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:49ab7a0805c81858a633aaf6022a63b74deaa0a9e2f7748ad9a4c25a0c6be37c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Floor division and modulo for floats\n\nVOC transpiles Python to Java by providing a runtime implementation of Python's\nbuilt-in types. Right now the `float` type is missing two arithmetic operators:\nfloor division (`//`) and modulo (`%`). Any program that uses them on a float\ncurrently blows up instead of computing a result.\n\nPlease implement both operators on `float` so they behave exactly like CPython.\n\nThe left operand is a `float`; the right operand may be an `int`, a `float`, or a\n`bool`. Concretely:\n\n- `a // b` returns a `float` equal to the floor of the true division `a / b`\n (e.g. `9.9 // 2.0` is `4.0`, `-9.9 // 2.0` is `-5.0`).\n- `a % b` returns a `float`. The result follows Python's convention that the\n modulo takes the **sign of the divisor**: `5.5 % -2` is `-0.5`, `-5.5 % 2` is\n `0.5`. When the remainder is exactly zero, the result is a signed zero whose\n sign matches the divisor, so `-7.0 % 7.0` is `0.0` while `7.0 % -7.0` and\n `0.0 % -3` are `-0.0`.\n- A `bool` right operand acts like the integer `1` (`True`) or `0` (`False`).\n\nEdge cases that must match CPython:\n\n- Dividing or taking the modulo by a zero divisor (whether `0`, `0.0`, or\n `False`) raises `ZeroDivisionError`. Floor division by zero reports\n `float divmod()`; modulo by zero reports `float modulo`.\n- Applying either operator with an unsupported right-hand type raises\n `TypeError` with the message\n `unsupported operand type(s) for //: 'float' and ''` (and the\n analogous message with `%` for modulo), where `` is the offending\n operand's type name.\n\nBoth the global scope and inside-a-function scope must produce identical results,\nmatching what CPython prints for the same source.\n"} {"task_id": "format-code-task-000544", "source_id": "format-code-task-000544", "domain": "code", "task_path": "tasks/format-code-task-000544", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:94646fd528d24761f772104842b90e19e6ac5ad0013120de20520817c1254367", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Make the Java output normalizer cope with all our stack-trace shapes\n\nOur test harness compares the output of transpiled programs run on the JVM against\nexpected output. Because raw JVM output is noisy (full Java stack traces, native\nmemory addresses, `E`-style float exponents), the harness first runs everything\nthrough `cleanse_java(output)` in the test utilities, which collapses a Java\nexception into a stable, canonical block before comparison.\n\n`cleanse_java` currently only understands a couple of trace layouts, and it\nmishandles others — for some traces the exception name and message come out\nempty, and newer runner setups emit traces it doesn't recognise at all. Make it\nrobust to every shape of exception output we actually see.\n\n## Canonical form\n\nWhen the input contains an `org.python.exceptions.*` exception, the whole\nexception region must be replaced with:\n\n```\n### EXCEPTION ###\n: \n :\n ...\n```\n\n- `` is the exception class with the leading `org.python.exceptions.`\n package stripped (e.g. `org.python.exceptions.KeyError` → `KeyError`).\n- `` is the text following the exception class on that line, verbatim.\n- One indented ` :` line follows per stack frame, but only for\n frames in the transpiled program's own classes (those whose class path begins\n with `python.`), and frames for synthesised Java constructors (the\n `.` frames) are dropped. The frames are listed outermost-first — i.e.\n the reverse of the order they appear in the raw trace.\n- Any program output printed before the exception is preserved unchanged, and\n unrelated normalizations the helper already performs (masking memory addresses\n like `0x1eb19f4e` to `0xXXXXXXXX`, and lowercasing float exponents such as\n `7.95E-6` → `7.95e-6`) must keep working.\n\n## Trace shapes that must be handled\n\n1. **Direct exceptions**, where the first line is the exception itself:\n `org.python.exceptions.SomeError: message`, followed by `at ...` frames.\n\n2. **Wrapped exceptions**, where a Java wrapper (e.g.\n `java.lang.ExceptionInInitializerError`) is reported first, optionally\n followed by some reflection `at ...` frames, then a\n `Caused by: org.python.exceptions.SomeError: message` line and its `at ...`\n frames. The trailing `... N more` continuation line that the JVM appends to a\n \"caused by\" trace must not leak into the output.\n\nIn both shapes the leading `Exception in thread \"\"` label is optional\n(some environments omit it), and the thread name may contain hyphens. The\nexception name, message, and the selected file/line frames must be reported\ncorrectly regardless of which shape and whether the label is present.\n"} {"task_id": "format-code-task-000545", "source_id": "format-code-task-000545", "domain": "code", "task_path": "tasks/format-code-task-000545", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a3de4a9f8f8825c93a836fbf547716f5ac0d4b0c25e48f6e1fffad7ad9d2db40", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\npyproject.toml (PEP 518) support\nHi! Is there a plan to support `pyproject.toml` from [PEP 518](https://www.python.org/dev/peps/pep-0518/)?\n\nSpecifically, it would be nice if `behave` could understand `[tool.behave]` section in the `pyproject.toml` file like other python libraries do (e.g. `black`, `coverage`, `isort`, `mypy`, `pytest`, ...)"} {"task_id": "format-code-task-000546", "source_id": "format-code-task-000546", "domain": "code", "task_path": "tasks/format-code-task-000546", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:053e8801ba423ba3fd17e186350a56b50fa4729c0871ba558b89b4ab32368c92", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a structured v2 connector API for opening Hive connections through Go's `database/sql` interfaces without building a DSN string. The API should expose a `Config` struct with `Host string`, `Port int`, `Auth string`, `Username string`, `Password string`, `Database string`, `TransportMode string`, `HTTPPath string`, `Service string`, `TLSConfig *tls.Config`, `SSLCertFile string`, `SSLKeyFile string`, `SSLCAFile string`, `SSLInsecureSkip bool`, and `HiveConfiguration map[string]string`.\n\nI need `NewConnector(cfg Config) *HiveConnector` to retain the provided config for later use, `OpenDB(cfg Config) *sql.DB` to return `sql.OpenDB(NewConnector(cfg))`, and `(*HiveConnector).Driver() driver.Driver` to return the package's `*Driver`. Calling `(*HiveConnector).Connect(ctx context.Context) (driver.Conn, error)` should translate the `Config` into the internal Hive connection settings: pass through username, password, database, Hive configuration, and auth/host/port, keep the default transport mode and HTTP path unless `Config.TransportMode` or `Config.HTTPPath` are non-empty, and default the service name to `\"hive\"` when `Config.Service` is empty.\n\nFor TLS, `Connect` should use `Config.TLSConfig` directly when it is provided. If no direct TLS config is provided and both `SSLCertFile` and `SSLKeyFile` are set, it should load that client certificate/key pair, apply `SSLInsecureSkip`, and return an error prefixed with `failed to configure SSL:` when the files cannot be loaded. If only `SSLCAFile` is provided, it should read that CA file into a root CA pool, apply `SSLInsecureSkip`, return `invalid ca path: ` when the file cannot be read, and return `invalid certification ` when the PEM cannot be appended. If no TLS config or files are provided but `SSLInsecureSkip` is true, it should create a TLS config with `InsecureSkipVerify` enabled.\n\nWhen the underlying Hive connection opens successfully, `Connect` should return a `driver.Conn` backed by that Hive session so callers can use `database/sql` query and exec methods. When the underlying connection fails, `Connect` should return the error wrapped with the prefix `gohive connect:`."} {"task_id": "format-code-task-000547", "source_id": "format-code-task-000547", "domain": "code", "task_path": "tasks/format-code-task-000547", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:df93c9f173b862ffda1d1d785ec550ce63048b96640963bfc2e5bae18a8e0564", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the GoHive root package `github.com/beltran/gohive` and the v2 package `github.com/beltran/gohive/v2` to expose a direct Hive metastore connection API with the same shape in both packages. The API should include `NewMetastoreConnectConfiguration() *MetastoreConnectConfiguration`, where the returned configuration defaults to `TransportMode: \"binary\"`, `Username: \"\"`, and `Password: \"\"`. It should include `ConnectToMetastore(host string, port int, auth string, configuration *MetastoreConnectConfiguration) (*HiveMetastoreClient, error)` and a `(*HiveMetastoreClient).Close()` method.\n\nA typical session should work like this: create a configuration, optionally set `configuration.TransportMode = \"binary\"` and username/password fields, call `ConnectToMetastore(\"hm.example.com\", 9083, \"KERBEROS\", configuration)`, receive a non-nil `*HiveMetastoreClient` with a non-nil public `Client` field, use that `Client` to call Hive metastore thrift operations such as listing databases, then call `Close()` to close the underlying transport. The returned `HiveMetastoreClient` should expose `Client *hive_metastore.ThriftHiveMetastoreClient` so callers can use the generated metastore API directly after the connection is established.\n\nFor `TransportMode == \"binary\"`, `ConnectToMetastore` should support `auth == \"KERBEROS\"` by opening a GSSAPI SASL transport for the Hive service, `auth == \"NONE\"` by opening a PLAIN SASL transport, and `auth == \"NOSASL\"` by opening a buffered Thrift transport. With `auth == \"NONE\"`, an empty password should be replaced with `\"x\"`; an empty username should be filled from the current OS username with spaces removed. If the host address cannot be resolved, return an error that includes `error resolving address `. If the socket open or SASL setup fails, return that error. Unsupported auth values should panic with `Unrecognized auth`, and unsupported transport modes should panic with `Unrecognized transport mode `."} {"task_id": "format-code-task-000549", "source_id": "format-code-task-000549", "domain": "code", "task_path": "tasks/format-code-task-000549", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:644b214be893106a3a0f2692e32ff8d93a44911f1524e46ba2170e7a5ff9b641", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nRegular expression does not match multi-line string\nRegular expression does not match multi-line string.\n\n```console\n$ goawk 'BEGIN{VAR=\"a\\nb\"; print match(VAR, /^a.*b$/)}'\n0\n$ goawk 'BEGIN{VAR=\"a\\nb\"; print match(VAR, /(?s)^a.*b$/)}'\n1\n\n$ gawk 'BEGIN{VAR=\"a\\nb\"; print match(VAR, /^a.*b$/)}'\n1\n$ mawk 'BEGIN{VAR=\"a\\nb\"; print match(VAR, /^a.*b$/)}'\n1\n```"} {"task_id": "format-code-task-000550", "source_id": "format-code-task-000550", "domain": "code", "task_path": "tasks/format-code-task-000550", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b7f0e4290b2b231549fa2ea19154aec4f140b0af1b2e60836aac3120b214e3fa", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `aws-sso-util` to provide a top-level `login` subcommand for signing in to an AWS IAM Identity Center instance without choosing an account or role. Users should be able to run `aws-sso-util login [SSO_START_URL] [SSO_REGION]`, or `python -m aws_sso_util.login [SSO_START_URL] [SSO_REGION]`, and the command should also accept `--profile PROFILE_NAME`, `--all`, `--force-refresh`, `--headless`, and `--verbose`/`-v`. The hidden compatibility flags `--sso-start-url`, `--sso-region`, and `--force` should behave like the positional start URL, positional region, and `--force-refresh` respectively.\n\nWhen `--all` is not set, the command should resolve exactly one Identity Center instance from `--profile`, the optional start URL and region inputs, and the login-specific environment defaults `AWS_SSO_LOGIN_DEFAULT_SSO_START_URL` and `AWS_SSO_LOGIN_DEFAULT_SSO_REGION`. If no matching instance exists, it should print `No Identity Center config found` when there are no configured instances, or `No Identity Center config matched ...` when configured instances exist but none match, then exit with status 1. If multiple instances match, it should print `Found N Identity Center configs, please specify one or use --all: ...` and exit with status 1.\n\nWhen `--all` is set, or when `AWS_SSO_LOGIN_ALL` is `true` or `1`, the command should log in to every discovered Identity Center instance using the same login-specific environment defaults. For each selected instance it should print `Logging in `, request or refresh the Identity Center token for that instance, and on success print `Login succeeded, valid until `, formatting the expiration in local time when possible and otherwise as UTC. With more than one selected instance it should first print `Logging in N Identity Center instances`.\n\n`--force-refresh` should force re-authentication instead of accepting an existing cached token, and `--headless` should disable automatic browser opening while still allowing the user-code login flow. A pending authorization timeout should print `Login window expired` and exit with status 2. An `InvalidGrantException` should print `Login failed; the login window may have expired: ` with the service message in parentheses when present and exit with status 3. Other AWS client errors and unexpected exceptions should print `Login failed: ...` and exit with status 4. `aws-sso-util login --help` should list the command, arguments, visible options, and the login-oriented help text with exit status 0."} {"task_id": "format-code-task-000551", "source_id": "format-code-task-000551", "domain": "code", "task_path": "tasks/format-code-task-000551", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:273b6d9b17e2ea8afdfef5afc4923db04bcf818726444953fc8eeabeabd737d1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a stateful `StrategyMonitor` API for keeping persisted trading strategies alive in the messaging runtime. The public API should be `new StrategyMonitor(dispatcher: PlatformDispatcher)`, `start(): Promise`, `stop(): Promise`, `subscribeToStrategy(row: StrategyAttributes): Promise`, `unsubscribeFromStrategy(strategyId: number): Promise`, and `restartForAccount(accountId: number): Promise`.\n\nWhen `start()` is called, it should load every persisted strategy row in id order, subscribe each one, and notify the owning account with `Strategy \"\" started.` after a successful subscription while continuing past rows that fail. `subscribeToStrategy(row)` should be idempotent for an already active row; otherwise it should load the row's account, parse the `BASE,COUNTER` pair and JSON config, construct the named strategy, load JSON state into the strategy when `row.state` exists, build the broker client for that account, start a live trading session, and remember the session under the row id. If the account no longer exists, it should skip starting that row.\n\nThe monitor should persist strategy mutations automatically: when the strategy requests a save, coalesce duplicate save requests in the same microtask and write the current strategy `state` and `config` back to the strategy row as JSON. When `stop()` is called, every active session should stop with `cancelOpenOrders: false` and the monitor should clear its active-session state. `unsubscribeFromStrategy(strategyId)` should stop only the active session for that id with `cancelOpenOrders: true`, remove it from the monitor, and do nothing when the id is not active.\n\nThe monitor should route live session events into user-facing notifications. A fill should send exactly `Order Filled!` followed by blank line and `Strategy: `, `Pair: `, `Side: `, `Price: `, `Size: `, `Fee: `, and `Order ID: ` lines. A strategy message should be sent as ` (): `, and the exported `formatStrategyMessage(strategyName: string, pair: string, text: string): string` helper should produce that same string while preserving the message text verbatim.\n\nWhen a strategy signals finish, the monitor should stop that session with `cancelOpenOrders: true`, remove it from active sessions, delete the persisted strategy row, and notify the account with `Strategy finished and removed.` plus the strategy id, strategy name, and pair. When a session emits an unrecoverable error, the monitor should tear that strategy down only once even if the error repeats, stop with `cancelOpenOrders: true`, keep the persisted row so the user can fix it, and notify the account with `Strategy stopped due to an error.` plus id, strategy name, pair, the error message, and `Fix the cause and restart it when ready.`\n\nWhen `restartForAccount(accountId)` is called, it should find strategy rows for that account, skip rows that are not currently active, persist each active strategy's current state/config, stop the session with `cancelOpenOrders: false`, remove it from active sessions, reload the fresh row by id, and subscribe it again if the row still exists."} {"task_id": "format-code-task-000552", "source_id": "format-code-task-000552", "domain": "code", "task_path": "tasks/format-code-task-000552", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2cab3c04b27282f16546f6975771299770b3a9685c7056cb6969f93e1c42ce4b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the existing `NSE` client object to expose corporate filings and exchange communications query methods that use the client session and return the NSE JSON response directly. The public instance methods should be `actions(segment: Literal[\"equities\", \"sme\", \"debt\", \"mf\"] = \"equities\", symbol: Optional[str] = None, from_date: Optional[datetime] = None, to_date: Optional[datetime] = None) -> List[Dict]`, `announcements(index: Literal[\"equities\", \"sme\", \"debt\", \"mf\", \"invitsreits\"] = \"equities\", symbol: Optional[str] = None, fno=False, from_date: Optional[datetime] = None, to_date: Optional[datetime] = None) -> List[Dict]`, and `boardMeetings(index: Literal[\"equities\", \"sme\"] = \"equities\", symbol: Optional[str] = None, fno: bool = False, from_date: Optional[datetime] = None, to_date: Optional[datetime] = None) -> List[Dict]`.\n\nFor `actions()`, call `/corporates-corporateActions` with `index` set from `segment`; include `symbol` only when supplied, and include `from_date` and `to_date` formatted as `dd-mm-YYYY` only when both dates are supplied. For `announcements()`, call `/corporate-announcements` with `index`, optional `symbol`, optional `fo_sec=True` when `fno` is true, and the same paired date-window behavior. For `boardMeetings()`, call `/corporate-board-meetings` with the same `index`, optional `symbol`, optional `fo_sec=True`, and paired date-window behavior. In all three of those methods, if both dates are supplied and `from_date > to_date`, raise `ValueError` with the exact message \"'from_date' cannot be greater than 'to_date'\".\n\nI also need `annual_reports(symbol: str, segment: Literal[\"equities\", \"sme\"] = \"equities\") -> Dict[str, List[Dict[str, str]]]` to call `/annual-reports` with `index=segment` and `symbol=symbol`, and `shareholding(symbol: str, index: Literal[\"equities\", \"sme\"] = \"equities\") -> List[dict]` to call `/corporate-share-holdings-master` with `index=index` and the uppercased symbol. The same feature should include `listPastIPO(from_date: Optional[datetime] = None, to_date: Optional[datetime] = None) -> List[Dict]`: when no dates are supplied it should use today as `to_date` and 90 days before that as `from_date`, send both as `dd-mm-YYYY` to `/public-past-issues`, and return the response JSON. If `listPastIPO` receives a `to_date` earlier than `from_date`, it should raise `ValueError` with the exact message \"Argument `to_date` cannot be less than `from_date`\".\n\nFinally, add `circulars(subject: Optional[str] = None, dept_code: Optional[str] = None, from_date: Optional[datetime] = None, to_date: Optional[datetime] = None) -> dict` for exchange circular searches. With no dates it should default `to_date` to today and `from_date` to seven days before that, format both as `dd-mm-YYYY`, call `/circulars`, include `sub=subject` only when a subject is given, and include `dept=dept_code.upper()` only when a department code is given. If `circulars` receives a `to_date` earlier than `from_date`, it should raise `ValueError` with the exact message \"Argument `to_date` cannot be less than `from_date`\". For every method above, if the NSE response body parses to a list like `[{\"symbol\": \"HDFCBANK\"}]` or a dict like `{\"data\": []}`, the method should return that parsed object unchanged rather than wrapping or transforming it."} {"task_id": "format-code-task-000553", "source_id": "format-code-task-000553", "domain": "code", "task_path": "tasks/format-code-task-000553", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:40985110a35556d52b852a059553895c544b6d4c3148f08f4c8d4429c63112bd", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the root Go package to expose a one-shot HTML-to-PDF conversion API. The simple call should be `HtmlToPdf(target io.Writer, htmlContent ContentInput, fontConfig text.FontConfiguration) error`, where `ContentInput` accepts the package's public `InputString`, `InputReader`, `InputFilename`, and `InputUrl` forms. For example, calling `HtmlToPdf(&buf, InputString(\"

Hello

\"), fontConfig)` should return `nil`, write non-empty bytes to `buf`, and those bytes should be a readable PDF document. Calling the same function with `InputReader(strings.NewReader(\"

Reader input

\"))` should likewise return `nil` and write a valid PDF to the supplied writer.\n\nI also need the configurable form `HtmlToPdfOptions(target io.Writer, htmlContent ContentInput, baseUrl string, urlFetcher utils.UrlFetcher, mediaType string, stylesheets []tree.CSS, presentationalHints bool, fontConfig text.FontConfiguration, zoom float64, attachments []backend.Attachment) error`. It should parse the HTML input with the provided base URL, resource fetcher, and media type, render it with the supplied stylesheets and presentational-hints setting, apply the requested zoom, include any supplied attachments, then serialize the resulting PDF to `target`. Passing empty strings, `nil` fetcher/stylesheets/attachments, `false` presentational hints, and zoom `1` should behave like `HtmlToPdf`. If parsing the HTML input or resolving its resources fails, the function should return that error and not pretend the conversion succeeded."} {"task_id": "format-code-task-000554", "source_id": "format-code-task-000554", "domain": "code", "task_path": "tasks/format-code-task-000554", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fb0c831acc70c766cde410085830cfee5271ee80f07001ca1f0fc7ec148e8c86", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI need the Cosmos-backed EVM state plugin to support Ethereum contract storage on the stateful Plugin returned by NewPlugin after Reset(ctx). The Plugin interface should expose GetState(common.Address, common.Hash) common.Hash, SetState(common.Address, common.Hash, common.Hash), GetCommittedState(common.Address, common.Hash) common.Hash, SetStorage(common.Address, map[common.Hash]common.Hash), ForEachStorage(common.Address, func(common.Hash, common.Hash) bool) error, and IterateState(func(common.Address, common.Hash, common.Hash) bool).\n\nA fresh plugin should return common.Hash{} for GetState(addr, slot) when no value exists. If I call SetState(addr, common.Hash{3}, common.Hash{1}), GetState(addr, common.Hash{3}) should return common.Hash{1}; if I then call SetState(addr, common.Hash{3}, common.Hash{2}), GetState should return common.Hash{2}. Calling SetState with common.Hash{} as the value should delete that slot so GetState returns common.Hash{} again.\n\nCommitted storage should be separate from dirty storage. Before Finalize(), GetCommittedState(addr, slot) should still return common.Hash{} for a slot changed only by SetState; after Finalize(), GetCommittedState(addr, slot) should return the finalized value. If I change the same slot again after Finalize(), GetState should show the new dirty value while GetCommittedState should continue to show the finalized value until the next finalize.\n\nSetStorage(addr, storage) should apply each slot/value pair in the map as if SetState was called for each entry. ForEachStorage(addr, cb) should visit the non-empty storage slots for that address, pass each slot key and value to cb, stop early when cb returns false, and return nil when iteration succeeds. IterateState(cb) should iterate committed storage across accounts and pass the account address, slot key, and value to cb, also stopping when cb returns true.\n\nI also need the exported storage key helpers for the durable KV layout. StorageKeyFor(address) should return 1 + common.AddressLength bytes containing types.StorageKeyPrefix followed by the address bytes, and AddressFromStorageKey should recover that address. SlotKeyFor(address, slot) should return 1 + common.AddressLength + common.HashLength bytes containing types.StorageKeyPrefix, then the address bytes, then the slot bytes; AddressFromSlotKey and SlotFromSlotKey should recover those two values."} {"task_id": "format-code-task-000555", "source_id": "format-code-task-000555", "domain": "code", "task_path": "tasks/format-code-task-000555", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3e644731d44c223a69d4d8edbfe5b3d346058b308428b3b8d0d030514da46878", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the Cosmos EVM historical plugin returned by historical.NewPlugin(...) to provide finalized Ethereum block, receipt, and transaction lookup storage once it has been prepared with a Cosmos SDK context. The public methods I need are StoreBlock(block *Block) error, StoreReceipts(blockHash common.Hash, receipts Receipts) error, StoreTransactions(blockNum uint64, blockHash common.Hash, txs Transactions) error, GetBlockByNumber(number uint64) (*Block, error), GetBlockByHash(blockHash common.Hash) (*Block, error), GetTransactionByHash(txHash common.Hash) (*TxLookupEntry, error), and GetReceiptsByHash(blockHash common.Hash) (Receipts, error). Here Block, Receipts, and Transactions are from github.com/ethereum/go-ethereum/core/types, TxLookupEntry is from github.com/berachain/polaris/eth/core/types, and common.Hash is from github.com/ethereum/go-ethereum/common.\n\nIn a concrete session, if I prepare the plugin at height 1, create a block numbered 1 containing one transaction, call StoreBlock(block), StoreReceipts(block.Hash(), receipts), and StoreTransactions(1, block.Hash(), block.Transactions()), then GetBlockByNumber(1) should return a block with the same hash and GetBlockByHash(block.Hash()) should return that same block. GetTransactionByHash(tx.Hash()) should return a lookup entry whose Tx is the transaction, TxIndex is 0, BlockNum is 1, and BlockHash is the containing block hash. GetReceiptsByHash(block.Hash()) should return receipts for that block with derived Ethereum receipt fields, including the transaction hash and block hash matching the stored block.\n\nGenesis should also be readable through the same block lookup API: after genesis initialization, GetBlockByNumber(0) should return the genesis block, and GetBlockByHash(genesisBlock.Hash()) should return a block with that hash. If GetBlockByHash is called with an unknown block hash, it should return core.ErrBlockNotFound. If GetTransactionByHash is called with an unknown transaction hash, it should return core.ErrTxNotFound. If GetReceiptsByHash is called for a block hash with no stored receipts, it should return a non-nil error; if receipts exist but the corresponding block cannot be found, it should return that block lookup error.\n\nStoreBlock should persist blocks by block number and by block hash, and it should update the historical store version to the stored block number. For non-genesis blocks, StoreBlock should reject an out-of-sequence write by panicking when the current historical store version is not exactly blockNum - 1. StoreReceipts should return any receipt marshaling error instead of silently succeeding, and StoreTransactions should return any transaction lookup entry encoding error instead of silently dropping a transaction."} {"task_id": "format-code-task-000556", "source_id": "format-code-task-000556", "domain": "code", "task_path": "tasks/format-code-task-000556", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:640c5397bdd25befddd9b94b410c0133d930f42281b38896bd522d2a28aca76b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the Go library to expose WAC access-point beacon metadata as a `waclib.WACBeacon` type with fields `MACAddress [6]byte`, `Flags []byte`, `Name string`, `Supplier string`, and `Type string`. A caller should be able to create `waclib.WACBeacon{MACAddress: [6]byte{0x00, 0x11, 0x22, 0x33, 0x44, 0x55}, Flags: []byte{0x70, 0x03}, Name: \"Name\", Supplier: \"waclib\", Type: \"Type\"}` and call `Marshal() []byte` to get the exact vendor information element bytes `dd2400a0400007060011223344550002700301044e616d6502067761636c6962030454797065`. With only `MACAddress: [6]byte{0xaa, 0xbb, 0xcc, 0xdd, 0xee, 0xff}` and `Flags: []byte{0xf0, 0x03}`, `Marshal()` should return `dd1000a040000706aabbccddeeff0002f003`, omitting the empty text fields. Calling `Marshal()` repeatedly on the same value should produce the same bytes and should not modify the struct fields.\n\nI also need `func (ie *WACBeacon) Unmarshal(data []byte) error` to parse these same beacon information element bytes into the receiver. For `dd2400a0400007060011223344550002700301044e616d6502067761636c6962030454797065`, it should set the MAC address to `00:11:22:33:44:55`, flags to `7003`, name to `Name`, supplier to `waclib`, and type to `Type`. For `dd1000a040000706aabbccddeeff0002f003`, it should set the MAC address and flags while leaving `Name`, `Supplier`, and `Type` empty. If the input does not contain the Apple WAC vendor record, `Unmarshal` should return an error whose message is `record not found`; if the MAC address field is present with the wrong length, it should return `MAC address has wrong length`. This API should do all encoding and decoding in memory without filesystem or network side effects."} {"task_id": "format-code-task-000557", "source_id": "format-code-task-000557", "domain": "code", "task_path": "tasks/format-code-task-000557", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2d945af1d2c10ca52f5475f74fa6e7f66d320f89d6fc91d6805bcf119893b314", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nHi, I have tested the newest version of Foolbox and it seems like it can't handle parallel batch attacks with estimated gradients models, any idea on when it will be released?\n\n(In case it helps: I'd also expect a public per-sample entry point like `gradient_one(...)` on the estimated-gradient wrapper alongside the batched `gradient(...)`.)"} {"task_id": "format-code-task-000558", "source_id": "format-code-task-000558", "domain": "code", "task_path": "tasks/format-code-task-000558", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2ac45dd57bd79029cbb06471b4fcecaddaee3fceb0d81a22d47b8fb18890cb71", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nComparison with empty string throws Invalid query: NO_COLUMN: null\n**Describe the bug**\nComparison with empty string throws Invalid query: NO_COLUMN: null (at least for Google Sheets sources)\n\n**To Reproduce**\nTake the query from the README:\n\n```\nSELECT country, SUM(cnt)\nFROM \"https://docs.google.com/spreadsheets/d/1_rN3lm0R_bU3NemO0s9pbFkY5LQPcuy1pscv8ZXPtg8/edit#gid=0\"\nWHERE cnt > 0\nGROUP BY country\n```\n\nAdd a comparison with empty string:\n\n```\nSELECT country, SUM(cnt)\nFROM \"https://docs.google.com/spreadsheets/d/1_rN3lm0R_bU3NemO0s9pbFkY5LQPcuy1pscv8ZXPtg8/edit#gid=0\"\nWHERE cnt > 0 AND country != ''\nGROUP BY country\n```\n\nRaises:\n\n```\nProgrammingError: (shillelagh.exceptions.ProgrammingError) Invalid query: NO_COLUMN: null\n```"} {"task_id": "format-code-task-000559", "source_id": "format-code-task-000559", "domain": "code", "task_path": "tasks/format-code-task-000559", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0015ca9c9da53a8c09981d2b006d2dbf25292be02169b8514e7fbd20c8c4179a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nIncorrect schema generation in the beta release\n> well it doesn't really work since it generate 2 primary keys :/\n> ```ts\n> export const account = pgTable(\"account\", {\n> id: serial(\"id\").primaryKey(),\n> accountId: text(\"account_id\").notNull(),\n> providerId: text(\"provider_id\").notNull(),\n> userId: serial(\"user_id\")\n> .primaryKey() // <-- this is wrong!\n> .notNull()\n> .references(() => user.id, { onDelete: \"cascade\" }),\n> accessToken: text(\"access_token\"),\n> refreshToken: text(\"refresh_token\"),\n> idToken: text(\"id_token\"),\n> accessTokenExpiresAt: timestamp(\"access_token_expires_at\"),\n> refreshTokenExpiresAt: timestamp(\"refresh_token_expires_at\"),\n> scope: text(\"scope\"),\n> password: text(\"password\"),\n> createdAt: timestamp(\"created_at\").notNull(),\n> updatedAt: timestamp(\"updated_at\").notNull(),\n> });\n> ``` \n\n _Originally posted by @body20002 in [#3308](https://github.com/better-auth/better-auth/issues/3308#issuecomment-3049875914)_"} {"task_id": "format-code-task-000560", "source_id": "format-code-task-000560", "domain": "code", "task_path": "tasks/format-code-task-000560", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a6a0a56c7b75541c9587d7ae84e5cea9885904ea80e617b9703e1131384892e7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `charging_start_time` shows wrong time for scheduled charging\n\nI'm using `bimmer_connected` from my Home Assistant setup to surface the planned charging start time of my BMW on a dashboard. The car has a charging window configured (e.g. starts at 23:00 local time) and is plugged in but not yet charging — so the state is `WAITING_FOR_CHARGING`.\n\nWhat I see on `fuel_and_battery.charging_start_time` does not match what the official MyBMW app shows:\n\n- In the app the planned start is, say, **today 23:00** (my local time).\n- In `bimmer_connected` I get a datetime that is several hours off from what the app shows. I'm not in UTC, and the offset matches my timezone difference, so the hour/minute coming from the API is being interpreted as if it were UTC.\n- Even worse, if I poll after the window's start hour has already passed (e.g. I check at 00:30, just after midnight), I still get a `charging_start_time` that lies in the past for today, instead of rolling over to tomorrow's window. A \"planned\" start time pointing into the past is obviously not useful — Home Assistant happily displays \"-2 hours\" relative to now.\n\nMinimal repro of how I use it:\n\n```python\n# in WAITING_FOR_CHARGING state\nprint(vehicle.fuel_and_battery.charging_start_time)\n# e.g. window is 23:00 local, current local time is 00:30 the next day\n# -> got a datetime for 'yesterday 23:00 UTC' (or similar)\n# -> expected: tomorrow's window start, in local time, matching the BMW app\n```\n\nExpected:\n- `charging_start_time` should match what the BMW app shows the user (i.e. the local-time hour/minute coming from the charging window).\n- It should always refer to the **next** occurrence of that hour/minute — if today's window is already in the past, it should be tomorrow.\n\n---\n\nWhile I'm at it, a smaller related observation: `climate.activity_end_time` seems to be computed relative to \"right now\" at parse time rather than the time the data was actually fetched from the server. If I look at the same fetched payload a few seconds later (or replay it), the end time drifts. It would be nicer if it were anchored to the fetch timestamp so the value is stable for a given response."} {"task_id": "format-code-task-000561", "source_id": "format-code-task-000561", "domain": "code", "task_path": "tasks/format-code-task-000561", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:8108a4b711bb2337134e3be02fc089039b6993de3d70148cbbba932ce455fb2e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Cannot force a sequence type when reading a FASTA file\n\nWhen I load a FASTA file with `fasta.get_sequence(...)` / `fasta.get_sequences(...)` / `fasta.get_alignment(...)`, biotite always tries to guess the sequence type by first attempting to parse the string as a `NucleotideSequence` and falling back to `ProteinSequence` only when that fails.\n\nThis works most of the time, but it breaks for short protein sequences that happen to consist only of letters that are also valid nucleotide symbols. For example, a peptide like `ACGTACGT` or a fragment containing only residues from `{A, C, G, T, N}` will silently come back as a `NucleotideSequence`, even though the file clearly contains proteins.\n\n```python\nfrom biotite.sequence.io import fasta\n\nf = fasta.FastaFile.read(\"my_proteins.fasta\")\nseq = fasta.get_sequence(f)\nprint(type(seq)) # NucleotideSequence -- but this is a protein file!\n```\n\nThere is currently no way to tell these functions \"this file contains proteins, don't guess\". I have to bypass the convenience functions entirely and build the `ProteinSequence` objects manually from the raw strings, which also means re-implementing the small normalisations biotite does (uppercasing, handling `U`, etc.).\n\nIt would be very useful if `get_sequence`, `get_sequences` and `get_alignment` accepted an optional argument letting the caller specify the expected `Sequence` subclass up front. When provided, the auto-detection should be skipped and the given type used directly; when omitted, the current guessing behaviour should be preserved so existing code keeps working.\n\nI'd expect to call it as a keyword argument, something like `seq_type=...`."} {"task_id": "format-code-task-000562", "source_id": "format-code-task-000562", "domain": "code", "task_path": "tasks/format-code-task-000562", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d567b4580037dba0e9ebd5e4d892f5d1abe9856300b3901446223b90e6bd2fd4", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nIt feels kind of clunky that whenever I have a `MnemonicCode` or a `BIP32Path` I have to reach into `.words` or `.path` before I can do anything useful with it — like `mnemonic.words.mkString(\" \")` or `path.path.foldLeft(...)`. Same story with `DeriveAddressesResult`, I just want to map/iterate over the addresses directly without going through `.addresses` every time. Could these wrapper types just behave like the sequence they're wrapping so I can call `.map`, `.mkString`, `.length`, indexing, etc. straight on them? Would save a lot of `.x.y` noise in calling code.\n\n## Expected outcomes\n\n- Sequence-backed wrapper values should support normal read-style Scala sequence operations directly on the wrapper, while preserving the same element order and values as their existing underlying sequence fields.\n- `MnemonicCode` should be usable directly as a sequence of words: operations such as `mkString`, `length`, indexing, iteration, and mapping should behave the same as they do on `mnemonic.words`.\n- `BIP32Path` should be usable directly as a sequence of `BIP32Node` values: operations such as `toList`, `foldLeft`, `length`, indexing, and iteration should behave the same as they do on `path.path`.\n- `DeriveAddressesResult` should be usable directly as a sequence of `BitcoinAddress` values: operations such as `length`, indexing, iteration, and mapping should behave the same as they do on `result.addresses`.\n- Existing construction, validation, serialization, and domain-specific behavior of the affected types should continue to work; this change is about reducing access boilerplate for sequence-like wrappers, not changing their underlying data.\n\n## Implementation notes\n\n- Prefer a reusable approach over one-off forwarding methods when several wrappers need the same sequence behavior, but the exact API shape, delegation strategy, data structures, and placement of shared code are up to the implementer.\n- Do not require callers to use the old inner-field access pattern when a direct sequence operation on the wrapper is sufficient."} {"task_id": "format-code-task-000563", "source_id": "format-code-task-000563", "domain": "code", "task_path": "tasks/format-code-task-000563", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6e236d66a62aab63af3c75af426b0c4c2b10fa4855b9634415b313d84e6c9bae", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI keep getting checksum mismatch errors on my managed order book for tXRPBTC at P0 precision — the data from the server looks fine but it errors out anyway. The issue seems to show up around very small price values like 0.00001, but I don't have a clean way to inspect the book with the original string values to confirm. Can you help me get past this so I stop getting false checksum errors, and also give me a way to read the book with prices/amounts as strings instead of floats?\n\n## Expected Outcomes\n\n- Managed order book checksum validation should not emit checksum-mismatch errors for valid server order book data whose numeric values are precision-sensitive.\n- `WSv2#getLosslessOB(symbol)` should return the currently managed order book for that symbol with order book prices and amounts preserved as the original string-form numeric values received from the websocket message.\n- `WSv2#getLosslessOB(symbol)` should return `null` when no string-preserved managed order book is available for the requested symbol, such as before a managed snapshot exists or when order book management is not enabled.\n- The existing managed order book API should continue to expose its normal numeric order book view for callers that use `WSv2#getOB(symbol)`.\n\n## Implementation Notes\n\n- The representation, parsing strategy, and update mechanism for preserving exact order book values are implementation details, as long as the public behavior above is satisfied.\n- Checksum handling should remain compatible with the existing managed order book flow and should avoid changing unrelated websocket message behavior.\n- Add or update public documentation for any new public API surface."} {"task_id": "format-code-task-000564", "source_id": "format-code-task-000564", "domain": "code", "task_path": "tasks/format-code-task-000564", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4f2847f415ddff59604fe4b2e0cc032874cbb7d28108ff9fb3039b33b3624fdf", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Setting values at a nested path — `GetPath` has no write-side equivalent\n\nI'm using simplejson to manipulate some nested JSON (a config blob, in my case). Reading deep into it is really clean with `GetPath`:\n\n```go\nhost := js.GetPath(\"server\", \"database\", \"host\").MustString()\n```\n\nBut when I want to **update** a value at a nested location, I can't find a clean way to do it. `Set` only operates on the top-level map, so to change something a few levels down I end up doing the type-assertion dance manually:\n\n```go\nm, _ := js.Map()\ninner, _ := m[\"server\"].(map[string]interface{})\ndb, _ := inner[\"database\"].(map[string]interface{})\ndb[\"host\"] = \"newhost\"\n```\n\n…and that's only the happy path — if any of the intermediate keys aren't there yet (e.g. I'm building the structure up from an empty `New()`), I also have to check for each one and create the missing maps myself before I can descend further. It gets ugly fast and there's no way it's the intended API.\n\nCould we get a write-side counterpart to `GetPath`? Something where I hand it the path and the value and it lands the value at the leaf, taking care of creating any intermediate structure that doesn't exist yet. The symmetry with `GetPath` would make the API a lot more usable for anything beyond flat objects.\n\nThe new method I'd expect is something like `SetPath(branch, val)`, mirroring `GetPath`. Handing it an empty path could just replace the whole root value."} {"task_id": "format-code-task-000566", "source_id": "format-code-task-000566", "domain": "code", "task_path": "tasks/format-code-task-000566", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4f5df719c15afd362f0bd6f3153563a5b7d35454eb0d560b92715e05c26d797b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[bug] fix the prompt when asking to document a model for the first time\nwhen we document a model for the first time we get this:\n```\n Installed dbt-sugar version: 0.0.0\n\n\n ____ __ \n ____/ / /_ / /_ _______ ______ _____ ______\n / __ / __ \\/ __/_____/ ___/ / / / __ `/ __ `/ ___/\n/ /_/ / /_/ / /_/_____(__ ) /_/ / /_/ / /_/ / / \n\\__,_/_.___/\\__/ /____/\\__,_/\\__, /\\__,_/_/ \n /____/ \n\nGetting sweetness out of the cupboard 🍬! \n\nThe model 'fct_orders' has not been docummented yet. Creating a new entry.\n? Do you want to change the model description of fct_orders (Y/n)\n```\n\nWe probably should change it to \"Do you want to write a description for fct_orders?\""} {"task_id": "format-code-task-000567", "source_id": "format-code-task-000567", "domain": "code", "task_path": "tasks/format-code-task-000567", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4ea076c193808753bd4aad9d650af062a0ece7be05e60cc6742fd81292900670", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Templater 在迭代 map / 嵌套 map 时缺少几个常用判断\n\n我在用 `bin-templater` 渲染 `.cfg` 模板，数据源是 map（有时候是嵌套 map）。在模板里 `range` 一个 map 时碰到两类问题，现有的 `isMap` / `isArray` / `isString` 这套帮不上忙：\n\n**1. 没法判断当前迭代到的 key 是 map 里的第一个还是最后一个**\n\n典型场景是渲染逗号分隔列表，最后一项后面不能有逗号；或者第一项前面不加分隔符。模板里 `range` 一个 map 拿到 `$k, $v`，想做这种边界判断时，没有可以直接调用的函数。Go 原生 `text/template` 又没有 `last` 这种 helper，只能在外面把数据先转成 slice，但这对模板用户很不友好——而且需要一个跟模板里实际迭代顺序（按 key 排序）一致的\"边界\"判断，否则首尾认错了。\n\n**2. 嵌套 map 里没法知道某个 value 处在第几层**\n\n我有形如 `map[string]interface{}` 嵌套若干层的数据，渲染时想按层做缩进或者根据深度切换格式。当前模板里完全没有获取\"某个 element 在这个嵌套 map 里的深度\"的途径。\n\n希望模板函数表里能补上这几类查询，让用户在模板里直接调用就行，不用把数据预先在 Go 代码里铺平。\n\nI'd expect the new template helpers to be named something like `IsMapFirst` / `IsMapLast` / `HowDeep`, each taking the map plus the element being checked (e.g. `IsMapFirst $data $key`, `HowDeep $data $value`)."} {"task_id": "format-code-task-000568", "source_id": "format-code-task-000568", "domain": "code", "task_path": "tasks/format-code-task-000568", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:196027ee69906f6f2f2424d29b0bf4fd12d3fe04b1470231a3214fbb5e1e9a79", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `youtube-dlc` command-line tool to support selecting exact playlist entries with `--playlist-items ITEM_SPEC URL`. `ITEM_SPEC` should be a comma-separated list of 1-based playlist positions, where each segment is either a single number like `2` or an inclusive range like `2-4`; for example, `youtube-dlc --playlist-items 2,4 PLAYLIST_URL` should process only playlist items 2 and 4, and `youtube-dlc --playlist-items 2-4 PLAYLIST_URL` should process items 2, 3, and 4.\n\nWhen the requested positions include duplicates or overlapping ranges, the tool should process each playlist position once in the order it first appears in the item spec, so `--playlist-items 2-4,3-4,3` processes items 2, 3, and 4. Requested positions outside the playlist length should be ignored without causing valid requested positions to fail, so `--playlist-items 3-10` on a four-item playlist processes items 3 and 4, while `--playlist-items 10` processes no videos. For an already loaded four-item playlist, `--playlist-items 2,4` should report that two videos are being downloaded rather than the full playlist. The selected entries should keep their original playlist positions in playlist metadata, so `--playlist-items 4,2` downloads the fourth item first with playlist index 4 and then the second item with playlist index 2.\n\nThis should work for playlists whose entries are already loaded, paged lazily, or exposed as a generic iterable. A valid selection that downloads the requested videos should exit with code 0 under the same conditions as a normal playlist download. If `ITEM_SPEC` is malformed, such as `abc` or `2-x`, the command should fail with a non-zero exit instead of silently downloading the whole playlist. `youtube-dlc --help` should list `--playlist-items ITEM_SPEC` in the video selection options and describe both comma-separated item numbers and inclusive ranges."} {"task_id": "format-code-task-000569", "source_id": "format-code-task-000569", "domain": "code", "task_path": "tasks/format-code-task-000569", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3e85fe8494a23c7bcb5bc5d11b9a76c23d5c4346229cab1bc62fde4a4ab4c955", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Reverse token filter mangles strings that contain combining characters\n\nI'm using the `reverse` token filter as part of an analyzer chain to index text in a few languages that make heavy use of diacritics (think Vietnamese, or French/Spanish text that ends up in NFD form after some upstream normalization). The idea is to reverse the token so I can do suffix-style matching.\n\nThe problem: tokens that contain combining characters come out of the reverse filter looking corrupted. Characters that the user sees as a single accented letter get broken apart, and the pieces end up in the wrong order relative to each other. So a word that should still be a readable (reversed) string of accented letters comes out as garbage where the accent marks are attached to the wrong base letters.\n\nFor plain ASCII tokens everything looks fine. The breakage only shows up once the input contains characters that are encoded as a base character followed by one or more combining marks.\n\nWhat I'd expect is that reversing a token treats each user-perceived character as a single unit — so an accented letter stays an accented letter after reversal, just in a new position in the string. Right now reversing visibly destroys those characters, which makes the filter unusable for anything beyond pure ASCII / pre-composed input.\n\nCould the reverse filter be made to handle these combining-character sequences correctly?"} {"task_id": "format-code-task-000570", "source_id": "format-code-task-000570", "domain": "code", "task_path": "tasks/format-code-task-000570", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5044a9042afc114f469f97ed35a752fd8c17464bdcb3fb72474500b36a0e6271", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## RangeIterator.Seek can return keys outside the requested range\n\nI'm using bleve's KV store `RangeIterator(start, end)` and calling `Seek` on it to jump around within the range. I noticed that if I pass `Seek` a key that's lexicographically before `start`, the iterator happily positions itself there and starts yielding keys that are outside the range I asked for.\n\nMinimal repro against the gtreap (in-memory) store — same thing happens with the boltdb store:\n\n```go\n// populate some keys: \"a\", \"b\", \"c\", \"d\", \"e\"\nreader, _ := store.Reader()\nit := reader.RangeIterator([]byte(\"c\"), []byte(\"e\"))\n\n// seek to something before the range\nit.Seek([]byte(\"a\"))\n\nfor k, _, ok := it.Current(); ok; it.Next() {\n k, _, ok = it.Current()\n fmt.Printf(\"%s\\n\", k)\n}\n```\n\nI expected the iterator to only ever yield keys in `[start, end)` regardless of what I pass to `Seek` — i.e. seeking to a key before `start` should behave the same as seeking to `start`. Instead I get `a`, `b`, `c`, `d` back.\n\nThis breaks the contract I'd expect from a range iterator: once I've constrained the iterator to `[start, end)`, no `Seek` should be able to escape that window on the low side. The high side (`end`) is already respected correctly."} {"task_id": "format-code-task-000571", "source_id": "format-code-task-000571", "domain": "code", "task_path": "tasks/format-code-task-000571", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:bac8b51006eef62a5f36ec02039688156570eb28624c3e5e2b445f9e125c0b30", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `pystack core` fails on core dumps from Python scripts launched via shebang\n\nI have a Python script that I run directly with a shebang (i.e. it starts\nwith `#!/usr/bin/env python3`, it's `chmod +x`'d, and I just execute it as\n`./myscript.py`). The script crashed and produced a core dump, and I wanted\nto use pystack to inspect what the interpreter was doing at the time.\n\nWhen I run\n\n```\npystack core ./core.12345\n```\n\nwithout specifying an executable (the second positional argument is\ndocumented as optional), pystack picks up an executable automatically and\nthen bails out complaining that what it picked up isn't a valid executable\n— it looks like it grabbed the path to my `.py` script itself, not the\n`python3` binary that was actually running it. That makes some sense\nbecause for shebang-launched scripts the kernel does record the script\npath as the \"executable\" of the process, but it's not what pystack\nactually needs to analyze the core.\n\nAs a workaround I can pass the interpreter explicitly:\n\n```\npystack core ./core.12345 /usr/bin/python3\n```\n\nand then everything works fine. But this is annoying — the core file\nclearly contains enough information about the real interpreter that was\nloaded (I can pass it by hand and it just works), so pystack shouldn't\nhave to give up and demand that I supply it manually for what's a pretty\ncommon way of running Python scripts.\n\nIt would be great if `pystack core`, when run without an explicit\nexecutable, could recover in this situation instead of immediately erroring\nout, so that the shebang-script case works out of the box."} {"task_id": "format-code-task-000573", "source_id": "format-code-task-000573", "domain": "code", "task_path": "tasks/format-code-task-000573", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:13ef01a4ec92ad82fb02bd3548dd0cea608095ec6a088118efa4caa7f45d368c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## LiveTable silently drops fields it can't display\n\nI'm using `LiveTable` with a list of fields I want to watch during a scan. Some of those fields end up not showing up in the printed table at all — no column header, no values, nothing. The scan otherwise runs fine.\n\nAfter poking around for a while I figured out it's happening for fields whose `dtype` in the descriptor isn't one of the simple scalar types LiveTable knows how to format (e.g. array-valued detector channels). That's fair enough — LiveTable obviously can't print an arbitrary array as a single table cell — but the fact that it just disappears with zero feedback is really confusing. The first time it happened I thought I'd typo'd the field name, or that my detector wasn't producing data, and spent a while chasing the wrong thing.\n\nIt would be much nicer if LiveTable told me when it's dropping a requested field because it doesn't know how to render its dtype, so I at least know to either remove that field from my list or send it to a different callback."} {"task_id": "format-code-task-000574", "source_id": "format-code-task-000574", "domain": "code", "task_path": "tasks/format-code-task-000574", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fbbddcb3bb0822a8804a69623c9767acdcdce4b9f6b893b8c9f8e0943a89949a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Block fetcher still accepts blocks at or below finalized height\n\nAfter Fast Finality was enabled on BSC, blocks at or below the finalized height are deterministically final — they cannot be reorged out. However, the block fetcher in `eth/fetcher` still accepts propagated blocks within the legacy side-chain window (roughly `[CurrentBlock - maxUncleDist, CurrentBlock + maxQueueDist]`, i.e. about `[head-11, head+32]`).\n\nThis means peers can announce / broadcast blocks whose number is `<=` our current finalized height, and the fetcher will happily queue and try to import them. Those blocks will never become part of the canonical chain, so the work spent announcing, queuing, and verifying them is wasted, and they show up as noise in fetcher debug logs on a node that already has finality information available locally (`chain.CurrentFinalBlock()`).\n\nCould the block fetcher be made aware of the local finalized height and skip any propagated block / announcement whose number is at or below it? On BSC with finality enabled there is simply no reason to fetch them."} {"task_id": "format-code-task-000575", "source_id": "format-code-task-000575", "domain": "code", "task_path": "tasks/format-code-task-000575", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d3bf5d27e5d07e202c2a48561808c8e9a43cc9039db9ee758e1524912f5974c3", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Guest can send packets but never receives anything\n\nI'm trying to get networking working in a Linux guest booted by gokvm, using the tap device that gokvm sets up. The outbound direction looks fine — anything the guest sends (ARP, DHCP discover, ping, etc.) shows up on the host's tap interface when I `tcpdump` it.\n\nThe reverse direction is broken though. Nothing from the host side ever makes it into the guest:\n\n- DHCP from the guest leaves and the DHCP server on the host replies, but the guest never sees the offer and the request eventually times out, so the interface never gets an address.\n- If I assign an address manually and `ping ` from the host, the echo requests are visible on the tap, but the guest doesn't reply. Running `tcpdump` inside the guest on the virtio NIC shows nothing at all — the packets simply never arrive at the guest's NIC.\n\nSo in practice the virtio-net device behaves as send-only: the guest can transmit, but nothing the tap delivers ever reaches it. It would be great if gokvm forwarded packets coming in on the tap into the guest's virtio-net device so the NIC actually works as a normal full-duplex interface.\n\n(For symmetry with the existing Tx path, it would be natural to expose the per-packet receive step as a method like `Rx()` on the virtio-net device so it can be driven both from a background thread and from tests.)"} {"task_id": "format-code-task-000576", "source_id": "format-code-task-000576", "domain": "code", "task_path": "tasks/format-code-task-000576", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:78d014c39b883f9ea463448a0f264a236954cc1cca6eb1e04dcdc6853a2c2c51", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add `exec` commands for running arbitrary executables with local bins on PATH\n\n`pyarn` can already `run` package scripts, but there's no way to invoke an arbitrary\nexecutable (for example a binary installed by a dependency) with the project's locally\ninstalled tools available on `PATH`. Add a family of `exec` commands that do this.\n\nThere should be four variants, mirroring how the existing `run` commands are organised:\n\n- a top-level `exec` that runs in the package closest to the current directory,\n- a `project` `exec` that runs in the project root package,\n- a `workspace` `exec` that runs inside a single named workspace,\n- a `workspaces` `exec` that runs in every workspace (respecting the usual workspace\n filters).\n\nExpose each one the same way the existing commands are exposed: a runner function plus a\nmatching `to…Options` builder, re-exported from the commands module, following the same\nnaming pattern already used by `run`/`toRunOptions`, `projectRun`/`toProjectRunOptions`,\n`workspaceRun`/`toWorkspaceRunOptions` and `workspacesRun`/`toWorkspacesRunOptions`.\n\n## Behaviour\n\n**Parsing the command.** The executable to run and its arguments come from the `--`\npassthrough array on the parsed flags (`flags['--']`): the first element is the executable\nname and the remaining elements are its arguments. As with the other commands, `cwd` comes\nfrom `flags.cwd` and defaults to the current working directory. For the single-workspace\nvariant the workspace name is the first positional argument.\n\n**Execution.** Unlike `run`, this does not go through a package script — it launches the\nnamed executable directly as a child process. The child must be spawned with:\n\n- its working directory set to the directory of the target package, and\n- an environment inherited from the current process **except** that `PATH` is prepended\n with the relevant local `node_modules/.bin` directories.\n\n**PATH precedence.** The augmented `PATH` is built from, in order:\n\n1. the target package's own `node_modules/.bin` directory — but only when the target\n package is not the project root itself (so the project root's bin directory is never\n added twice),\n2. the project root package's `node_modules/.bin` directory,\n3. the pre-existing `PATH` from the environment (when set).\n\nThese entries are joined with `:`. So when running inside a workspace, that workspace's\nlocal bin directory comes first, followed by the project root's bin directory, followed by\nthe inherited `PATH`. When running at the project root, the project root bin directory\ncomes first.\n\n**Unknown workspace.** The single-workspace variant must throw an error when no workspace\nwith the given name exists in the project.\n"} {"task_id": "format-code-task-000577", "source_id": "format-code-task-000577", "domain": "code", "task_path": "tasks/format-code-task-000577", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d015967776544d515512c54637565f0bdcd30e45f714fd1f0c3e94fa3d6458e8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI've run into two issues while configuring wsl for our codebase.\n\n## 1. `AllowCuddleDeclaration` doesn't seem to cover the common cases\n\nI turned on `AllowCuddleDeclaration` because I want to be able to put a `var` declaration right next to the line that uses it. But I still get warnings on patterns like:\n\n```go\nvar foo string\nfoo = \"some value\"\n```\n\n→ `assignments should only be cuddled with other assignments`\n\n```go\nvar server = NewServer()\nserver.Start()\n```\n\n→ `expressions should not be cuddled with declarations or returns`\n\nI'd expect that once I've opted in to cuddling declarations, follow-up lines that initialize or use the just-declared variable would be allowed too. Otherwise the option only helps for stacking multiple `var` lines together, which is a pretty narrow use case.\n\n## 2. `ForceCuddleErrCheckAndAssign` is awkward with multi-line assignments\n\nWith `ForceCuddleErrCheckAndAssign` enabled I get pushed toward writing:\n\n```go\nresult, err := someFunc(\n \"argument one\",\n \"argument two\",\n \"argument three\",\n)\nif err != nil {\n return err\n}\n```\n\nThis is fine for short assignments, but when the call producing the error spans many lines I really want a blank line between the closing paren of the call and the `if err != nil` — visually the assignment block and the error handling block are two distinct things, and gluing them together makes the code harder to read.\n\nSo in this case:\n\n```go\nresult, err := someFunc(\n \"argument one\",\n \"argument two\",\n \"argument three\",\n)\n\nif err != nil {\n return err\n}\n```\n\nwsl complains because the err check isn't cuddled with the assignment, but I don't think the rule should fire when the assignment itself is multi-line. For single-line assignments I'm happy to keep the strict cuddling requirement.\n\nCould both of these be addressed?"} {"task_id": "format-code-task-000578", "source_id": "format-code-task-000578", "domain": "code", "task_path": "tasks/format-code-task-000578", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2271de202b950d5f897769e750f81d0d22a97a68865a7e8b8c7713b1eb63e06b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## OpenTSDB `dropcounter` rate option not supported\n\nOpenTSDB supports a `dropcounter` rate option (in addition to `counter`) which behaves like `counter` but drops data points where the counter appears to have reset, instead of emitting a huge negative/wrap-around value. This is very useful for metrics from processes that restart and reset their counters — `counter` with a `counterMax` doesn't always cut it because we don't know the real max.\n\nWhen I write a bosun query using it, e.g.\n\n```\nsum:rate{dropcounter}:os.cpu\n```\n\nbosun doesn't handle it as a `dropcounter` — it just treats the rate options block as if `dropcounter` weren't a counter at all, and round-tripping the parsed `Query` back to a string loses the option entirely. So I can't take advantage of this OpenTSDB feature from bosun.\n\nWould be great if `opentsdb.ParseQuery` (and the corresponding `Query.String()`) understood `dropcounter` the same way it understands `counter`, and carried the information through on the `RateOptions` so it ends up in the JSON sent to OpenTSDB.\n\nI'd expect a new boolean field on `RateOptions` (something like `DropResets`) to carry this through alongside the existing `Counter` flag."} {"task_id": "format-code-task-000579", "source_id": "format-code-task-000579", "domain": "code", "task_path": "tasks/format-code-task-000579", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:cdd226fdfdf2ef1f48f73a677b194f1313cd04af1f0ece85f1de2faabd535bb2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `||` and `&&` don't short-circuit when one side is NaN\n\nI'm writing alerts in Bosun and trying to use `||` as a fallback when a query returns no data. Something like:\n\n```\n$value = q(\"sum:my.metric{}\", \"5m\", \"\")\n$result = $value || 0\n```\n\nThe idea is: if the query returns NaN (no data points in the window), fall back to 0 so the alert doesn't go unknown. But what I'm seeing is that `$result` is also NaN whenever `$value` is NaN — the `|| 0` part doesn't help at all.\n\nSame thing the other direction with `&&`. I have a guard like `$hasData && $value > 100`, and when `$hasData` is 0 I'd expect the whole thing to be 0 regardless of what `$value` is. Instead, if `$value` happens to be NaN, the result is NaN and the alert misbehaves.\n\nThis makes both operators pretty useless for the most common reason you'd reach for them in a monitoring context — handling missing data. In every other language I know, `1 || anything` is true and `0 && anything` is false without ever looking at the right side, so NaN on the unused side shouldn't poison the result. Bosun's logical operators should behave the same way: if the left side alone is enough to determine the result, the right side (NaN or otherwise) shouldn't matter."} {"task_id": "format-code-task-000580", "source_id": "format-code-task-000580", "domain": "code", "task_path": "tasks/format-code-task-000580", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a773e1f4319675498f034a48600e6f37abb9c8274a120e139979a4f6d75ba47e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want DynamoDB resources created with `boto3.resource('dynamodb', region_name=...)` to automatically provide a high-level parameter and response interface for the generated resource and table operation methods. A typical session should work like `dynamodb = boto3.resource('dynamodb', region_name='us-east-1')`, `table = dynamodb.Table('Users')`, then `table.put_item(Item={'pk': 'u#1', 'age': 42, 'active': True, 'tags': {'admin', 'beta'}, 'profile': {'name': 'Ada'}, 'scores': [1, 2], 'empty': None})`; when the request is built, every operation member modeled as a DynamoDB `AttributeValue` should be serialized to the corresponding DynamoDB wire dictionary, including nested maps and lists. For read responses, operation output members modeled as `AttributeValue` should be deserialized back to ordinary Python values, so a response item like `{'pk': {'S': 'u#1'}, 'age': {'N': '42'}, 'active': {'BOOL': True}, 'empty': {'NULL': True}}` is returned to the caller as `{'pk': 'u#1', 'age': Decimal('42'), 'active': True, 'empty': None}`.\n\nI also want table operations that accept DynamoDB expressions to accept condition objects directly. For example, `table.query(KeyConditionExpression=Key('pk').eq('u#1'), FilterExpression=Attr('age').gte(21))` should convert those objects into expression strings, add generated `ExpressionAttributeNames` and `ExpressionAttributeValues`, and merge those generated placeholders with any placeholders I supplied myself. A key condition must only use `Key(...)` attributes; using `Attr(...)` inside a key condition should raise the same DynamoDB key-condition validation exception that the condition builder raises.\n\nThe automatic interface should preserve caller-owned input structures. If the same Python item object is passed twice in one request, each occurrence should be transformed independently, and the original object I passed should not be mutated or partially serialized. Operations with no output shape, such as a successful tag call returning `{}`, should still complete normally without attempting response conversion. The generated DynamoDB resource documentation should describe `AttributeValue` parameters as accepting valid Python DynamoDB values and expression parameters as accepting `Key(...)` or `Attr(...)` condition objects."} {"task_id": "format-code-task-000581", "source_id": "format-code-task-000581", "domain": "code", "task_path": "tasks/format-code-task-000581", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b9f84884e5211828f85d25df4d54268e53313a1779887751865a9207a9aeaae8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nEnvironment variables set to empty cause silent failing of auth\nHello,\n\nrecently, we stumbled upon an interesting 'bug' that took us some time to resolve - we had set `AWS_ACCESS_KEY_ID` env variable set to empty, which resulted of having it in `os.environ` as empty string. If it's there as an empty string, it causes config resolver to use EnvProvider - even if it's clearly invalid. While i realize that this issue is caused by us, I would like to propose a small improvement to EnvProvider - which will not only check if env variables are set - but also if they're valid and log/error out when that's not the case, allowing users for quicker debugging. \n\nIf I can help with the implementation, please let me know, I would gladly take a stab at it.\n\nRegards,\nPiotr"} {"task_id": "format-code-task-000582", "source_id": "format-code-task-000582", "domain": "code", "task_path": "tasks/format-code-task-000582", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d7a109a80cfc2fecea80a1235d05b6f1dc5ace200ac2050924909290c562b066", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### No way to abort an in-progress chunked upload from `ChunkedUploader`\n\nI'm using the automatic chunked uploader to push large video files to Box:\n\n```python\ntest_file_path = '/path/to/large_file.mp4'\ncontent_stream = open(test_file_path, 'rb')\ntotal_size = os.stat(test_file_path).st_size\nchunked_uploader = client.upload_session('56781').get_chunked_uploader_for_stream(content_stream, total_size)\nuploaded_file = chunked_uploader.start()\n```\n\nOccasionally `start()` blows up partway through (flaky network, the connection drops on one of the parts, etc.). At that point I've already uploaded a bunch of parts on the server side and I'd like to cleanly cancel the whole thing so I can start over with a fresh session.\n\nThe problem is the `ChunkedUploader` object doesn't expose any way to do that. I can see that `UploadSession` itself has an abort, and the manual chunked-upload flow documents using it, but when you go through the automatic uploader you're holding a `ChunkedUploader` and there's no equivalent on it — so wrapping `start()` in a try/except doesn't really help me recover, because I have nothing to call in the `except` branch to tear the session down.\n\nIt would be great if `ChunkedUploader` itself supported cancelling — something I can call from an `except` clause to abort the upload and clean up the parts that were already uploaded. Ideally after that the uploader instance is \"dead\" and trying to resume/start it again would fail loudly rather than do something undefined, so callers are pushed toward creating a fresh upload session."} {"task_id": "format-code-task-000583", "source_id": "format-code-task-000583", "domain": "code", "task_path": "tasks/format-code-task-000583", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:da9438bf7587d2ce9c0b14ce59d266879c1f7bf7e3f921c27f9f4a028e1dfd81", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Annotations API endpoints don't enforce permissions\n\nI'm wiring up the activity feed to include annotations, and I noticed that the methods on the `Annotations` API class (`createAnnotation`, `deleteAnnotation`, `getAnnotation`, `getAnnotations`) just fire off the HTTP request without first checking whether the caller actually has permission to perform that action on the file.\n\nCompare this with the other feed-related APIs in the codebase — for example `CommentsAPI.getComments` and `AppActivityAPI.getAppActivity` both take the file's permissions as an argument and short-circuit through the error callback (via the shared `checkApiCallValidity` helper on the base class) when the relevant permission is missing. That way the UI gets a consistent error code back instead of us blindly hitting the server and relying on the backend to reject the call.\n\nThe annotations endpoints should behave the same way: before making the network request, they should verify that the file (or the annotation itself, in the delete case) grants the appropriate permission, and surface a sensible error through `errorCallback` when it doesn't. The `BoxItem` permission shape will also need to grow whatever new permission flags are needed to express \"may this user view annotations on this file\" and \"may this user create annotations on this file\", since those don't exist on `BoxItemPermission` today.\n\n`Feed.fetchAnnotations` (the only current caller inside this repo) should be updated to thread the file's permissions through to the new annotations call so the gate actually runs in practice.\n\nThis is purely about parity with how comments / app activity already work — no behavior change for callers that already have the right permissions, just a fast, local rejection (and a proper error code) for callers that don't."} {"task_id": "format-code-task-000584", "source_id": "format-code-task-000584", "domain": "code", "task_path": "tasks/format-code-task-000584", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0d486c30272bcba040f95841ba87631028e0f648eaa2a25b0b59da38e20a78fa", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nPlease add metadata-cascade-policies:create\n### Is your feature request related to a problem? Please describe.\n\nWe need to create metadata cascade policy for many folders. \nWe want to use CLI but it looks like it doesn’t have `metadata-cascade-policies:create` command. \nI guess `metadata-cascade-policies:force-apply` command doesn’t work for us as it requires metadata id that has already cascade enabled.\n\n### Describe the solution you'd like\n\nPlease simply add `box metadata-cascade-policies:create` command which calls the following API.\nhttps://developer.box.com/reference/post-metadata-cascade-policies/"} {"task_id": "format-code-task-000585", "source_id": "format-code-task-000585", "domain": "code", "task_path": "tasks/format-code-task-000585", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:06a2d9b635c894e3839b5b3699b6237e41224941e1f6242ef1fb52fa6917a425", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `rerun_filter` under pytest always receives `None` for err and name\n\nI'm using flaky with pytest and trying to set up a `rerun_filter` so that only certain kinds of failures get retried (e.g. retry on a flaky network exception, but fail immediately on `AssertionError`).\n\nMinimal repro:\n\n```python\nimport pytest\nfrom flaky import flaky\n\ndef my_filter(err, name, test, plugin):\n print(\"rerun_filter called with err=%r name=%r\" % (err, name))\n # I want to inspect err[0] (the exception type) here to decide\n return True\n\n@flaky(max_runs=3, rerun_filter=my_filter)\ndef test_something():\n assert False\n```\n\nWhen I run this under pytest, the filter does get invoked between reruns, but the output is always:\n\n```\nrerun_filter called with err=(None, None, None) name=None\n```\n\nNo matter what the test actually raises, `err` is `(None, None, None)` and `name` is `None`. That makes it impossible to write a filter that branches on the exception type or on which test failed — the only signal the callback gets is \"something failed, decide now\", with no information about *what* failed.\n\nI'd expect the same callback to receive the real exception info and test name (the way it does for the nose integration), so that filters like \"retry only on `ConnectionError`\" can actually be written.\n\nflaky version: 3.0.1, pytest as test runner."} {"task_id": "format-code-task-000586", "source_id": "format-code-task-000586", "domain": "code", "task_path": "tasks/format-code-task-000586", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9b5a7b1bbfa689de74725904036b206b9f425edd07bdc04fb055d68dcffd0b65", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `%+v` on errtrace-wrapped errors doesn't include the return trace\n\nI'm using errtrace to track how errors propagate through my service, and\nI log errors at the top of my handlers the same way I would with most\nother error libraries:\n\n```go\nif err := doWork(ctx); err != nil {\n log.Printf(\"request failed: %+v\", err)\n return\n}\n```\n\n`doWork` and everything it calls wraps returned errors with\n`errtrace.Wrap`, so by the time the error reaches my handler it already\ncarries a full return trace internally.\n\nThe problem: the `%+v` output looks identical to plain `%v` — I just get\nthe underlying error's message on a single line, with no trace at all.\nTo actually see the return path I have to remember to call\n`errtrace.Format(os.Stderr, err)` (or `errtrace.FormatString(err)`) as a\nseparate step, which is easy to forget and clutters every error-logging\nsite.\n\nI'd expect that when I format an errtrace-wrapped error with `%+v`, the\nreturn trace comes along automatically — that's the whole reason I\nreached for errtrace over plain `fmt.Errorf`. The other common verbs\n(`%v`, `%s`, `%q`, ...) should still behave the way they do today so\nexisting logs don't suddenly change shape; only `%+v` needs to opt into\nthe richer output.\n\nCould errtrace render the trace itself when the caller asks for the\nverbose form, instead of requiring a separate `Format` call?"} {"task_id": "format-code-task-000587", "source_id": "format-code-task-000587", "domain": "code", "task_path": "tasks/format-code-task-000587", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:693941370708e2e630765b01aaf56bc8d4a97cab19a31b816804c254eeaa6793", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: a decorator that wraps a function in a Redlock\n\nRight now if I want to make sure a function only runs one-at-a-time across\nmy workers, I have to wrap every call site (or the body of the function)\nwith the `Redlock` context manager:\n\n```python\nfrom pottery import Redlock\n\ndef refresh_cache():\n with Redlock(key='refresh-cache'):\n # ... actual work ...\n ...\n```\n\nThis works, but it's boilerplate I end up repeating in a lot of places, and\nindenting the whole function body just to hold a lock feels off — the lock\nis really a property of the function itself, not of the body.\n\nIt would be nice if pottery shipped a decorator form so I could just write\nsomething like:\n\n```python\n@some_decorator(key='refresh-cache')\ndef refresh_cache():\n # ... actual work ...\n ...\n```\n\nand have every call to `refresh_cache()` transparently acquire the Redlock\nbefore running and release it after. The decorator should accept the same\nknobs `Redlock` already accepts (the lock key, the redis masters to use,\nthe auto-release timeout) so I don't lose any of the existing flexibility.\n\nI've been carrying a private version of this in my own app for a while and\nit's been useful enough that I think it belongs in the library proper.\n\nNaming-wise I'd expect the new helper to be importable from `pottery` as\nsomething like `redlock` (the lowercase counterpart of the existing\n`Redlock` class), so usage reads `from pottery import redlock`."} {"task_id": "format-code-task-000588", "source_id": "format-code-task-000588", "domain": "code", "task_path": "tasks/format-code-task-000588", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d98da8cd14f2bbd384a1ab7d7a2b5cd275dafad401301f8aa576aa659ca3f1ac", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n'Notify me about token promotions' switch doesn't work if you've already said 'Maybe Later'\n\n\n### Description\nIf you are prompted for a promotion and you select 'Maybe Later', the remindTimestamp is set to remind you 24h later. If you then go into Adv Settings and toggle 'Notify me about token promotions' off, you will still be reminded 24h later.\n\n\n### Steps to Reproduce\n\n\n 1. Clean install of 0.20 using ledger_environment=staging.\n 2. Get notification.\n 3. Select Maybe Later.\n 4. Go to Adv Settings, toggle 'Notify me about token promotions' off.\n 5. Close Brave.\n 6. Open session-store-1 and locate remindTimestamp. Using https://www.epochconverter.com/ you can see that it is set to 24h later.\n 7. Change this value to 5-10 minutes from now.\n 8. Relaunch Brave (make sure you still use ledger_env staging)\n\n\n\n**Actual result:**\nIn 30-60 minutes you will be notified about the promotion again. \n\n\n**Expected result:**\nYou should not be notified about the promotion since you toggled the switch to off.\n\n**Reproduces how often:**\nEasily\n\n\n### Brave Version\n\n**about:brave info:**\n0.20.30\n\n\n**Reproducible on current live release:**\nYes\n\n\n### Additional Information\nMissed when testing https://github.com/brave/browser-laptop/issues/12313"} {"task_id": "format-code-task-000589", "source_id": "format-code-task-000589", "domain": "code", "task_path": "tasks/format-code-task-000589", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1d9872c1d78beae42d4cb19d0c4e9cbfa5e5a4eccbe926103f5a8024f17f0718", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Password reset is broken on Django 1.6+\n\nAfter upgrading my project to Django 1.6, the userena password reset flow no longer works end-to-end. On Django 1.5 the same setup was fine.\n\nSteps to reproduce on a stock userena install:\n\n1. Go to `/password/reset/`, submit a valid email.\n2. Receive the password reset email and click the link inside.\n3. Try to set a new password.\n\nWhat actually happens:\n\n- After submitting the email on the reset form, the request errors out instead of landing on the \"we've sent you an email\" page.\n- If I bypass that and click the confirm link from the email directly, the confirm view doesn't match the URL / can't find the user, so I can't actually pick a new password.\n- Even when I work around the above, finishing the flow blows up trying to render the \"your password has been changed\" step.\n\nEverything is plain userena URLs (`include('userena.urls')`) — I haven't overridden the reset views or the email template. The same project works correctly on Django 1.5, so this looks like fallout from changes to `django.contrib.auth`'s password reset views in 1.6.\n\nIt would be great to get the password reset flow working on Django >= 1.6 again, while still working on 1.5 (please don't drop 1.5 just to fix this). Ideally without renaming any of the existing `userena_password_reset*` URL names, since I (and presumably others) have `{% url %}` tags and reverse() calls pointing at them from custom templates and code."} {"task_id": "format-code-task-000590", "source_id": "format-code-task-000590", "domain": "code", "task_path": "tasks/format-code-task-000590", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d4f1da048710c82bf6bca28783b01edb74b531cb0baef1ed75b54be7eff3ec10", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the Lisp environment's built-ins to include a `(load path)` form for loading another Lisp source file during evaluation. The built-in loader should initialize a public `load-path` variable from the `SHADEN_LISP_PATH` environment variable, splitting it with the platform path-list separator, so Lisp code can inspect or replace that list before calling `load`.\n\nWhen I create an environment, call `builtin.Load(env)`, and evaluate `(load \"testdata/load-example.lisp\")`, it should search each directory in `load-path`, open the matching file, parse it as Lisp, evaluate it in the same environment, and return the file's final expression value. For example, if `testdata/load-example.lisp` contains `(+ 1 1)`, evaluating `(load \"testdata/load-example.lisp\")` should return `2`. If I first evaluate `(set! load-path (list \"testdata/load-path/\"))` and then `(load \"load-path-example.lisp\")`, the loader should find the file relative to that configured directory and return the loaded file's final value.\n\nThe form should require exactly one argument, evaluate that argument before using it, and require the evaluated value to be a string path. Missing files should fail with a non-nil error such as `doesntexist.lisp not found in load-path`. File open, parse, and evaluation failures should also surface as errors that include the file path context."} {"task_id": "format-code-task-000591", "source_id": "format-code-task-000591", "domain": "code", "task_path": "tasks/format-code-task-000591", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:feca071fe2b64f71bb933e8d807d4054acc20bee1dd27b2405cfc87398a5a369", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Make the init callback optional\n\nWhen constructing a Lawnchair instance with a synchronous adapter (e.g. localStorage), there's no real reason to pass a callback — the store is ready immediately. But right now Lawnchair refuses to construct without one:\n\n```js\n// throws: \"No callback was provided\"\nvar store = new Lawnchair({ name: 'things' });\n\n// also throws: \"Incorrect # of ctor args!\"\nvar store = new Lawnchair();\n```\n\nSo in practice a lot of code ends up looking like:\n\n```js\nnew Lawnchair({ name: 'things' }, $.noop);\nnew Lawnchair({ name: 'things' }, Prototype.emptyFunction);\nnew Lawnchair({ name: 'things' }, function(){});\n```\n\n…just to get past the constructor check. That's noise — the user clearly doesn't care about being notified, especially against a synchronous backend.\n\nCould the callback be made optional? I'd expect all of these to just work:\n\n```js\nnew Lawnchair();\nnew Lawnchair({ name: 'things' });\nnew Lawnchair(function (ref) { /* ... */ });\nnew Lawnchair({ name: 'things' }, function (ref) { /* ... */ });\n```\n\ni.e. you can pass options, a callback, both, or nothing, and Lawnchair still constructs cleanly."} {"task_id": "format-code-task-000592", "source_id": "format-code-task-000592", "domain": "code", "task_path": "tasks/format-code-task-000592", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0268a66dc2ef4ca4eaa0bc36d95d39ee06ca8f1433faef9f2dd476276ff15cf0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## checkov misses resources in deeply nested Terraform child modules when scanning a tfplan JSON\n\nI'm using checkov to scan the JSON output of `terraform show -json` against a Terraform project that has nested modules — my own module composes a couple of versioned upstream modules from the registry, so the resulting plan has child modules inside child modules.\n\nWhen I scan the plan JSON, checkov picks up resources that are defined directly in the root module, and it also picks up resources from the immediate child modules. But the resources that live in the *next* level down — the ones declared inside a module that was itself instantiated by one of my child modules — never appear in the scan results at all. They aren't reported as passing, and they aren't reported as failing; checkov just behaves as if those resources don't exist.\n\nReproduction is roughly:\n\n```hcl\n# root main.tf\nmodule \"my_team_module\" {\n source = \"./modules/my_team_module\"\n ...\n}\n\n# modules/my_team_module/main.tf\nmodule \"upstream\" {\n source = \"terraform-aws-modules/s3-bucket/aws\"\n version = \"x.y.z\"\n ...\n}\n```\n\n```\nterraform plan -out plan.out\nterraform show -json plan.out > plan.json\ncheckov -f plan.json\n```\n\nThe S3 bucket and related resources that the upstream module actually creates are visible in `plan.json` (under `planned_values.root_module.child_modules[...].child_modules[...].resources`), but they never show up in checkov's output. If I refactor the same code so that the upstream module is called directly from the root (i.e. only one level of nesting), those same resources are scanned correctly.\n\nComposing modules out of other modules is a pretty normal pattern in Terraform, so this is a real blind spot when relying on checkov for plan-file scanning. Could `checkov -f plan.json` be made to find resources at any depth of module nesting, not just the first level?"} {"task_id": "format-code-task-000593", "source_id": "format-code-task-000593", "domain": "code", "task_path": "tasks/format-code-task-000593", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d9e45f6cf16588abd06a7a84b06f3ae11d11a0669e1a66e41423525e9446bc6e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a public `dynamize_value(val)` function for DynamoDB condition-building code that converts one native Python scalar or homogeneous set into a DynamoDB AttributeValue dictionary. For example, `dynamize_value('foo')` should return `{'S': 'foo'}`, `dynamize_value(54)` should return `{'N': '54'}`, and `dynamize_value(Binary(b'\\x01'))` should return `{'B': 'AQ=='}`. It should also handle sets: numeric sets become an `NS` value containing stringified numbers, string sets become an `SS` value containing the same strings, and binary sets become a `BS` value containing base64-encoded entries.\n\nThe function should reject unsupported values by raising `TypeError`, matching the type detection rules used by the DynamoDB type utilities. Calling it should be a one-shot pure conversion: it should not mutate the input value or set, should not touch the filesystem or network, and repeated calls with the same supported input should produce the same equivalent AttributeValue output."} {"task_id": "format-code-task-000595", "source_id": "format-code-task-000595", "domain": "code", "task_path": "tasks/format-code-task-000595", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:de1023dc7115aafb207bb6ae002b385e6d347dbaae695c3fc100cb3f4ee9e884", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: E-Prime check\n\nI write a fair amount of technical prose in [E-Prime](https://en.wikipedia.org/wiki/E-Prime) style, which means avoiding all forms of the verb \"to be\" (is, was, been, etc.). I'd love to use `write-good` as part of my workflow to flag those verbs, but looking at the current list of checks (`passive`, `weasel`, `cliches`, etc.) there's nothing that catches \"to be\" forms specifically — passive voice catches some of them but misses plenty (e.g. \"The sky is blue.\").\n\nWould you consider adding an E-Prime check to write-good?\n\nOne thing to keep in mind: most users probably *don't* write in E-Prime, and flagging every \"is/was/are\" by default would be incredibly noisy and almost certainly unwanted. So this check should be off by default, and only run when the user explicitly asks for it — both from the JS API (passing it in `opts`) and from the CLI.\n\nThe current CLI only seems to support turning checks *off* (`--no-passive`); there's no obvious way to opt *into* a check that ships disabled by default. So enabling an opt-in check from the command line would need some thought too.\n\nProgrammatic usage I'd expect to look something like:\n\n```js\nvar writeGood = require('write-good');\n// default — no E-Prime noise\nwriteGood('The sky is blue.'); // []\n\n// opt in\nwriteGood('The sky is blue.', { /* enable eprime */ });\n// -> suggestion flagging \"is\"\n```\n\nHappy to help review if someone wants to take a stab at it."} {"task_id": "format-code-task-000596", "source_id": "format-code-task-000596", "domain": "code", "task_path": "tasks/format-code-task-000596", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:db7f2fd412d739f573a9658c381f4d4feb6f179a06873f396367941be9670fc7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Feature request: support `clean` as a top-level option in `buf.gen.yaml` (v2)\n\nRight now the only way to wipe plugin output directories before code generation is the `--clean` flag on `buf generate`. There's no equivalent in the config file.\n\nIn a v2 `buf.gen.yaml` like this:\n\n```yaml\nversion: v2\nplugins:\n - local: custom-gen-go\n out: gen/go\n opt: paths=source_relative\n strategy: directory\n - protoc_builtin: java\n out: gen/java\n```\n\nI always want `gen/go` and `gen/java` to be cleared before regeneration so stale files from renamed/removed protos don't linger. Today I either have to remember to pass `--clean` on every invocation, or wrap `buf generate` in a script that does `rm -rf` first. Both are easy to forget, especially in CI / Makefile setups shared across a team.\n\nIt would be much nicer to declare this once in the config, e.g.:\n\n```yaml\nversion: v2\nclean: true\nplugins:\n - local: custom-gen-go\n out: gen/go\n ...\n```\n\nand have `buf generate` honor it the same way `--clean` does today (delete the directories / jar / zip that each plugin's `out` points at, before generation runs).\n\n### Interaction with the existing `--clean` flag\n\nThe CLI flag should still win when the user explicitly passes it, in either direction:\n\n- `clean: true` in config + `buf generate --clean=false` → don't clean\n- `clean: false` (or unset) in config + `buf generate --clean` → clean\n\nThat way the config sets the project default, and someone running locally can still override it for a single invocation without editing the file."} {"task_id": "format-code-task-000598", "source_id": "format-code-task-000598", "domain": "code", "task_path": "tasks/format-code-task-000598", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:cd33f0c609f0d1575e1c573f41b18b21766448ba8114d9b2fa1531e19b369cf8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Custom notify / sessions endpoints set via environment variables are ignored\n\nI'm running bugsnag-go against our self-hosted (on-premise) Bugsnag deployment and configuring everything through environment variables (we don't want internal URLs baked into the Go binary). My env looks like:\n\n```\nBUGSNAG_API_KEY=...\nBUGSNAG_NOTIFY_ENDPOINT=https://errors.bugsnag.internal.example\nBUGSNAG_SESSIONS_ENDPOINT=https://sessions.bugsnag.internal.example\n```\n\nand in code I just do:\n\n```go\nbugsnag.Configure(bugsnag.Configuration{\n ReleaseStage: \"production\",\n AppVersion: buildVersion,\n // intentionally NOT setting Endpoints in code -\n // I want them to come from the env vars above\n})\n```\n\nI'd expect error reports and session data to be POSTed to my internal hosts. Instead, nothing shows up on our internal Bugsnag dashboard. When I look at outbound traffic from the app, the notifier is actually calling the public SaaS endpoints (`notify.bugsnag.com` / `sessions.bugsnag.com`) — my env-var values appear to have no effect at all.\n\nAs a workaround, if I read the env vars myself and stuff them onto the `Endpoints` field of `Configuration`:\n\n```go\nbugsnag.Configure(bugsnag.Configuration{\n Endpoints: bugsnag.Endpoints{\n Notify: os.Getenv(\"BUGSNAG_NOTIFY_ENDPOINT\"),\n Sessions: os.Getenv(\"BUGSNAG_SESSIONS_ENDPOINT\"),\n },\n ReleaseStage: \"production\",\n AppVersion: buildVersion,\n})\n```\n\nthen everything works correctly — events arrive at the right place. So configuring via the `Configuration` struct directly is fine; it's specifically the \"endpoints configured *only* through environment variables\" path that doesn't take effect.\n\nSince `BUGSNAG_NOTIFY_ENDPOINT` / `BUGSNAG_SESSIONS_ENDPOINT` are documented as a supported configuration channel, I'd expect setting them in the env to behave the same as setting `Endpoints` in code — including in the common case where the app only calls `Configure` with non-endpoint options (release stage, app version, etc.) and leaves the `Endpoints` field zero-valued."} {"task_id": "format-code-task-000599", "source_id": "format-code-task-000599", "domain": "code", "task_path": "tasks/format-code-task-000599", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:994c37646758ba7e20bf2558802d607829e6d2bf8cf691490aa3c08311d51470", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Allow hooks to skip checkout by clearing `BUILDKITE_REPO`\n\n### Use case\n\nWe have a few buildkite pipelines that don't actually need source code—things like deployment jobs, maintenance tasks that hit external APIs, jobs that just orchestrate other jobs, etc. For those, doing a full git clone is pointless and just slows things down.\n\nWhat I'd like to do is decide in a hook (typically `environment`) whether the job needs source for this particular run, and if not, skip checkout entirely. The most natural way to express that is from the hook itself, e.g. something like:\n\n```bash\n# .buildkite/hooks/environment\nif [[ \"$BUILDKITE_PIPELINE_SLUG\" == \"deploy-prod\" ]]; then\n export BUILDKITE_REPO=\"\"\nfi\n```\n\n### What actually happens\n\nClearing or unsetting `BUILDKITE_REPO` from a hook doesn't skip checkout—the agent still goes into the checkout phase and tries to clone, which then fails (since there's no repo URL to clone from) and the whole job errors out before the command phase ever runs.\n\n### What I'd expect\n\nIf `BUILDKITE_REPO` ends up empty by the time the checkout phase starts, the agent should just skip checkout and continue on to run the command. The command still needs somewhere reasonable to execute from, so the working directory should be set up sensibly even when nothing was checked out (and ideally cleaned up afterwards so we don't leak directories on the agent host across many such jobs).\n\nThis would let people opt individual jobs out of checkout from a hook, without having to maintain a separate agent / queue / pipeline config for \"no-checkout\" jobs."} {"task_id": "format-code-task-000600", "source_id": "format-code-task-000600", "domain": "code", "task_path": "tasks/format-code-task-000600", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2fa2fa459165e7eb90f0d85857196f6846332ce8eb1a7fcdd505379e872709bd", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI switch between a few different builders depending on the project I'm working on, and right now there's no easy way to manage my default builder from the `pack config` command. I'd love to be able to check what my current default builder is, set a new one, and clear it out when I don't want a default anymore — all through `pack config`. Ideally when I set one it'd make sure the builder actually exists first so I don't end up with a bogus default. Also, I keep typo-ing `pack config trusted-builders` as `trust-builder`, so it'd be nice if those just worked too.\n\n## Expected Outcomes\n\n- `pack config default-builder` reports the saved default builder when one exists, and reports that none is set with guidance toward suggested builders when it does not.\n- `pack config default-builder ` only saves a builder after confirming it can be found; if it cannot be found, the command fails clearly and leaves the previous default unchanged.\n- `pack config default-builder --unset` and `pack config default-builder -u` clear a saved default builder, or report that there was nothing to clear.\n- `pack config trust-builder` and `pack config trust-builders` behave as aliases for the existing `pack config trusted-builders` command, including its subcommands.\n\n## Implementation Notes\n\nKeep the behavior observable through the CLI and persisted pack configuration. Preserve existing `pack config` behavior while adding the new default-builder management flow and trusted-builder aliases; command wiring, validation placement, helper names, and internal data organization are up to the implementation."} {"task_id": "format-code-task-000601", "source_id": "format-code-task-000601", "domain": "code", "task_path": "tasks/format-code-task-000601", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:82addb509c36014eb750472393765187bfb4b28b3929912f0608495e1f4f5864", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI'm using react-window with React.memo to wrap my row component, but my items are re-rendering on every scroll even when their data hasn't changed. Could you provide a built-in comparison helper I can drop into React.memo, and ideally something equivalent for class-based item components too, so my memoized rows actually skip these pointless re-renders?\n\n# Expected outcomes\n\n- Functional item renderers can import a top-level `areEqual` helper from `react-window` and use it as a `React.memo` comparison function.\n- The functional helper skips updates when item renderer props are unchanged in shallow terms, including the rendered positioning/style values, even if wrapper objects are recreated between renders.\n- The functional helper allows updates when any shallow style/positioning value or any other shallow prop value actually changes.\n- Class-based item renderers can import a top-level `shouldComponentUpdate` helper from `react-window` and use it as an instance update check.\n- The class helper skips updates when props are unchanged by the helper’s item-prop comparison and state is shallowly unchanged.\n- The class helper allows updates when item props differ or when state shallowly differs.\n\n# Implementation notes\n\n- The helpers should be optional utilities and should not change the default rendering behavior of existing list or grid components.\n- The exact internal organization, helper functions, file layout, and reuse between helpers are up to the implementation.\n- Comparisons should remain shallow; deep equality is not required."} {"task_id": "format-code-task-000602", "source_id": "format-code-task-000602", "domain": "code", "task_path": "tasks/format-code-task-000602", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5b073219bb1a57efd3cc226a1dc468cc788e38ee98cc915fcb29bc84db2f8ca5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: per-call field overrides for `FakeData`\n\nI'm using `faker` to populate structs in unit tests. Most of the time the default tag-based generation is exactly what I want, but I keep running into two cases where I need finer control on a **per-call** basis:\n\n1. **Some fields shouldn't be randomized at all.** Typical example is a primary-key-ish field that I want to leave at its zero value (or set myself afterwards), or a field whose zero value carries semantic meaning for the test. Today `FakeData` fills every exported field, so I end up zeroing them out again right after the call, which is noisy.\n\n2. **A specific field needs a value from a constrained set.** For example a `Status` field that must be one of a handful of valid enum strings, or an `Age` that must satisfy some test-specific invariant. The `oneof` tag covers some of this, but I don't want to bake test-only constraints into the struct tags of my production types. I'd like to supply a generator function for that one field, just for this call.\n\nWhat I'd like is something roughly like:\n\n```go\ntype User struct {\n ID int64\n Name string\n Status string\n}\n\nvar u User\nerr := faker.FakeData(&u, /* tell faker to skip ID and to use my fn for Status */)\n```\n\n…where the \"skip these fields\" list and the \"use this function for that field\" mapping are passed in at the call site, so they don't leak into the struct definition and don't affect other tests / other calls to `FakeData`.\n\n### Why not the existing extension points?\n\n- `AddProvider` / `RemoveProvider` register a provider against a **tag name**, globally. That means I'd have to (a) put a custom tag on my struct field and (b) mutate global state in tests, which is ugly and racy when tests run in parallel.\n- Struct tags like `faker:\"-\"` work, but again they live on the type, not on the call. I don't always want that field skipped — only in this particular test.\n\n### Expected behavior\n\n- A way to pass, at call time, a set of field names that `FakeData` should leave alone (no randomization, original/zero value preserved).\n- A way to pass, at call time, a mapping from field name to a user-supplied function that produces the value for that field. If the function returns an error, `FakeData` should surface it rather than silently swallowing it.\n- Existing call sites (`faker.FakeData(&x)` with no extra arguments) should keep working unchanged.\n\nHappy to help review if someone picks this up.\n\nThe shape I have in mind is something like `faker.FakeData(&u, faker.WithFieldsToIgnore(\"ID\"), faker.WithCustomFieldProvider(\"Status\", myStatusFn))`."} {"task_id": "format-code-task-000603", "source_id": "format-code-task-000603", "domain": "code", "task_path": "tasks/format-code-task-000603", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:21816500b0e2140886df4134ebdb0e7fbeb8680774a5d25a50b21635edeceed0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the zkSync and Scroll chain adapters to each expose `getTransactions(address: string): Promise` so wallet analysis can fetch a complete transaction-history array for a single address in one async call.\n\nFor zkSync, `getTransactions('0xabc')` should call `https://block-explorer-api.mainnet.zksync.io/api` with account `txlist` requests using `address: '0xabc'`, `startblock` beginning at `0`, `endblock: 99999999999999`, `sort: 'asc'`, and `offset: 1000`. If the first mocked explorer response has `status: 200`, `data.status: '1'`, and `data.result` containing one transaction object `txA`, the promise should resolve to `[txA]`. If a page contains exactly 1000 transactions, the next request should use the last transaction's `blockNumber` as the new `startblock`, and the resolved array should concatenate the pages in fetch order until a page shorter than 1000 is returned.\n\nFor Scroll, `getTransactions('0xdef')` should first page through `https://api.scrollscan.com/api` account `txlist` responses with the same `startblock`, `endblock`, `sort`, and `offset` behavior. After the normal transaction pages finish, it should also page through account `txlistinternal` responses for the same address. If the normal result is `[normalTx]` and the internal result contains one transaction with the same `hash`, the resolved array should remain `[normalTx]`. If the internal result contains a different `hash`, the resolved array should append a transaction-shaped object copied from the internal transaction for `contractAddress`, `from`, `gas`, `gasUsed`, `hash`, `input`, `isError`, `timeStamp`, `to`, `value`, and `blockNumber`, while filling unsupported normal-transaction fields with zero, empty string, or an empty `transfers` array.\n\nWhen an explorer response has `data.status: '0'`, the function should log `Error occurred while retrieving transactions:` with the explorer message and resolve with the transactions accumulated so far. When an HTTP request throws, it should log `Error occurred while making the request:` with the thrown error, stop paging, and resolve with the transactions accumulated so far."} {"task_id": "format-code-task-000604", "source_id": "format-code-task-000604", "domain": "code", "task_path": "tasks/format-code-task-000604", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fb6f69a1e254fd4ab41800b037e966fd94649ca31205c19a663e4ca48f470350", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI'm running into a weird issue with `useModel`: when I do `setState((state) => { state.count += 1 })`, the model state ends up as `undefined`, and when I do `setState({ count: 5 })` my other fields like `name` disappear. I also noticed an action middleware that calls `action()` around this setter seems to trigger my updater logic more than once. I'm hitting this while moving code from `4.0.x` to `4.1.x`, and the README doesn't really show how the old `Model`/registry/`useStore` setup maps to the new `createStore` style.\n\n## Expected Outcomes\n\n- `useModel` setters created inside `createStore` should support function updaters that mutate the provided state object without returning a value; the mutation should be reflected in the store and the state should not become `undefined`.\n- When the current model state is an object and the setter receives an object, the update should behave like a shallow partial update: changed keys are updated and unrelated existing keys remain available.\n- Action middleware that invokes the setter action should observe the completed setter result for function updaters without causing the updater callback to execute more than once for a single setter call.\n- The README FAQ should include a migration guide for moving from `4.0.x` to `4.1.x`, covering the shift from the old `Model`/registry/`useStore` pattern to the `createStore`-based pattern.\n\n## Implementation Notes\n\n- Preserve the existing public `createStore`, `useModel`, middleware, and store usage style while adjusting the observable setter behavior.\n- The exact internal state-update mechanism, validation location, and documentation wording are up to the implementation as long as the public behavior and migration guidance are clear."} {"task_id": "format-code-task-000605", "source_id": "format-code-task-000605", "domain": "code", "task_path": "tasks/format-code-task-000605", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a2068a858ad998e64c86bcf473355f4f669a82b7f4225aa617bb819353d81504", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a filesystem-backed metadata key/value store\n\nOur data-join components persist their metadata through a small key/value\ninterface (the existing etcd and MySQL clients in `fedlearner.common` are\nexamples). We now need to be able to keep that metadata on a distributed\nfilesystem (HDFS or an NFS mount) instead of a database, so deployments that\nalready have shared storage don't have to stand up a separate KV service.\n\nAdd a new client, importable as\n\n```python\nfrom fedlearner.common.dfs_client import DFSClient\n```\n\n`DFSClient(base_dir)` is constructed with a base directory under which all\nmetadata lives, and it must expose the same key/value surface the other\nclients already provide:\n\n- `set_data(key, data)` — store `data` (`bytes` or `str`) under `key`,\n creating any intermediate directories as needed. Returns a truthy value on\n success. Storing an existing key overwrites its value.\n- `get_data(key)` — return the stored value **as `bytes`**, or `None` if the\n key was never set (values stored as `str` come back as `bytes`).\n- `delete(key)` — remove the value stored at `key`. Afterwards `get_data(key)`\n is `None`.\n- `delete_prefix(key)` — recursively remove everything stored at or below\n `key`. Returns a truthy value when something was removed and a falsy value\n when there was nothing to delete.\n- `cas(key, old_data, new_data)` — atomic compare-and-swap. Compare the\n currently stored value against `old_data` **by textual content** (so a\n value previously stored as `b'x'` matches an expected `'x'` and vice\n versa); if they match, write `new_data` and return a truthy value, otherwise\n leave the value untouched and return a falsy value. Comparing against a key\n that has never been set succeeds only when `old_data` is `None`, in which\n case the key is created.\n- `get_prefix_kvs(prefix, ignore_prefix=False)` — return a list of\n `(key, value)` pairs, both `bytes`, for every key stored at or below\n `prefix`, sorted in ascending key order. The returned keys are relative to\n the client's base directory. When `ignore_prefix` is `True`, the pair for\n the exact `prefix` key itself is excluded. An unknown prefix yields an empty\n list.\n\nKeys use `/` as a separator and form a hierarchy: the **same** key may both\nhold a value and serve as the prefix of other, nested keys (e.g. `foo` can\nhold a value while `foo/a` and `foo/b` hold their own values), and\n`get_prefix_kvs('foo')` then returns the value at `foo` followed by the values\nat `foo/a` and `foo/b`.\n\nBecause the data lives on a shared filesystem under the base directory, the\nstore is not in-process state: a freshly constructed `DFSClient` pointed at the\nsame base directory must observe everything a previous client wrote there.\n\n"} {"task_id": "format-code-task-000606", "source_id": "format-code-task-000606", "domain": "code", "task_path": "tasks/format-code-task-000606", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0d5f89c4af24df16e0595d0b752716f2f88331d4836a619fdac10fb1992bc0dc", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Some paragraphs fail to render with \"unexpected type of element while serializing a paragraph\"\n\nI'm using libasciidoc to render some `.adoc` files to HTML. On certain documents (regular-looking paragraphs, nothing fancy), the conversion bails out instead of producing output. The error I get bubbles up from the substitution stage:\n\n```\nunexpected type of element while serializing a paragraph: '[]interface {}'\n```\n\nSo the serializer for a paragraph is rejecting a line because it's not a `RawLine`, but apparently the lines that make it down to `serializeParagraph` aren't always raw — sometimes a line shows up as `[]interface{}` (presumably a line that already went through some other processing step / contains nested elements). When that happens, the whole paragraph never gets reparsed, and the document errors out.\n\nThis should just work — a paragraph whose lines have been turned into nested element slices is still a valid paragraph, and the serializer should be able to flatten it back to its textual form so the normal-paragraph substitution can reparse it like any other paragraph. It shouldn't be an error condition.\n\nCould the paragraph serialization be made to handle that case (and ideally any reasonable nesting of raw lines inside it) instead of refusing?"} {"task_id": "format-code-task-000607", "source_id": "format-code-task-000607", "domain": "code", "task_path": "tasks/format-code-task-000607", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4f33d287c6284088c946f79499d022fdf89ed4ab3c583fdcfede9bc213781a36", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `devcockpit cleanup list` to be a direct CLI command that reports how much space is used by common developer cache directories under my home directory, without launching the interactive TUI. When I run `devcockpit cleanup list`, it should exit 0 on a successful scan and print a cache-size table to stdout headed `Dev Cockpit — Cache Sizes`, with `CACHE` and `SIZE` columns, one row per existing cache location, a separator line, and a final `Total` row that sums the listed sizes.\n\nOn Linux, the command should check these home-relative locations: `.cache`, `.npm/_cacache`, `.cache/yarn`, `go/pkg/mod/cache`, `.cache/pip`, `.gradle/caches`, `.m2/repository`, and `.docker`, using labels like `User Cache`, `npm Cache`, `Yarn Cache`, `Go Module Cache`, `pip Cache`, `Gradle Cache`, `Maven Cache`, and `Docker Data`. On macOS, it should check `Library/Caches/Homebrew`, `.npm/_cacache`, `Library/Caches/Yarn`, `go/pkg/mod/cache`, `Library/Caches/pip`, `Library/Caches/CocoaPods`, `Library/Developer/Xcode/DerivedData`, `.gradle/caches`, `.m2/repository`, and `Library/Containers/com.docker.docker/Data`, using labels like `Homebrew Cache`, `npm Cache`, `Yarn Cache`, `Go Module Cache`, `pip Cache`, `CocoaPods Cache`, `Xcode DerivedData`, `Gradle Cache`, `Maven Cache`, and `Docker Temp`.\n\nDirectories that do not exist should be skipped rather than printed as errors. Existing directories should be measured with `du -sk` using a short timeout; if measuring a directory fails, that row should still be included with a zero size instead of failing the whole command. Sizes should be human readable as `B`, `KB`, `MB`, or `GB` with one decimal for KB and above, and rows over 1 GB should be visually warned while rows over 5 GB should be visually critical in the terminal styling.\n\nIf the home directory cannot be determined, the command should print `Error: Cannot determine home directory` to stderr and exit 1. The global `devcockpit --help` output should list `devcockpit cleanup list` as the cache-size command and include it in the examples, and disk/storage diagnostics should suggest `devcockpit cleanup list` when low disk space or large caches are detected."} {"task_id": "format-code-task-000609", "source_id": "format-code-task-000609", "domain": "code", "task_path": "tasks/format-code-task-000609", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:30ed34ae12acafacf63956e750963607d87a4559895885891ab05772281ea73d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### What can be improved?\n\nBeing able to load tabular data is a wonderful new feature.\nBut it is currently limited by forcing the data to 'match' calliope's dimension names in certain cases. Enforcing strict naming by default makes a lot of sense: you avoid ambiguity and you also avoid the risks of relying on column position.\n\nHowever, it will often be too inflexible.\n\n## Reasoning\n\nSome of our names might lead to files being less human readable, or finicky: \n- if a model is national resolution, `nodes` is less informative than `country`\n- people will often go for `technology` or `tech` instead of `techs`, since it can be intuitive to name columns in singular\n- people might prefer `time` or `utc_timestamp`, over `timesteps`...\n\nThis will lead to a lot of 'boilerplate' code that just shapes the data to fit Calliope's naming. See the following 3 examples for different names used for timeseries in Euro Calliope with v6.10:\n\n![image](https://github.com/user-attachments/assets/d3e2f638-3a36-4005-81da-781653cd9f7e)\n![image](https://github.com/user-attachments/assets/73b95df4-59f0-42a3-b9b8-c9e8c4c64833)\n![image](https://github.com/user-attachments/assets/fed11440-0a8b-425b-88a0-345ba2a49edc)\n\nAll 3 are equally 'human' readable, but since they do not specify `timesteps`, they won't load into Calliope.\n\n# Proposal\n\nAn option to use mappings would solve this issue.\nFor example, you could load one of the timeseries above this way:\n\n```\ndata_sources:\n demand_elec_timeseries:\n source: timeseries/demand/electricity.csv\n columns: nodes\n rows: {timesteps: time}\n add_dims:\n techs: demand_elec\n parameters: sink_use_equals\n```\n\nThis is still strict, but more flexible.\n\n### Version\n\nv0.7.0.dev3\n\n---\n\nA new top-level data-source option (something like `rename_dims`) that takes a `{old_name: new_name}` mapping would be a natural fit here."} {"task_id": "format-code-task-000610", "source_id": "format-code-task-000610", "domain": "code", "task_path": "tasks/format-code-task-000610", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:315e0dd06f9f96994807519db0306097d0d3bcb2c1e91e67e17aedcc8e9eabf5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `fontVariant` style on `TextInput` is ignored\n\nI'm using `TextInput` from react-native-paper and trying to apply a `fontVariant` (e.g. `small-caps` for a username field, or `tabular-nums` for an amount field). It works fine on a plain RN `` but has no visible effect when I apply it through paper's `TextInput`.\n\n```tsx\nimport { TextInput } from 'react-native-paper';\n\n\n```\n\nThe text inside the input renders with the default glyphs as if `fontVariant` wasn't set at all. Other text style props on the same `style` object (like `fontSize` and `fontWeight`) do come through correctly, so it seems specific to `fontVariant`.\n\nThis happens with both `mode=\"flat\"` and `mode=\"outlined\"`, on both iOS and Android.\n\nIt would be great if `TextInput` respected `fontVariant` from `style` the same way it already respects the other font-related text styles."} {"task_id": "format-code-task-000611", "source_id": "format-code-task-000611", "domain": "code", "task_path": "tasks/format-code-task-000611", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ebc70fb841229c11b8b55398442e5069afd9af574869967876322961ba5cacba", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## README map-reduce example doesn't run\n\nJust trying communist out for the first time — looks great for what I want to do (crunch some numbers across a few workers). I copy-pasted the map-reduce example straight from the README:\n\n```js\nvar worker = communist(4);\n//pass it the number of map workers\nworker.data([1,2,3]);\nworker.map(function(x){return x*x;});\nworker.reduce(function(a,b){return a+b;});\nworker.data([4,5,6]);\nworker.fetch().then(function(a){console.log(a)});\n// README says this should print 91\nworker.data([6,7,8]).fetch().then(function(a){console.log(a)});\n// and this should print 240\nworker.close().then(function(a){console.log(a)});\n// and this should print 389\n```\n\nWhen I run this, it blows up on the `worker.fetch()` line — the object I get back from `communist(4)` doesn't seem to have a `fetch` (or a `close`). I only get `.data`, `.map`, `.reduce` on it. So none of the lines after the initial three setup calls actually work.\n\nIs the README out of date, or is `communist(N)` supposed to give back the object the README describes (the one with `data` / `map` / `reduce` / `fetch` / `close`, where you can keep feeding it data and ask for results whenever)? That's the workflow I'm after — incrementally pushing chunks in, occasionally fetching the running result, and closing manually when I'm done."} {"task_id": "format-code-task-000613", "source_id": "format-code-task-000613", "domain": "code", "task_path": "tasks/format-code-task-000613", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:655b2c7aa9681d1ddb7c9acea0344ccd689e9dc0fd258204c782e4bc1741b3b2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Camelot table results to be exportable from Python after `read_pdf` returns them. A single `Table` should provide `to_csv(path, **kwargs)`, `to_json(path, **kwargs)`, `to_excel(path, **kwargs)`, `to_html(path, **kwargs)`, `to_markdown(path, **kwargs)`, and `to_sqlite(path, **kwargs)`, each writing the table's `df` to the requested filesystem path and returning `None`.\n\nFor a table whose `df` contains rows `[[\"A\", \"B\"], [\"1\", \"2\"]]`, `page == 1`, and `order == 2`, `table.to_csv(\"out.csv\")` should create UTF-8 CSV without pandas index or header columns, using pandas CSV quoting mode 1 by default. `table.to_json(\"out.json\")` should write records-oriented JSON such as `[{\"0\":\"A\",\"1\":\"B\"},{\"0\":\"1\",\"1\":\"2\"}]` by default, and `mode=\"a\"` should append instead of overwriting. `table.to_excel(\"out.xlsx\")` should create a workbook sheet named `page-1-table-2` without pandas index or headers by default, while still allowing caller kwargs like `index=True` or `header=True` to override those defaults. `table.to_html(\"out.html\")` and `table.to_markdown(\"out.md\")` should write pandas-generated HTML and Markdown text using UTF-8, and both should consume a `mode` kwarg for overwrite or append. `table.to_sqlite(\"out.sqlite\")` should create or replace a SQLite table named `page-1-table-2` with no pandas index column by default.\n\nA `TableList` should provide `export(path: str, f=\"csv\", compress=False)` for exporting all contained tables at once. For `f` equal to `csv`, `html`, `json`, or `markdown`, it should write one file per table beside `path`, using the output stem and extension as a template: `tables.export(\"reports/out.csv\", f=\"csv\")` with tables from page 1/table 1 and page 2/table 1 should create `reports/out-page-1-table-1.csv` and `reports/out-page-2-table-1.csv`. For `f=\"excel\"`, it should create one workbook at the exact `path` with one sheet per table named `page--table-`. For `f=\"sqlite\"`, it should create one SQLite database at the exact `path` with one database table per Camelot table using the same `page--table-` names. When `compress=True`, CSV/HTML/JSON/Markdown exports should be generated in a temporary directory and bundled into `.zip` in the requested output directory; Excel and SQLite exports should similarly be zipped as a single workbook or database file."} {"task_id": "format-code-task-000614", "source_id": "format-code-task-000614", "domain": "code", "task_path": "tasks/format-code-task-000614", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2d3e3e6c602a2d5acb6a6e8948458f8c5905e93babab71b8ad1219ec51a788c8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `TableList.stack_contiguous(match: str = \"column_count\", keep_first_header: bool = False) -> TableList` so code that extracts one physical table per PDF page can collapse adjacent continuation tables into longer logical tables. With the default `match=\"column_count\"`, a `TableList` containing two adjacent two-column tables like `[[\"1\", \"Alice\"], [\"2\", \"Bob\"]]` and `[[\"3\", \"Charlie\"], [\"4\", \"Diana\"]]` should return a new `TableList` with one table whose dataframe has four rows and two columns in that order. If an adjacent table has a different number of columns, such as a two-column table followed by a three-column table, the result should keep them as two separate tables.\n\nI also need `match=\"first_row\"` for reports that repeat headers across page breaks: two adjacent tables with first row `[\"ID\", \"Name\"]` should stack only when the first-row values match and their column counts match. In that mode, continuation headers should be dropped by default, so `[[\"ID\", \"Name\"], [\"1\", \"Alice\"], [\"2\", \"Bob\"]]` followed by `[[\"ID\", \"Name\"], [\"3\", \"Charlie\"]]` returns one table with a single header row and three body rows. Passing `keep_first_header=True` should keep repeated header rows instead. Passing an unsupported `match` value, for example `match=\"bogus\"`, should raise `ValueError`.\n\nThe returned tables should preserve the page and order from the first table in each stacked run, average the run's `accuracy` and `whitespace` values, and leave `confidence` consistent with those averaged quality metrics. Cell and row geometry for continuation tables should be shifted so the stacked table remains coherent for downstream coordinate-based use. The method must not mutate the input `TableList` or its `Table` objects; changing a returned table must not change the input table. Calling it repeatedly with equivalent inputs should produce equivalent outputs, and it should not perform filesystem, network, or global-state side effects."} {"task_id": "format-code-task-000616", "source_id": "format-code-task-000616", "domain": "code", "task_path": "tasks/format-code-task-000616", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:cd55ba42b2ec8f0b504d93be9cee3d8a902740aa499d909b2b7f18d15d5bbf5c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Opening a broken BPMN file gives no error feedback\n\nIf I try to open a BPMN file that's malformed (e.g. it got corrupted on disk, or I hand-edited the XML and accidentally broke it), Camunda Modeler doesn't tell me anything went wrong. The tab opens, the loading spinner shows up, and then... that's it. It just sits in the loading state and never finishes — no dialog, no message, nothing to suggest the file failed to parse.\n\nFrom the user side, this is pretty confusing: I can't tell whether the modeler is just slow, frozen, or whether the file itself is the problem.\n\nI'd expect that when a diagram can't be opened, the modeler surfaces a proper error dialog so I know the file is broken, instead of leaving me staring at an indefinite loading state."} {"task_id": "format-code-task-000617", "source_id": "format-code-task-000617", "domain": "code", "task_path": "tasks/format-code-task-000617", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1d0a28a19a54ed21b967217155cd6eff2ebc313a07c2f73823dbe9ad7dccea58", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `snapcraft assemble` shows no error info when the build fails\n\nWhen I run `snapcraft assemble` on a snap that has a problem, I get the \"Snapping ...\" progress line and then the command exits with a non-zero status — but **nothing** is printed about what actually went wrong. I'm left with just an exit code and no way to diagnose the failure.\n\nSteps:\n1. Set up a `snapcraft.yaml` that will cause `snappy build` to reject the result (for example, something that violates snappy's checks on the snap dir).\n2. Run `snapcraft assemble`.\n3. Watch the \"Snapping ...\" line spin, then the command quits with a non-zero status and no further output.\n\nIf I run `snappy build ` directly on the same directory, I get a clear error message explaining what's wrong. Going through `snapcraft assemble` seems to throw that information away, which makes failures very hard to debug.\n\nI'd expect a failing assemble to surface whatever the underlying build tool said about the failure so I can fix it."} {"task_id": "format-code-task-000619", "source_id": "format-code-task-000619", "domain": "code", "task_path": "tasks/format-code-task-000619", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:06abe268ccbfae08e92c3d16c0184e94c39d42bf20c2d35a8652aaee2590920e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI'm using `osutil.EnsureFileState` in snapd to keep some files in sync with their desired state, and it works great for regular file content, but I've got a case where what I actually want to manage is a symlink pointing at a specific target. Right now there doesn't seem to be a way to express \"this path should be a symlink to X\" through EnsureFileState — I end up having to drop down and manage the symlink manually outside of it, which is annoying because I lose the same \"create if missing, fix if drifted, no-op if already correct\" behavior I get for regular files. Could we get symlinks supported as a FileState too, so I can hand EnsureFileState a target and let it do the right thing?\n\n## Expected outcomes\n\n- Symlink desired state:\n - `osutil.SymlinkFileState` should be available as a `FileState` implementation for describing a symbolic link target.\n - Calling `State()` on `osutil.SymlinkFileState{Target: target}` should represent the requested link target as a symlink state.\n\n- Ensuring symlinks:\n - `osutil.EnsureFileState` should accept a symlink `FileState` and create the requested symlink when the path is missing.\n - If the path already is a symlink to the requested target, `osutil.EnsureFileState` should report `osutil.ErrSameState`.\n - If the path exists but does not match the requested symlink target, `osutil.EnsureFileState` should update it so that it points to the requested target.\n - Symlink states should also work when used through directory-state synchronization APIs that accept `FileState` values.\n\n- Unsupported file-state kinds:\n - `osutil.EnsureFileState` should reject `FileState` implementations that report neither a regular-file state nor a symlink state, with an internal-error message indicating that the reported type is unsupported.\n - Existing built-in regular-file `FileState` implementations should reject non-regular modes with an internal-error message indicating that only regular files are supported.\n\n## Implementation notes\n\nThe concrete comparison and update strategy is up to the implementation, as long as the externally observable filesystem state and errors match the outcomes above. Keep existing regular-file behavior intact while adding symlink support and validation for unsupported state kinds."} {"task_id": "format-code-task-000620", "source_id": "format-code-task-000620", "domain": "code", "task_path": "tasks/format-code-task-000620", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fd27ace5ab77702fb1e07f6491001bd5eb76f8e06252d21949a2af21eb6314cd", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Session agent doesn't work when snapd snap is installed on core\n\nWhen the snapd snap is installed on a core system, the session agent it ships isn't picked up by the desktop session.\n\nLooking at what the snapd snap actually provides under `usr/share/applications/`, there are `.desktop` files in there (the SessionAgent one in particular is the reason I noticed this — without its desktop entry the session agent isn't launched). But on the running core system, `/var/lib/snapd/desktop/applications/` doesn't gain those entries after the snapd snap is set up, so the desktop environment never sees them.\n\nThis looks inconsistent with how other things shipped by the snapd snap are handled on core. For example the D-Bus session service activation files (also under `usr/share/dbus-1/services/` inside the snap) do get materialized into the corresponding system directory when `AddSnapdSnapServices` runs, and removed again on the undo path. The `.desktop` files coming from the snapd snap don't seem to get the same treatment — they're never written out on install, and (consequently) there's nothing cleaning them up on removal/revert either.\n\nThis only affects core; on classic the distro packaging is responsible for those files, so they're already in place there.\n\nCould the snapd-snap-on-core install path be extended so that the desktop files the snapd snap ships are deployed into the system desktop applications directory (and torn down again when the snapd snap is removed / its install is rolled back)? That would let the session agent (and anything else exposed via a desktop entry from the snapd snap) actually function."} {"task_id": "format-code-task-000621", "source_id": "format-code-task-000621", "domain": "code", "task_path": "tasks/format-code-task-000621", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9f798609e6b1910a9e0d805ad51f94a56cb8f02e8f20d92d34b165c701c86a6c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## snap-bootstrap: systemd mount helper has no way to request `noexec` / `nodev`\n\nIn preparation for tightening mount options on ubuntu core partitions\n(ubuntu-save in particular), I went looking at the systemd-mount helper\nin `cmd/snap-bootstrap` (`doSystemdMount` / `systemdMountOptions`) and\nnoticed there is no way to ask for the partition to be mounted with\n`noexec` or `nodev`.\n\nToday `systemdMountOptions` exposes `NoSuid`, `ReadOnly`, `Private`,\n`Bind`, etc., but nothing equivalent for the other two hardening\noptions that are commonly paired with `nosuid` for data partitions\nthat aren't supposed to hold executables or device nodes. As a result\ncallers from snap-bootstrap can't express \"mount this partition with\nnoexec,nodev,nosuid\" without going around this helper.\n\nIt would be good to extend the helper so callers can opt into `noexec`\nand `nodev` the same way they opt into `nosuid` today, so the actual\nmount invocation ends up passing those options through to\n`systemd-mount`."} {"task_id": "format-code-task-000622", "source_id": "format-code-task-000622", "domain": "code", "task_path": "tasks/format-code-task-000622", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:41c0de22a34d8c5e1496970845916f87a5dc5799b72803796c144e570424e38d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### snap login tests are unreliable on ppc64el / powerpc\n\nWhen building and running the `cmd/snap` test suite on ppc64el (and also powerpc), the tests covering `snap login` are flaky / fail outright. On amd64 they pass.\n\nLooking at how the login tests work, they spin up a pty via `/dev/ptmx` to feed a fake password into `terminal.ReadPassword`, because `requestLogin` reads the password directly from the terminal fd. That pty-based setup just doesn't behave consistently across architectures — what works on amd64 doesn't necessarily work on ppc64el/powerpc, and the test ends up either hanging or failing to deliver the input.\n\nIt would be much better if the login tests didn't depend on a real pty at all. The actual thing under test is the login flow logic (prompting for password, handling 2FA retry, trimming, etc.), not the terminal plumbing — so the password-reading step should be substitutable in tests without going through `/dev/ptmx`. That way the suite runs reliably on every architecture we care about.\n\nI'd expect the swappable seam to look something like a package-level `ReadPassword` variable that tests can reassign to a fake."} {"task_id": "format-code-task-000624", "source_id": "format-code-task-000624", "domain": "code", "task_path": "tasks/format-code-task-000624", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:dcfd71de34a59fbc329fce8763f8555b0fde809924fc22506b68164103f91c17", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want configured DataProfiler data labelers to expose a stateful `predict(data, batch_size=32, predict_options=None, error_on_mismatch=False, verbose=True) -> dict` method. After I create or load a structured, unstructured, regex, or column-name data labeler with a preprocessor, model, and postprocessor, calling `predict` should run the prediction pipeline on that instance and return the postprocessor's prediction dictionary.\n\nFor example, if I build a labeler with custom components where the preprocessor turns `['abc', 'def']` into one batch, the model returns `{'pred': [1, 2]}`, and the postprocessor turns that into `{'pred': ['A', 'B']}`, then `labeler.predict(['abc', 'def'])` should return `{'pred': ['A', 'B']}`. If I call `labeler.predict(['abc'], batch_size=8, predict_options={'show_confidences': True}, verbose=False)`, the preprocessor should receive the requested batch size, the model should receive `show_confidences=True` and `verbose=False`, and the returned dictionary should include the postprocessor output for those model results.\n\nThe method should accept prediction inputs as lists, NumPy arrays, pandas Series/DataFrames, and DataProfiler data objects, and should reject unsupported inputs such as a plain string with a `TypeError`. Before running the model it should validate that the preprocessor, model, and postprocessor parameters are compatible; by default mismatches should emit a runtime warning and continue, while `error_on_mismatch=True` should raise `RuntimeError` instead of producing predictions."} {"task_id": "format-code-task-000625", "source_id": "format-code-task-000625", "domain": "code", "task_path": "tasks/format-code-task-000625", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ecc7de8ea3f4a35a710c75416ab99f551404fc049ee457bb071d80174b9c5c19", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a PagerTree notification service\n\nWe'd like Apprise to be able to send alerts to [PagerTree](https://pagertree.com)\nthrough its webhook integration API. Please add a new notification service that is\nreachable with the `pagertree://` URL scheme and behaves as described below.\n\n## URL format\n\n```\npagertree://{integration_id}\npagertree://{integration_id}?action=resolve&thirdparty=abc123&urgency=high&tags=prod,server\n```\n\n* `integration_id` is the only required component. A valid integration id begins\n with `int_` followed by between 7 and 14 additional characters, each of which\n is a letter, digit, hyphen, or underscore. If the integration id is missing or\n does not satisfy this pattern, constructing the object must raise a `TypeError`.\n* `thirdparty` (optional) — an externally supplied id used to correlate alerts.\n If supplied it must be a non-empty string, otherwise a `TypeError` is raised.\n* `action` (optional) — one of `create`, `acknowledge`, or `resolve`. The default\n is `create`. Any value that is not one of these three is ignored and the default\n (`create`) is used instead.\n* `urgency` (optional) — one of `silent`, `low`, `medium`, `high`, or `critical`.\n Any other value (or omitting it) means no urgency is sent at all.\n* `tags` (optional) — a comma separated list of tags.\n\nThe service must also support three kinds of prefixed URL arguments, mirroring the\nconvention used by other webhook-style services in this project:\n\n* arguments prefixed with `+` are added as **HTTP headers** on the outgoing request,\n* arguments prefixed with `:` are merged into the top level of the **JSON payload**,\n overriding any value already there,\n* arguments prefixed with `-` are collected into a **`meta`** object inside the payload.\n\nThe object's `url()` must round-trip: re-instantiating from the value returned by\n`url()` must yield an equivalent service.\n\n## Delivery behavior\n\nA notification performs an HTTP `POST` to\n`https://api.pagertree.com/integration/{integration_id}`. The request body is JSON\nand the request always carries a `Content-Type: application/json` header (plus any\nheaders supplied with the `+` prefix, which take precedence).\n\nThe JSON body always contains:\n\n* `id` — the `thirdparty` value when one was provided, otherwise a generated,\n non-empty unique identifier (a fresh one per notification).\n* `event_type` — the resolved action (`create`, `acknowledge`, or `resolve`).\n\nWhen (and only when) the action is `create`, the body additionally contains:\n\n* `title` — the notification title, falling back to the application description\n when no title is given,\n* `description` — the notification body,\n* `meta` — a JSON object built from the `-` prefixed arguments (an empty object\n when none were given),\n* `tags` — the list of tags (an empty list when none were given),\n* `urgency` — present only when a valid urgency was provided.\n\nFor the `acknowledge` and `resolve` actions, none of `title`, `description`,\n`meta`, `tags`, or `urgency` are included — only `id`, `event_type`, and whatever\nwas merged in through `:` prefixed payload arguments.\n\nAny `:` prefixed payload arguments are applied last, so they can add to or override\nthe fields above.\n\n## Response handling\n\nHTTP responses with a 2xx status code (specifically `200`, `201`, and `202`) are\ntreated as success and the notification reports success. Any other status code\n(for example `402`, `403`, `404`, `405`, or `429`) is treated as a delivery\nfailure and the notification reports failure. A connection-level error is also a\nfailure.\n"} {"task_id": "format-code-task-000626", "source_id": "format-code-task-000626", "domain": "code", "task_path": "tasks/format-code-task-000626", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:24ffd1b1598fd03c690b25a00443f13d6c3177689771a154ba473f59123efab0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Stale binary symlinks after a package's binaries change\n\nI'm using hermit to manage tooling for my project. When a package gets\nupgraded (either via a channel update or because I bump it manually), I\nsometimes end up with broken or wrong-version commands afterwards — even\nthough `hermit` itself reports the new version is installed and ready.\n\n### What I'm seeing\n\nSteps that reproduce it for me:\n\n1. Install some package `foo` whose extracted layout puts a binary at\n e.g. `pkg/foo-1.0.0/bin/foo`. hermit creates a symlink under\n `state/binaries/foo-1.0.0/foo` pointing there. Everything works.\n2. Upgrade the package to a new version where the on-disk binary now\n lives at a different path inside the extracted package (different\n directory inside the archive, renamed, etc.). The package gets\n re-extracted under a new `pkg/...` directory.\n3. Run `foo` again.\n\nWhat I get: the command either fails with a \"no such file or directory\"\nsort of error, or runs something stale. If I `ls -l` the entries in the\n`state/binaries/...` directory, the symlinks for that package are still\npointing at the old extracted location, not the new one. The link target\nis just wrong now.\n\nIt also happens in the reverse direction — if a package adds a brand new\nbinary, or removes/renames one — the symlink folder for that package\ndoesn't reflect what the package actually ships post-upgrade. Old links\nhang around, and links that *do* exist sometimes point at paths that no\nlonger make sense.\n\n### What I'd expect\n\nAfter hermit decides a package is installed and its binaries are linked,\nthe symlinks under `state/binaries//` should be a faithful view of\nthat package's *current* binaries — every link should point at the path\nthe package actually expects right now, with nothing left over from a\nprevious install of the same package.\n\nRight now hermit seems to treat \"a symlink with the right filename\nexists\" as good enough and skips re-linking, which is what's leaving the\nstale targets in place."} {"task_id": "format-code-task-000627", "source_id": "format-code-task-000627", "domain": "code", "task_path": "tasks/format-code-task-000627", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:12d80dcb57f1a399b578d782c33635db29bb74810f5442742d21d907f7c64bf5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: a `with*`-style helper to scope DOM activities to a CSS/XPath selector\n\nI'm writing CasperJS scripts against pages that contain several visually similar \"panels\" (think modals, cards, repeated widgets). Each panel has the same internal structure — same class names, same input names, same buttons — just inside a different container. So when I do something like:\n\n```js\ncasper.then(function () {\n this.click('.submit-button');\n this.fill('form.editor', { name: 'foo' }, true);\n});\n```\n\n…the selectors are ambiguous and I have to remember to manually prefix every single one with the container, e.g. `'#panel-2 .submit-button'`, `'#panel-2 form.editor'`, and so on. It's verbose, easy to forget on one of the calls, and it makes the test code harder to read because the panel I'm \"in\" is repeated on every line instead of being said once.\n\nCasperJS already ships a really nice family of helpers for the equivalent problem on a different axis:\n\n- `withFrame(frameInfo, then)` — switches the DOM context to a given frame, runs the step, and automatically reverts when the step finishes.\n- `withPopup(popupInfo, then)` — same idea but for popups.\n\nBoth of them solve exactly the \"say it once, scope everything inside, auto-revert when done\" problem — but only for frames and popups. There doesn't seem to be an equivalent for \"scope everything inside this element on the current page\", which is by far the most common case I run into.\n\nWhat I'd like is a sibling helper in that same family that takes a CSS3 / XPath selector instead of a frame name or popup info, so I could write something like:\n\n```js\ncasper.start('http://example.com/dashboard');\n\n// everything inside this `then` is implicitly scoped to #panel-2\ncasper.('#panel-2', function () {\n this.click('.submit-button'); // resolves inside #panel-2 only\n this.fill('form.editor', { name: 'foo' }, true);\n this.test.assertExists('.success-message');\n});\n\n// back to the full document automatically\ncasper.then(function () {\n this.test.assertExists('#global-toolbar');\n});\n\ncasper.run();\n```\n\nThe behavioural contract I'm hoping for is the same as `withFrame` / `withPopup`:\n\n- the scope switch lasts only for the duration of the step passed in;\n- once that step is done, the scope is restored to whatever it was before (the document by default), so the next `casper.then(...)` is back to operating on the whole page;\n- it should accept the same kind of selector argument the rest of CasperJS already takes (CSS3 string or XPath object).\n\nThis would make tests against pages with repeated structures dramatically cleaner and remove a whole class of \"I forgot to prefix one selector\" bugs. Would you be open to adding something like this to the `with*` family?\n\nA natural name for this new helper, to stay consistent with `withFrame` / `withPopup`, would be something like `withSelectorScope(selector, then)`."} {"task_id": "format-code-task-000628", "source_id": "format-code-task-000628", "domain": "code", "task_path": "tasks/format-code-task-000628", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5a88b4407c5d2637848dd1c8e6ab44f69c5452e9684baf8cf54dc3804f58acf8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want AngularCSS to treat `media` and `breakpoint` on stylesheet declarations as responsive activation rules in the injected `$css` service. In an AngularJS app I can configure `$cssProvider.defaults.breakpoints = { mobile: '(max-width: 480px)' }`, then call `$css.add({ href: 'mobile.css', breakpoint: 'mobile' })`; the service should resolve `mobile` to that media query and not append the `` while `$window.matchMedia('(max-width: 480px)').matches` is false. When the same MediaQueryList later changes to `matches === true`, after Angular's timeout/digest the link should be present with `media=\"(max-width: 480px)\"`; when it changes back to false, that link should be removed again. Direct `media` declarations should behave the same way: `$css.add({ href: 'tablet.css', media: '(min-width: 768px)' })` with a matching query immediately appends one link, and a later non-matching event removes it. Calling `$css.remove({ href: 'tablet.css', media: '(min-width: 768px)' })` should remove the link and unregister the media-query listener so later media changes do not re-add it. For `media: 'print'`, and for browsers where `$window.matchMedia` is unavailable, `$css.add` should fall back to normal stylesheet behavior and append the link immediately instead of trying to install a responsive listener."} {"task_id": "format-code-task-000629", "source_id": "format-code-task-000629", "domain": "code", "task_path": "tasks/format-code-task-000629", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f875e48ea810e5c85822dc14df75628a4975e7f2b8238ea80b98ff911e60ad5d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `ccache` command-line tool to support a compression report command via `ccache -x` and `ccache --show-compression`. The command should read the configured local cache directory, honoring normal global options such as `--dir`, `--config-path`, and `--threads`, scan all cache subdirectories in parallel, and exit with status 0 after printing the report.\n\nFor each readable ccache result or manifest entry, it should add the file's disk usage to `Compressed data` and add the entry header's original uncompressed size, rounded to disk-block accounting, to `Original size`. Files in cache subdirectories that cannot be parsed as ccache cache entries should not make the command fail; their disk usage should be counted under `Incompressible data` instead. The command must not remove, rewrite, or update timestamps for cache contents.\n\nOn non-TTY stdout, the report should be a human-readable table with these rows and labels: `Total data:`, `Compressed data:`, ` Original size:`, ` Compression ratio:`, and `Incompressible data:`. `Total data` is compressed disk usage plus incompressible disk usage. `Compressed data` includes a parenthesized percentage of original size, `Compression ratio` prints a three-decimal ratio followed by `x` and a parenthesized space-savings percentage, and all sizes use the configured decimal or binary size unit style. For an empty cache, the command should still exit 0 and print the same table shape with zero-byte sizes and zero ratios.\n\nWhen stdout is a terminal, the command should show a `Scanning...` progress bar while it works, then separate the completed progress display from the final table with blank lines. `ccache --help` should list `-x, --show-compression` under common options as the way to show compression statistics."} {"task_id": "format-code-task-000630", "source_id": "format-code-task-000630", "domain": "code", "task_path": "tasks/format-code-task-000630", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:08d7c018e5669046cfd8cef27186b9b75aab14a9f2603b5bb70467f2e72629d3", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Bitso integration: all REST calls fail\n\nI'm trying to use ccxt to pull market data from Bitso (Python). Pretty\nstandard setup — instantiate the exchange, call `fetch_markets()` or\n`fetch_ticker('BTC/MXN')`. Every call fails immediately: I'm getting\nconnection errors / non-2xx responses back from ccxt and nothing comes\nthrough.\n\nWhen I dug into the request that ccxt is actually firing, the host it's\nhitting doesn't appear to be reachable anymore. Trying the same base\nhost straight from a browser / curl doesn't load either, so it looks like\nBitso has moved their REST API to a different URL and the one shipped\ninside ccxt is stale.\n\nCould the Bitso URLs in ccxt be updated to whatever the currently-live\nendpoint is? Same goes for the sandbox/test endpoint if that also moved\n— I haven't been able to reach the test environment either.\n\nMinimal repro:\n\n```python\nimport ccxt\nex = ccxt.bitso()\nprint(ex.fetch_markets()) # fails\n```\n\nHappy to test once it's updated."} {"task_id": "format-code-task-000631", "source_id": "format-code-task-000631", "domain": "code", "task_path": "tasks/format-code-task-000631", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3de6b06390f6027afd8ef0a4e332adbc39b477fb4e4813ba175057a266bd20ff", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `HBondAcceptorCountDescriptor` undercounts carbonyl oxygens attached to aromatic rings\n\nI'm using `HBondAcceptorCountDescriptor` to compute `nHBAcc` for a screening set and noticed it gives 0 for molecules that obviously have a hydrogen-bond-accepting carbonyl, as long as the carbonyl carbon happens to be part of an aromatic ring.\n\nMinimal example — acetophenone (`CC(=O)c1ccccc1`):\n\n```java\nIAtomContainer mol = new SmilesParser(SilentChemObjectBuilder.getInstance())\n .parseSmiles(\"CC(=O)c1ccccc1\");\n\nHBondAcceptorCountDescriptor d = new HBondAcceptorCountDescriptor();\nd.setParameters(new Object[]{ true }); // checkAromaticity\nDescriptorValue v = d.calculate(mol);\nSystem.out.println(v.getValue()); // -> 0\n```\n\nI'd expect the carbonyl oxygen to be counted, so `nHBAcc = 1`. Same story for things like benzaldehyde, methyl benzoate, or 4-pyridone — the C=O oxygen is a textbook H-bond acceptor but the descriptor returns 0.\n\nLooking at the descriptor's own javadoc, the intent for oxygen seems to be:\n\n> any oxygen with formal charge ≤ 0, **except**\n> 1. an aromatic ether oxygen (an ether oxygen adjacent to an aromatic carbon)\n> 2. an oxygen adjacent to a nitrogen\n\nSo aromatic *ether* oxygens (like the ring O in furan, or an –O– bridging into a phenyl) should be excluded — that part is fine. But a carbonyl O that just happens to be doubly bonded to an aromatic-ring carbon isn't an ether oxygen at all and shouldn't fall under that exclusion.\n\nRight now the descriptor seems to treat any O adjacent to an aromatic C the same way, so all aromatic carbonyls (ketones, aldehydes, amides, esters fused to aromatic rings, pyridone-type tautomers, …) silently get dropped from the count. Could the oxygen rule be tightened so that only the aromatic-ether case is excluded, and aromatic-ring carbonyl oxygens are counted as acceptors like the docs imply?"} {"task_id": "format-code-task-000632", "source_id": "format-code-task-000632", "domain": "code", "task_path": "tasks/format-code-task-000632", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:bce5817d70816d5970da430c32e16acf11a609d0b32669e0eecff0e87c9ff170", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n[custom] Add `width` property for `CustomMetricGroup`\n### Discussed in https://github.com/cdklabs/cdk-monitoring-constructs/discussions/450\n\n
\n\nOriginally posted by **ericxinzhang** November 2, 2023\nMy code looks as following. I can specify `height` in `CustomMonitoringProps`. I am wondering why `CustomMetricGroup` does not have a `width` property?\n\nI understand that for high level APIs (e.g. `monitor`) it's easier for the library to calculate width dynamically. However for `CustomMetricGroup` may we have a `width` property?\n\n\n```Javascript\nconst monitoring = new MonitoringFacade(this, 'MyDashboard');\n\nmonitoring.monitorCustom({\n height: 5,\n humanReadableName: 'API',\n metricGroups: [\n {\n title: 'API Errors',\n // I wish I could specify `width` here\n graphWidgetLegend: cw.LegendPosition.RIGHT,\n metrics:[\n new cloudwatch.Metric({...}),\n new cloudwatch.Metric({...}),\n ],\n },\n {\n title: 'API Latency',\n graphWidgetLegend: cw.LegendPosition.RIGHT,\n metrics:[\n new cloudwatch.Metric({...}),\n new cloudwatch.Metric({...}),\n ],\n }\n ],\n});\n```
"} {"task_id": "format-code-task-000633", "source_id": "format-code-task-000633", "domain": "code", "task_path": "tasks/format-code-task-000633", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7d13b956a7ab2d7e924a962fb4d305d8436733771659285315f30d9f81de047e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `celery.schedules.crontab.from_string(crontab: str) -> crontab` so application code can build a Celery crontab schedule from a standard five-field cron expression string. The string fields should be interpreted in this order: minute, hour, day of month, month of year, day of week, and the returned object should be equal to constructing `crontab` with those same fields in Celery's normal constructor order.\n\nFor example, `crontab.from_string('* * * * *')` should return the same schedule as `crontab()`. `crontab.from_string('* * * * SUN')` should return the same schedule as `crontab(day_of_week='SUN')`. `crontab.from_string('0 8 5 * *')` should return the same schedule as `crontab(minute='0', hour='8', day_of_month='5')`.\n\nIf the input does not contain exactly five space-separated fields, such as `crontab.from_string('*')`, it should raise `ValueError`. Calling it repeatedly with the same string should produce equal schedule objects, and it should not perform filesystem or network side effects."} {"task_id": "format-code-task-000634", "source_id": "format-code-task-000634", "domain": "code", "task_path": "tasks/format-code-task-000634", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:771eaa2125b009adf2aaa97bb014816045010f5548c11a07d763da96426fc2a1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Celery task result handles to support dependency-tree inspection from an `AsyncResult` instance. Given `result = app.AsyncResult(task_id)`, `result.iterdeps(intermediate=False)` should walk the result tree breadth-first and yield `(parent_result, child_result)` pairs beginning with `(None, result)`, following each ready result's `children`. If it encounters a result that is not ready and `intermediate` is false, it should raise `celery.exceptions.IncompleteStream`; with `intermediate=True`, it should yield the incomplete node and stop descending through that branch instead of raising.\n\nI also need `result.collect(intermediate=False, **get_options)` to iterate over that same dependency tree and yield `(result_handle, value)` tuples, where each value is obtained by calling `get(**get_options)` on the yielded handle. For example, if a root result has three ready child results with values `0`, `1`, and `2`, `list(root.collect())` should include the root's value followed by the child values in breadth-first tree order. `result.get_leaf()` should return the value from the last result visited by `iterdeps()`, so a simple chain ending in a task returning `2` returns `2`.\n\nFinally, I want graph support for the same dependency tree: `result.build_graph(intermediate=False, formatter=None)` should return a `DependencyGraph` containing an arc for every visited result and an edge from each parent to each child, using a default graph formatter rooted at the task id when no formatter is provided. The `result.graph` property should expose the default built graph and cache it on the result instance."} {"task_id": "format-code-task-000635", "source_id": "format-code-task-000635", "domain": "code", "task_path": "tasks/format-code-task-000635", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:46cfd3b86af6e4b869f75bfc15351c595b41c039a54596898798af165a1d5b85", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Celery canvas to support chunked task execution through a `chunks(task, it, n, **options)` signature and the convenience method `Task.chunks(it, n)`. A user should be able to write `sig = add.chunks(zip(range(4), range(4)), 2)`, then call `sig.group()` and get a group made from two `starmap`-style task signatures: one for `[(0, 0), (1, 1)]` and one for `[(2, 2), (3, 3)]`.\n\nThe chunked signature should also be executable: `add.chunks(zip(range(4), range(4)), 2)()` should apply the generated group immediately and return the group result, while `.apply_async(countdown=10)` should publish the generated group with the supplied execution options and route it using the underlying task name. Serialized chunk signatures should round-trip with `chunks.from_dict(...)` using the stored task, iterable, chunk size, and options.\n\nWorkers also need a built-in `celery.chunks` task that accepts `(task, it, n)`, creates the same chunked signature in the current app, executes it, and returns its result. The top-level `from celery import chunks` API and the interactive `celery shell` locals should expose this primitive alongside `group`, `chain`, `xmap`, and `xstarmap`."} {"task_id": "format-code-task-000636", "source_id": "format-code-task-000636", "domain": "code", "task_path": "tasks/format-code-task-000636", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:57480741f65491f78b93acc71cbe8bf8b62a6e9e133ce87dc51384329933b3a8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nSSLTransport error: unexpected keyword argument 'ca_certs'\nUsing amqp version 5.0.1, I discovered an SSLTransport error. When I use amqp version 5.0.0b1, SSL works just fine.\n\nHere's the error I get when using amqp==5.0.1:\n\n`/usr/local/lib/python3.8/dist-packages/kombu/connection.py:283: in channel\n chan = self.transport.create_channel(self.connection)\n/usr/local/lib/python3.8/dist-packages/kombu/connection.py:858: in connection\n return self._ensure_connection(\n/usr/local/lib/python3.8/dist-packages/kombu/connection.py:435: in _ensure_connection\n return retry_over_time(\n/usr/local/lib/python3.8/dist-packages/kombu/utils/functional.py:325: in retry_over_time\n return fun(*args, **kwargs)\n/usr/local/lib/python3.8/dist-packages/kombu/connection.py:866: in _connection_factory\n self._connection = self._establish_connection()\n/usr/local/lib/python3.8/dist-packages/kombu/connection.py:801: in _establish_connection\n conn = self.transport.establish_connection()\n/usr/local/lib/python3.8/dist-packages/kombu/transport/pyamqp.py:128: in establish_connection\n conn.connect()\n/usr/local/lib/python3.8/dist-packages/amqp/connection.py:314: in connect\n self.transport.connect()\n/usr/local/lib/python3.8/dist-packages/amqp/transport.py:77: in connect\n self._init_socket(\n/usr/local/lib/python3.8/dist-packages/amqp/transport.py:188: in _init_socket\n self._setup_transport()\n/usr/local/lib/python3.8/dist-packages/amqp/transport.py:323: in _setup_transport\n self.sock = self._wrap_socket(self.sock, **self.sslopts)\n_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ \n\nself = \nsock = \ncontext = None\nsslopts = {'ca_certs': '/home/cameron/Defense/tls-gen/basic/result/ca_certificate.pem', 'cert_reqs': return self._wrap_socket_sni(sock, **sslopts)\nE TypeError: _wrap_socket_sni() got an unexpected keyword argument 'ca_certs'\n\n/usr/local/lib/python3.8/dist-packages/amqp/transport.py:330: TypeError\n----------------------------------------------------------- Captured stdout call -----------------------------------------------------------\n2020-10-28 16:46:54:\tINFO:\tConsumer successfully connected to kombu server at localhost:5671\n2020-10-28 16:46:54:\tERROR:\tFailed to initialize rabbit consumer connection:\nTraceback (most recent call last):\nTypeError: _wrap_socket_sni() got an unexpected keyword argument 'ca_certs'`\n\nLooking at `py-amqp/amqp/transport.py`, the parameters for the method `_wrap_socket_sni` for amqp==5.0.1, it doesn't include `ca_certs`. On amqp==5.0.0b1, the method _wrap_socket_sni` does include `ca_certs` as a parameter. Is this a bug or was this left out by accident?\n\nHere are the links to amqp version 5.0.1 and 5.0.0b1 of `py-amqp/amqp/transport.py` with the line numbers highlighting the method:\namqp==5.0.1: https://github.com/celery/py-amqp/blob/93e4f3a2990f2ed1a6da861c99c7f0a3b0d32160/amqp/transport.py#L337\namqp==5.0.0b1: https://github.com/celery/py-amqp/blob/c5fe7daaf379cfbcccbe81fcd1ea12807274f8fb/amqp/transport.py#L339"} {"task_id": "format-code-task-000638", "source_id": "format-code-task-000638", "domain": "code", "task_path": "tasks/format-code-task-000638", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b52b55443b48a16b5c03e15aaf4ce77632875da1d022903264573a7304a5bf18", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI'm porting the Election contract tests over to Foundry, and I keep needing a few test helpers that don't exist yet in our Solidity test setup. Could we get some shared utilities for things like checking two values are approximately equal (within some margin), comparing whether two address arrays contain the same set, and generating a pseudo-random number in a range for test inputs? Also, I'm trying to read the `DAY`/`WEEK`/`YEAR` constants from a test contract but I can't get at them externally right now — it'd be great if those were reachable from outside.\n\n# Expected outcomes\n\n- Public test constants:\n - `Constants.DAY()`, `Constants.WEEK()`, and `Constants.YEAR()` are readable from outside the contract through the usual Solidity public constant getters.\n - The durations represented by those constants remain unchanged: one day, one week, and one 365-day year respectively.\n\n- Approximate equality helper:\n - `Utils.assertAlmostEqual(uint256 actual, uint256 expected, uint256 margin)` succeeds when `actual` and `expected` differ by no more than `margin`.\n - It fails when the absolute difference is greater than `margin`, and the failure message reports the difference using the `\"Difference is \"` prefix.\n\n- Address-array set comparison helper:\n - `Utils.arraysEqual(address[] arr1, address[] arr2) -> bool` returns whether the two address arrays represent the same address set.\n - Arrays of different lengths are not considered equal.\n - For arrays of the same length, ordering does not affect equality.\n\n- Pseudo-random test input helper:\n - `Utils.generatePRN(uint256 min, uint256 max, uint256 salt) -> uint256` returns a pseudo-random number suitable for tests.\n - For valid ranges, the returned value is always within the inclusive range `[min, max]`.\n - Across a non-degenerate range, varying the salt should be able to produce varied outputs rather than a fixed boundary or constant value.\n\n# Implementation notes\n\n- These utilities are intended for the Solidity test support layer, not for production contract behavior.\n- The concrete data structures, entropy sources, and validation placement are implementation choices, as long as the public helper behavior above is satisfied.\n- Existing test utilities should continue to work while the new helpers and externally readable constants are added."} {"task_id": "format-code-task-000639", "source_id": "format-code-task-000639", "domain": "code", "task_path": "tasks/format-code-task-000639", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:47795eefd4733740b5b10a3aa3e16d3ca2f40d16a19fde33348a0e9b9461cf42", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a machine-readable JSON view to the zpages trace endpoints\n\nThe zpages exporter serves two debugging pages over HTTP: `/tracez` (a summary of\ncollected spans, grouped by span name) and `/traceconfigz` (the current trace\nsampling configuration). Today both endpoints only render HTML, which makes them\nhard to consume programmatically (for example, from tests or external tooling).\n\nAdd an opt-in JSON view to both endpoints, selected with the query parameter\n`json=1`. When that parameter is present the endpoint must respond with the same\ninformation it would have rendered as HTML, but serialized as a JSON document\n(HTTP 200). When the parameter is absent (or anything other than `1`), the\nendpoints keep serving their existing HTML pages unchanged.\n\n## `/tracez?json=1`\n\nResponds with a JSON object describing the per-span-name summary. It has two\nfields:\n\n- `spanCells`: an array with one entry per span name currently known to the\n exporter, **ordered ascending by name**. Each entry reports the aggregate\n counts for that name and contains at least:\n - `name`: the span name.\n - `RUNNING`: how many spans with that name have started but not yet ended.\n - `ERRORS`: how many spans with that name finished with a non-OK status (a\n status code other than `0`).\n - `latencies`: an object that buckets the *completed, OK* spans by their\n duration. It has one numeric counter per latency bucket, keyed by these\n exact bucket names: `ZERO_MICROSx10`, `MICROSx10_MICROSx100`,\n `MICROSx100_MILLIx1`, `MILLIx1_MILLIx10`, `MILLIx10_MILLIx100`,\n `MILLIx100_SECONDx1`, `SECONDx1_SECONDx10`, `SECONDx10_SECONDx100`,\n `SECONDx100_MAX`. Every completed OK span lands in exactly one bucket, so the\n sum of the counters equals the number of completed OK spans for that name.\n\n- `selectedTraces`: details about an individual span name selected with the\n `tracename` query parameter, filtered by the `type` query parameter (the same\n parameters the HTML page already understands; `type` may be `RUNNING`,\n `ERRORS`, or one of the latency bucket names). When `tracename` is **not**\n provided this field is `null`. When it is provided, the field is an object\n with:\n - `name`: the requested span name.\n - `traces`: an array of the matching spans. Note that spans contain circular\n references, so they cannot be serialized directly — each entry must be a\n plain, JSON-safe representation that still carries the span's identifying\n information, including its `traceId` and span `id`.\n\n## `/traceconfigz?json=1`\n\nResponds with a JSON object describing the sampling configuration:\n\n- `samplingProbability`: the sampling probability currently in effect on the\n tracer (e.g. `1` when sampling is always on, `0` when it is never on).\n- `defaultConfig`: an object whose `samplingRate` is a number.\n\nThe HTML responses, and any existing behavior of these endpoints when `json=1`\nis not supplied, must continue to work exactly as before.\n"} {"task_id": "format-code-task-000640", "source_id": "format-code-task-000640", "domain": "code", "task_path": "tasks/format-code-task-000640", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:127245b11fb8d15a75600c68ababd60247971e4eb32e395075728bb7a08cdddc", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Allow per-client ping/pong configuration\n\nRight now the application-level ping/pong settings (ping interval and pong timeout) can only be configured at the transport / handler level — every client served by the same handler ends up with identical ping behavior.\n\nIn our setup we'd like to vary this per connection. For example:\n\n- Mobile clients on flaky networks: we want a more aggressive ping interval to detect dead connections faster.\n- Long-lived desktop clients on stable networks: we'd prefer a much longer ping interval to save bandwidth/battery on the server side.\n\nWe have all the information needed to make that decision inside `OnConnecting` (we look at the token claims / client name / version), so it would be very natural to make the decision there and return it as part of the `ConnectReply`. Today there is no way to do that — once the client is created it just picks up whatever the transport reports, and we'd have to register a separate handler/endpoint per client class just to get a different ping interval, which is awkward.\n\nWould it be possible to let `OnConnecting` optionally override the ping/pong settings for the connection being established? If nothing is returned, the existing transport-level default should still apply, so existing code keeps working.\n\nOne small side note while you're in that area: the transport interface currently exposes this as `AppLevelPing()` returning an `AppLevelPing` struct, while the handler configs already embed a `PingPongConfig` for the same concept. Having two names for the same thing is a bit confusing when reading the code — it would be nice if these were unified."} {"task_id": "format-code-task-000641", "source_id": "format-code-task-000641", "domain": "code", "task_path": "tasks/format-code-task-000641", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e916c0668386c7f393f665e9f01f2fdd714debdb61801b590f10dbfba372eea4", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Make the Redis engine's PUB/SUB subscriptions concurrency-safe\n\nThe Redis engine drives all of its channel fan-out through a single Redis PUB/SUB\nconnection. Under load this falls apart: many client connections call into the\nengine's `subscribe` / `unsubscribe` paths from different goroutines at the same\ntime, and they all poke that one shared PUB/SUB connection directly. The result is\na corrupted protocol stream — most of the calls come back with errors (you'll see\nthings like `redigo: connection closed`) and the channels never actually end up\nsubscribed, so messages published to them are silently dropped.\n\nFix the engine so that subscribing and unsubscribing are safe to call\nconcurrently from any number of goroutines, while keeping the existing\nsingle-threaded behavior intact.\n\nRequired observable behavior:\n\n- `subscribe(chID)` and `unsubscribe(chID)` may be called concurrently from many\n goroutines, including for different channels at the same time and for the same\n channel from several goroutines. Each call must return only after its\n subscription change has been applied to Redis, returning a `nil` error on\n success.\n- After a set of `subscribe` calls return successfully, every one of those\n channels must really be subscribed on the engine's PUB/SUB connection: a\n `publish` to such a channel must report that it has active subscribers.\n- After a successful `unsubscribe`, the channel must no longer be subscribed: a\n `publish` to it must report no active subscribers.\n- Subscribing the same channel is idempotent — a single `unsubscribe` for a\n channel returns it to the unsubscribed state regardless of how many times it was\n subscribed.\n- The engine must continue to receive and dispatch messages published to\n subscribed channels exactly as before, and all of the engine's other operations\n (publish, presence, history, the API listener) keep working unchanged.\n\nThe engine is started via its `run` method before any subscribe/unsubscribe calls\nare made.\n"} {"task_id": "format-code-task-000642", "source_id": "format-code-task-000642", "domain": "code", "task_path": "tasks/format-code-task-000642", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e3f71ec1f49e4b741a8e0efc3d5db42d747351f93e0260f84029f6a4f708cf42", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `metadata` KMS reads the user passphrase secret too eagerly at init time\n\nWe're using ceph-csi RBD with at-rest encryption, KMS type `metadata`, and we point it at a per-tenant K8s Secret via `secretName` / `secretNamespace` in the StorageClass (the `userSecret` style of configuration, where `encryptionPassphrase` lives in a Kubernetes Secret in the tenant's namespace).\n\nTwo problems with how this currently behaves:\n\n1. **KMS init fails hard if the user secret isn't reachable at that moment.**\n When the driver initializes the `metadata` KMS, it immediately goes to the K8s API to read the tenant's `encryptionPassphrase` secret. If that secret isn't there yet (e.g. the tenant namespace / secret is being provisioned around the same time as the workload, or the API server is briefly unreachable), the whole KMS initialization errors out and we can't proceed with the volume operation at all. The provider doesn't actually need the passphrase yet at init time — it only needs it later when a DEK has to be encrypted or decrypted — so failing this early feels too strict.\n\n2. **Updates to the passphrase in the K8s Secret don't take effect.**\n Once the KMS is initialized, the passphrase value seems to be held onto by the driver. If we rotate / change `encryptionPassphrase` in the underlying K8s Secret afterwards, ceph-csi keeps using the old value until we restart the driver pods. That's not great operationally — we'd expect the driver to always pick up the current value of the configured Secret when it actually needs to encrypt/decrypt.\n\nWhat we'd like is for the `metadata` KMS to defer reading the passphrase Secret until it actually needs it (i.e. at the point of encrypting/decrypting a DEK or fetching the secret for a volume), instead of doing it once during initialization and caching the result. That way, transient unavailability of the Secret at init time doesn't break things, and a later update to the Secret is picked up without a driver restart.\n\nThe same fallback path that's there today should still work — if `userSecret` (`secretName` / `secretNamespace`) isn't configured, fall back to `encryptionPassphrase` in the StorageClass secrets like before."} {"task_id": "format-code-task-000643", "source_id": "format-code-task-000643", "domain": "code", "task_path": "tasks/format-code-task-000643", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3a78c604c41a5cbf4b9dfaaafa130abe67cd8a026d46af08c81e3f1cadb14cb3", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add explicit `sequence` and `parallel` flow primitives\n\nRight now the only way to express concurrency in a function tree is by nesting a\nplain array inside the tree — an \"array inside an array\" means \"run these in\nparallel\". This is implicit and surprising: people import chains from other\nfiles and accidentally change execution semantics, and there is no way to label\na group of steps.\n\nI'd like to introduce two first-class primitives for describing execution flow\nand make them part of the package's public API (named exports alongside the\nexisting default factory):\n\n- `sequence(...)` — runs its items one after the other.\n- `parallel(...)` — runs its items at the same time.\n\nBoth should accept an **optional leading string** used as a human-friendly name\nfor the group, followed by any number of items. An item may be a function, a\nplain array, or another `sequence`/`parallel` value. The name is purely a label\n— it must never be executed as a step.\n\nThe runnable tree you hand to `execute(...)` should now accept a value produced\nby these primitives just like it already accepts a plain array.\n\nBehavior that must hold:\n\n- **Sequence is sequential.** `sequence(a, b)` runs `a` then `b`, threading the\n payload through and merging each step's returned object into the payload, the\n same way a plain top-level array already behaves. A plain array must keep\n working and must keep meaning \"a sequence\".\n\n- **Parallel is concurrent.** `parallel(a, b)` runs its items concurrently. The\n `parallelStart` event must report the number of items being run as the number\n of branches, and the final payload must be the merge of every branch's\n resulting payload.\n\n- **A group is one branch.** When a parallel's item is a plain array (or a\n `sequence`), that group is treated as a *single* parallel branch and runs as\n one unit: its inner functions execute sequentially (a later step only runs\n after the earlier one has resolved), and the group counts as exactly one\n branch in the parallel — not one branch per inner function.\n\n- **Nesting an array no longer means parallel.** A plain array nested inside\n another array (or sequence) is just a nested sequence and runs sequentially;\n it must not trigger parallel execution.\n\n- **Parallel flag.** The per-function details already emitted on the function\n lifecycle events expose an `isParallel` indicator. It must be `true` only for\n functions that are *direct* branches of a `parallel`. Functions that live\n inside a grouped sequence — even one nested within a `parallel` — are not\n themselves parallel branches and must report `false`.\n\nKeep everything else (paths/outputs, providers, payload merging, existing\nevents) working as before.\n"} {"task_id": "format-code-task-000644", "source_id": "format-code-task-000644", "domain": "code", "task_path": "tasks/format-code-task-000644", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3940638a1f030f31a88f4121a74085a1c35b10cfc7868ae9fc27d301ea8e347e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the Workflow DSL to support Argo artifact wiring through decorators on workflow task and template methods. Users should be able to import `artifact` from `argo.workflows.dsl.tasks`, and use `inputs.artifact` and `outputs.artifact` from `argo.workflows.dsl.templates`; each decorator should accept the same artifact keyword fields users pass to `V1alpha1Artifact`, including at least `name`, `path`, and `_from`.\n\nFor a workflow class, a DAG task decorated with `@artifact(name=\"message\", _from=\"{{tasks.generate-artifact.outputs.artifacts.hello-art}}\")` should receive a `V1alpha1Artifact` argument named `message` when `Workflow.compile()` runs, so the task method can call a template like `return self.print_message(message=message)`. If an artifact name contains separators such as `input-file`, the injected Python keyword should use the snake_case form, such as `input_file`.\n\nThe compiled manifest from `wf.to_dict()` should include task artifact arguments under the DAG task's `arguments.artifacts` list. In a concrete session, after `wf = ArtifactPassing()` compiles a class with a `consume_artifact` task decorated as above, the `consume-artifact` DAG task should contain one artifact argument with `name == \"message\"` and `_from == \"{{tasks.generate-artifact.outputs.artifacts.hello-art}}\"`.\n\nTemplate artifact decorators should advertise file inputs and outputs in the generated template definitions. A template decorated with `@outputs.artifact(name=\"hello-art\", path=\"/tmp/hello_world.txt\")` should compile to `outputs.artifacts` containing that artifact, and a template decorated with `@inputs.artifact(name=\"message\", path=\"/tmp/message\")` should compile to `inputs.artifacts` containing that artifact. Stacking multiple artifact decorators on the same task or template should preserve all of them in the resulting artifact list."} {"task_id": "format-code-task-000645", "source_id": "format-code-task-000645", "domain": "code", "task_path": "tasks/format-code-task-000645", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fb02dde8faa4ecdd53ab86d8a5ff9660ca21117931b50398b6dd52c07ee15b38", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Bot gets stuck retrying messages that violate the harmonization\n\nI have a parser bot in my pipeline that occasionally receives messages where one of the fields contains a value that doesn't match the harmonization rules (e.g. something that should be an IP address but isn't, or an enum field with a value that's not in the allowed list).\n\nWhen that happens, the bot doesn't move on. It logs the harmonization error, then retries the same message, logs the same error again, retries again, and so on — the pipeline effectively halts on that one bad message and just keeps spinning on it. The only way out is to stop the bot and manually remove the message from the queue.\n\nThat's not what I'd expect: if the value in the message is invalid according to the harmonization, retrying isn't going to make it valid. The bot should give up on that message immediately, dump it somewhere so I can look at it later, and continue processing the next message in the queue — same as it would for other \"this message is broken, no point retrying\" situations.\n\nCould the harmonization-violation case be treated as a non-retryable error, with the offending message dumped for inspection instead of being retried indefinitely?"} {"task_id": "format-code-task-000646", "source_id": "format-code-task-000646", "domain": "code", "task_path": "tasks/format-code-task-000646", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:98bfeb663013bef6113ff5e6e7d8dc537d0bb1c49a9b1b8819365effec436102", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Console errors when scrolling through certain iRegs interpretations pages\n\nI was browsing the interpretations pages on the regulations site and noticed that the dev tools console fills up with errors as soon as I start scrolling. The wayfinder component (the little floating thing at the top that shows the current section/paragraph) seems to be involved.\n\n### Steps to reproduce\n\n1. Visit `https://www.consumerfinance.gov/policy-compliance/rulemaking/regulations/1002/Interp-9/` (or run it locally).\n2. Open the dev tools console.\n3. Scroll down through the page.\n\n### What I see\n\nThe console starts spitting out errors like:\n\n```\nUncaught TypeError: Cannot read property 'split' of undefined\n```\n\nThey keep firing as I scroll past the early paragraphs on the page. Other interpretations pages I've poked at seem fine, but this one (and I suspect a few others) reliably blows up.\n\n### What I expect\n\nScrolling through any interpretations page should not throw errors in the console. The wayfinder should keep working — or at minimum fail gracefully — even on pages whose content has unusual paragraph structure."} {"task_id": "format-code-task-000647", "source_id": "format-code-task-000647", "domain": "code", "task_path": "tasks/format-code-task-000647", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fa1eabf23e5b2fb2b67636af2be441a6a3a1f08d236b1cb0ff2f9ab097829075", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n**Bug: \"Estimated years attending\" dropdown doesn't update the view**\n\nIn the Paying for College disclosures tool, changing the \"Estimated years attending\" dropdown does nothing — the rest of the disclosure view doesn't react to the new selection. The sections of the page that should reflect a different program length stay exactly as they were.\n\nIf I open the browser dev tools and watch the console while changing the dropdown, an error is logged each time. So it looks like the change handler is blowing up rather than just silently doing nothing.\n\n**Steps to reproduce**\n1. Open the Paying for College disclosures tool.\n2. Open the browser developer console.\n3. Change the \"Estimated years attending\" dropdown to a different value.\n\n**Expected**\nThe view re-renders to reflect the newly selected program length.\n\n**Actual**\nNothing visible changes, and an error shows up in the console."} {"task_id": "format-code-task-000651", "source_id": "format-code-task-000651", "domain": "code", "task_path": "tasks/format-code-task-000651", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:65a455484d275ff35b7fd4f5fdaf9e9341c33935f83abcdb43a214c8cbbacda3", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n`Variable.__repr__` raises error when underlying data is not initialized\nIn v2, the following code raises an error.\n\n```\nIn [1]: import chainer\nIn [2]: a = chainer.Variable()\nIn [4]: a.data is None\nOut[4]: True\nIn [5]: repr(a)\n---------------------------------------------------------------------------\nAttributeError Traceback (most recent call last)\n in ()\n----> 1 repr(a)\n\n/Users/oonokenta/dev/chainer/chainer/variable.py in __repr__(self)\n 305 \n 306 def __repr__(self):\n--> 307 return variable_repr(self)\n 308 \n 309 def __str__(self):\n\n/Users/oonokenta/dev/chainer/chainer/variable.py in variable_repr(var)\n 75 prefix = 'variable'\n 76 \n---> 77 if arr.size > 0 or arr.shape == (0,):\n 78 lst = numpy.array2string(arr, None, None, None, ', ', prefix + '(')\n 79 else: # show zero-length shape unless it is (0,)\n\nAttributeError: 'NoneType' object has no attribute 'size'\n```\n\nIt is because `variable_repr(var)` assumes that `var.data` is not `None`."} {"task_id": "format-code-task-000652", "source_id": "format-code-task-000652", "domain": "code", "task_path": "tasks/format-code-task-000652", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:98683347f9717caae1ec75549f83d753004b86ed5884e363bcd2fc2477c1bdeb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nZeroDivisionError in softmax_cross_entropy when I pass masked array\n- Chainer version: `origin/master`\n - CuPy version: `2.0.0b1`\n - OS/Platform: `Ubuntu 14.04`\n - CUDA/cuDNN version: `CUDA-8.0.61-1/cudnn5.1.5-1`\n\n## Problem\n\nWhen I mask two arrays and calculate `softmax_cross_entroy`, `ZeroDivisionError` occurs when masked array's data is `[]`\nI thought it should be checked in ndim and shape check at the beginning of the function, but the array's shape was something like `(0, 2, 2)`\n\n## Reproduction\n\n```python\nimport chainer\nimport cupy\nimport numpy as np\n\n\ndata = np.random.random((10, 2, 2))\nt = cupy.ones((data.shape[0], data.shape[2]), dtype=cupy.int32)\nx = chainer.Variable(data)\nx.to_gpu()\nmask = t[:, 0] == 0\nx_masked = x[mask]\nprint('=================')\nprint('x_masked')\nprint(' data: {}'.format(x_masked.data))\nprint(' shape: {}'.format(x_masked.shape))\nprint('=================')\nt_masked = t[mask]\nprint('=================')\nprint('t_masked')\nprint(' data: {}'.format(t_masked))\nprint(' shape: {}'.format(t_masked.shape))\nprint('=================')\nchainer.functions.softmax_cross_entropy(x_masked, t_masked)\n```\n* Shape\n```\n=================\nx_masked\n data: []\n shape: (0, 2, 2)\n=================\n=================\nt_masked\n data: []\n shape: (0, 2)\n=================\n```\n* Error\n```bash\nTraceback (most recent call last):\n File \"spam.py\", line 25, in \n chainer.functions.softmax_cross_entropy(x_masked, t_masked)\n File \"/home/shingo/chainer/chainer/functions/loss/softmax_cross_entropy.py\", line 275, in softmax_cross_entropy\n normalize, cache_score, class_weight, ignore_label, reduce)(x, t)\n File \"/home/shingo/chainer/chainer/function.py\", line 212, in __call__\n ret = node.apply(inputs)\n File \"/home/shingo/chainer/chainer/function_node.py\", line 219, in apply\n outputs = self.forward(in_data)\n File \"/home/shingo/chainer/chainer/function.py\", line 117, in forward\n return self._function.forward(inputs)\n File \"/home/shingo/chainer/chainer/function.py\", line 319, in forward\n return self.forward_gpu(inputs)\n File \"/home/shingo/chainer/chainer/functions/loss/softmax_cross_entropy.py\", line 102, in forward_gpu\n log_y = log_softmax._log_softmax(x)\n File \"/home/shingo/chainer/chainer/functions/activation/log_softmax.py\", line 36, in _log_softmax\n x.reshape(x.shape[:2] + (-1, 1)))\n File \"cupy/core/core.pyx\", line 461, in cupy.core.core.ndarray.reshape (cupy/core/core.cpp:11315)\n File \"cupy/core/core.pyx\", line 434, in cupy.core.core.ndarray._reshape (cupy/core/core.cpp:10994)\n File \"cupy/core/internal.pyx\", line 139, in cupy.core.internal.infer_unknown_dimension (cupy/core/internal.cpp:3060)\nZeroDivisionError: integer division or modulo by zero\n```"} {"task_id": "format-code-task-000654", "source_id": "format-code-task-000654", "domain": "code", "task_path": "tasks/format-code-task-000654", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:41113b6d2b5bec8ad147d172362764f88705f6b48e78b285b04a9adcbd0f72c7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI'm hitting an issue with `chainercv.transforms.resize` on CHW grayscale images: when I pass a NumPy array shaped like `(1, H, W)`, the result seems to get squeezed into 2D or sometimes errors out. In my no-`cv2` setup I also noticed non-RGB inputs behave oddly, like a 1-channel image failing and extra channels disappearing.\n\n## Expected outcomes\n\n- Grayscale CHW inputs passed to `chainercv.transforms.resize` should resize successfully and preserve their single channel dimension, returning an array shaped `(1, new_H, new_W)`.\n- When `chainercv.transforms.resize` runs without `cv2` available, non-RGB CHW inputs should resize according to their actual number of channels rather than assuming exactly three channels.\n- The resized output should preserve the input channel count for both fewer-than-three-channel and more-than-three-channel inputs in the no-`cv2` fallback path.\n- Resizing should operate on the image data in the preserved channels rather than replacing valid non-constant input with a constant placeholder.\n- Existing RGB CHW resize behavior should continue to return the expected channel-first shape.\n\n## Implementation notes\n\n- The implementation may choose any appropriate internal handling for channel order, backend-specific image resizing behavior, and per-channel processing, as long as the public resize behavior above is satisfied.\n- Keep the public API behavior centered on `chainercv.transforms.resize`; no particular private helper structure or internal code organization is required."} {"task_id": "format-code-task-000655", "source_id": "format-code-task-000655", "domain": "code", "task_path": "tasks/format-code-task-000655", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:73f5c9fc3bf815afa9a8c353a73eae8c62a3c9224bcab3e5d5076348df65c369", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI'm seeing weird samples when I train DQN/DDPG with batched environments and replay: n-step sequences and episodic replay entries sometimes contain transitions from different parallel envs stitched together. It also looks like when one batch item hits done/reset, the replay state for the other still-running envs gets disturbed. This shows up with interleaved env steps where sampled trajectories have state/action markers from multiple envs.\n\n## Expected outcomes\n\n- Sampled n-step replay sequences never combine transitions from different parallel environment sources, including when environment steps are interleaved.\n- Calling `stop_current_episode(env_id=...)` ends only the replay trajectory for that environment and does not flush, clear, or mix trajectories from other environments.\n- Stored and sampled episodic replay entries never combine transitions from different parallel environment sources, including when episodes are completed or interrupted in a different order than their steps were observed.\n- In DQN/DDPG batch training with replay, each batch position is handled as a distinct replay environment identity, and a done/reset flag for one batch position does not disturb the replay trajectory for other batch positions.\n- `ReplayBuffer.append` accepts additional keyword fields and preserves them in stored experiences so they can be observed after sampling.\n\n## Implementation notes\n\nThe exact internal data structures, storage layout, and validation locations are up to the implementation. Preserve existing single-environment behavior by treating omitted environment identity as the default environment.\n\n## Required public API surface (mechanical binding)\n\nThe following names / signatures MUST be implemented exactly as listed; downstream test harness mechanically binds these:\n\n- `ReplayBuffer.append(..., env_id=..., **kwargs)` and equivalent concrete replay-buffer append methods - tests must pass a named public environment identity to create multiple replay streams through the buffer API, while still verifying arbitrary extra keyword fields are stored.\n"} {"task_id": "format-code-task-000656", "source_id": "format-code-task-000656", "domain": "code", "task_path": "tasks/format-code-task-000656", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4db676212c8422eae67a09655f566b101527eb7cdb7179e36dc829f69a279c93", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `HanziWriter` instances to support `updateDimensions(options: Partial<{ width: number; height: number; padding: number }>): void` after creation. In a session like `const writer = HanziWriter.create('target', '人', { width: 200, height: 200, padding: 10, charDataLoader }); await writer.setCharacter('人'); writer.updateDimensions({ width: 300, height: 350, padding: 20 });`, the rendered SVG target should have `width=\"300\"` and `height=\"350\"`, and the currently displayed character should stay visible using the new size and padding rather than resetting to an unloaded state. The same method should work for canvas-backed writers: after `HanziWriter.create('target', '人', { renderer: 'canvas', width: 200, height: 200, padding: 10, charDataLoader }).updateDimensions({ width: 300, height: 350, padding: 20 })`, the canvas element should expose `width=\"300\"` and `height=\"350\"`.\n\nThe options object is partial: omitted `width`, `height`, or `padding` values should keep their current values while provided values take effect. If a quiz is active, calling `updateDimensions` should keep the quiz active and align subsequent pointer input with the resized character, so strokes are graded in the same character coordinate space after the resize. Updating dimensions for an already-rendered character should apply the new geometry without discarding what is currently visible."} {"task_id": "format-code-task-000657", "source_id": "format-code-task-000657", "domain": "code", "task_path": "tasks/format-code-task-000657", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ebcb07f9e12de9521aad43ba339d163c819d74fda67d65f5051487aa2853e92b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nCurrently `opnieuw` will only raise the last caught exception. This creates two issues:\n1. We lose any exceptions that were previously raised. This is especially problematic when the first caught exception is different from the second caught exception.\n2. It is hard to tell whether `opnieuw` actually retried in a Sentry stack trace as the retrying isn't explicitly in the trace.\n\n```python\nfrom opnieuw import retry\n\ncounter = 0\n\n\n@retry(\n retry_on_exceptions=(TypeError, ValueError),\n max_calls_total=3,\n retry_window_after_first_call_in_seconds=1,\n)\ndef raise_depending_on_counter() -> None:\n global counter\n counter += 1\n if counter == 1:\n raise TypeError\n elif counter == 2:\n raise ValueError\n else:\n raise IndexError\n\n\nraise_depending_on_counter()\n```\n\nThis currently raises:\n```console\n❯ python test.py\nTraceback (most recent call last):\n File \"/home/daniel/channable/requestmachine/test.py\", line 22, in \n raise_depending_on_counter()\n File \"/nix/store/mzflaklx9497sffvm5k28cf6m0zdi0yd-python3-3.11.4-env/lib/python3.11/site-packages/opnieuw/retries.py\", line 240, in wrapper\n return f(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^\n File \"/home/daniel/channable/requestmachine/test.py\", line 19, in raise_depending_on_counter\n raise IndexError\nIndexError\n```\n\nIdeally it would be something like:\n```console\n❯ python test.py\nTraceback (most recent call last):\n File \"/nix/store/mzflaklx9497sffvm5k28cf6m0zdi0yd-python3-3.11.4-env/lib/python3.11/site-packages/opnieuw/retries.py\", line 241, in wrapper\n return f(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^\n File \"/home/daniel/channable/requestmachine/test.py\", line 15, in raise_depending_on_counter\n raise TypeError\nTypeError\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/nix/store/mzflaklx9497sffvm5k28cf6m0zdi0yd-python3-3.11.4-env/lib/python3.11/site-packages/opnieuw/retries.py\", line 241, in wrapper\n return f(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^\n File \"/home/daniel/channable/requestmachine/test.py\", line 17, in raise_depending_on_counter\n raise ValueError\nValueError\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"/home/daniel/channable/requestmachine/test.py\", line 22, in \n raise_depending_on_counter()\n File \"/nix/store/mzflaklx9497sffvm5k28cf6m0zdi0yd-python3-3.11.4-env/lib/python3.11/site-packages/opnieuw/retries.py\", line 265, in wrapper\n raise last_exception\n File \"/nix/store/mzflaklx9497sffvm5k28cf6m0zdi0yd-python3-3.11.4-env/lib/python3.11/site-packages/opnieuw/retries.py\", line 241, in wrapper\n return f(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^\n File \"/home/daniel/channable/requestmachine/test.py\", line 19, in raise_depending_on_counter\n raise IndexError\nIndexError\n```\n\nA draft PR is here: https://github.com/channable/opnieuw/pull/21."} {"task_id": "format-code-task-000658", "source_id": "format-code-task-000658", "domain": "code", "task_path": "tasks/format-code-task-000658", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:af7ea3afb72308b57b186bbcc1488fbac77216accbc8f8a4cd99d2f1ef316a65", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI'm still using `docker: true` in some fogg configs, and right now `fogg plan` / `fogg apply` don't give me any heads-up that Docker support is going away. Can we have those commands show a non-fatal deprecation warning when Docker is enabled, and have config validation surface that kind of warning instead of only fatal errors?\n\n## Expected outcomes\n\n- **Docker-enabled CLI runs**\n - When a configuration enables Docker with `docker: true`, both `fogg plan` and `fogg apply` should emit the warning `Docker support is deprecated and will be removed in a future version of fogg.`\n - That warning is non-fatal: it should not by itself cause `fogg plan` or `fogg apply` to fail.\n\n- **Docker-disabled CLI runs**\n - When Docker is not enabled in the configuration, `fogg plan` and `fogg apply` should not emit the Docker deprecation warning.\n\n- **Configuration validation warnings**\n - `config/v2.Config.Validate` should surface non-fatal validation warnings as well as fatal validation errors.\n - Calling `Validate` on a config whose `Docker` setting is true should return a warning containing `Docker support is deprecated and will be removed in a future version of fogg.`\n - Fatal validation failures should still be reported as errors; non-fatal warnings should remain inspectable even when fatal validation errors are also present.\n\n## Implementation notes\n\nThe exact mechanism for collecting, propagating, and displaying warnings is up to the implementation. Keep existing fatal validation behavior intact while adding a way for callers and CLI commands to observe non-fatal warnings."} {"task_id": "format-code-task-000659", "source_id": "format-code-task-000659", "domain": "code", "task_path": "tasks/format-code-task-000659", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3fb8e74eb82b816b31d9087dcc6ff8fe32441f25947e67441e5b73295285eb45", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nImprove the start_test-log to JUnit XML converter so CI consumers receive a complete, standards-compatible report from both ordinary logs and logs wrapped in a subtest. The converter must discover every test block in input order even when there is no [Starting subtest] envelope, and retain each case name, directory classname, original output, elapsed time, and skipped status. Emit a single testsuite with numeric tests, failures, errors, skipped, and total time attributes. A test with an [Error...] line is a failure (serialized as a JUnit failure element with the first error line as its message), while warnings remain diagnostic output and do not count as failures. Preserve existing prefix-removal behavior and XML escaping, and continue to exit successfully while producing the report for valid logs."} {"task_id": "format-code-task-000663", "source_id": "format-code-task-000663", "domain": "code", "task_path": "tasks/format-code-task-000663", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:61d5b329d4f1d172a8e14762abdb2379728b74d12860206575b8aa555432ef11", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nProvide a way to lookup chart instance given canvas element\n### Feature Proposal\nUpon instantiation of the Chart, it would be great if a reference to the Chart instance could be tagged to the `HTMLCanvasElement` for later usage, rather than [having to use hacky ways to access them](https://github.com/airblade/chartjs-ror/issues/34#issuecomment-351558852).\n\n### Feature Use Case\nI'm looking for a way to reference the instance of one of my charts which is defined in a different JS module to the one I want to reference it in. Exporting the instance doesn't seem to work, and I took a look at @vijayj's [implementation](https://github.com/airblade/chartjs-ror/issues/34#issuecomment-351558852), and it would've worked fine, but I push the Chart instances to the `window.charts` AFTER importing my other JS module, so when I came to access `window.charts`, it's empty until my other JS module is loaded and executed.\n\nChartJS already adds a `$chartjs` property onto the `HTMLCanvasElement` once instantiated, so it would make sense to add the instance to that property.\n\n## Possible Implementation\n### index.html\n```html\n\n```\n### module1.js\n```js\nconst chartEl = document.querySelector('#chart');\nnew Chart(chartEl, {\n type: 'pie',\n data: {\n labels: [\"Red\", \"Blue\", \"Yellow\", \"Green\", \"Purple\", \"Orange\"],\n datasets: [{\n data: [15, 1, 1, 1, 45, 1],\n backgroundColor: [\n 'rgba(255, 99, 132, 0.2)',\n 'rgba(54, 162, 235, 0.2)',\n 'rgba(255, 206, 86, 0.2)',\n 'rgba(75, 192, 192, 0.2)',\n 'rgba(153, 102, 255, 0.2)',\n 'rgba(255, 159, 64, 0.2)'\n ],\n borderColor: [\n 'rgba(255,99,132,1)',\n 'rgba(54, 162, 235, 1)',\n 'rgba(255, 206, 86, 1)',\n 'rgba(75, 192, 192, 1)',\n 'rgba(153, 102, 255, 1)',\n 'rgba(255, 159, 64, 1)'\n ],\n borderWidth: 1\n }]\n }\n});\n```\n\n### module2.js\n```js\nconst chartInstance = document.querySelector('#chart').$chartjs.instance;\nchartInstance.getDatasetMeta(0).hidden = true;\nchartInstance.update();\n```"} {"task_id": "format-code-task-000664", "source_id": "format-code-task-000664", "domain": "code", "task_path": "tasks/format-code-task-000664", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5f62ad2db89f1117100a68d486ac7404535b01303e6405726633be3df7e04abe", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## SuppressWarningsHolder: align Javadoc with xdoc\n\nPart of #5750.\n\nWhile going through the checks to align Javadoc with the corresponding xdoc page, I noticed `SuppressWarningsHolder` is out of sync. Its Javadoc currently is just a one-liner:\n\n> Maintains a set of check suppressions from {@link SuppressWarnings} annotations.\n\n…whereas the xdoc page (`config_annotation.xml`, `SuppressWarningsHolder` section) has a full description, the `aliasList` property documented in the properties table, and several usage examples covering the `checkstyle:` prefix, the alias mechanism, and the `\"all\"` argument.\n\nA user reading the class Javadoc in their IDE gets almost nothing, while a user reading the website gets the real documentation. The two should match, the same way we've been doing it for other checks in #5750.\n\nWhile we're at it, the xdoc entry for this check could also use a pass to bring it in line with how the other check pages in `config_annotation.xml` are written (the `aliasList` row, the surrounding examples section, etc., feel a bit inconsistent with the rest of the page).\n\nNo code behavior changes — this is a documentation alignment only."} {"task_id": "format-code-task-000665", "source_id": "format-code-task-000665", "domain": "code", "task_path": "tasks/format-code-task-000665", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:267ba0898f010ed4434a4191898d6b0e6e36b4dd35e240ba65751448431bf3a0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `chocolatey_package` still shells out to `choco list` even with `use_choco_list false`\n\nAfter rolling Chef 18 out across our Windows fleet, chef-client runs got noticeably slower on nodes that use a lot of `chocolatey_package` resources. Querying installed package versions through `choco list` is the expensive part, and the resource has a `use_choco_list` property exactly so we can opt out of that and let the provider read versions from the local chocolatey lib directory instead.\n\nI set `use_choco_list false` on our packages (and we don't set `Chef::Config[:always_use_choco_list]`), expecting the provider to use the on-disk fast path. Run times didn't improve at all — when I watch what chef is doing on a node, it's still invoking `choco list` to determine currently-installed versions for every chocolatey_package resource. Toggling `use_choco_list` true vs. false makes no observable difference.\n\nReproducer (any Windows node with chocolatey installed):\n\n```ruby\nchocolatey_package 'git' do\n action :install\n use_choco_list false\nend\n```\n\nExpected: with `use_choco_list false` and no global override, the provider determines installed versions by inspecting the local chocolatey package directory, not by shelling out to `choco list`.\n\nActual: `choco list` is invoked anyway, and the run is just as slow as it was before the property existed. Setting the property to `true` produces the same behavior, which is what tipped me off that the selection logic between the two code paths isn't doing what its name says.\n\nThis looks like a regression — the on-disk path was the whole point of `use_choco_list false` for us, and right now there's no way to actually reach it."} {"task_id": "format-code-task-000666", "source_id": "format-code-task-000666", "domain": "code", "task_path": "tasks/format-code-task-000666", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b23a0ff9eb97a0c35f128f161cb86fb49f40f78a5607fa3c7d380b3e475c1bf2", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `pyromat.get(idstr: str) -> object` so application code can retrieve one already-loaded substance from the PYroMat catalog by its exact substance ID. For example, after `import pyromat as pm`, `pm.get(\"ig.O2\")` should return the oxygen substance object whose `sid()` is `\"ig.O2\"` and whose `pmclass()` is `\"ig2\"`; `pm.get(\"mp.H2O\")` should return the multiphase water substance object whose `sid()` is `\"mp.H2O\"` and whose `pmclass()` is `\"mp1\"`.\n\nThe function should be a lookup over the current loaded catalog: it should not scan files itself, create a new catalog, write to disk, or mutate the catalog. If the ID is not present, such as `pm.get(\"not.a.real.substance\")`, it should report that no substance with that ID was found and raise `pyromat.utility.PMParamError` with an invalid-substance-ID error."} {"task_id": "format-code-task-000667", "source_id": "format-code-task-000667", "domain": "code", "task_path": "tasks/format-code-task-000667", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0b2bbd30f406bc237ac619b1d4d7fae3348a10b56fabd7d4f34e9873e2b8932f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `MarketManager` to support manual execution of an existing resting order through `ErrorCode ExecuteOrder(uint64_t id, uint64_t quantity)` and `ErrorCode ExecuteOrder(uint64_t id, uint64_t price, uint64_t quantity)`. In a session where a `MarketManager` already has a symbol, an order book, and a resting limit order with id `10`, price `101`, quantity `100`, executed quantity `0`, and leaves quantity `100`, calling `ExecuteOrder(10, 25)` should return `ErrorCode::OK`, emit `onExecuteOrder` with price `101` and quantity `25`, update the order to executed quantity `25` and leaves quantity `75`, keep the order available through `GetOrder(10)`, update the affected book level volumes, and emit an order update. Calling `ExecuteOrder(10, 99, 80)` after that should cap execution to the remaining `75`, emit execution at price `99` for quantity `75`, update market prices for the book from that execution price, emit order deletion, remove the order from the manager, and make `GetOrder(10)` return `nullptr`. Both overloads should reject `id == 0` with `ErrorCode::ORDER_ID_INVALID`, reject `quantity == 0` with `ErrorCode::ORDER_QUANTITY_INVALID`, return `ErrorCode::ORDER_NOT_FOUND` for an unknown order id, and return `ErrorCode::ORDER_BOOK_NOT_FOUND` if the stored order no longer has an order book. If automatic matching is enabled, a successful manual execution should run matching on the affected order book before returning."} {"task_id": "format-code-task-000668", "source_id": "format-code-task-000668", "domain": "code", "task_path": "tasks/format-code-task-000668", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:52ce716ca04569fb9d7dd0a7a797e705c4caa0aeb51adf9168f3c9e086ebcc41", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI'm hitting CRD validation errors when Cilium tries to create identity resources in my cluster — the API server rejects them complaining about label names being longer than 63 characters. Looking at the rejected payload, the offending names all seem to come from namespace labels that Cilium is propagating onto the identity (stuff under `io.cilium.k8s.namespace.labels.`), and some of our namespace labels have pretty long keys. New identities aren't getting created because of this. Can you take a look?\n\n## Expected outcomes\n\n- Identity resource creation should not fail merely because a namespace has label keys that would produce overlong Kubernetes label names when propagated onto the identity object.\n- Namespace-derived label information should be omitted from the identity object's Kubernetes metadata labels rather than submitted to the API server.\n- Normal Kubernetes-sourced labels that are valid for identity metadata should continue to be included on the identity resource.\n- Non-Kubernetes label sources should continue to be excluded from the identity resource's Kubernetes metadata labels.\n- Existing identity label semantics outside the Kubernetes metadata-label projection should remain unchanged.\n\n## Implementation notes\n\n- Prefer preserving existing behavior for labels unrelated to namespace metadata propagation.\n- The fix should be based on externally observable identity creation and label projection behavior, not on a specific refactor shape."} {"task_id": "format-code-task-000669", "source_id": "format-code-task-000669", "domain": "code", "task_path": "tasks/format-code-task-000669", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f12878572466ca1291e88678d3872a9511ff35846840b401a503afa5fcc6cb2a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add IPv6 support to the ip-masq-agent\n\nThe eBPF ip-masq-agent watches a configuration file and provisions the set of\n\"non-masquerade\" CIDRs (destinations that pod traffic should reach without\nbeing source-NAT'd). Today this only works for IPv4: the `nonMasqueradeCIDRs`\nlist rejects any IPv6 prefix, and the only link-local handling is for IPv4.\n\nExtend the agent so it can manage IPv6 destinations as well.\n\n## What it should do\n\n- The `nonMasqueradeCIDRs` list must accept IPv6 prefixes in addition to IPv4.\n Both families are provisioned together from a single configuration, and each\n CIDR is stored in its canonical network form (host bits cleared), exactly as\n IPv4 entries already are — e.g. `2001:db8::1/32` is provisioned as\n `2001:db8::/32`.\n\n- Add a new optional boolean configuration key `masqLinkLocalIPv6` that mirrors\n the existing `masqLinkLocal` key but for the IPv6 link-local prefix\n `fe80::/10`. When `masqLinkLocalIPv6` is explicitly set to `false`, then\n `fe80::/10` is appended to the non-masquerade set. When the key is absent or\n set to `true`, `fe80::/10` is not added. (Defaulting the absent case to \"not\n added\" keeps the behavior of existing IPv4-only deployments unchanged.)\n\n- The existing IPv4 link-local behavior is unchanged and independent of the new\n option: when `masqLinkLocal` is absent or `false`, `169.254.0.0/16` is part of\n the non-masquerade set; when it is `true`, it is not.\n\nThe configuration may be written as YAML or JSON, as it is today. When the\nconfiguration file is missing or empty, the agent keeps provisioning the\nexisting default IPv4 non-masquerade CIDRs (plus the IPv4 link-local prefix),\nand does not add the IPv6 link-local prefix.\n"} {"task_id": "format-code-task-000670", "source_id": "format-code-task-000670", "domain": "code", "task_path": "tasks/format-code-task-000670", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f1359b4b1e5a0c5961538151881bc38a9834cbff404ea3239e95a490bc76ed0e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want an initialized `Document` to provide `guess_notes(horizontal: str, vertical: str, current_page: int = 0) -> PdfPage` so Pympress can decide which notes mode to use when a PDF is opened. The method should inspect the document's page labels, page count, renderability, and aspect ratios without changing the document's current notes mapping.\n\nIf any page label starts with `notes:`, `guess_notes(...)` should return `PdfPage.MAP`. If `current_page` does not resolve to a renderable page, it should return `PdfPage.NONE`. For an even-page document where the first half of pages all share one aspect ratio, the second half all share another aspect ratio, and the two ratios differ, it should return `PdfPage.AFTER`.\n\nFor a renderable current page with aspect ratio at least 2, it should return the `PdfPage` named by the `horizontal` preference, case-insensitively; if that preference is not a valid `PdfPage` name, it should fall back to `PdfPage.RIGHT`. For classic US letter ratio `8.5 / 11` and ISO paper ratio `1 / sqrt(2)`, it should return `PdfPage.NONE`. For a renderable current page with aspect ratio below 1, it should return the `PdfPage` named by the `vertical` preference, case-insensitively; if that preference is not valid, it should fall back to `PdfPage.BOTTOM`. Other ordinary slide ratios should return `PdfPage.NONE`."} {"task_id": "format-code-task-000672", "source_id": "format-code-task-000672", "domain": "code", "task_path": "tasks/format-code-task-000672", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4be6b712592987a368d7d547f95b411f15387dd33ca3db23c002f978e783315f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nIt seems `changesInclude` is not supported in Starlark.\n\n(I'd expect the Starlark-side builtin to be something like `changes_include`, with a way to feed in the list of affected files via a larker option such as `WithAffectedFiles`.)"} {"task_id": "format-code-task-000673", "source_id": "format-code-task-000673", "domain": "code", "task_path": "tasks/format-code-task-000673", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:735da8886e318f8078418cad6e829b9958532351bcb03b227bd8876685b6f6e7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the installed `bsdetector` shell command to generate batch feature reports from a newline-delimited text file. The command should be invoked as `bsdetector -i FILE [-o json|tsv|html]`, where `-i` is required and `-o` defaults to `json`.\n\nOn a successful run, the command reads the input file, splits it on newline characters, treats each non-empty line as one statement, extracts the existing bias features for each statement, and exits with status 0. With the default JSON output, stdout should be an indented JSON array; each item should contain the extracted feature fields for one non-empty input line plus a `text` field containing the original line. For example, if the file contains `The cat sucks.\\n\\nThis is neutral.`, the JSON report should contain two objects, one with `\"text\": \"The cat sucks.\"` and one with `\"text\": \"This is neutral.\"`, and no object for the blank line.\n\nWith `-o tsv`, stdout should be a tab-separated table whose header starts with `sentence` followed by the feature names, and each non-empty input line should produce one row starting with the original sentence followed by the corresponding feature values. The TSV mode should also write the Excel import hint to stderr. With `-o html`, stdout should include the HTML usage hint and then an HTML document containing a bordered table; rows should be emitted only for input lines longer than three characters, with the first column containing the original sentence and the remaining columns containing feature values.\n\nThe command should provide argparse-style help: `bsdetector --help` exits 0 and prints usage including `-i FILE` and the `-o {json,tsv,html}` choices. Missing `-i` or an unsupported output type such as `bsdetector -i input.txt -o csv` should fail before analysis, print the parser error to stderr, and exit with status 2. If the input file cannot be opened, the command should fail with a non-zero exit instead of emitting a partial report."} {"task_id": "format-code-task-000675", "source_id": "format-code-task-000675", "domain": "code", "task_path": "tasks/format-code-task-000675", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e4f9bb893ce1dcfc1d2db4bc600553b2306a81459181f87d22237b601bab3aad", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want ckb-daemon's command FIFO to support a shorthand live RGB command where a bare six-digit hex color after `rgb` applies to every LED in the currently selected mode. For example, after a device is active, writing `rgb ff0000` should make every key or lighting zone red, and writing `rgb 000000` should turn every key or lighting zone off. The shorthand should behave like applying that same color across the whole device, so users do not have to enumerate individual keys or use a longer all-LED selector just to set a single solid color."} {"task_id": "format-code-task-000676", "source_id": "format-code-task-000676", "domain": "code", "task_path": "tasks/format-code-task-000676", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:da9791a2c317987b8bc7933892e51796d13d251244306cd47ef50d54f41c5fda", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want lrzip to support a gzip/zlib backend compression mode for archive stream blocks. When I run `lrzip -g input.txt` or `lrzip --gzip input.txt`, it should select gzip as the second-stage backend, compress each eligible rzip stream block with zlib at the configured compression level, and write gzip-marked block headers so the resulting `.lrz` file can be decompressed later by the normal `lrzip -d input.txt.lrz` path. If zlib cannot make a block smaller, lrzip should store that block uncompressed instead of growing the archive; if zlib compression fails for another reason, compression should fail non-zero with an error.\n\nThe same backend should be selectable from `lrzip.conf` with `CompressionMethod = gzip`, subject to the same “only one compression method” rule as the other backends. `lrzip --help` should list `-g, --gzip` as gzip compression using zlib, and verbose compression summaries should report `Compression mode is: GZIP` when this mode is selected. `lrzip -i archive.lrz` should identify gzip-compressed blocks as `gzip` and summarize such archives as `rzip + gzip`. During decompression, gzip-marked blocks should be inflated with zlib, their output length must match the uncompressed length stored in the block header, and corrupted gzip data or a length mismatch should make decompression fail non-zero."} {"task_id": "format-code-task-000677", "source_id": "format-code-task-000677", "domain": "code", "task_path": "tasks/format-code-task-000677", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2f21ea0c74e39623a76fd9afb9cd8fab1149fbc1ef770a96e2c98e78af7c70ef", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a pure `getRoutesForLang(currentLang: string, page: string, isHomepage?: boolean = false)` helper that returns alternate-language route objects for the site's supported languages. It should support exactly English (`en`, label `in english`), Spanish (`es`, label `en español`), and Chinese (`zh`, label `中文`), and it should omit the current language from the returned list.\n\nFor `getRoutesForLang(\"en\", \"why\")`, return `[ { \"label\": \"en español\", \"lang\": \"es\", \"path\": \"/es/why.html\" }, { \"label\": \"中文\", \"lang\": \"zh\", \"path\": \"/zh/why.html\" } ]`. For `getRoutesForLang(\"es\", \"resources\")`, return `[ { \"label\": \"in english\", \"lang\": \"en\", \"path\": \"/resources.html\" }, { \"label\": \"中文\", \"lang\": \"zh\", \"path\": \"/zh/resources.html\" } ]`.\n\nFor homepage calls, `isHomepage=true` should treat the page as `index`, so `getRoutesForLang(\"zh\", \"\", true)` returns English and Spanish route objects with paths `/index.html` and `/es/index.html`. The function must be deterministic for the same inputs, must not mutate its arguments, and must not perform filesystem or network side effects."} {"task_id": "format-code-task-000678", "source_id": "format-code-task-000678", "domain": "code", "task_path": "tasks/format-code-task-000678", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:35c2bcb36c7d14b29a60a88ace1f3fa40a6e79d64ef6993a270ff868a529e6a0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Describe the bug\n\nUsing `gh` to `browse` a file with given line number does not work with Markdown files. The URL the user is taken to is the rendered Markdown view without line numbers on the left-hand side.\n\n### Steps to reproduce the behavior\n\n```\n$ gh browse README.md:3\nnow opening https://github.com/cli/cli/tree/trunk/README.md#L3 in browser\n$ gh --version\ngh version 2.0.0 (2021-08-24)\nhttps://github.com/cli/cli/releases/tag/v2.0.0\n```\n\n### Expected vs actual behavior\n\nLooks like a `?plain=1` parameter was [added to GitHub Web UI](https://github.blog/changelog/2021-06-30-parameter-to-disable-markdown-rendering/) to specifically help navigate to Markdown lines.\n\nSo, instead of:\n* https://github.com/cli/cli/tree/trunk/README.md#L3\n\nI'd expect the user to be taken to:\n* https://github.com/cli/cli/blob/trunk/README.md?plain=1#L3\n\n_note: the new URL has `blob` in the path because `?plain=1` does not work with `tree` in the path_\n\n### Logs\n\nN/A"} {"task_id": "format-code-task-000679", "source_id": "format-code-task-000679", "domain": "code", "task_path": "tasks/format-code-task-000679", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:981bd4c9883cc7ebf20f51ed8603d965944b2d9e1a264051021fdd2d7636c0be", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n`gh release create`: \"Write using git tag message as template\" option is hidden sometimes even when annotations exist\n### Describe the bug\n\n`gh --version`:\ngh version 2.4.0 (2021-12-21)\nhttps://github.com/cli/cli/releases/tag/v2.4.0\n\nIn earlier versions, I would *always* see the option to create release notes from a tag message if the tag was annotated. Currently, that option is hidden when I own the repository.\n\n### Steps to reproduce the behavior\n\n```shell\ngh repo clone me/my-repo\ncd my-repo\ngit tag -am \"foo\" bar\ngh release create bar # don't see option\n```\n```shell\ngh repo clone another-user/repo-i-do-no-maintain\ncd repo-i-do-not-maintain\ngit tag -am \"foo\" bar\ngh release create bar # see tag message option\n```\n\n### Expected vs actual behavior\n\nI expect the \"Write using git tag message as template\" option to *always* appear, assuming the annotated tag is available.\n\n### Extra Info\n\nI think these lines are doing it. It looks like, as long as `generatedNotes != nil`, `tagDescription` is not set, so it equals `\"\"`.\nhttps://github.com/cli/cli/blob/ad8d7bb02e36e66e28c78837f5392afd1a5187a3/pkg/cmd/release/create/create.go#L237-L239\nhttps://github.com/cli/cli/blob/ad8d7bb02e36e66e28c78837f5392afd1a5187a3/pkg/cmd/release/create/create.go#L261-L263"} {"task_id": "format-code-task-000680", "source_id": "format-code-task-000680", "domain": "code", "task_path": "tasks/format-code-task-000680", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:521e16b22928d0710c4264c7e226c92e9964a67b318635df7d6211729e8365cf", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `auto_ml.Predictor` to support regression uncertainty prediction as part of the normal object lifecycle. A user should be able to construct `Predictor(type_of_estimator='regressor', column_descriptions=...)`, call `train(...)` with the opt-in keyword arguments `train_uncertainty_model=True`, `uncertainty_data=...`, `uncertainty_delta=None`, `uncertainty_delta_units=None`, `uncertainty_delta_direction=None`, `calibrate_uncertainty=False`, `uncertainty_calibration_settings=None`, and `uncertainty_calibration_data=None`, and then call `predict_uncertainty(prediction_data)` or `score_uncertainty(X, y, advanced_scoring=True, verbose=2)` on the same instance.\n\nWhen `train_uncertainty_model=True` is used for a regressor, training should clean and transform `uncertainty_data`, append the base regression prediction as an additional feature, label each row as certain or uncertain from the observed prediction error, and train a classifier that predicts the probability that a future regression prediction will be uncertain. If no `uncertainty_delta` is supplied, the threshold should default to one half of the standard deviation of the original regression training targets, measured in absolute units. If `uncertainty_delta_units='absolute'`, compare either `abs(prediction - actual)` for `uncertainty_delta_direction='both'` or the signed delta for `uncertainty_delta_direction='directional'`; if `uncertainty_delta_units='percentage'`, make the same comparisons after dividing the delta by the actual value.\n\n`predict_uncertainty(prediction_data)` should return the base regression prediction together with the uncertainty probability. For batch prediction input, it should return a pandas `DataFrame` with at least `base_prediction` and `uncertainty_prediction` columns, where `uncertainty_prediction` is the probability of the uncertain class. For a single row dict, it should return a dict with the same fields. `score_uncertainty(X, y, advanced_scoring=True, verbose=2)` should call `predict_uncertainty`, derive the true uncertain labels from `y` using the configured threshold rules, print the normal classifier probability diagnostics, and return the Brier score.\n\nIf `calibrate_uncertainty=True`, training should use `uncertainty_calibration_data` plus `uncertainty_calibration_settings` to bucket uncertainty probabilities and attach calibration details to later batch `predict_uncertainty` results. The default calibration settings should be `{'num_buckets': 10, 'percentiles': [25, 50, 75]}`. With custom settings like `{'num_buckets': 3, 'percentiles': [25, 50, 75]}`, batch uncertainty predictions should include `bucket_num`, `percentile_25_delta`, `percentile_50_delta`, and `percentile_75_delta` columns.\n\nInvalid training configurations should fail early with `ValueError`: `train_uncertainty_model=True` on a classifier, `train_uncertainty_model=True` without `uncertainty_data`, an `uncertainty_delta` without `uncertainty_delta_units`, or an `uncertainty_delta_direction` other than `both` or `directional`."} {"task_id": "format-code-task-000681", "source_id": "format-code-task-000681", "domain": "code", "task_path": "tasks/format-code-task-000681", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fa7998e14b2b38495e9e7b32a86b955c0d642cbe38dfda63f0b27a3726a204bf", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## CI `wasm_bindings` job is broken and `cargo test --all` blows up in `crates/testing`\n\nI was trying to put up a PR and noticed the `wasm_bindings` job on GitHub Actions fails right away. The \"Build rust-wasm-test\" step is invoking\n\n```\ncargo run -p spacetimedb-cli -- build crates/modules/rust-wasm-test\n```\n\nbut there is no `crates/modules/` directory anymore — the modules live under the top-level `modules/` directory now (`modules/rust-wasm-test`, `modules/benchmarks`, `modules/spacetimedb_quickstart`, all listed as workspace members in the root `Cargo.toml`). So the CLI bails out because it can't find the manifest. Looks like the CI workflow was just never updated when modules were moved.\n\nI tried reproducing locally with `cargo test --all` and ran into a separate but related mess: the integration tests in `crates/testing` (`with_module` / `with_module_async` → `load_module`) panic when they try to read the compiled wasm. From poking at it, it seems the test harness is still resolving module paths and wasm artifact paths as if modules were under `crates/modules/`, so it ends up pointing at a file that doesn't exist and `std::fs::read(...).unwrap()` blows up.\n\nA second thing I noticed while looking at this: the harness assumes every module's wasm artifact has the same fixed filename. But each module is its own crate with its own name, and `cargo build --target=wasm32-unknown-unknown --release` produces a wasm whose name comes from the crate name — so different modules produce differently-named `.wasm` files. As soon as you point it at, say, `benchmarks` instead of `rust-wasm-test`, the hardcoded filename won't match.\n\nAlso minor, but `modules/spacetimedb_quickstart` uses `snake_case` while the other entries under `modules/` are `kebab-case` — would be nice to make that consistent while we're cleaning this up.\n\nIt would be great to get CI green again and have the testing crate actually able to load module wasm from wherever the modules currently live."} {"task_id": "format-code-task-000682", "source_id": "format-code-task-000682", "domain": "code", "task_path": "tasks/format-code-task-000682", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:841295b1106d2046ba7c991b2942bbb464308aaddd35de60fd3eef2a715617c7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: KMS encryption filter and action for SNS topics\n\nSNS now supports server-side encryption with KMS, and I'd like to manage it with Cloud Custodian, but the SNS resource doesn't seem to expose anything for it.\n\nMy use cases:\n\n1. **Audit which SNS topics are encrypted and with which key.** I want to write a policy that selects SNS topics whose KMS master key is (or isn't) one of our approved CMKs — same kind of filtering that other resources support against a KMS key (matching by id / arn / alias). Right now I can only get at it by writing a raw `value` filter against `KmsMasterKeyId`, which doesn't help me match against alias names or filter through KMS key attributes.\n\n2. **Remediate unencrypted topics.** Once I've found topics that aren't encrypted (or are encrypted with the wrong key), I want an action that flips encryption on for them, ideally letting me pick the KMS key. I'd also like to be able to turn encryption back off on a topic if needed.\n\nA rough example of the kind of policy I'd like to be able to write:\n\n```yaml\npolicies:\n - name: sns-encrypt-with-approved-cmk\n resource: sns\n filters:\n - \n actions:\n - \n```\n\nWould it be possible to add KMS-aware filtering and an encryption-management action to the `sns` resource?"} {"task_id": "format-code-task-000683", "source_id": "format-code-task-000683", "domain": "code", "task_path": "tasks/format-code-task-000683", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e5596d132e87a4f0d7ee86858e17734e49140ce7c56f0de10dc0179bea486643", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## iam-role resource has no way to delete roles via a policy\n\nI'm trying to write a Custodian policy that cleans up IAM roles that haven't been used in a long time. The `usage` filter on `iam-role` works well for finding them — e.g.:\n\n```yaml\npolicies:\n - name: iam-delete-unused-role\n resource: iam-role\n filters:\n - type: usage\n match-operator: all\n LastAuthenticated: null\n actions:\n - delete\n```\n\n…but when I run this Custodian rejects the policy because `iam-role` doesn't actually have a `delete` action registered. Other IAM resources have actionable lifecycle support, so it's surprising that role deletion isn't there.\n\nCould `iam-role` get a `delete` action so unused/stale roles can be cleaned up directly from a policy?\n\nA couple of things I'd want the action to handle gracefully, since they come up routinely when sweeping a large account:\n\n- Some roles will still be attached to an instance profile and AWS refuses to delete them in that state. The action shouldn't silently pretend it worked, but it also shouldn't be the kind of thing where one stuck role aborts the whole sweep before the other deletable roles are processed.\n- When iterating over a batch, individual roles may have already been removed (e.g. by another process between list and delete) — that case should just be skipped rather than treated as a failure."} {"task_id": "format-code-task-000684", "source_id": "format-code-task-000684", "domain": "code", "task_path": "tasks/format-code-task-000684", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2cb176d7b83af98bfe7d5a85f0cae0eed9ddce69905c57e48610740a60ca9869", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nThe S3 `data-events` filter is used to find buckets whose object-level activity is covered by CloudTrail, but its selector parsing is too literal for the shapes returned by the CloudTrail API. A selector may identify an S3 bucket with the canonical bucket ARN (`arn:aws:s3:::bucket`) or with the object resource form (`arn:aws:s3:::bucket/*`), and a wildcard bucket resource must mean every bucket in the policy input. The filter also needs to remain safe when a trail contains management-event selectors or data resources for another AWS service.\n\nUpdate the filter's observable behavior without changing its policy-facing shape: `state: present` must return each input bucket covered by at least one S3 object data resource from any of the discovered trails, while `state: absent` must return exactly the complementary input buckets. Coverage from trails in different home regions must be combined; each trail's event selectors must be read from its own `HomeRegion`, so a trail in one region cannot hide or replace selectors from another. Preserve the input bucket order. Existing policy configurations and the `state` values remain unchanged."} {"task_id": "format-code-task-000685", "source_id": "format-code-task-000685", "domain": "code", "task_path": "tasks/format-code-task-000685", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:068190ba8a09e9af2393f069dfedc74a1684da291ad9a25eb545a1906919b8f7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `GET /v3/service_plans` ignores `fields[service_offering.service_broker]`\n\nAccording to the CF API docs, listing service plans supports a `fields` query parameter that lets you sideload related broker info alongside the plans, e.g.\n\n```\nGET /v3/service_plans?fields[service_offering.service_broker]=guid,name\n```\n\nThe expected response has the plans in `resources` and the related brokers under `included.service_brokers` (with only the requested fields).\n\nAgainst korifi I get the plans back fine, but the `included` section never contains any broker entries no matter what I pass under `fields[service_offering.service_broker]`. It looks like the parameter is just being dropped — the response is identical to a request without it.\n\nFor comparison, `include=service_offering` on the same endpoint does work and populates `included.service_offerings`, so plan-side sideloading is partly there, just not the broker fields variant.\n\nCould korifi honour `fields[service_offering.service_broker]` on the service plans list endpoint and return the matching broker resources under `included`?"} {"task_id": "format-code-task-000686", "source_id": "format-code-task-000686", "domain": "code", "task_path": "tasks/format-code-task-000686", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7eb06ed7e263c7323bbaff9f13242cb5e4c0395d5871f2820af20138985105bd", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Service instance response is missing the `service_plan` link\n\nWhen I fetch a managed service instance via the API, the `links` block in the response doesn't include a link to the service plan it was created from.\n\nRepro:\n\n```\ncf create-service my-offering my-plan my-instance\ncf service my-instance --guid\n# -> \ncf curl /v3/service_instances/\n```\n\nIn the returned JSON, under `links` I get `self`, `space`, `credentials`, `service_credential_bindings`, `service_route_bindings` — but there's no entry pointing at the plan. On regular CF this response includes a link that lets you navigate from a service instance to its plan, and tooling I'm porting over relies on following that link to fetch plan details (e.g. to display which plan an instance is on, or to look up plan metadata) without having to know the plan GUID up front.\n\nIt would be good for korifi's `/v3/service_instances/` response to expose the service plan as a link too, so clients can traverse from an instance to its plan the same way they do against cloud_controller."} {"task_id": "format-code-task-000691", "source_id": "format-code-task-000691", "domain": "code", "task_path": "tasks/format-code-task-000691", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2ea15999c2a296f58d2dddfbcb152f8d44a89daedf541cae88f4ddaa970f539c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\noptimize: add a option to controls whether debugPrintRoute executes\n**Is your feature request related to a problem? Please describe.**\n\n- When the server is running, it will print a lot of logs of the routes and i wish I could control its printing\n\n**Describe the solution you'd like**\n\n- add a option"} {"task_id": "format-code-task-000692", "source_id": "format-code-task-000692", "domain": "code", "task_path": "tasks/format-code-task-000692", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ef3ce3b9d856adf00342e6e65dfe30de4c2ebd58a60e9a69042431248de63c89", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nHarden the URL rewriting performed by the client-go transport used to talk to Clusternet's shadow API. The transport receives ordinary Kubernetes API requests and forwards them through the hub's configured base path, but the current pattern matching is permissive and can rewrite paths that are not actually addressed to the configured endpoint or that belong to Clusternet's own APIs. That can corrupt proxy URLs and make otherwise valid API groups unreachable.\n\nKeep the existing shadow target (/apis/shadow/v1alpha1) and the public constructors/transport interfaces. For a request whose path is under the base path supplied to NewClusternetTransport, rewrite core Kubernetes paths beginning with /api/v1 and grouped paths beginning with /apis// to the corresponding shadow path, retaining the resource/subresource suffix. A version segment is valid only when it is v followed by digits, optionally followed by alpha or beta and digits (for example v1, v1alpha1, or v2beta3); API-like paths with other version text must pass through unchanged. Exact API roots (/api/v1 and /apis//) and roots with a trailing slash are valid and must normalize while preserving whether that trailing slash was present.\n\nGroups owned by Clusternet (any DNS group ending in .clusternet.io, including groups with hyphens) must never be rewritten, even when their path contains a valid version. Matching the configured base path must be prefix- and segment-aware: a request outside that prefix, or merely containing the prefix in a later path segment, must remain byte-for-byte unchanged apart from normal URL handling by the underlying transport. Requests that do match retain the configured prefix and all path suffix segments. Query strings and other URL components must be forwarded unchanged. Non-API paths continue to pass through untouched. Existing behavior for ordinary core/grouped API resources and Clusternet proxy payloads must remain compatible."} {"task_id": "format-code-task-000693", "source_id": "format-code-task-000693", "domain": "code", "task_path": "tasks/format-code-task-000693", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0e83dc0e800451fd62ba1974bcb39281db991aaedc4e27a8b533b79a284327f7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Reusable partial conversations via INCLUDE\n\nOur text scripting format lets you write conversations (the `.convo.txt` files) made up of\n`#me` / `#bot` steps. As the test suites grow, the same building blocks (logins, greetings,\nteardown flows, …) get copy-pasted into many conversations. We want a way to factor those\nshared fragments out into reusable pieces and pull them into a conversation on demand.\n\nPlease add support for **partial conversations** and an **`INCLUDE` directive**.\n\n## What it should do\n\n- A new script kind, the *partial conversation*, lives in files ending in `.pconvo.txt`. They use\n the exact same text syntax as a normal conversation (a name on the first line, then `#me` /\n `#bot` steps with their asserters/logic hooks). The name of a partial conversation is taken from\n its header (falling back to the file name when no header name is given).\n- Partial conversations are **not** runnable conversations on their own: loading a `.pconvo.txt`\n file must not add anything to the regular set of convos. They are kept aside, keyed by name, so\n that other scripts can reference them.\n- Inside any conversation step, an `INCLUDE ` directive (written like the other logic-hook\n lines, e.g. `PAUSE`) references a partial conversation by name. When that conversation runs, the\n referenced partial conversation's steps are spliced into the conversation in place — i.e. the\n conversation behaves exactly as if the partial conversation's steps had been written inline at\n that point. Both the original and the included steps are executed and appear in the run's\n transcript.\n- Includes may be **nested**: a partial conversation can itself `INCLUDE` another one, to arbitrary\n depth, and all of them are expanded.\n\n## Error handling\n\nThe following situations must cause a clear failure rather than silently doing the wrong thing:\n\n- **Circular includes.** If expanding the includes would loop back on a partial conversation that\n is already being expanded (directly or transitively), running the conversation must fail with an\n error instead of looping forever.\n- **Unknown reference.** If an `INCLUDE` names a partial conversation that does not exist, running\n the conversation must fail with an error.\n- **Bad argument.** `INCLUDE` takes exactly one argument (the name). Anything else (zero arguments,\n or more than one) must make the run fail with an error.\n- **Duplicate names.** Two partial conversations may not share the same name; loading a second one\n with a name that is already taken must fail with an error.\n- **Invalid name.** A partial conversation name may not contain the `|` character; loading one with\n such a name must fail with an error.\n\nLoading the existing `.convo.txt`, `.utterances.txt` and `.xlsx` scripts must keep working exactly\nas before.\n"} {"task_id": "format-code-task-000694", "source_id": "format-code-task-000694", "domain": "code", "task_path": "tasks/format-code-task-000694", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:53f0f4383d81f6753f08bd0427abe6b51320360c9136be25c8272c0a5d9a1357", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Workspace `last_used_at` keeps advancing even when nobody is connected, breaking autostop / dormancy\n\nWe rely on inactivity-based automation (autostop, and a dormancy policy that\neventually marks unused workspaces for deletion). After turning those on we\nnoticed they essentially never fire — workspaces look perpetually \"just used\"\neven when we know for a fact nobody is connected to them.\n\n### What we see\n\nPick any workspace where the agent is running, then make sure there's nothing\nactually interacting with it:\n\n- close VS Code (remote / coder extension)\n- close JetBrains Gateway\n- exit every `coder ssh` session\n- close any web terminal / reconnecting PTY tab in the dashboard\n\nLeave it like that for a while and watch the workspace in the dashboard (or\nquery `last_used_at` directly). The \"last used\" timestamp keeps ticking\nforward on its own, roughly in sync with the agent's stats reporting cadence,\neven though there are zero active sessions of any kind against the agent.\n\nBecause `last_used_at` never stops moving, the inactivity clock that\nautostop / dormancy depend on never accumulates, and those policies just sit\nthere doing nothing. Effectively any workspace whose agent is healthy is\nconsidered \"in use\" forever.\n\n### What we expect\n\n`last_used_at` should track real user activity against the workspace, not\njust the fact that the agent process is alive and phoning home. If the user\nisn't actually connected through any of the supported session types, the\ntimestamp should hold steady so the inactivity-based policies can do their\njob."} {"task_id": "format-code-task-000695", "source_id": "format-code-task-000695", "domain": "code", "task_path": "tasks/format-code-task-000695", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b60bfdd7e70630922e86b21822a959756f45d7bab235e38fef184572b14f747b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Filtering users by their GitHub user ID\n\nWe have a Coder deployment where most users sign in via GitHub OAuth. The `users` table already stores each user's GitHub user ID (it gets populated when they log in), but there doesn't seem to be a way to actually search/filter users by it from the outside.\n\nA couple of concrete situations where this matters for us:\n\n- Someone leaves the org, their GitHub username gets renamed/transferred, but the numeric GitHub user ID is stable. We'd like to look up the Coder account from that stable ID.\n- We have some automation that, given a GitHub user ID from a webhook, needs to find the matching Coder user to take action on (audit, suspend, list workspaces, etc.). Right now we have to fetch all users and filter client-side, which doesn't scale.\n\nI tried `coder users list` and the `GET /api/v2/users` endpoint — neither seems to accept any GitHub-ID-based filter. The data is clearly there in the DB, it's just not queryable.\n\nCould we get a first-class way to filter the users list by GitHub user ID, both from the CLI (`coder users list`) and from the HTTP API? It would be enough for our use case if exact-match filtering is supported (one GitHub user ID in, the matching Coder user out, or empty if none)."} {"task_id": "format-code-task-000696", "source_id": "format-code-task-000696", "domain": "code", "task_path": "tasks/format-code-task-000696", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ccda66981b5e8a499675cd82b61ba898b84ee818e01193bb768b18c561e4f760", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI’m trying to see all projects in my current org from the CLI, but `coder projects` doesn’t seem to have a way to list them. Could you add a `coder projects list`/`ls` command that shows the basics like name, source, last updated, and how many developers are using each project? It’d also be helpful if the projects API exposed that usage count so the CLI can show it.\n\n## Expected Outcomes\n\n- Project listing from the CLI: `coder projects list` lists projects in the current organization and displays a tabular summary with `Project`, `Source`, `Last Updated`, and `Used By` columns.\n- Short alias: `coder projects ls` performs the same listing behavior as `coder projects list`.\n- Empty organization output: when there are no projects in the current organization, the list command prints a friendly empty-state message and points users to `coder projects create ` instead of showing an empty table.\n- API usage count: project listing API responses include a numeric `workspace_owner_count` field for each project.\n- CLI usage count display: the `Used By` column reflects each project’s usage count, using singular wording for exactly one developer and plural wording otherwise.\n\n## Implementation Notes\n\n- The command should use the existing CLI authentication, organization selection, and project retrieval conventions.\n- The API and CLI should remain consistent: the CLI should display the usage count exposed by the project listing API.\n- Specific data structures, helper functions, query organization, and formatting implementation details are up to the implementer as long as the observable behavior above is satisfied."} {"task_id": "format-code-task-000697", "source_id": "format-code-task-000697", "domain": "code", "task_path": "tasks/format-code-task-000697", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:32fd726a7176272f278fe7ed7ecd535e083165de1f49ede2f4ace3169a9fe99c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `Lines.Bytes()` doesn't include a trailing newline\n\nI'm embedding the texteditor widget in my app to let users edit source\nfiles. To save changes I grab the buffer contents with `Lines.Bytes()`\nand write the result to disk.\n\nThe saved files never end with a newline character, even though\neverything else in the editor is preserved byte-for-byte. This causes\n`git` to flag \"No newline at end of file\" on every save, and a few\ncommand-line tools (wc, some linters/formatters) complain similarly.\n\nBy convention text files end with a newline (per POSIX). Could\n`Lines.Bytes()` return the content with a final `\\n` so the output is\nsuitable for writing straight to a `.go`/`.txt`/etc. file without any\npost-processing on my end?"} {"task_id": "format-code-task-000698", "source_id": "format-code-task-000698", "domain": "code", "task_path": "tasks/format-code-task-000698", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fff6a733f0ef2496bc18a2aa35674ece2ae98fbb93666cc19750a3520ed98279", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `udiskie-umount` to support an all-devices mode. When I run `udiskie-umount -a` or `udiskie-umount --all`, it should not require any DEVICE argument; instead it should find every handleable root device known to udiskie and attempt to remove each one.\n\nFor this mode, udiskie should use the same automatic removal behavior it uses for device trees: recursively unmount mounted filesystems and lock unlocked encrypted volumes below each root. The `--detach`, `--no-detach`, `--eject`, `--no-eject`, `--lock`, and `--no-lock` strategy flags should apply to each selected root in all-devices mode, so `udiskie-umount --all --detach --eject --no-lock` attempts those requested follow-up operations for every selected device.\n\nThe command should return exit code 0 when every attempted root removal succeeds, and exit code 1 when any selected root removal fails. `udiskie-umount --help` should document the invocation as accepting either `-a`/`--all` or one or more device paths, and the shell completions and manpage should expose `-a, --all` as the option for unmounting all handleable devices."} {"task_id": "format-code-task-000699", "source_id": "format-code-task-000699", "domain": "code", "task_path": "tasks/format-code-task-000699", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d2d3a2ed53cbccf49dd9ab66ffc59f902a503820cd000bbf609214bd194f4f05", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: built-in base64 string validator\n\nI'm using Zod to validate API payloads, and a few of my endpoints accept fields that are expected to be base64-encoded strings (think things like binary blobs uploaded as base64, signed tokens, webhook payloads, etc.).\n\nWhen writing the schema for one of those endpoints, I reached for the obvious thing:\n\n```ts\nconst Schema = z.object({\n payload: z.string().base64(),\n});\n```\n\n…and TypeScript immediately complained that `base64` doesn't exist on `ZodString`. Looking at the docs, I see Zod ships built-in string format validators for a bunch of common formats — `email`, `url`, `uuid`, `cuid`, `cuid2`, `ulid`, `ip`, `datetime`, `date`, `time`, etc. — but base64 isn't one of them, which surprised me a bit since it's such a common format on the wire.\n\nRight now my workaround is to drop down to a `.refine()` with a hand-rolled regex, but that means:\n\n- Every project I work on ends up with its own slightly-different base64 regex copy-pasted in.\n- The resulting `ZodIssue` doesn't look like the ones from the other built-in string formats (different `code` / `validation` shape), so my shared error-formatting / i18n layer has to special-case it.\n\nIt feels pretty natural for this to live alongside the other string format validators in core Zod. Would you accept a PR that adds `z.string().base64()` as a first-class string validator, behaving consistently with the existing ones (chainable on `ZodString`, accepts an optional custom error message, fails with a regular `ZodError` whose issue is shaped like the other string-format issues)?\n\nHappy to put up the PR if you're open to it."} {"task_id": "format-code-task-000700", "source_id": "format-code-task-000700", "domain": "code", "task_path": "tasks/format-code-task-000700", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:554e37a1a15a78324365b226b4aff733274027c4603617a704490019d4bed8ea", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Problem Statement\nI’m working with CIE Lab values and need to compare colours using the HyAB metric from Abasi 2020, but I can’t find any way to do that through `colour.difference` or `delta_E`. Could you add HyAB as one of the supported colour-difference methods so I can use it like the existing metrics?\n\n### Expected outcomes\n- `colour.difference.delta_E_HyAB(Lab_1, Lab_2)` is available for CIE L*a*b* inputs and returns the HyAB colour-difference value as defined by Abasi 2020.\n- `colour.difference.delta_E(Lab_1, Lab_2, method=\"HyAB\")` selects the same HyAB result, and HyAB is discoverable through the existing public `colour.difference.DELTA_E_METHODS` method registry.\n- The `colour.difference` documentation includes a HyAB entry and lists `delta_E_HyAB` in the public API summary.\n\n### Implementation notes\n- Match the existing colour-difference API style used in this package.\n- The exact internal structure, helper layout, dispatch mechanism, and validation placement are up to the implementation, as long as the public outcomes above hold."} {"task_id": "format-code-task-000701", "source_id": "format-code-task-000701", "domain": "code", "task_path": "tasks/format-code-task-000701", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6e1728ed7a8533904096b98ad51905841520ea37ee74c0a0cd7baeda798983ae", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nMake the template-file utilities safe to use as the generator's filesystem boundary. Today they create or truncate the destination before template execution, discard execution errors, and treat symlinks or regular files as output directories. A bad render can therefore destroy an existing file or publish partial output, while callers cannot tell that the operation failed.\n\nKeep the existing parameter lists of TemplateFileAndOverwrite and TemplateFileIfDoesNotExist, but have each return a receive-only <-chan error. Existing statement-style calls that ignore the result must remain source-compatible. Every invocation returns its own non-nil channel, reports exactly one terminal value, and then closes it. The value is nil for a successful write and for an intentional create-only skip, and non-nil for validation, directory, or rendering failures. Once the supplied wait group has finished, the result must already be readable and the channel closed; ignoring the result channel must not prevent wait-group completion. Multiple invocations sharing a wait group must not share results or let one render's error affect another.\n\nRender before publishing. During template execution an overwrite destination must still contain its old bytes, and a new destination must not exist yet. Only a fully successful render may be published. A successful overwrite preserves the destination's existing permission bits; a newly created file uses 0644 permissions. If execution emits some bytes and then returns an error, propagate an error containing the underlying reason, leave an existing destination's bytes and permissions unchanged, leave an absent destination absent, and clean up any temporary artifacts in the output directory. TemplateFileIfDoesNotExist must continue to skip an existing destination without executing its template and report nil.\n\nTreat fileName as a basename, not another path. Both helpers must reject empty names, ., .., absolute names, or names containing a platform path separator, without changing the filesystem. Output directories must be real directories: do not follow a symlink used as fileDir or anywhere in its existing ancestor chain, and reject an existing non-directory. Apply the same directory safety to CreateDirIfDoesNotExist: it may create missing real nested directories and accept an existing real directory, but it must return an error for a symlink target, a symlinked ancestor, or an existing non-directory, without modifying the symlink target. The template helpers surface these directory failures through their completion channels rather than panicking or publishing output."} {"task_id": "format-code-task-000702", "source_id": "format-code-task-000702", "domain": "code", "task_path": "tasks/format-code-task-000702", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0d18ef08aef755be3928f68f9cdfdc63b6b247b72121c5110e96463871ca7ecb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Support JSON as a configuration file format\n\nRight now Commitizen only reads its configuration from TOML files (`pyproject.toml` / `.cz.toml`). For Python projects that's fine, but I'm using Commitizen on a non-Python project (a JS codebase) where TOML is pretty foreign — none of our other tooling uses it, and contributors have to learn yet another format just to tweak commit rules.\n\nJSON, on the other hand, is already everywhere in these ecosystems (`package.json`, tsconfig, eslint, prettier, …). It would be great if Commitizen could pick up a JSON config file the same way it picks up `.cz.toml` today, so I can drop something like `.cz.json` at the repo root and have it Just Work.\n\nConcretely, I'd expect:\n\n- All the same settings that work in `.cz.toml` (name, version, version_files, style, the whole `customize` block, …) to be expressible in JSON and behave identically.\n- `cz init` to be a viable way to bootstrap a JSON config, not only a TOML one — i.e. if I'm in a JS project I shouldn't be forced to end up with a `.cz.toml`.\n- `cz bump` and friends that write back to the config (e.g. updating the version) to keep working when the config happens to be JSON.\n\nThe docs currently say customization \"is only supported when configuring through toml\" — ideally that restriction would also go away once JSON is supported.\n\nRelated: #120"} {"task_id": "format-code-task-000703", "source_id": "format-code-task-000703", "domain": "code", "task_path": "tasks/format-code-task-000703", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e894c0540912bb1dab8f58908eccd8ba554b0ee97213a611628cfd2b3ed63fa3", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `del self.options.fPIC` fails when `shared` is set from the command line\n\nI have a pretty standard recipe that follows the shared/fPIC pattern that's all over the conan docs:\n\n```python\nclass Pkg(ConanFile):\n name = \"pkg\"\n settings = \"os\", \"arch\", \"compiler\", \"build_type\"\n options = {\"shared\": [True, False], \"fPIC\": [True, False]}\n default_options = {\"shared\": False, \"fPIC\": True}\n\n def configure(self):\n if self.options.shared:\n del self.options.fPIC\n```\n\nIf I just run `conan install .` everything works fine — `shared` defaults to `False`, the `if` is false, nothing gets removed.\n\nBut the moment I try to build the shared variant from the command line:\n\n```\nconan install . -o pkg*:shared=True --build=missing\n```\n\nit blows up with:\n\n```\nConanException: Incorrect attempt to remove option 'fPIC' with current value 'True'\n```\n\nIt looks like as soon as `shared` has been given a value from outside the recipe, `del self.options.fPIC` is no longer allowed — even though removing fPIC *based on* the shared value is exactly the documented pattern and the whole reason this `configure()` body exists. The conditional removal is the only way to express \"fPIC doesn't make sense when shared=True\", and it has to happen in `configure()`, which runs after the command-line / profile options have been applied.\n\nI'd expect `del self.options.fPIC` inside `configure()` to just work in this case, regardless of whether `shared` got its value from `default_options` or from `-o`."} {"task_id": "format-code-task-000705", "source_id": "format-code-task-000705", "domain": "code", "task_path": "tasks/format-code-task-000705", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ed5d608d51ddd631c6596e57aaf194db6c4794bdaad697cf5c47ed110f73d8b6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Conan crashes when a cached recipe/package is missing its manifest.txt\n\nI keep a local conan cache shared across a few CI machines, and occasionally one of the cached entries ends up without a `manifest.txt` (manual cleanup gone wrong, partial copy between machines, an interrupted download, etc.). The recipe / package folder itself is still there, just the manifest file is gone.\n\nWhen I then run `conan install` with integrity checking enabled against that cache, conan blows up instead of doing something reasonable. From the user's point of view the situation is recoverable — the recipe/package can simply be re-fetched from the remote — but right now I get a hard crash partway through and have to manually `conan remove` the entry to get unstuck.\n\nI'd expect that if the integrity check finds no manifest at all in the cache for a given recipe or package, conan treats that entry the same way it treats one whose checksums don't match: warn that it's bad, drop it, and refetch from the configured remote. Having a missing manifest take down the whole install is pretty unfriendly, especially because the fix from the user side is exactly the same as for a corrupted manifest."} {"task_id": "format-code-task-000706", "source_id": "format-code-task-000706", "domain": "code", "task_path": "tasks/format-code-task-000706", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:91fa923a0fd51d9407e9f0d4eaa2f6585e42b19ba4c86154629fe8e55d408f1a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### `cmake_find_package_multi` doesn't honor the kebab-case lowercase config file naming\n\nAccording to the CMake docs for [`find_package` Config Mode](https://cmake.org/cmake/help/latest/command/find_package.html#config-mode-search-procedure), CMake recognizes two naming conventions for the package config files:\n\n```\nConfig.cmake\n-config.cmake\n```\n\n…and similarly for the version file (`ConfigVersion.cmake` / `-config-version.cmake`).\n\nIn my recipes I've been setting a lowercase filename for the multi generator, e.g.\n\n```python\ndef package_info(self):\n self.cpp_info.filenames[\"cmake_find_package_multi\"] = \"mylib\"\n # ...\n```\n\nbecause downstream I want to do `find_package(mylib CONFIG REQUIRED)` and follow the lowercase-with-dash style that CMake documents. But after `conan install` the only files I see in the generators folder are:\n\n```\nmylibConfig.cmake\nmylibConfigVersion.cmake\nmylibTargets.cmake\nmylibTarget-release.cmake\n```\n\nI'd expect that when the filename I'm asking for is all lowercase, the generator produces the `mylib-config.cmake` / `mylib-config-version.cmake` form instead, since that's the convention CMake associates with lowercase package names. When the filename has uppercase characters (e.g. `MyLib`) the current `MyLibConfig.cmake` style is fine and shouldn't change.\n\nCould `cmake_find_package_multi` be taught to pick the right naming style based on the case of the filename?"} {"task_id": "format-code-task-000707", "source_id": "format-code-task-000707", "domain": "code", "task_path": "tasks/format-code-task-000707", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2952a0bbfb9958537d1be60c3bda4e6c7f2a0a5935c5e74951d60f93421acdaf", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nOut of memory crash when objects have a large length field unrelated to array size\nWhile writing tests for [GraphBrainz](https://github.com/exogen/graphbrainz) I noticed that my latest test was consistently crashing AVA. Running the single test alone was enough to cause an OOM error, always in `concordance.serialize`, and the object I was snapshotting was not large by any means.\n\nI narrowed it down to the `length` field on my objects. This field has nothing to do with array lengths, rather the meaning in this context is a recording length in milliseconds.\n\n```js\n[\n {\n title: \"Airbag\",\n length: 284400\n },\n {\n title: \"Paranoid Android\",\n length: 383493\n },\n // …more…\n]\n```\n\nConcordance seems to be assigning some special meaning to these `length` fields and allocating memory based on them.\n\nTry this and you'll notice it takes quite a long time and uses a lot of memory:\n\n```\n> concordance = require('concordance');\n> concordance.serialize(concordance.describe({ length: 12345678 }));\n```\n\nIt's possible this could be the cause of several of the OOM bug reports opened for AVA.\n\n
\nStack trace\n\n```\n==== JS stack trace =========================================\n\n 0: ExitFrame [pc: 0x3241121dc01d]\nSecurity context: 0x3f5ee919e681 \n 1: encode [0x3f5ecb13e4a1] [/Users/brianbeck/Projects/graphbrainz/node_modules/concordance/lib/encoder.js:~177] [pc=0x3241125889c2](this=0x3f5ecfec66f9 ,serializerVersion=2,rootRecord=0x3f5ecb13e771 ,usedPlugins=0x3f5ecb145bf9 )\n 2: serialize [0x3f5ecb102e21] [/Users/b...\n\nFATAL ERROR: Ineffective mark-compacts near heap limit Allocation failed - JavaScript heap out of memory\n 1: 0x100039dbf node::Abort() [/Users/brianbeck/.nvm/versions/node/v10.10.0/bin/node]\n 2: 0x100039fc9 node::OnFatalError(char const*, char const*) [/Users/brianbeck/.nvm/versions/node/v10.10.0/bin/node]\n 3: 0x1001d1375 v8::internal::V8::FatalProcessOutOfMemory(v8::internal::Isolate*, char const*, bool) [/Users/brianbeck/.nvm/versions/node/v10.10.0/bin/node]\n 4: 0x10059c572 v8::internal::Heap::FatalProcessOutOfMemory(char const*) [/Users/brianbeck/.nvm/versions/node/v10.10.0/bin/node]\n 5: 0x10059f045 v8::internal::Heap::CheckIneffectiveMarkCompact(unsigned long, double) [/Users/brianbeck/.nvm/versions/node/v10.10.0/bin/node]\n 6: 0x10059aeef v8::internal::Heap::PerformGarbageCollection(v8::internal::GarbageCollector, v8::GCCallbackFlags) [/Users/brianbeck/.nvm/versions/node/v10.10.0/bin/node]\n 7: 0x1005990c4 v8::internal::Heap::CollectGarbage(v8::internal::AllocationSpace, v8::internal::GarbageCollectionReason, v8::GCCallbackFlags) [/Users/brianbeck/.nvm/versions/node/v10.10.0/bin/node]\n 8: 0x1005995c5 v8::internal::Heap::CollectAllAvailableGarbage(v8::internal::GarbageCollectionReason) [/Users/brianbeck/.nvm/versions/node/v10.10.0/bin/node]\n 9: 0x1005a5a21 v8::internal::Heap::AllocateRawWithRetryOrFail(int, v8::internal::AllocationSpace, v8::internal::AllocationAlignment) [/Users/brianbeck/.nvm/versions/node/v10.10.0/bin/node]\n10: 0x100574da6 v8::internal::Factory::NewFixedArrayWithFiller(v8::internal::Heap::RootListIndex, int, v8::internal::Object*, v8::internal::PretenureFlag) [/Users/brianbeck/.nvm/versions/node/v10.10.0/bin/node]\n11: 0x10051c2de v8::internal::(anonymous namespace)::ElementsAccessorBase >::GrowCapacity(v8::internal::Handle, unsigned int) [/Users/brianbeck/.nvm/versions/node/v10.10.0/bin/node]\n12: 0x1007b6f9f v8::internal::Runtime_GrowArrayElements(int, v8::internal::Object**, v8::internal::Isolate*) [/Users/brianbeck/.nvm/versions/node/v10.10.0/bin/node]\n13: 0x3241121dc01d\nAbort trap: 6\n```\n\n"} {"task_id": "format-code-task-000708", "source_id": "format-code-task-000708", "domain": "code", "task_path": "tasks/format-code-task-000708", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4c8ef0bb6b551b8c495c59da057655b332bfc5b0ce294d07ba8e72c6ff01e7d7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI'm hitting a bare `AssertionError` (no message at all) when conda tries to install into my env — the traceback just ends in `PrefixData.insert` with an empty assert, so I have no idea which package is the problem. It seems to happen after the solver finishes, during the actual write-out step. I dug a bit and it looks like the same package name shows up twice in what the solver decided to install, but I can't tell from the error which one it is. Could you take a look?\n\nExpected outcomes:\n- When conda computes the final set of package records for an environment operation, the result should not contain more than one record for the same package name.\n- If insertion of a package record still encounters a duplicate package name, the raised `AssertionError` should include enough context to identify the conflicting package name and make the failure actionable instead of appearing as an empty assertion.\n\nImplementation notes:\n- The specific data structures, filtering location, and validation mechanism are up to the implementer.\n- Preserve the existing solver and prefix write-out semantics except for eliminating duplicate-name records and improving the duplicate insertion failure message."} {"task_id": "format-code-task-000710", "source_id": "format-code-task-000710", "domain": "code", "task_path": "tasks/format-code-task-000710", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c29214b29f8847210de763473da5d912981e5843bb3c96fb82deb033331231bd", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Hooks directory monitor gets out of sync with the filesystem\n\nI'm using podman's hook directory monitoring (dropping `*.json` files into a hooks.d directory and letting podman pick them up at runtime). I've run into a few situations where what podman thinks is in the directory doesn't match what's actually there.\n\n### 1. One bad hook file hides the other valid hooks\n\nIf I drop two hook files into the directory and one of them is malformed (bad JSON / fails validation), the valid one next to it doesn't get picked up either. I'd expect podman to log/report the error for the bad file but still register the good one — instead the good hook is silently missing.\n\nThis shows up both at startup and while the monitor is running: as soon as something invalid is in the directory, working hooks alongside it stop being honored.\n\n### 2. `mv`-ing a hook file out of the directory leaves it active\n\nIf I `mv somehook.json /tmp/` (instead of `rm`-ing it), podman keeps acting like `somehook.json` is still installed and keeps running it on container start. The file is no longer in the hooks directory, so it shouldn't be applied anymore.\n\n`rm` works correctly; `mv` out of the directory does not.\n\n### 3. Renaming a hook file makes it run twice\n\nIf I rename a hook file in place — e.g. `mv myhook.json myhook-renamed.json` within the hooks directory — podman ends up running the hook twice on the next container: once under the old name (which it never dropped) and once under the new name. Only the renamed file actually exists on disk, so it should run once.\n\n### 4. `chmod` on a hook file isn't reflected\n\nChanging permissions on a hook file (e.g. so it becomes unreadable, or fixing it back) doesn't seem to cause the monitor to re-evaluate the directory — the in-memory state just stays whatever it was.\n\n---\n\nIn all of these cases, what I'd like is simple: the set of active hooks should match the set of valid `*.json` files currently in the configured hook directories. Errors on individual files should be reported but shouldn't prevent the rest from loading, and filesystem operations like rename/move/chmod should be reflected the same way create/delete already (mostly) are."} {"task_id": "format-code-task-000712", "source_id": "format-code-task-000712", "domain": "code", "task_path": "tasks/format-code-task-000712", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:47b3532699382de8ea4583a2f5fe1856a03ed60a13c8a63c4e61ef1a27004d87", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `netassert run --input-file tests.yaml --kubeconfig kubeconfig` to support Kubernetes workload references in each test's `src.k8sResource` and, when present, `dst.k8sResource`. A `k8sResource` with `kind: pod`, `name`, and `namespace` should resolve to that exact Pod only when it exists, is in the Running phase, and has a non-empty Pod IP; otherwise the test should fail before any scanner or sniffer container is launched, with a reason that says the Pod is missing, not running, or has no IP.\n\nFor `kind: deployment`, NetAssert should find the named Deployment in the requested namespace, choose a ReplicaSet owned by that Deployment with at least one replica, then choose one running owned Pod whose labels match the ReplicaSet selector and whose Pod IP is set. If no such ReplicaSet or Pod exists, the test should fail with a reason tied to the named Deployment and namespace.\n\nFor `kind: statefulset`, NetAssert should find the named StatefulSet, reject StatefulSets scaled to zero, then choose one running owned Pod matching the StatefulSet selector with a Pod IP. For `kind: daemonset`, NetAssert should find the named DaemonSet, reject DaemonSets with zero available Pods, then choose one running owned Pod matching the DaemonSet selector with a Pod IP.\n\nIn a TCP test where the source is a Deployment and the destination is a Pod, the scanner should be launched in the resolved source Pod and should target the resolved destination Pod IP. In a UDP test where both source and destination are Kubernetes resources, the sniffer should be launched in the resolved destination Pod and the scanner should be launched in the resolved source Pod, using the destination Pod IP as the scanner target."} {"task_id": "format-code-task-000714", "source_id": "format-code-task-000714", "domain": "code", "task_path": "tasks/format-code-task-000714", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d1ee762a50b3ecb158d0983a79d86cf04991fc13d92c09bd284df55398d5ae3e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nThe JSON request-body processor currently flattens some object-shaped payloads into ARGS, but JSON variables are not actually inspectable by rules and array-shaped documents can be rejected even though JSON arrays are part of the supported body shape. Make JSON body inspection consistent and useful to WAF policies.\n\nWhen a body is processed as JSON, recursively flatten both object and array containers into the established dotted keys (for example, `json.user.name`, `json.items.0`, and deeper nested indexes). Every array element must be addressable, including string, number, boolean, null, object, and nested-array elements. Convert scalar values to their textual form (`null` becomes an empty string) without dropping them. The resulting flattened values must remain available through both `ARGS_POST` and the combined `ARGS` collection using those keys.\n\nThe public `BodyProcessor` lookup path used by `SecRule JSON:` must resolve existing nested object and array paths to their scalar value. A path that is absent must produce no match rather than matching another field or the whole document. Malformed JSON and valid JSON documents whose root is a scalar (rather than an object or array) must be reported as processing errors. Keep the existing object flattening behavior and normal request-body error propagation intact."} {"task_id": "format-code-task-000715", "source_id": "format-code-task-000715", "domain": "code", "task_path": "tasks/format-code-task-000715", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1636d3ac9ed0f4ee8b6cc2fae6dbe52df28e4fe8dd1f5cc8276b181203a4a0c7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Client doesn't retry other machines when a request to one machine fails at the network layer\n\nI'm using go-etcd against a multi-node etcd cluster and relying on the\nclient's built-in retry/fail-over to keep things working when one node\nbecomes unavailable (machine down, connection refused, transient network\nhiccup, etc.).\n\nWhat I expected: if the client can't even get a response back from one\nof the machines in the cluster, it should move on and try another one,\nup to the usual retry limit.\n\nWhat actually happens: the very first network-level failure bubbles\nstraight back out of `SendRequest` to my code. The other machines in\nthe cluster are never tried, even though they're healthy and listed in\n`cluster.Machines`. Effectively a single bad node takes the whole call\ndown instead of being failed over.\n\nMinimal repro: point a `Client` at a cluster where the first machine in\nthe list is unreachable (e.g. nothing listening on that port, or pull\nits network), and do any normal `Get`/`Set`. The call returns the\nunderlying dial/connection error immediately instead of retrying the\nother members.\n\nThe HTTP-status retry path (e.g. 500 from the server) seems to behave\ncorrectly — only the \"no response at all\" path is broken. Since this is\nexactly the case where retrying matters most, I'd expect the default\nretry policy to keep going here rather than surface the error on the\nfirst attempt."} {"task_id": "format-code-task-000716", "source_id": "format-code-task-000716", "domain": "code", "task_path": "tasks/format-code-task-000716", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b0089a302ec032a4a8be3c00e32a842d993faa23108cf13bea8fba363931b20a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI've got a bunch of Ignition configs floating around that still declare `\"ignition\": { \"version\": \"2.0.0\" }`, and when I feed them through `config.Parse` they blow up because it tries to read them as the latest schema. I don't really want to hand-rewrite them all to the newest version — can the parser just recognize 2.0.0 configs and handle them properly, ideally giving me back the same internal config type so the rest of my code doesn't care which version the input was?\n\n# Expected outcomes\n\n- `config.Parse` should accept valid Ignition configs that declare version `2.0.0`.\n- Parsing a valid `2.0.0` config should return the repository’s current `types.Config` representation, so callers can consume it like configs parsed from newer supported versions.\n- Supported user-visible configuration data from valid `2.0.0` configs should be preserved in the current config representation.\n- Invalid `2.0.0` configs should still fail with an error and validation report rather than being silently accepted.\n- Existing behavior for other supported versions should continue to work.\n\n# Implementation notes\n\n- Do not require callers of `config.Parse` to know whether the input was originally version `2.0.0` or a newer supported version.\n- Prefer behavior-compatible integration with the existing config parsing and validation conventions.\n- The exact parsing organization and compatibility implementation are up to the implementer."} {"task_id": "format-code-task-000717", "source_id": "format-code-task-000717", "domain": "code", "task_path": "tasks/format-code-task-000717", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c3de659162347d153f06da7c7fa2c847d4224ebba1278aad00455e424f6ee187", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a reusable pan & zoom controller for the plotting views\n\nOur VisPy-based views need a small, self-contained controller that owns the\n**pan** (2D translation) and **zoom** (2D scale) state of a scene and keeps any\nattached GPU programs in sync. Right now the pan/zoom logic is tangled up inside\nthe canvas code and there is no plain object we can unit-test or reuse.\n\nPlease add a `PanZoom` class, importable as `from phy.plot import PanZoom`,\nthat implements the following behavior.\n\n## Construction & state\n\n- `PanZoom(pan=(0.0, 0.0), zoom=(1.0, 1.0), zmin=..., zmax=...)`.\n- By default the controller starts at the origin (`pan == (0, 0)`) with unit\n zoom (`zoom == (1, 1)`). `zmin`/`zmax` have sensible finite defaults.\n- `pan` and `zoom` are readable/writable properties. Each holds a pair of\n values (one per axis); reading either returns a length-2 value. A scalar\n assigned to `zoom` applies to both axes.\n\n## Zoom limits\n\n- `zoom` is always kept within `[zmin, zmax]` per axis: assigning a value\n outside the range clamps it to the nearest bound.\n- `zmin` and `zmax` are readable/writable and may never cross: setting `zmin`\n above the current `zmax` leaves `zmin` capped at `zmax`, and setting `zmax`\n below the current `zmin` leaves `zmax` raised to `zmin`.\n- Changing either bound immediately re-clamps the current `zoom` so it still\n lies within the new range.\n\n## Coordinate transforms\n\n- `map(coords)` converts data coordinates to scene coordinates, applying pan\n first and then zoom about the origin: a point `(x, y)` maps to\n `(zoom_x * (x + pan_x), zoom_y * (y + pan_y))`.\n- `imap(coords)` is the exact inverse of `map`.\n- Both accept either a single `(x, y)` point or an array of points shaped\n `(n, 2)`, and return a result of the same shape.\n\n## Interactive updates\n\n- `pan_delta(d)` shifts the current pan by the 2D amount `d`.\n- `zoom_delta(d, center=(0.0, 0.0))` applies an incremental zoom. The magnitude\n of the change grows with `|d|`; a positive `d` zooms in (increases zoom) and a\n negative `d` zooms out, with the result clamped to `[zmin, zmax]`. The zoom is\n **centered** at `center` (given in scene coordinates): the data point located\n under `center` before the operation stays under it afterwards — that is,\n `imap(center)` is unchanged by the call.\n- `reset()` returns the controller to the identity state (`pan == (0, 0)`,\n `zoom == (1, 1)`).\n\n## GPU program binding\n\n- `add(programs)` registers one program or an iterable of programs with the\n controller. A program is any object supporting item assignment (e.g. a VisPy\n `gloo` program, or a plain dict).\n- On registration, and on every subsequent change to pan or zoom (through any\n of the operations above), each registered program receives the current state\n via the uniforms `u_pan` and `u_zoom`, each a length-2 value.\n"} {"task_id": "format-code-task-000718", "source_id": "format-code-task-000718", "domain": "code", "task_path": "tasks/format-code-task-000718", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c5dc2d5560531975482a1f31664f0096c99fcf27ceb15a85afeff1d556aa3750", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\n我现在在 Cortex 里上传 Alertmanager 配置时，receiver 的 HTTP 通知鉴权基本只能用 basic auth 或 bearer token，但我们这边的 webhook 接口要求走 OAuth2 client credentials。能不能让 `http_config` 这类通知配置直接支持 `oauth2`，比如用 client id、inline secret 和 token URL 去拿 token？另外这个配置是租户自己传的，最好不要允许它通过 `client_secret_file` 去读服务器上的本地文件。\n\n# Expected outcomes\n\n- **OAuth2 notification authentication**\n - Alertmanager receiver HTTP notification configuration accepts an `oauth2` block under `http_config` where HTTP client configuration is supported.\n - A valid OAuth2 client-credentials configuration using `client_id`, inline `client_secret`, and `token_url` is accepted and used to obtain an access token for outgoing notification requests.\n - Optional OAuth2 settings such as `scopes` and `endpoint_params` are preserved and applied when obtaining the token.\n\n- **Tenant-uploaded config safety**\n - Tenant-uploaded Alertmanager configurations must reject any OAuth2 configuration that sets `client_secret_file`.\n - Rejection must happen during configuration validation, before the server can read the referenced local file.\n - The validation failure should clearly state that setting OAuth2 `client_secret_file` is not allowed.\n\n- **OAuth2 validation**\n - When an `oauth2` block is present, `client_id` and `token_url` are required.\n - Exactly one OAuth2 client secret source is allowed: inline `client_secret` or `client_secret_file`.\n - Missing required OAuth2 fields, missing a secret source, or setting both secret sources should produce configuration validation errors.\n\n- **Authentication mutual exclusion**\n - OAuth2 authentication must participate in the existing HTTP authentication mutual-exclusion rules.\n - `oauth2` must not be accepted together with any other existing HTTP authentication mechanism in the same HTTP client configuration.\n\n# Implementation notes\n\n- The exact internal data structures, validation placement, and HTTP client wiring are up to the implementation.\n- Preserve existing Alertmanager configuration behavior unrelated to OAuth2.\n- Do not introduce a path that lets tenant-uploaded Alertmanager configuration read OAuth2 client secrets from server-local files."} {"task_id": "format-code-task-000719", "source_id": "format-code-task-000719", "domain": "code", "task_path": "tasks/format-code-task-000719", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3c1f061d22a1afdc6306c274bcc9c8348193eed7847d7105b950182f429c1a37", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `max_fetched_series_per_query` is not enforced for the `/series` API\n\nWe rely on `max_fetched_series_per_query` to protect our cluster from queries that fan out into too many series. With the limit configured, running a range query with a very loose matcher (e.g. `{__name__=~\".+\"}`) is correctly rejected once the limit is exceeded — which is exactly what we want.\n\nHowever, the `/api/v1/series` endpoint doesn't seem to apply the same limit. Hitting `/series` with the same kind of broad matcher happily returns well above `max_fetched_series_per_query` series, and we can see noticeable memory pressure on the distributors when this happens. The series response itself can also be huge.\n\nIt feels inconsistent that one read path enforces the limit while another silently ignores it. From an operator's point of view we'd like a single configured value to protect *all* query APIs uniformly, so it would be great if `/series` honored `max_fetched_series_per_query` too and rejected the request (rather than continuing to accumulate) once the limit is reached."} {"task_id": "format-code-task-000720", "source_id": "format-code-task-000720", "domain": "code", "task_path": "tasks/format-code-task-000720", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1b38d7d38acc451dfdca3aab0a7a73dffe9b9a92e6fbc28c5d53f8d9969580ea", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## S3 bucket client keeps retrying after the request context is canceled\n\nWhen running Cortex against S3-backed object storage, I noticed that if an in-flight operation's context gets canceled (either explicitly, or because an upstream deadline expires), the bucket client doesn't actually stop right away. It keeps going around the retry loop and re-issuing the operation, even though the context is already done so every attempt is guaranteed to fail immediately.\n\nThe two visible symptoms:\n\n1. **Extra latency unwinding canceled requests.** After cancellation/timeout I'd expect the bucket call to return essentially immediately, but instead it sits there going through retry iterations before finally giving up.\n2. **Misleading error logs.** Each canceled operation ends up producing a `bucket operation fail after retries` error in the logs, which makes it look like S3 is unhealthy and the retries got exhausted, when in reality the caller had already canceled the request and nothing was going to succeed regardless.\n\nThe retry wrapper already knows how to short-circuit and return immediately for some error classes (object-not-found and access-denied) — those don't get retried because retrying them is pointless. The same logic should apply when the operation failed because the caller's context was canceled or its deadline was exceeded: there's no point spinning through more attempts, we should just propagate the context error back to the caller right away and not log it as a \"fail after retries\" outcome."} {"task_id": "format-code-task-000721", "source_id": "format-code-task-000721", "domain": "code", "task_path": "tasks/format-code-task-000721", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6de8fe5e951562653770909a2828bac93e3ffc6ad124fe9d2847dc1146754e59", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI turned on resource monitoring with `MonitoredResources` set to track heap, but the heap utilization always reads as 0 even when the process is clearly using a lot of memory, so the heap-based limiting never kicks in. On top of that, I'm scraping `/metrics` and I see `cortex_resource_based_limiter_limit` only carries a `component` label — when I configure both cpu and heap on the same component I can't tell which limit belongs to which resource, they look like they're stomping on each other. I also tried putting a resource name that isn't cpu or heap in the config just to test and it started up fine without complaining, which felt off. Can you take a look?\n\n## Expected outcomes\n\n- Resource monitoring should report heap utilization from the process's actual heap usage rather than remaining at zero or otherwise unusable when heap monitoring is enabled.\n- Heap-based resource limiting should be able to make decisions from the reported heap utilization, so configured heap limits are not silently ineffective.\n- The `cortex_resource_based_limiter_limit` metric should distinguish limits for different monitored resources on the same component, including separate cpu and heap entries.\n- Configuring an unsupported monitored resource name should fail startup/configuration with a clear error identifying the unknown resource.\n- Leaving monitored resources unset or effectively empty should continue to skip resource monitor initialization without error.\n- Enabling resource monitoring and allowing it to run should not panic while recording resource utilization.\n\n## Implementation notes\n\n- The exact validation location, data structures, and internal monitor/limiter organization are up to the implementer.\n- Prefer externally observable behavior for validation: configuration outcomes, exported metrics, resource utilization behavior, and absence of runtime panics.\n- Keep existing supported resource names and existing public configuration/metric surfaces compatible while fixing the incorrect behavior."} {"task_id": "format-code-task-000722", "source_id": "format-code-task-000722", "domain": "code", "task_path": "tasks/format-code-task-000722", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:813d39ad573e8a0294490e08496e97a6b12000b1cd6f8d02c53531d672686fe5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Record import: a file decoder for compose\n\nWe want to support importing records into compose from uploaded data files. Two\nfile shapes need to be handled uniformly: **flat** files (CSV-style, where the\nfirst row is a header naming the columns) and **structured** files (JSONL —\none JSON object per line). The rest of the import pipeline shouldn't care which\nshape it's dealing with; it just wants to know how many records there are, what\nthe columns/field names are, and to walk the records turning each one into a\n`compose/types.Record`.\n\nPlease add a decoder package under `compose/decoder` (import path\n`github.com/cortezaproject/corteza-server/compose/decoder`) that exposes two\nconstructors:\n\n- `NewFlatReader(r, f)` — wraps a row reader `r` whose only requirement is a\n `Read() ([]string, error)` method (a `*csv.Reader` satisfies this) together\n with a seekable handle `f` (`io.ReadSeeker`) to the same raw data.\n- `NewStructuredDecoder(d, f)` — wraps a streaming decoder `d` exposing\n `Decode(interface{}) error` and `More() bool` (a `*json.Decoder` satisfies\n this) together with a seekable handle `f` (`io.ReadSeeker`) to the same raw\n data.\n\nBoth constructors return a value exposing the same three methods:\n\n- `Header() []string` — the field/column names.\n - For flat input the header is the first physical row, returned in its\n original column order.\n - For structured input there is no dedicated header row; the names are the\n keys of the first object (order is not significant).\n\n- `EntryCount() (uint64, error)` — the number of **data** records in the file,\n determined from the seekable handle and leaving that handle rewound to the\n start afterwards.\n - For flat input the header row is not a data record, so a file with a header\n and N data rows reports N; an empty file reports 0.\n - For structured input every line is a data record (a file with N objects\n reports N).\n\n- `Records(fields map[string]string, create RecordCreator) error` — iterates\n over the data records, builds a `types.Record` for each, and invokes the\n `create` callback once per record (a `RecordCreator` is\n `func(*types.Record) error`). `fields` maps a source column/key name to the\n target record field name.\n - Only the columns/keys present in `fields` are imported; anything else in\n the source is ignored.\n - A target name that matches a known system field is set directly on the\n record struct: `recordID`/`ID` → ID, `moduleID`, `namespaceID`, `ownedBy`,\n `createdBy`, `createdAt`, `updatedBy`, `updatedAt`, `deletedBy`,\n `deletedAt`. (The numeric IDs parse as base-10 unsigned integers; the\n timestamps parse as RFC3339.)\n - Every other target name produces a `types.RecordValue` (with that `Name`\n and the cell's string value) appended to the record's `Values`.\n - If any target name in `fields` is the empty string, importing fails with an\n error and the callback must not be relied upon to have succeeded.\n\nOne subtlety: obtaining the header must not consume any data record. After\ncalling `Header()`, a subsequent full iteration via `Records()` must still visit\nevery data record — including the first object of a structured file, which is a\nreal data record even though its keys were used to derive the header.\n"} {"task_id": "format-code-task-000723", "source_id": "format-code-task-000723", "domain": "code", "task_path": "tasks/format-code-task-000723", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:837362be0a1fa7f3999ac2c8e694d14b73ab7220fca342ff69ffd4bd84ae465d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Role REST API silently drops `handle` on create / update\n\nI'm wiring up a small admin tool against the Corteza system API and I want to give each role a stable, machine-friendly `handle` next to its display `name` (similar to what users and the signup endpoint already accept). The `Role` type itself looks like it carries a handle, and other resources expose it on their endpoints, so I assumed `POST /roles/` and `PUT /roles/{roleID}` would too.\n\nWhat I'm seeing:\n\n- I `POST` to `/roles/` with a body containing `name`, `handle`, and `members`. The role is created, but when I `GET` it back the `handle` is empty. Same thing if I send the field at create time and then read the role list — handle is never populated.\n- I tried patching it in afterwards via `PUT /roles/{roleID}` with `handle` in the body. Also a no-op — the field appears to be ignored, the stored role still has no handle.\n- Looking at `docs/system/README.md` under \"Create role\" / \"Update role details\", the request parameter tables only list `name` and `members`. Compare that to the users / signup sections where `handle` is listed explicitly. So it looks like the role endpoints just never wired `handle` through, even though the underlying model supports it.\n\nExpected: `POST /roles/` and `PUT /roles/{roleID}` should accept `handle` in the request body, persist it on the role, and the API spec + docs for those two endpoints should reflect that the parameter exists. Required-ness of `handle` should follow whatever convention the existing `name` parameter uses on each of those endpoints (so that create and update stay consistent with each other)."} {"task_id": "format-code-task-000724", "source_id": "format-code-task-000724", "domain": "code", "task_path": "tasks/format-code-task-000724", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:828bab10aae3f32ddab8644bebcb0ec542b7d8eff3f9c7bf95e9f9d09718acaa", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\n我想通过 REST API 动态改某个 db 的 audit 配置，按理说 PUT/POST `/{db}/_config/audit` 应该能做这事吧？但我试了下，不管发什么请求体过去，配置都没变，GET 回来还是原样，感觉这俩接口压根没接上。能不能让 PUT 整体替换、POST 增量改 audit events 真正生效？另外 GET 现在直接返回一坨 events map，我想顺便知道 audit 整体是不是开着的，最好能把全局 enabled 也带出来。\n\n# Expected outcomes\n\n- `GET /{db}/_config/audit` returns a JSON object that includes both the database audit enabled state and the audit event configuration, rather than returning the event map as the top-level response.\n- The audit enabled state and event enabled states returned by `GET /{db}/_config/audit` reflect the current persisted database configuration after defaults/runtime configuration are applied, including changes made through the configuration API without requiring a database restart.\n- `GET /{db}/_config/audit` includes an ETag corresponding to the current database configuration version when the audit configuration is available.\n- If database audit configuration cannot be read because the server is running without persistent configuration support, `GET /{db}/_config/audit` returns a `503 Service Unavailable` response with an audit-configuration-unavailable error.\n- `PUT /{db}/_config/audit` treats the request body as a full replacement of the database audit configuration: the global audit enabled value is replaced from the request body, and the enabled event set is rebuilt from the events explicitly enabled in the request.\n- `POST /{db}/_config/audit` treats the request body as an incremental update: explicitly supplied global audit enabled values are updated, events set to enabled are added, events set to disabled are removed, and omitted fields/events are left unchanged.\n- `PUT /{db}/_config/audit` and `POST /{db}/_config/audit` reject malformed or unknown audit event IDs in the submitted events object with a `400 Bad Request` response that identifies the audit configuration update failure and distinguishes invalid event ID syntax from unknown event IDs.\n- When `GET /{db}/_config/audit` is requested in verbose form, each event entry includes the available event metadata and enabled/filterable state using the documented event object fields, while omitting unavailable metadata fields from the JSON response.\n\n# Implementation notes\n\nThe implementation may choose where to perform parsing, validation, persistence updates, and response construction, as long as the externally observable REST API behavior above is satisfied. Keep the behavior compatible with existing audit event identifiers and existing admin API conventions for JSON responses, errors, and configuration versioning."} {"task_id": "format-code-task-000725", "source_id": "format-code-task-000725", "domain": "code", "task_path": "tasks/format-code-task-000725", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:287c8900834a792784570c8f2373d3036fe83f3945c5ca4065a82e8a650315a9", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\n我在用 ALL 的 SAC 跑连续控制时发现，只把环境 action space 从 `[-1, 1]` 线性放大到比如 `[-2, 2]`，`SoftDeterministicPolicy` 返回的 `log_prob`、entropy 和 policy loss 就会明显跟着尺度变，训练表现也飘得很厉害。另外我开了 `clip_grad` 后，有次梯度范数已经是 NaN/Inf，`step()` 还是继续往下跑了；我这边环境用的是 torch 1.9，但包依赖看起来还卡在 1.8。\n\n## Expected outcomes\n\n- `SoftDeterministicPolicy` 在连续动作空间边界不是 `[-1, 1]` 时，返回的 `log_prob` 应与缩放后的动作分布一致；只线性放大同一任务的动作范围时，`log_prob` 只应体现动作尺度带来的常数偏移，不应出现逐维概率量级被错误放大的现象。\n- 基于该 `log_prob` 的 entropy 与 SAC policy loss 不应因为动作空间做等价线性缩放而产生额外的异常尺度变化。\n- `Approximation.step()` 在启用非零 `clip_grad` 时，如果待裁剪的梯度范数为 NaN 或 Inf，应抛出 `RuntimeError` 并阻止优化步骤静默继续。\n- 包元数据版本应更新为 `0.7.2`，文档配置中的 release 也应更新为 `0.7.2`。\n- 安装依赖应面向 torch 1.9，声明为 `torch~=1.9.0`；项目 CI 中固定安装的 CPU 版 torch 也应更新到 `torch==1.9.0+cpu`。\n\n## Implementation notes\n\n具体校正位置、辅助函数组织方式、内部命名和测试覆盖方式由实现者决定；只要公开行为、错误处理语义、版本元数据与依赖声明满足上述结果即可。"} {"task_id": "format-code-task-000726", "source_id": "format-code-task-000726", "domain": "code", "task_path": "tasks/format-code-task-000726", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7e0f88882137b27ca0eba428ad39b6aff3db7be8ecc7d0e4f606d9d5f5418b33", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Fix double-backprop through shared feature networks\n\nIn our actor-critic setup, a single `FeatureNetwork` produces a feature\nrepresentation that is shared by several downstream heads (e.g. a value\nfunction and a policy). Each head computes its own loss and calls\n`.backward()`. Today, because the features the network hands out are still\nwired into the feature model's computation graph, every head's backward pass\nruns all the way back through the shared feature model. With two heads that\nmeans we backpropagate through the shared model twice per update even though we\nonly take one optimizer step — wasteful, and it forces callers to juggle\n`retain_graph=True` to avoid \"backward through the graph a second time\" errors.\n\nRework `FeatureNetwork` so the shared model is only backpropagated through\nonce per training step, while keeping its existing public surface\n(`__call__`, `eval`, `reinforce`).\n\nRequired observable behavior:\n\n- Calling the network on a state returns a `State` whose mask and info are\n carried over from the input, and whose feature tensor holds the model's\n output for that state. That feature tensor must be a gradient-tracking leaf\n that is **disconnected** from the feature model's graph: any number of\n downstream computations may derive a loss from it and call `.backward()`\n independently, in sequence, without passing `retain_graph` and without\n raising. The gradient each head produces accumulates on this returned\n feature tensor.\n\n- `reinforce()` takes no loss of its own — the training signal is exactly the\n gradient that downstream heads have accumulated on the feature tensors handed\n out since the last `reinforce()`. It pushes that accumulated gradient back\n through the shared model in a single backward pass, honoring the optional\n gradient clipping, then performs one optimizer step and clears the gradients.\n\n- The network may be called multiple times before a single `reinforce()`; all\n of those forward passes contribute their accumulated gradients to that one\n optimizer step. After `reinforce()` returns, the pending state is cleared:\n the next forward/reinforce cycle is independent and must not re-apply\n gradients from a previous cycle.\n\n- `eval(state)` returns a `State` (mask/info preserved) whose features are the\n model output computed without building a graph (no gradient tracking), and\n it must not disturb the model's training mode.\n\nThe net effect: for a set of downstream losses, the optimizer step applied to\nthe shared model is exactly the single SGD-equivalent step you would get by\nbackpropagating the summed downstream gradient through the model once.\n"} {"task_id": "format-code-task-000727", "source_id": "format-code-task-000727", "domain": "code", "task_path": "tasks/format-code-task-000727", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:18e9bfe40fa52adb11f38f307a785b06057c4b1b9db827bb718a774bf3939170", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: drop-in configuration files\n\nI'm packaging CRI-O for our distribution and want to ship sensible defaults in `/etc/crio/crio.conf` while letting users (and our config management) layer in machine-specific tweaks — custom registries, an extra runtime entry, a different cgroup manager — without ever touching that base file.\n\nToday the only knob we have is `--config, -c` pointing at one TOML file, which forces us into two awkward options:\n\n- Have users edit `/etc/crio/crio.conf` directly: conflicts on every package update, painful for config management to template.\n- Have users replace it with their own copy: they lose any new defaults we ship in future releases.\n\nMost other system daemons (systemd, networkd, sshd, …) solve this with a \"drop-in\" directory pattern: a directory of partial config files that are merged on top of the base config in a predictable order. I'd like CRI-O to support the same idea — a directory of partial TOML configs that get layered on top of the main `crio.conf` at startup, with ordering determined by filename so admins can pick priority via prefixes like `00-…`, `10-…`, `99-…`.\n\nWhat matters for this to be useful:\n\n- The base config file is applied first, then each fragment is layered on top in deterministic name order. Each fragment is just a partial TOML — only the keys it sets should take effect, the rest fall through to the lower-priority layers.\n- Command-line flags passed to `crio` continue to have the highest priority, same as today, so a CLI override always beats both the base config and any fragments.\n- If the drop-in directory doesn't exist, startup should not fail — that's the expected state for users who don't use drop-ins at all.\n\nThis would let distros and config-management tools layer overrides cleanly without ever rewriting `crio.conf`.\n\nA natural shape would be a new CLI flag like `--config-dir` (short `-d`) on `crio` to point at the directory, plus a corresponding method on the `Config` type (something like `UpdateFromPath`) that walks the directory and layers each file on top via the existing per-file update path."} {"task_id": "format-code-task-000728", "source_id": "format-code-task-000728", "domain": "code", "task_path": "tasks/format-code-task-000728", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:731da94fe8533fb0c088ca906bf1349eefa96607cd84e8467f3c2d62c3b51b82", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Closing a WAMP session raises errors when a pending call's awaiting task was cancelled (asyncio)\n\nI'm using autobahn-python over asyncio. In my app I make WAMP calls\nfrom inside tasks and put timeouts on them with `asyncio.wait_for`, so\nsome calls end up with their awaiter cancelled before `session.call(...)`\nhas actually returned. Roughly:\n\n```python\nasync def fetch():\n return await session.call(u'com.example.fetch')\n\ntry:\n result = await asyncio.wait_for(fetch(), timeout=2.0)\nexcept asyncio.TimeoutError:\n pass # give up and move on\n```\n\nAfter running like this for a while I eventually shut the session down —\neither explicitly via `leave()`, or because the underlying transport\ndrops. When that happens and there are still-pending requests whose\nawaiting task had previously been cancelled / timed out, autobahn itself\nerrors out during the teardown of the session.\n\nI'd expect closing a session to be a clean operation: even if some of\nthe still-outstanding request futures have already been abandoned or\ncancelled by their callers, shutting the session down shouldn't blow up."} {"task_id": "format-code-task-000729", "source_id": "format-code-task-000729", "domain": "code", "task_path": "tasks/format-code-task-000729", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b16ba0d70ea7cbf7806c3ca638668b8217adefd8ff6b7bf6870ef858b94eb165", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\n我用 terrajet 把资源导出成 tfstate 之后，发现 refresh 阶段完全不理会我在 main.tf 里配的 operation timeouts，每次还是按默认超时走。我特地配了 read/create 这些超时，apply 的时候看着是生效的，但 refresh 还是该慢慢慢、该卡卡，感觉像是 refresh 根本没拿到我配的超时值。这是不是 tfstate 里少存了点什么？\n\n## Expected outcomes\n\n- `WriteTFState` should preserve configured non-zero operation timeout settings in the generated tfstate using a Terraform-compatible state representation observable by refresh.\n- `WriteTFState` should omit timeout state metadata when no operation timeout is configured, and should not disturb any existing state metadata in that case.\n- When existing state metadata is present, `WriteTFState` should preserve unrelated existing metadata while adding or updating the configured timeout information.\n- `WriteMainTF` should keep its existing behavior for operation timeouts: configured non-zero timeout settings are emitted through the existing `timeouts` parameter in main.tf, and no `timeouts` parameter is emitted when all operation timeouts are unset.\n\n## Implementation notes\n\nThe exact structure of helper code, data types, and validation placement is up to the implementer. The important requirement is that the generated tfstate and main.tf remain compatible with Terraform’s observable behavior for operation timeouts and preserve unrelated existing metadata.\n\n## Required output literals (exact-match contract)\n\nWhen `WriteTFState` emits configured operation timeouts into Terraform state private metadata, the decoded private metadata MUST contain Terraform's resource timeout metadata key exactly:\n\n- State private metadata key: `e2bfb730-ecaa-11e6-8f88-34363bc7c4c0` — this is Terraform's resource timeout private-state compatibility key consumed by refresh.\n\nWithin that metadata object, configured operations MUST use Terraform's operation names exactly: `read`, `create`, `update`, and `delete`. Values MUST be Terraform-compatible duration nanosecond values."} {"task_id": "format-code-task-000730", "source_id": "format-code-task-000730", "domain": "code", "task_path": "tasks/format-code-task-000730", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2c51da9e32a5fc98c20f4247d71a806158f25fb307275a31744c9b7ca89ed3fe", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Open a memiavl database at a historical version\n\nThe embedded `memiavl` database (the `memiavl` Go module in this repository) can\ncurrently only be opened at its latest committed version: loading the database\nreplays the entire write-ahead log, so you always end up at the tip.\n\nWe need to be able to open the database **at a specific earlier committed\nversion**, so that callers can reconstruct the exact state of a past version\n(for serving versioned queries, state-sync, etc.) without having to mutate or\nroll back the database.\n\nAdd a load option that selects the version to open. The observable contract:\n\n- The load options gain a way to request a target version, expressed as an\n unsigned integer.\n- When the target version is left at its zero value (the default), loading\n behaves exactly as it does today and opens the latest committed version.\n- When a positive target version `v` that exists in the database's committed\n history is requested, the loaded database must reflect exactly the committed\n state as of version `v`: the version it reports is `v`, and its commit hash\n (and therefore its per-store contents) is identical to what it was right after\n version `v` was committed — as if the commits made after `v` had never been\n applied.\n- Opening at a target version must be non-destructive: it must not delete or\n rewrite the existing on-disk history, so the same database can afterwards be\n reopened at any other version, including the latest.\n\nThe existing latest-version load path and all other database behavior must be\nunaffected.\n"} {"task_id": "format-code-task-000732", "source_id": "format-code-task-000732", "domain": "code", "task_path": "tasks/format-code-task-000732", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:266f46b4c3c222f8bffe8ae3f61ee156f94169c3d4b552f2ad2746679688791c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI'm parsing GDB MI output with `gdbmiparser.parse_response`, and the escape sequences in the strings aren't coming through right. For example, when I parse `~\"a\\nb\"`, the payload comes back as the literal `a\\nb` with the backslash still in it, instead of an actual newline. Same kind of thing with `\\t`. And when there's an error like `^error,msg=\"some error\\non multiple lines\\twith escapes\"`, the `msg` I get back is even worse — the backslashes just vanish, so `\\n` turns into a literal `n` and `\\t` into `t`. I'm also seeing octal escapes like `\\040` left as-is. Makes it pretty hard to log or display GDB's output cleanly.\n\nExpected outcomes:\n- Result records parsed through `gdbmiparser.parse_response` should preserve their normal parsed structure, and escaped string fields should contain the characters represented by GDB MI escape sequences rather than losing backslashes or leaving escapes literal.\n- Textual stream records parsed through `gdbmiparser.parse_response` should return payload text with GDB MI escape sequences interpreted into the corresponding characters.\n- GDB MI string escapes, such as the newline/tab and octal examples above, should be handled consistently wherever escaped MI strings are parsed.\n\nImplementation notes:\n- The parsing API and existing response shapes should remain compatible with current callers.\n- The internal organization of the unescaping logic is up to the implementer; focus on externally observable parsed values rather than any particular helper, module layout, or parsing strategy."} {"task_id": "format-code-task-000733", "source_id": "format-code-task-000733", "domain": "code", "task_path": "tasks/format-code-task-000733", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b41d3727a5af6a39173793a03ee9f8317ab16fa395c7e3edc7ee21cb488f76c0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `compressjs.BWT.bwtransform2(T, U, n, alphabetSize)` as a one-shot cyclic Burrows-Wheeler transform helper. `T` and `U` are typed-array-like buffers, `n` is the number of input symbols to transform from the start of `T`, `alphabetSize` is optional for `Uint8Array` and `Uint16Array`, and the function writes the transformed sequence into `U[0..n)` while returning the primary index as a number.\n\nFor ASCII input `bcababa` with `n == 7` and `alphabetSize == 256`, it should write `cbbaaab` into `U` and return `5`. For `ABCDEFGHIJKLMNOPQRSTUVWXYZ`, it should write `ZABCDEFGHIJKLMNOPQRSTUVWXY` and return `0`. For `ZYXWVUTSRQPONMLKJIHGFEDCBA`, it should write `BCDEFGHIJKLMNOPQRSTUVWXYZA` and return `25`. For `n == 1`, it should copy the single symbol to `U[0]` and return `0`.\n\nWhen `alphabetSize` is omitted, `Uint8Array` inputs should default to an alphabet size of 256 and `Uint16Array` inputs should default to 65536; inputs without a byte width that allows a default should raise `Error` with the message `Need to specify alphabetSize`. Repeated calls with the same first `n` input symbols and fresh output buffers should produce the same output contents and primary index, without filesystem, network, or module-level side effects."} {"task_id": "format-code-task-000734", "source_id": "format-code-task-000734", "domain": "code", "task_path": "tasks/format-code-task-000734", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:53b70a39b9886e2f0685ef61132c73b30986e012739cbfca056fdbbb2a6cf282", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `linspace` is missing from `cubed.array_api`\n\nI'm trying to generate an evenly-spaced sequence with a fixed number of points (typical use cases: axis coordinates, sampling grids), and reached for `linspace` since it's part of the standard Array API:\n\n```python\nimport cubed.array_api as xp\n\nxs = xp.linspace(0.0, 1.0, 50)\n```\n\nBut `linspace` doesn't exist on `cubed.array_api`. Looking at the API coverage table in `api_status.md`, it's listed under Creation Functions but the \"Implemented\" column is empty (it's flagged as difficulty 2, \"Like `arange`\").\n\n`arange` works fine for stepping by a known increment, but it's awkward when what I actually want is \"give me exactly N points between a and b\" — that's what `linspace` is for, and it's also what the spec requires.\n\nCould `linspace` be implemented so cubed's array API surface matches the standard? It would be nice to have it behave consistently with the numpy/dask version when running large lazy arrays."} {"task_id": "format-code-task-000735", "source_id": "format-code-task-000735", "domain": "code", "task_path": "tasks/format-code-task-000735", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:da0b444734acba72989249de9935692e465c7719c771b57aa93845d1b3275f04", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the TM1py cells service to support an async DataFrame cube write method on an existing service object: `tm1.cells.write_dataframe_async(cube_name: str, data: pandas.DataFrame, slice_size_of_dataframe: int = 250_000, max_workers: int = 8, dimensions: Iterable[str] = None, increment: bool = True, sandbox_name: str = None, deactivate_transaction_log: bool = False, reactivate_transaction_log: bool = False, **kwargs)`.\n\nFor a session like `tm1 = TM1Service(...)`, calling `tm1.cells.write_dataframe_async(\"Sales\", df, dimensions=[\"Version\", \"Month\", \"Measure\"], slice_size_of_dataframe=2, max_workers=4, increment=False, sandbox_name=\"Plan\")` should split the DataFrame into row chunks of that size and write each chunk through the existing DataFrame write path using blob mode, preserving the cube name, dimension order, increment flag, sandbox name, and any extra keyword arguments. If `dimensions` is not provided, the method should infer the cube dimension order before validating or writing. A valid DataFrame must have exactly one value column in addition to the cube dimension columns; otherwise it should raise `ValueError(\"Number of columns in 'data' DataFrame must be number of dimensions in cube + 1\")`. Passing a non-DataFrame as `data` should raise `ValueError(\"argument 'data' must of type DataFrame\")`.\n\nWhen `increment=True`, numeric duplicate intersections in the DataFrame should be aggregated before the chunks are submitted, so duplicate rows increment the cube once with their summed value. If every chunk write succeeds, the method should return `None`. If one or more chunk writes fail with `TM1pyWriteFailureException` or `TM1pyWritePartialFailureException`, the method should raise one `TM1pyWritePartialFailureException` whose statuses and error log filenames are flattened from all failed chunks and whose attempt count sums each partial failure's attempts, counting a total failure without an attempts attribute as one attempt. The transaction-log deactivate/reactivate options should apply around the overall async DataFrame write call in the same way as other cell write methods."} {"task_id": "format-code-task-000736", "source_id": "format-code-task-000736", "domain": "code", "task_path": "tasks/format-code-task-000736", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:dc983040855ad4f5a9676504a906eb7b87ecac46ff812e615a2773f154292da9", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nPlease answer these questions before submitting a bug report.\n\n### What version of godog are you using?\nGodog version is: v0.10.0\n\n### What version of Go are you using?\ngo version go1.14.7 linux/amd64\n\n### What did you do?\nIf possible, provide a recipe for reproducing the error.\nI added a new features/foo.feature to my project and ran `godog .` for the first time in that project.\n\n### What did you expect to see?\ngodog suggesting the new framework functions listed at `// godog v0.10.0 (latest)` in the README.md with the correct regex for the steps. I just defined.\n```\nYou can implement step definitions for undefined steps with these snippets:\n[...]\n\nfunc InitializeTestSuite(ctx *godog.TestSuiteContext) {\n\tctx.BeforeSuite(func() { \n\t\t// clean state for the first rune\n\t})\n}\n\nfunc InitializeScenario(ctx *godog.ScenarioContext) {\n\tctx.BeforeScenario(func(*godog.Scenario) {\n\t\t// clean the state before every scenario\n\t})\n\n\tctx.Step(`regex`, stepcall)\n}\n```\n### What did you see instead?\nexample code for the deprecated calls\n```\nYou can implement step definitions for undefined steps with these snippets:\n[...]\n\nfunc FeatureContext(s *godog.Suite) {\n\ts.Step(`regex`, stepcall)\n}\n```\n\n### Additional context\nI use a project with Go modules.\nfeature file at features/happydogs.feature\n```gherkin\nFeature: Happy dogs\n As a dog owner\n I want my dogs to be happy\n So that it will make me happy as well\n\n Scenario: Initially the dog is sad\n Given 2 dogs\n When they play with no balls\n Then they are still sad\n\n Scenario: If they have at least one ball they are happy\n Given there are 2 dogs\n When they play with 1 ball\n Then then all dogs will be happy\n```\nFull output of godog .\n```plain\n$ godog .\nFeature: Happy dogs\n As a dog owner\n I want my dogs to be happy\n So that it will make me happy as well\n\n Scenario: Initially the dog is sad # features/happydogs.feature:6\n Given 2 dogs\n When they play with no balls\n Then they are still sad\n\n Scenario: If they have at least one ball they are happy # features/happydogs.feature:11\n Given there are 2 dogs\n When they play with 1 ball\n Then then all dogs will be happy\n\n2 scenarios (2 undefined)\n6 steps (6 undefined)\n246.175µs\n\nYou can implement step definitions for undefined steps with these snippets:\n\nfunc dogs(arg1 int) error {\n\treturn godog.ErrPending\n}\n\nfunc thenAllDogsWillBeHappy() error {\n\treturn godog.ErrPending\n}\n\nfunc thereAreDogs(arg1 int) error {\n\treturn godog.ErrPending\n}\n\nfunc theyAreStillSad() error {\n\treturn godog.ErrPending\n}\n\nfunc theyPlayWithBall(arg1 int) error {\n\treturn godog.ErrPending\n}\n\nfunc theyPlayWithNoBalls() error {\n\treturn godog.ErrPending\n}\n\nfunc FeatureContext(s *godog.Suite) {\n\ts.Step(`^(\\d+) dogs$`, dogs)\n\ts.Step(`^then all dogs will be happy$`, thenAllDogsWillBeHappy)\n\ts.Step(`^there are (\\d+) dogs$`, thereAreDogs)\n\ts.Step(`^they are still sad$`, theyAreStillSad)\n\ts.Step(`^they play with (\\d+) ball$`, theyPlayWithBall)\n\ts.Step(`^they play with no balls$`, theyPlayWithNoBalls)\n}\n```"} {"task_id": "format-code-task-000737", "source_id": "format-code-task-000737", "domain": "code", "task_path": "tasks/format-code-task-000737", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b171898f032c24ec228230be3bf0aa5dcef40afd579921eb5be45ac8a1091907", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n**Problem Statement**\n\nI’m trying to use CuPy sparse matrices the same way I use SciPy, where I can create an empty CSR or CSC matrix just by passing a shape, but `cupy.sparse.csr_matrix((m, n))` and `csc_matrix((m, n))` don’t seem to accept that. It would be really helpful if those constructors could handle a shape-only empty matrix, including when I pass a dtype.\n\n**Expected Outcomes**\n- `cupy.sparse.csr_matrix((M, N))` accepts a two-dimensional shape tuple and creates an empty CSR matrix with shape `(M, N)`.\n- `cupy.sparse.csc_matrix((M, N))` accepts a two-dimensional shape tuple and creates an empty CSC matrix with shape `(M, N)`.\n- When no `dtype` is supplied for this shape-only form, the resulting empty CSR or CSC matrix defaults to `float64`.\n- When a `dtype` is supplied for this shape-only form, the resulting empty CSR or CSC matrix preserves that dtype.\n- The documented constructor forms for these APIs include shape-only empty-matrix construction, including `csr_matrix((M, N), [dtype])` and `csc_matrix((M, N), [dtype])`.\n\n**Implementation Notes**\n\nAny implementation that preserves the current sparse matrix API and makes the public constructor behavior match these outcomes is acceptable. The exact validation flow, helper structure, and internal sparse-storage construction strategy are left to the implementer."} {"task_id": "format-code-task-000738", "source_id": "format-code-task-000738", "domain": "code", "task_path": "tasks/format-code-task-000738", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a2c225360a82d3b9e82c3bd5efe509a9e997e26e5a5665f8f1d98e700e00b57f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nNo validation for kernel name\nAlthough `ElementwiseKernel`'s `name` argument is directly used as a function name in generated CUDA code, there are no validation process to raise an exception when invalid characters are used in `name`.\nThat causes CUDA compile error, which is a bit difficult to debug."} {"task_id": "format-code-task-000739", "source_id": "format-code-task-000739", "domain": "code", "task_path": "tasks/format-code-task-000739", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3beff124086e1f9265dab78d18879d7501a9be0f363fcf0a788c1f190efb0e80", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want an exported pure TypeScript function `insertVariableToCurl({ curl, index, variableName }: { curl: string; index: number; variableName: string }) => string` that inserts a shell variable reference into cURL command text at a cursor offset. For `insertVariableToCurl({ curl: \"\", index: 0, variableName: \"TOKEN\" })`, it should return `$TOKEN`. For `insertVariableToCurl({ curl: \"curl example.com\", index: 16, variableName: \"TOKEN\" })`, it should append at the end and return `curl example.com$TOKEN`. When insertion occurs inside an unquoted word character, it should use brace syntax so the variable name is not merged with surrounding text: `insertVariableToCurl({ curl: \"curl example.com\", index: 6, variableName: \"HOST\" })` should return `curl e${HOST}xample.com`. In double-quoted values, inserting before a word character should also use braces, while inserting before punctuation or whitespace should use `$NAME` directly; for example `insertVariableToCurl({ curl: \"curl \\\"hello world\\\"\", index: 12, variableName: \"NAME\" })` should return `curl \\\"hello ${NAME}world\\\"`. In single-quoted values, inserting inside the quoted text should close the quote, insert `$NAME`, and reopen the quote so shell expansion can occur, such as returning `curl 'hello'$NAME' world'` for `insertVariableToCurl({ curl: \"curl 'hello world'\", index: 11, variableName: \"NAME\" })`. Inserting into whitespace or a newline should place `$NAME` at the cursor while preserving the surrounding characters. If the cursor points inside a cURL option token such as `-H`, the function should throw an `Error` with message `Cannot insert variable to option`. Calling it multiple times with the same arguments should produce the same string, and it should not mutate the argument object, touch the filesystem, use the network, or depend on hidden global state."} {"task_id": "format-code-task-000740", "source_id": "format-code-task-000740", "domain": "code", "task_path": "tasks/format-code-task-000740", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ae37155dd0276a0947addab78c235509f317faed1b434cdf19589caadc9fb6b0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nInheritance works with task parameters e.g:\n\n```ini\n[runtime]\n [[FAM]]\n [[task]]\n inherit = FAM\n```\n\nHowever, when mixing parameters across inheritance it only partially works.\n\n```ini\n[runtime]\n [[FAM]]\n [[task]]\n inherit = FAM\n```\n\n### Example\n\n```ini\n[task parameters]\n a = 1..2\n b = 1..2\n\n[scheduling]\n [[graph]]\n R1 = \"\"\"\n \n \"\"\"\n\n[runtime]\n [[]]\n inherit = \n script = test $b -eq 1\n\n [[]]\n [[[environment]]]\n b = %(b)d\n```\n\nThe inheritance does indeed work as shown by the output of `cylc config`:\n\n```console\n$ cylc config --sparse param-test \n[task parameters]\n a = 1, 2\n b = 1, 2\n[scheduling]\n cycling mode = integer\n initial cycle point = 1\n final cycle point = 1\n [[graph]]\n R1 = \n[runtime]\n [[root]]\n [[_a1]]\n inherit = _b1\n script = test $b -eq 1\n [[[environment]]]\n b = %(b)d\n [[_a2]]\n inherit = _b1\n script = test $b -eq 1\n [[[environment]]]\n b = %(b)d\n [[_b1]]\n [[[environment]]]\n b = %(b)d\n [[_b2]]\n [[[environment]]]\n b = %(b)d\n[visualization]\n [[node attributes]]\n```\n\nHowever, the parameter in the environment variable is not expanded in the job script:\n\n```bash\n# job\n\ncylc__job__inst__user_env() {\n # TASK RUNTIME ENVIRONMENT:\n export b\n b=\"%(b)d\"\n}\n```\n\n### Use Case\n\nExample use case which involves mapping one parameter onto another:\n\n```ini\n[runtime]\n # define environments\n [[environment]]\n\n # map models onto environments\n [[model]]\n inherit = environment<1>\n [[model]]\n inherit = environment<2>\n [[model]]\n inherit = environment<1>\n```"} {"task_id": "format-code-task-000741", "source_id": "format-code-task-000741", "domain": "code", "task_path": "tasks/format-code-task-000741", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5a33ea1ce8a36076d293421f3a3c957e76ce71ae79f31340bf0cdc63aaa1654c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI’m using Cython with C++20/23 code and keep having to write little wrappers because `libcpp` doesn’t expose things like `std::gcd`/`std::lcm`, container `.contains()`, or `std::string` helpers like `starts_with`, `ends_with`, and `contains`. It would be really nice if I could just cimport/use these standard library APIs directly from Cython.\n\n# Expected outcomes\n\n- Cython C++ code can cimport and call modern numeric helpers from `libcpp.numeric`, including `gcd`, `lcm`, and `midpoint`, with the corresponding standard-library behavior.\n- The standard ordered and unordered associative containers exposed by `libcpp` support their C++ membership query through `.contains(...)`, returning whether the requested key or value is present.\n- `libcpp.string.string` supports the standard prefix, suffix, and substring-style helpers `starts_with(...)`, `ends_with(...)`, and `contains(...)` for normal Cython byte/character string inputs, returning the corresponding boolean result.\n\n# Notes\n\nThese APIs should be available through normal Cython `cimport` usage in C++ mode and should work under the appropriate C++ standard level for the underlying standard-library feature."} {"task_id": "format-code-task-000742", "source_id": "format-code-task-000742", "domain": "code", "task_path": "tasks/format-code-task-000742", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b475ed90808c9daf0090115b083fafae7bffdeaba577cd6c48162e9c8a511ad9", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Support PEP 484 type hints in generated docstrings\n\nPyment builds docstrings from a function/method definition, but it currently\nignores the type annotations written in the signature. If a function is defined\nwith type hints, those types should be carried into the docstring it produces.\n\nPlease make the docstring generation read PEP 484 annotations from the\nsignature and reflect them in the output.\n\nConcretely, for a function such as:\n\n```python\ndef func(param1, param2: str = 'default val') -> int:\n```\n\ngenerating a reStructuredText docstring should produce, in addition to the\nusual `:param ...:` fields:\n\n- a `:type param2: str` field for the annotated parameter, and\n- an `:rtype: int` field for the return annotation.\n\nRequirements:\n\n- A parameter that carries an annotation gets a corresponding type field; a\n parameter with no annotation gets no type field (the existing type-stub\n behaviour is unchanged).\n- A return annotation (`-> ...`) becomes the return type of the docstring.\n- An annotated parameter that also has a default value keeps its\n `(Default value = ...)` note; its type comes from the annotation.\n- Annotations whose text contains commas inside brackets — e.g.\n `Dict[str, int]`, `Tuple[int, ...]`, `Optional[List[str]]` — must be treated\n as a single type and not split apart on the inner commas.\n- A type that is already written in the source docstring wins: when the input\n docstring already specifies a parameter's type (or the return type), the\n signature annotation must not overwrite it.\n- Methods still skip `self`/`cls`, and `async def` is handled like `def`.\n\nThis should work the same way for the styles Pyment already supports (reST,\njavadoc, numpydoc, google): whenever a style renders parameter types and a\nreturn type, the values inferred from the signature are used.\n"} {"task_id": "format-code-task-000743", "source_id": "format-code-task-000743", "domain": "code", "task_path": "tasks/format-code-task-000743", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:46552bf88c588cf3bdd43fc4703e81ea87b81b0eb55a3d99339d77894fe307c1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Dagger fails to load git config when `~/.gitconfig` contains a multi-line value\n\nI have a few entries in my `~/.gitconfig` whose values span multiple lines (e.g. an `insteadOf` rewrite with an embedded newline, plus a signing key block). Standard `git` tooling handles these fine, but as soon as Dagger needs to read the git config (for example when resolving a private repo via an `insteadOf` rule), the operation fails.\n\nThe error surfaces as something like:\n\n```\nFailed to parse git config invalid format: line \"...\" doesn't match key=value pattern\n```\n\n…where the quoted line is the *second* line of one of my multi-line values — so it has no `=` in it, and the parser bails out on the whole config.\n\nRunning `git config -l` directly in the same shell prints the config without complaint, so the values themselves are valid; it just looks like Dagger's parsing of the output doesn't handle entries whose value contains a newline. The end effect is that any feature relying on `GetConfig` (insteadOf URL rewriting, etc.) is unusable for users with this kind of `.gitconfig`.\n\nCould Dagger be made to read git config in a way that tolerates multi-line values? Stripping or rejecting those entries isn't really an option on my side — they're legitimate git configuration."} {"task_id": "format-code-task-000744", "source_id": "format-code-task-000744", "domain": "code", "task_path": "tasks/format-code-task-000744", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:bb4312ccc9ae6d957bc2770b9da31829ed6d8b0fdbbc25cf3c9f501df911c42d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI'm calling `instance.get_event_records` with `EventRecordsFilter(tags=...)` on an older Dagster instance and it just errors telling me to run `dagster instance migrate`, even though I'm not using a limit and I'd be fine with it scanning/filtering. Can we make tag filtering still work in that case? Also, when I do get the events back, it’d be really helpful if `EventLogEntry` exposed the materialization/observation tags and the observation directly instead of making me dig through `dagster_event`.\n\n## Expected Outcomes\n\n- On older, unmigrated instances, tag-filtered event record queries should still be usable when no limit is requested, and the results should respect the requested tag filters.\n- On older, unmigrated instances, tag-filtered event record queries that also request a limit should fail with a clear invocation error explaining that this combination requires migrating the instance.\n- `EventLogEntry.asset_observation` should provide direct access to an observation event’s asset observation, while non-observation events should not report one.\n- `EventLogEntry.tags` should provide direct access to tags recorded on asset materialization and asset observation events, while unrelated event types should not report asset-event tags.\n\n## Implementation Notes\n\n- Preserve the existing public APIs for event record retrieval and filtering; the specific storage-query strategy and fallback location are up to the implementation.\n- Existing migrated-instance behavior should remain compatible with the current event log storage paths."} {"task_id": "format-code-task-000745", "source_id": "format-code-task-000745", "domain": "code", "task_path": "tasks/format-code-task-000745", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:90fe44865c66aa25a08edc33c21e48e1f103a742229d119cae68a8794a1d924f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## AMP sensor doesn't see cross-repository dependencies when evaluating\n\nI have my assets split across two code locations / repositories:\n\n- `repo_upstream` defines `raw_data` (an upstream asset, materialized on some external schedule).\n- `repo_downstream` defines `processed_data`, which depends on `raw_data`. I have an auto-materialize policy on `processed_data` that's supposed to gate on the parent — i.e. don't materialize `processed_data` if `raw_data` is missing / hasn't been materialized yet.\n\nI set up an AMP sensor in `repo_downstream` whose `asset_selection` covers `processed_data` (and a few other downstream assets in the same repo), and let the asset daemon run it.\n\nWhat I expected: the daemon evaluates `processed_data`, sees that its parent `raw_data` (which lives in the other repo) is missing, and skips materialization — same behavior I'd get if both assets lived in the same repo.\n\nWhat actually happens: the AMP sensor behaves as if `raw_data` doesn't exist at all. The cross-repo parent relationship seems invisible to the evaluation, so the \"wait for parent\" rule never fires the way it should, and `processed_data` gets materialized at times that don't make sense relative to its real parent.\n\nIf I move both assets into the same repository, the AMP rules behave correctly and the parent dependency is respected. It only goes wrong once the parent and child live in different repos / code locations.\n\nIt seems like the AMP sensor evaluation is only looking at the assets within the sensor's own repository, instead of considering the full asset graph across the workspace when figuring out parent/child state. The selection of *which* assets the sensor is responsible for should still come from the sensor's `asset_selection`, but the dependency graph used to actually evaluate the policies needs to include assets from other code locations too.\n\nRepro sketch:\n\n```python\n# repo_upstream/repo.py\nfrom dagster import asset, Definitions\n\n@asset\ndef raw_data():\n ...\n\ndefs = Definitions(assets=[raw_data])\n```\n\n```python\n# repo_downstream/repo.py\nfrom dagster import asset, AssetKey, AssetSelection, Definitions, AutoMaterializePolicy\nfrom dagster._core.definitions.auto_materialize_sensor_definition import (\n AutoMaterializeSensorDefinition,\n)\n\n@asset(\n deps=[AssetKey(\"raw_data\")],\n auto_materialize_policy=AutoMaterializePolicy.eager(),\n)\ndef processed_data():\n ...\n\namp_sensor = AutoMaterializeSensorDefinition(\n name=\"downstream_amp\",\n asset_selection=AssetSelection.assets(processed_data),\n)\n\ndefs = Definitions(assets=[processed_data], sensors=[amp_sensor])\n```\n\nWith both repos loaded into the same workspace and `raw_data` never materialized, the AMP sensor still issues runs for `processed_data` as if it had no parent — instead of correctly waiting on `raw_data`."} {"task_id": "format-code-task-000746", "source_id": "format-code-task-000746", "domain": "code", "task_path": "tasks/format-code-task-000746", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:36af50c59f533e59af375817ee01a8cc98d04e5bd6b051c0482a94d735422003", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n我在用 GraphQL 管 sensor / schedule 的时候有点别扭：查出来的 `Sensor.id` 和 `sensorState.id` 看起来不是同一种 id，后面想调用 `stopSensor` / `stopRunningSchedule` 还得自己把 id 拆成两段。能不能让 state 里的 id 跟外层对象的 id 对齐，并且停止的时候直接拿查到的那个 id 就能用？\n\nExpected outcomes:\n- Sensor 查询结果中，同一个 sensor 的 `Sensor.id` 与 `Sensor.sensorState.id` 应表示同一个可直接复用的 id；客户端不需要从 state id 另行推导或拆分才能停止该 sensor。\n- Schedule 查询结果中，同一个 schedule 的 `Schedule.id` 与 `Schedule.scheduleState.id` 应表示同一个可直接复用的 id；客户端不需要从 state id 另行推导或拆分才能停止该 schedule。\n- `stopSensor` mutation 应支持通过 `id: String` 传入查询得到的 sensor id 来停止对应 sensor。\n- `stopRunningSchedule` mutation 应支持通过 `id: String` 传入查询得到的 schedule id 来停止对应 schedule。\n- 旧的两参数调用方式仍应可用：`stopSensor(jobOriginId: String, jobSelectorId: String)` 与 `stopRunningSchedule(scheduleOriginId: String, scheduleSelectorId: String)` 不应被移除。\n- 为兼容已有调用方，使用旧参数名但传入查询得到的完整 sensor/schedule id 的停止调用，也应无需客户端拆分该 id 即可定位并停止对应实体。\n- `stopSensor` 的停止参数应允许客户端只提供可用的单个查询 id，或提供完整的旧式两参数信息；当缺少可用 id 组合时，应返回错误信息 `Must specify id or jobOriginId and jobSelectorId`。\n- `stopRunningSchedule` 的停止参数应允许客户端只提供可用的单个查询 id，或提供完整的旧式两参数信息；当缺少可用 id 组合时，应返回错误信息 `Must specify id or scheduleOriginId and scheduleSelectorId`。\n\nImplementation notes:\n- 保持 GraphQL API 的向后兼容性：已有使用旧参数名的客户端应继续工作，新客户端可以直接复用查询返回的 id。\n- 具体如何表示、解析、校验这些 id，以及校验逻辑放在哪一层，由实现者根据现有代码结构决定。\n- 不要求改变 sensor 或 schedule 的启动、重置、权限检查语义；本任务只关注查询返回的 id 一致性和停止 mutation 的入参兼容性。"} {"task_id": "format-code-task-000747", "source_id": "format-code-task-000747", "domain": "code", "task_path": "tasks/format-code-task-000747", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a8c39809b30d23882e9d279021a41d8dd6325f455edd52bcb25d30ffcc5c8a86", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Components can only be constructed from Python; want a config-driven path\n\nRight now the only way to get a `PythonScript` component (and Components in general) is to instantiate them directly in Python, e.g.\n\n```python\nPythonScript(\n path=\"scripts/my_script.py\",\n specs=[\n AssetSpec(key=\"foo\", deps=[\"bar\"], group_name=\"scripts\"),\n AssetSpec(key=\"baz\"),\n ],\n)\n```\n\nThis is fine for one-offs, but for the components workflow we want to be able to declare component instances from configuration (a dict that could be loaded from YAML or similar) instead of writing Python for every instance. Something like:\n\n```yaml\n- key: foo\n deps: [bar]\n group_name: scripts\n- key: baz\n```\n\npassed alongside the script path and turned into a `PythonScript` automatically.\n\nToday there's no entry point on `Component` (or `PythonScript`) for this — `PythonScript.__init__` takes already-constructed `AssetSpec` objects, so anything wanting to drive component creation from data has to build the `AssetSpec`s itself and call the constructor manually. That defeats the purpose of having a declarative components layer.\n\nIt would be great if a Component subclass could:\n\n1. Declare what configuration shape it accepts (so callers / tooling can validate against it).\n2. Be constructed from a plain configuration value matching that shape, with the framework doing the conversion into the real domain objects (e.g. turning `{\"key\": \"foo\", \"deps\": [\"bar\"], ...}` entries into `AssetSpec`s with proper `AssetKey`s).\n\n`PythonScript` would be the first component to support this path: given the script's path plus a list of spec descriptions like the YAML above, produce an equivalent `PythonScript` to what you'd get by hand-writing the `AssetSpec(...)` calls. String fields like `key` and entries in `deps` should be parsed into `AssetKey`s, and the usual spec fields (description, metadata, group_name, skippable, code_version, owners, tags) should all be supported. If no specs are supplied in the config, behavior should match `PythonScript(path=...)` with no specs (i.e. the existing default).\n\nThe new entry point on the component side I'd expect is something like `PythonScript.from_component_params(path, component_params=...)`."} {"task_id": "format-code-task-000748", "source_id": "format-code-task-000748", "domain": "code", "task_path": "tasks/format-code-task-000748", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4c593ccd07662b6c0b5da1261ed0d89c5e612b90545f213296bcab507cc60610", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI am getting a panic when a request with a query parameter that is defined as an int array is sent."} {"task_id": "format-code-task-000749", "source_id": "format-code-task-000749", "domain": "code", "task_path": "tasks/format-code-task-000749", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f42ed5af3ab25c45023ab7985c9732e05702a5187719f2da4220e763b61593fe", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\n我在给 daos 写一个新的 test SConscript，需要用 mpicc 编一部分东西，但又不想让所有用 base_env 的子模块都被强行塞上 mpicc 的路径——现在基础环境在初始化时会被加入 MPI 编译器相关 PATH，我没法拿到一个\"干净的 base_env\"，也没法单独 Import 一个配好 MPI 的环境。能不能把这俩拆开，让 base_env 保持基础环境，再单独给我一个配好 MPI 的环境可以 Import？顺便如果机器上根本没装 mpich/openmpi，最好 scons 跑的时候能明确提示一下要跳过，不然现在静默不动我都不知道是没生效还是配漏了。\n\n## Expected outcomes\n\n- Environment export behavior:\n - SConscripts must be able to `Import('base_env_mpi')` in addition to the existing exported build environments.\n - When MPI is available for a normal build, `base_env_mpi` should be usable by MPI-dependent SConscripts as an MPI-capable environment.\n - When MPI is not available for a normal build, `base_env_mpi` should be `None` so MPI-dependent SConscripts can skip themselves through that public convention.\n - `base_env` should remain the non-MPI base environment, so SConscripts that only import and use `base_env` are not forced to inherit MPI compiler PATH changes.\n\n- MPI-unavailable behavior:\n - During normal builds, if MPI cannot be found or configured, the build output should clearly state that MPI-dependent tests are being skipped.\n - The user-facing output should include the guidance strings `Skipping compilation for tests that need MPI` and `Install and load mpich or openmpi`.\n - MPI-dependent test build logic should be able to skip itself by observing that `base_env_mpi` is `None`.\n\n- MPI-available behavior:\n - When MPI is available, MPI-dependent SConscripts should be able to use MPI without changing the observable environment of `base_env`.\n - Existing non-MPI build paths should continue to use `base_env` without requiring MPI-specific setup.\n\n## Implementation notes\n\nThe exact placement of the MPI configuration logic and the way SConscripts organize their local environments are up to the implementer. Prefer behavior-preserving changes for non-MPI build paths, and avoid coupling unrelated submodules to MPI-specific environment changes."} {"task_id": "format-code-task-000750", "source_id": "format-code-task-000750", "domain": "code", "task_path": "tasks/format-code-task-000750", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c3b88880ebedca568e07df266b974e44f0d13b04f60a8068490e05de824c2eff", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `RequestModel.to_curl(extra_args: str = \"\") -> str` to serialize a request model into a reproducible multiline `curl` command string. For a GET request with URL `https://example.com/api`, it should return exactly `curl \\\n 'https://example.com/api'`. For a POST request with JSON body content, it should include `-X POST`, a `-d ''` argument, and the quoted URL in that order.\n\nThe method should include enabled headers as `-H 'Name: value'`, enabled query parameters merged with any query string already present in the URL, and percent-encode query values using normal URL encoding. It should preserve URL fragments after adding query parameters. Enabled form data items should each become `-d 'name=value'`; enabled cookies should each become `--cookie 'name=value'`.\n\nBasic auth should serialize as `-u 'username:password'`, and digest auth should serialize as `--digest -u 'username:password'`. Request options should map `follow_redirects=False` to `--no-location`, `verify_ssl=False` to `--insecure`, non-default timeout values to `--max-time `, and proxy URLs to `--proxy ''`. If `extra_args` is provided, it should be inserted immediately after `curl`. Disabled headers, query params, form fields, and cookies should be skipped. Calling the method should not mutate the `RequestModel`, should have no filesystem or network side effects, and the same model state plus the same `extra_args` should always produce the same string."} {"task_id": "format-code-task-000751", "source_id": "format-code-task-000751", "domain": "code", "task_path": "tasks/format-code-task-000751", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:54dfdba6a0f4c0587b2549cd2db96c8606a7b70b393c247e6c181e16d55ed325", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### `missing-test-assertion` doesn't catch missing assertions in custom test wrappers\n\nWe use `dart-code-metrics` and recently enabled the `missing-test-assertion` rule. It works fine when we write tests directly with `test(...)` or `testWidgets(...)`, but it doesn't help us at all on the tests that go through our own helpers.\n\nIn our codebase (and I think this is pretty common in larger Flutter projects) we have a thin wrapper around `test`/`testWidgets` that sets up some shared boilerplate — something like:\n\n```dart\nvoid companyTest(String description, FutureOr Function() body) {\n test(description, () async {\n // shared setup for our company's tests\n await body();\n });\n}\n```\n\nand then test files use it like:\n\n```dart\nvoid main() {\n companyTest('does the thing', () {\n final a = 1;\n final b = a + 1;\n // oops, forgot the expect(...)\n });\n}\n```\n\nThe whole reason we wanted this rule on was to catch exactly this kind of mistake, but the linter stays silent on `companyTest` calls. From the rule docs it looks like only `test` and `testWidgets` are recognized as test methods, with no way to extend that list. `include-assertions` already lets us teach the rule about extra assertion functions (we use it for `verify` from mocktail), so it'd be very natural to have an analogous knob on the test-method side.\n\nCould the rule be made configurable so we can tell it which additional functions should be treated as test methods? That way teams using custom wrappers, or test helpers from third-party packages, can actually benefit from the rule. The new config key would presumably be something like `include-methods`, mirroring `include-assertions`."} {"task_id": "format-code-task-000752", "source_id": "format-code-task-000752", "domain": "code", "task_path": "tasks/format-code-task-000752", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:47ea975c0ee695295a139e8cde98209dc5003342d803b8a98399985150a1b426", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nWhen I call `dask.array.from_array()` on a large `numpy.memmap`, Dask seems to touch the whole mapped file while just building the graph, and on big files it can blow up memory before any computation starts. Could this avoid reading the memmap contents during tokenization and just treat it like a disk-backed array?\n\nExpected outcomes:\n- `dask.base.tokenize(...)` handles `numpy.memmap` inputs without materializing or hashing the full mapped array contents.\n- `dask.array.from_array(...)` can build a graph from a large `numpy.memmap` without paging the whole backing file into memory during graph construction.\n- Tokens for file-backed memmaps reflect the backing file and array/view metadata rather than only the array values, so distinct backing files, changed backing-file metadata, or distinct file-backed views are not accidentally treated as the same object just because their contents match.\n- Tokenization remains deterministic for the same file-backed memmap metadata across independent openings.\n- Existing tokenization behavior for ordinary in-memory NumPy arrays remains unchanged.\n\nImplementation notes:\n- The concrete token representation, metadata collection strategy, and validation location are up to the implementer.\n- The solution should preserve Dask’s public tokenization semantics while avoiding content reads specifically for disk-backed memmap arrays."} {"task_id": "format-code-task-000753", "source_id": "format-code-task-000753", "domain": "code", "task_path": "tasks/format-code-task-000753", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:86b609ec073ed5f7c8701b6a0c7b835f3f95f6b0da4c9215d16ecebb7dc3a5c9", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Fuse linear chains of tasks in a dask graph\n\nA dask graph is a plain dictionary that maps keys to *tasks*. A task is a tuple whose first\nelement is a callable and whose remaining elements are arguments; an argument may itself be a\nkey (referring to another entry in the graph), a literal value, a nested task, or a (possibly\nnested) list of any of these. For example:\n\n```python\n{'a': 1,\n 'b': 2,\n 'z': (add, 'a', 'b'),\n 'y': (inc, 'z'),\n 'x': (inc, 'y'),\n 'w': (inc, 'x')}\n```\n\nAs an optimization, long *linear* chains of tasks waste scheduler overhead: each intermediate\nkey has to be stored and looked up even though it is used exactly once. I'd like two public\nhelpers added to the core graph module.\n\n## `subs(task, key, val)`\n\nReturn a copy of `task` in which every occurrence of `key` has been replaced by `val`. The\nsubstitution must reach into nested tasks and into (arbitrarily nested) lists, while leaving the\ncallable in the leading position of a task untouched. A `task` that is itself exactly `key`\nreturns `val`; anything that neither matches `key` nor contains it is returned unchanged. For\nexample:\n\n```python\nsubs((inc, 'x'), 'x', 1) == (inc, 1)\nsubs((sum, [1, 'x']), 'x', 2) == (sum, [1, 2])\nsubs((sum, [1, ['x']]), 'x', 2) == (sum, [1, [2]])\n```\n\n## `fuse(dsk)`\n\nReturn a **new** graph (the input must not be mutated) that computes exactly the same results as\n`dsk` but with linear chains of tasks collapsed into single tasks.\n\nCollapsing works by inlining: a key `b` is substituted directly into the body of another key `a`\n(replacing the reference to `b` with `b`'s value) and then dropped as a separate entry. Inline\n`b` into `a` **if and only if**:\n\n- `a` is the *only* entry in the graph whose task refers to `b`, and\n- `a` refers to exactly one key in total — that single reference being `b`.\n\nThe second condition counts references with multiplicity: a task that mentions the same key twice\n(e.g. `(add, 'b', 'b')`) refers to two keys and therefore never has anything fused into it, and a\nkey that is referenced twice from one place is likewise not inlined — inlining there would\nduplicate work.\n\nApply this rule transitively, so a maximal linear chain `w -> x -> y -> z` collapses entirely into\nits topmost task `w`, whose value becomes the fully nested expression. Keys that are not inlined\nkeep their original key name; keys that are inlined disappear from the result. A key holding a\nplain value (not a task) is inlined under the same rule. A graph with no fusable chain comes back\nequal to the input.\n\nFor example:\n\n```python\nfuse({'a': 1, 'b': 2,\n 'z': (add, 'a', 'b'),\n 'y': (inc, 'z'),\n 'x': (inc, 'y'),\n 'w': (inc, 'x')})\n==\n{'a': 1, 'b': 2,\n 'w': (inc, (inc, (inc, (add, 'a', 'b'))))}\n```\n"} {"task_id": "format-code-task-000754", "source_id": "format-code-task-000754", "domain": "code", "task_path": "tasks/format-code-task-000754", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4e83316b461cd39065331878c7493f213e2f8ace1a94ac51235f00d5b5e3ed11", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a public `selectNext(nextStep, getState)` helper for wizard schemas that resolves the next wizard step in one call. `nextStep` can be a string, a function that receives `{ values: getState().values }`, or an object with `when` and `stepMapper`; `getState` returns the current form state object with a `values` property.\n\nWhen I call `selectNext('shipping', () => ({ values: { choice: 'x' } }))`, it should return `'shipping'`. When I call `selectNext({ when: 'choice', stepMapper: { business: 'business-details', personal: 'personal-details' } }, () => ({ values: { choice: 'business' } }))`, it should return `'business-details'`. It should also support lodash-style nested lookup, so `selectNext({ when: ['account', 'type'], stepMapper: { paid: 'billing' } }, () => ({ values: { account: { type: 'paid' } } }))` returns `'billing'`.\n\nWhen `nextStep` is a function, the helper should call it with the current values and return that function's result, including a Promise if the function returns one. For example, `selectNext(({ values }) => values.route, () => ({ values: { route: 'review' } }))` returns `'review'`. Given the same `nextStep` and the same `getState()` result, repeated calls should return the same value, and the helper should not mutate the `nextStep` definition or the state object returned by `getState`. If an object-style `nextStep` is malformed, such as missing `stepMapper`, calling the helper should fail with a normal JavaScript `TypeError` rather than silently choosing a step."} {"task_id": "format-code-task-000755", "source_id": "format-code-task-000755", "domain": "code", "task_path": "tasks/format-code-task-000755", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2db8ad32f4c4146aaa3553cfaf9042a40366cbcf0f175cc97c56ae06c6267e8a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nTests in koalas currently rely on three different helpers on `ReusedSQLTestCase` for \"is this thing equal to that thing?\":\n\n- `assert_eq`\n- `assert_array_eq`\n- `assert_list_eq`\n\nWhen I'm writing a new test it's not obvious which one to reach for, and I keep guessing wrong — the choice depends on whether the values happen to be a DataFrame/Series, a numpy-style array, or a plain Python list, which often isn't even something I care about at the call site. I just want to say \"these two results should match\" and let the helper figure out the rest.\n\nIt would be much nicer to only have `assert_eq`, and have it handle the cases that `assert_array_eq` / `assert_list_eq` were covering today, so test code doesn't have to pick between three near-duplicates."} {"task_id": "format-code-task-000756", "source_id": "format-code-task-000756", "domain": "code", "task_path": "tasks/format-code-task-000756", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f260c8214727f3f08fc053b66c619246cf5928fb049b3c217c5bb91f307ce16a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `getValueNode` / `getValueNodes` can't address individual items inside a repeated field\n\nI'm using the TS `lilac` library to render rows in our web UI. After `deserializeRow`, I want to look up specific cells in the tree by path — including elements inside repeated (array) fields.\n\nSuppose I have a schema with a repeated field `items` where each item has a `name`. After deserializing a row, I'd expect to be able to do something like:\n\n```ts\nconst row = deserializeRow(rawRow, schema);\n\n// First item's name\ngetValueNode(row, ['items', '0', 'name']);\n\n// All items' names\ngetValueNodes(row, ['items', '*', 'name']);\n```\n\nWhat actually happens:\n\n1. `getValueNode(row, ['items', '0', 'name'])` returns `undefined`. In fact any concrete-index path into a repeated field returns `undefined`.\n2. If I dump every value node with `listValueNodes(row)` and inspect their `L.path(...)`, every element under `items` reports the same path — something like `['items', '*', 'name']`. There's no way to tell two array elements apart, so even iterating the list manually doesn't help me find \"the 0th item\".\n3. `getValueNodes(row, ['items', '*', 'name'])` also returns nothing useful, because the comparison is strict equality and (separately) once I fix the per-element paths I'd still want the `*` form to work as a query that pulls every element.\n\nSo I'm stuck — there's no path I can pass to `getValueNode` that addresses a single element of a repeated field, and a wildcard query doesn't pull all of them either.\n\nThe behavior I'd expect:\n\n- Each element inside a repeated field is individually addressable in the deserialized row.\n- A path that uses `*` as the index segment (matching how schema field paths are written) should match every element of that repeated field when querying values / fields.\n\nSame issue applies to `getField(schema, path)` if I pass a concrete-index path — it doesn't find the field, because schema field paths use `*` and the lookup is strict equality.\n\nWhile we're touching this area, it would also be handy to expose a small helper like `listFieldParents(field, schema)` that returns the chain of parent schema fields for a given field — useful when walking up the tree from a leaf."} {"task_id": "format-code-task-000757", "source_id": "format-code-task-000757", "domain": "code", "task_path": "tasks/format-code-task-000757", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fb4a597b36fe1b1a69d70fbfa608673a13588ed3722823411b3ec02e4eb58f4a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Updating `databricks_service_principal` on Azure fails\n\nI'm managing service principals on an Azure Databricks workspace through Terraform. The service principal is backed by an existing Azure AD application (I pass the app GUID as `application_id`). Initial creation works fine, but any subsequent update to the resource fails on `terraform apply`.\n\nRoughly what my config looks like:\n\n```hcl\nresource \"databricks_service_principal\" \"sp\" {\n application_id = \"xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx\" # Azure AD app\n display_name = \"my-sp\"\n allow_cluster_create = true\n}\n```\n\nRepro:\n\n1. `terraform apply` — succeeds, the service principal shows up in the workspace.\n2. Change `display_name` (or flip one of the entitlement flags, e.g. `allow_cluster_create`).\n3. `terraform apply` again — the update step errors out against the Databricks SCIM endpoint. The SP is left in a broken state for me.\n\nIf I then taint and recreate it, the create path works again — so it's specifically the update path.\n\nI'm only seeing this on Azure workspaces; the same provider against an AWS workspace handles updates to service principals without complaints, so this looks Azure-specific.\n\nExpected: once a service principal is created via this resource, I should be able to keep editing `display_name` / `active` / entitlements through normal `terraform apply` cycles without the update blowing up."} {"task_id": "format-code-task-000758", "source_id": "format-code-task-000758", "domain": "code", "task_path": "tasks/format-code-task-000758", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6fe4110d8ec93eac3f0e76fc75f39c6f191133f8b9d9481cc2e543c5930b9e35", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nWhen I list SCIM users or service principals through the Go APIs, the responses are huge because every item drags along all its roles, even though I usually don’t need them for filtering/listing. Can we avoid fetching roles for those list/filter calls by default, and give me a way to do the same for group filtering when I don’t need roles? I still need looking up a group by display name to give me the full group object, though.\n\n## Expected outcomes\n\n- User list and filter calls through the Go SCIM user API should request list results without per-user roles by default.\n- Service principal list and filter calls through the Go SCIM service principal API should request list results without per-principal roles by default.\n- The public Go group filtering API should let callers explicitly choose whether group list/filter results include roles. For this task, the public contract is `GroupsAPI.Filter(filter string, includeRoles bool) (GroupList, error)`.\n- Calling `GroupsAPI.Filter` with `includeRoles` set to `false` should request group list/filter results without roles; calling it with `true` should preserve the previous behavior of requesting roles.\n- Looking up a group by display name should still return the complete group object, including fields that are omitted from the initial list/filter lookup.\n\n## Implementation notes\n\n- Except for the public API contract explicitly named above, the exact internal structure, helper functions, and call sites are up to the implementation.\n- Use the repository’s existing SCIM request patterns and error-handling conventions.\n- Preserve existing filtering semantics apart from omitting roles where the behavior above requires it.\n\n## Required output literals (exact-match contract)\n\nWhen a SCIM user, service principal, or group list/filter request must omit roles, the outgoing SCIM request MUST include the query parameter `excludedAttributes=roles`. Query parameter order is not part of the contract.\n"} {"task_id": "format-code-task-000761", "source_id": "format-code-task-000761", "domain": "code", "task_path": "tasks/format-code-task-000761", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0ef32f44a215da2a4e5762b54c501ee80b35e4b30fbca1c700640336cabcd2f3", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the Go config package to expose pure migration helpers for converting between the v2 and v3 runtime configuration data structs: `UpgradeV2ToV3(*v2.Data, fs.Filesystem) (*config.Data, error)`, `MergeV2toV3(*config.Data, *v2.Data) (*config.Data, error)`, and `DowngradeV3toV2(*config.Data) (*v2.Data, error)`.\n\nFor `UpgradeV2ToV3`, when a v2 config has `Version == 2`, `ID == \"core-a\"`, `Name == \"studio\"`, `Host.Name == []string{\"example.com\"}`, `Router.Routes == map[string]string{\"/live\":\"/memfs/live\"}`, and `Storage.Disk.Cache.Types == []string{\".ts\", \".mp4\"}`, the returned v3 config should have `Version == 3`, preserve those copied values, set `Storage.Disk.Cache.Types.Allow` to `[]string{\".ts\", \".mp4\"}`, keep the v3 default blocked cache types `[]string{\".m3u8\", \".mpd\"}`, set `Debug.MemoryLimit` to `0`, set `Storage.S3` to an empty slice, and return a nil error. For `MergeV2toV3`, I need the same field mapping applied into the destination `*config.Data` supplied by the caller, returning that same destination pointer so callers can start from v3 defaults and overlay v2 values.\n\nFor `DowngradeV3toV2`, when a v3 config has `Version == 3`, `Storage.Disk.Cache.Types.Allow == []string{\".jpg\", \".png\"}`, TLS settings filled in, storage mime types set, memory storage settings set, and debug profiling/force-gc values set, the returned v2 config should have `Version == 2`, `Storage.Disk.Cache.Types == []string{\".jpg\", \".png\"}`, preserve the TLS, storage, and debug values that exist in v2, and return a nil error. These helpers should not mutate their input structs, and slice/map fields such as host names, API access lists, Auth0 tenants, SRT log topics, router prefixes, and router routes should be copied so changing the returned config does not change the source config. Calling the helpers twice with equivalent inputs should produce equivalent outputs, with no filesystem writes, network calls, environment-variable reads, or hidden global state changes."} {"task_id": "format-code-task-000762", "source_id": "format-code-task-000762", "domain": "code", "task_path": "tasks/format-code-task-000762", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:db65aeb314ce957b8a10a38457bd668fb8a3d11752a590a6c34418fa14103b00", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `no-ref-siblings` doesn't check referenced files, and results don't tell me which file the violation is in\n\nI have an OpenAPI spec split across several yaml files using `$ref`s to pull in pieces from sibling files (a fairly large API, so keeping it all in one file got unmanageable). I'm running vacuum with the `no-ref-siblings` / `oasRefSiblings` rules enabled.\n\nTo sanity-check the linter I deliberately put a violation in one of the referenced files — a `$ref` with a `description` sitting next to it, something like:\n\n```yaml\n# components/schemas/Pet.yaml\nMyThing:\n $ref: './other.yaml#/components/schemas/Thing'\n description: this shouldn't be here\n```\n\nWhen I lint the root spec, vacuum only flags violations that live in the root file itself. The same kind of violation in any of the `$ref`-pulled files is silently ignored. If I lint that sub-file directly it gets caught, so the rule logic works — it just isn't being applied to the files that get loaded through references.\n\nRelated: even for the violations vacuum *does* report, the result doesn't tell me which file the offending node came from. With everything split across files I end up grep-ing through the whole tree to figure out where the flagged `$ref` actually lives. It would be really helpful if the rule result carried the origin file of the node it's complaining about.\n\nBoth `no-ref-siblings` (the generic one) and `oasRefSiblings` (the OAS-specific one) behave the same way here.\n\nCould the rule walk the whole set of indexed files rather than just the root one, and surface enough info in each result to identify the source file?"} {"task_id": "format-code-task-000764", "source_id": "format-code-task-000764", "domain": "code", "task_path": "tasks/format-code-task-000764", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e2908e033651b2395b60ca82ecd352ec9b14549ed5301525d2b47c65329e1979", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## iframeResizer crashes on load when `window.jQuery` is declared but undefined\n\nI'm using iframe-resizer on a page that doesn't depend on jQuery — I'm just using the native `window.iFrameResize` API. The page does, however, load a couple of other third-party scripts, and one of them declares `window.jQuery` on the global object but leaves it as `undefined` (it's a shim that conditionally exposes jQuery only if some other condition is met).\n\nAs soon as `iframeResizer.js` is included on the page, it blows up during the initial script execution, before I ever get a chance to call `iFrameResize(...)`. Nothing on the page works after that point.\n\nIf I remove the offending shim (so that `jQuery` isn't a property on `window` at all), iframeResizer loads fine and the native API works as expected. So it really does look like the script is unhappy specifically when `window.jQuery` exists as a name but doesn't actually point at a usable jQuery object.\n\nI'd expect the library to just skip its jQuery integration in this situation and keep working through the native API. Folks using iframeResizer without jQuery (or in environments where some other code has nulled it out / set it to undefined) shouldn't have to monkey-patch the global scope just to load the script.\n\nRepro is basically:\n\n```html\n\n\n\n```\n\nCould the jQuery detection be relaxed so that \"the name exists on window but the value isn't actually a jQuery\" is treated the same as \"no jQuery at all\"?"} {"task_id": "format-code-task-000766", "source_id": "format-code-task-000766", "domain": "code", "task_path": "tasks/format-code-task-000766", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:50b7e55c5e0d1e52a28d430cc7afec806f14c2c229aa30af67502ebd11dbf746", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a file-backed cache API for workflow code. The public API should provide `NewCache(dir string) *Cache`, a `Cache` value exposing its `Dir string`, and these methods: `Store(name string, data []byte) error`, `Load(name string) ([]byte, error)`, `StoreJSON(name string, v interface{}) error`, `LoadJSON(name string, v interface{}) error`, `LoadOrStore(name string, maxAge time.Duration, reload func() ([]byte, error)) ([]byte, error)`, `LoadOrStoreJSON(name string, maxAge time.Duration, reload func() (interface{}, error), v interface{}) error`, `Exists(name string) bool`, `Expired(name string, maxAge time.Duration) bool`, and `Age(name string) (time.Duration, error)`.\n\nA concrete byte-cache session should work like this: `c := NewCache(dir)` creates the directory if needed; after `c.Store(\"answer.txt\", []byte(\"42\"))`, `c.Exists(\"answer.txt\")` returns true, `c.Load(\"answer.txt\")` returns `[]byte(\"42\")`, and `c.Age(\"answer.txt\")` returns a non-negative duration. Calling `c.Store(\"answer.txt\", nil)` should remove that cached file, after which `c.Exists(\"answer.txt\")` is false and `c.Load(\"answer.txt\")` returns an error from the filesystem. Cache names should be interpreted as file names relative to `Cache.Dir`.\n\nFor JSON values, `c.StoreJSON(\"repo.json\", value)` should marshal the value as indented JSON and save it, and `c.LoadJSON(\"repo.json\", &out)` should decode the saved JSON into `out`. Passing `nil` to `StoreJSON` should delete that JSON cache entry. If the value cannot be marshaled, `StoreJSON` should return an error; if a saved file is unreadable or contains malformed JSON, `LoadJSON` should return an error.\n\nFor reload behavior, `LoadOrStore(\"payload.bin\", 0, reload)` should call `reload` when the file is missing, save the returned bytes, and return those bytes. A second call with the same name and `maxAge == 0` should return the cached bytes without calling `reload`. If `maxAge` is greater than zero and the saved file is older than that duration, `LoadOrStore` should call `reload`, replace the file, and return the new bytes; if `reload` returns an error, that error should be returned and no successful value should be reported. `LoadOrStoreJSON` should apply the same missing/fresh/expired behavior while marshaling the reload result to JSON, saving it, and decoding the selected JSON into the destination value."} {"task_id": "format-code-task-000767", "source_id": "format-code-task-000767", "domain": "code", "task_path": "tasks/format-code-task-000767", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9e6879c43570fdd48058f4175304642187411b420d622824a3be7c9776b502fa", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n**Describe the bug**\n- dbt `>1.5` introduced [model versions](https://docs.getdbt.com/docs/collaborate/govern/model-versions)\n- the precommit hooks `Check the model has properties file` and `Check the model has description` fail\n- probably because it expects description for each version, while there is none\n\n**To Reproduce**\nSteps to reproduce the behavior:\n1. Create two versions of the same model `fancy_model_v1.sql` and `fancy_model_v2.sql` \n2. Adapt the `yml` file\n```yml\n- name: fancy_model\n latest_version: 1\n description: fancy model\n config:\n contract:\n enforced: true\n columns:\n - name: tool_raw\n description: raw name of tool\n data_type: varchar(20)\n constraints:\n - type: not_null\n - type: primary_key\n warn_unenforced: False\n - name: tool\n description: tool name, \"jpg-to-pdf\"\n data_type: varchar(20)\n\n versions:\n - v: 1\n - v: 2\n columns:\n - include: all\n exclude: [tool]\n - name: new_column\n data_type: varchar(6)\n description: \"I am an experimental column\"\n```\n3. Run the `Check the model has properties file` and `Check the model has description`\n4. The error is:\n```\n#14 [dbt_precommit 4/4] RUN pre-commit run --all-files\n...\n#14 1.186 Check the model has properties file......................................Failed\n#14 8.483 - hook id: check-model-has-properties-file\n#14 8.483 - exit code: 1\n#14 8.483 models/fancy_model_v1.sql: does not have model properties defined in any .yml file.\n#14 8.483 models/fancy_model_v2.sql: does not have model properties defined in any .yml file.\n...\n#14 8.484 Check the model has description..........................................Failed\n#14 34.62 - hook id: check-model-has-description\n...\n#14 34.62 models/fancy_model_v1.sql: does not have defined description or properties file is missing.\n#14 34.62 models/fancy_model_v2.sql: does not have defined description or properties file is missing.\n```\n\n**Expected behavior**\n- `precommit` should understand that `fancy_model_v1.sql` and `fancy_model_v2.sql` are the same and have description under `fancy_model` in `yml` file\n\n**Version:**\nv0.1.0\n\n**Additional context**\nAdd any other context about the problem here."} {"task_id": "format-code-task-000768", "source_id": "format-code-task-000768", "domain": "code", "task_path": "tasks/format-code-task-000768", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:819a9f56a535769153d6d2a7b85503aeb0880b5aa256ec61384fcb3caeba280d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the Snowflake relation library API to expose `SnowflakeRelation.dynamic_table_config_changeset(relation_results: RelationResults, relation_config: RelationConfig) -> Optional[SnowflakeDynamicTableConfigChangeset]` as a pure comparison helper for dynamic tables. It should normalize the existing warehouse metadata and the requested model config, compare only the dynamic-table settings that can affect materialization decisions, and return `None` when there are no relevant changes.\n\nFor example, if the existing dynamic table and requested config both describe `target_lag='1 minute'`, warehouse parameter `WH_X`, scheduler `ENABLE`, refresh mode `AUTO`, no initialization warehouse, no `immutable_where`, no `cluster_by`, and no explicit transient difference, the method should return `None`. If the existing table has `target_lag='1 minute'` and warehouse parameter `WH_X`, while the requested config has `target_lag='5 minutes'` and `refresh_warehouse='WH_REFRESH'`, it should return a `SnowflakeDynamicTableConfigChangeset` whose `target_lag` is a `SnowflakeDynamicTableTargetLagConfigChange(action='alter', context='5 minutes')`, whose `snowflake_warehouse` is a `SnowflakeDynamicTableWarehouseConfigChange(action='alter', context='WH_REFRESH')`, whose `has_changes` is true, and whose `requires_full_refresh` is false.\n\nDifferences in `target_lag`, `snowflake_warehouse`, `snowflake_initialization_warehouse`, `scheduler`, `immutable_where`, and `cluster_by` should produce `alter` change objects with the requested value as `context`, except `target_lag` should be ignored when the requested value is `None`. A `refresh_mode` difference should produce `SnowflakeDynamicTableRefreshModeConfigChange(action='create', context=)` and require a full refresh only when the requested mode differs and is not `AUTO`; requesting `AUTO` should not produce a refresh-mode change. A `transient` difference should produce `SnowflakeDynamicTableTransientConfigChange(action='create', context=)` and require a full refresh, but only when both existing and requested transient values are explicitly known; if either value is `None`, no transient change should be reported.\n\n`SnowflakeDynamicTableConfigChangeset.has_changes` should be true only when at least one change field is populated, and `requires_full_refresh` should be true if any contained change requires full refresh. Calling `SnowflakeRelation.dynamic_table_config_changeset` twice with the same inputs should produce equal results, should not mutate `relation_results` or `relation_config`, and should not perform filesystem access, network access, or hidden global-state updates."} {"task_id": "format-code-task-000769", "source_id": "format-code-task-000769", "domain": "code", "task_path": "tasks/format-code-task-000769", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:669d8be6ea4011150e20ac472a2cd92d3521f3670e661d485e21aa5174930758", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI'm trying to trace exactly which version of a dbt node ran just from the structured log events, but `node_info` doesn't tell me the node's source checksum, so I have to go look it up somewhere else. Could dbt include the node checksum in the `node_info` event data when it's available?\n\nExpected outcomes:\n- Structured log event data that includes `node_info` should expose a `node_info.node_checksum` field.\n- When the logged node has a source checksum available, `node_info.node_checksum` should contain that checksum string so log consumers can identify the exact node source version from the log event alone.\n- When the logged node does not have checksum information available, `node_info.node_checksum` should still be present in the Python dictionary representation and should have a null/`None` value rather than a source hash.\n- The protobuf event contract should expose the same information through `NodeInfo.node_checksum`.\n\nImplementation notes:\n- The specific code path, data structure, and validation location used to populate the checksum are up to the implementer, as long as the existing structured logging behavior is preserved and the checksum is consistently represented in both the dictionary event data and the protobuf event data."} {"task_id": "format-code-task-000771", "source_id": "format-code-task-000771", "domain": "code", "task_path": "tasks/format-code-task-000771", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2f397d0d9ea5615c433ecbe345a4e1b5a81da5d4dacc02ad96c39ec73d063971", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `dcos-launch` library API is awkward to use — every launcher method demands the config/info dict be passed in again\n\nI'm trying to use `dcos-launch` as a library (not just through the CLI) to spin up a cluster, wait for it, run some checks, and tear it down. The current launcher API feels really clunky:\n\n```python\nimport launch, launch.config\n\nconfig = launch.config.get_validated_config(config_path)\nlauncher = launch.get_launcher(config)\ninfo = launcher.create(config)\nlauncher.wait(info)\nlauncher.describe(info)\nlauncher.test(info, 'py.test')\nlauncher.delete(info)\n```\n\nA couple of things bother me here:\n\n1. I already handed `config` to `get_launcher(...)` so it could build the right `BotoWrapper` / `AzureWrapper`. Then I have to hand the same `config` to `create(...)` again. Why does the launcher need it twice?\n\n2. After `create`, every subsequent call (`wait`, `describe`, `test`, `delete`) needs me to thread the `info` dict back in as the first argument. The launcher object itself knows which deployment it's responsible for — at least it should — so making me carry `info` around on every call feels redundant. If I forget and call `launcher.wait()` I get a `TypeError`; if I accidentally pass a stale dict I get really confusing behavior.\n\nThe natural way to use a launcher feels like:\n\n```python\nlauncher = launch.get_launcher(config_or_info)\nlauncher.create()\nlauncher.wait()\nlauncher.describe()\nlauncher.test('py.test')\nlauncher.delete()\n```\n\ni.e. the launcher should own the configuration / deployment state it needs, and each method just acts on that state. `create()` still returns the info dict so I can persist it to disk for a later process to pick back up via `get_launcher(info)`.\n\nCould the AWS and Azure launchers (and the CLI that drives them) be refactored so callers don't have to keep feeding the same dict back in on every method call?"} {"task_id": "format-code-task-000772", "source_id": "format-code-task-000772", "domain": "code", "task_path": "tasks/format-code-task-000772", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:854cbf4d1d46b8d2f9d19cbb3bfbd96ef31da40245efeac8abc96cff8b824390", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI'm writing a plugin and want my own store that just lets consumers subscribe via addChangeListener, but the only BaseStore I can see in the codebase comes bundled with this get/set thing that auto-dispatches APP_STORE_CHANGE through PluginSDK — which I really don't want my plugin store doing.\n\nCould we expose a lean BaseStore through PluginSDK that's basically just the change-listener part, and keep the get/set+dispatch behavior as a separate thing for stores that actually need it?\n\nWould make it way easier to build plugin-side stores without inheriting the whole app-store machinery."} {"task_id": "format-code-task-000773", "source_id": "format-code-task-000773", "domain": "code", "task_path": "tasks/format-code-task-000773", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9ca362dbcaf0bb71ed920ba92dacaa7125611c40ebf6a218eed783b1c923df44", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a `RandomForest` model class for NumPy arrays that I can use as `RandomForest(n_trees, max_depth, n_feats, classifier=True, criterion=\"entropy\")`, followed by `fit(X, Y)` and then `predict(X)`. Calling `fit` should build `n_trees` independent decision trees, each trained on a bootstrap sample of the training rows drawn with replacement, while passing through `max_depth`, `n_feats`, `criterion`, and `classifier` to each tree.\n\nAfter fitting, `predict(X)` should return a one-dimensional NumPy array with one prediction for each input row. In classification mode, it should aggregate the trees by majority vote over integer class labels, choosing the smallest class label when vote counts tie. For example, after `np.random.seed(0)`, fitting `RandomForest(n_trees=3, max_depth=2, n_feats=1, classifier=True, criterion=\"entropy\")` on `X=np.array([[0.0], [1.0], [2.0]])` and `Y=np.array([1, 1, 1])` should make `predict(np.array([[10.0], [-5.0]]))` return `np.array([1, 1])`.\n\nIn regression mode, the same class should be usable with `classifier=False` and `criterion=\"mse\"`, and `predict` should average the numeric predictions from all trees. For example, after fitting `RandomForest(n_trees=4, max_depth=2, n_feats=1, classifier=False, criterion=\"mse\")` on `X=np.array([[0.0], [1.0], [2.0]])` and `Y=np.array([2.5, 2.5, 2.5])`, `predict(np.array([[10.0], [-5.0]]))` should return `np.array([2.5, 2.5])`."} {"task_id": "format-code-task-000775", "source_id": "format-code-task-000775", "domain": "code", "task_path": "tasks/format-code-task-000775", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4fde5d5ac1e4781d3b5521659679f624a299fa0aa35563a9af9ad37c049563cb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a stateful Go API facade for managing iCode deployment and execution. The package should expose an ICodeApi type and NewICodeApi(containerService ivm.ContainerService, gitService ivm.GitService, eventService common.EventService) ICodeApi so callers can wire their container, Git, and event implementations once and then use the same API value across deploy, execute, list, and undeploy calls.\n\nICodeApi needs Deploy(baseSaveUrl string, gitUrl string, sshPath string, password string) (ivm.ICode, error) and DeployFromRawSsh(baseSaveUrl string, gitUrl string, rawSsh []byte, password string) (ivm.ICode, error). Deploy should call the Git service clone path with those arguments, start the container for the returned ivm.ICode, publish an icode.created event containing the iCode ID, path, version, commit hash, Git URL, and repository name, then return that same ivm.ICode with a nil error. DeployFromRawSsh should do the same flow but use the raw SSH clone path and preserve the raw key bytes and password when it calls the Git service. If cloning fails, either deploy method should return the zero ivm.ICode and that error without starting a container or publishing an event; if container start fails, it should return the zero ivm.ICode and that error without publishing an event; if publishing icode.created fails after the container starts, it should return the zero ivm.ICode with a nil error.\n\nICodeApi also needs UnDeploy(id ivm.ID) error. It should stop the matching container first; if stop fails, return that error and do not publish anything. When stop succeeds, publish the iCode removal event carrying the iCode ID and return the publish result.\n\nFor execution, ICodeApi needs ExecuteRequest(request ivm.Request) (ivm.Result, error), which directly forwards the request to the configured container service, and ExecuteRequestList(requestList []ivm.Request) []ivm.Result, which processes requests in order and returns one result per input request. If an individual request fails, the list result for that position should be ivm.Result{Err: err.Error()} and later requests should still run. Finally, GetRunningICodeList() []ivm.ICode should return the running iCodes reported by the configured container service, so a session can deploy an iCode, observe it in the running list, execute requests against it, undeploy it, and observe that it is no longer running."} {"task_id": "format-code-task-000776", "source_id": "format-code-task-000776", "domain": "code", "task_path": "tasks/format-code-task-000776", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:41fa11cf0ab0f63f7d77a03e91167fbfc9a39d26e213fae0496eaeb26ea39213", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: fallback / conditional transforms for summary string templates\n\nI'm using summary string templates with filter syntax to format entry summaries. Right now I can use `upper`, `lower` and `date('')`, which is great, but I keep running into two cases that aren't covered.\n\n### 1. Fallback when a field is empty\n\nSome entries in my collection don't have every field filled in. For example, I have an optional `subtitle` field, and when it's missing my summary ends up with a dangling separator or an empty section:\n\n```yaml\ncollections:\n - name: 'posts'\n label: 'Posts'\n folder: '_posts'\n summary: \"{{title}} — {{subtitle}}\"\n fields:\n - { label: 'Title', name: 'title', widget: 'string' }\n - { label: 'Subtitle', name: 'subtitle', widget: 'string', required: false }\n```\n\nI'd like a way, from within the template syntax, to say \"if this field is empty, use this literal string instead\" — so I can write something like `{{subtitle | }}` and get `Untitled` when the field is blank, without having to write a custom preview/summary component just for that.\n\n### 2. Pick one of two values based on whether a field is set\n\nRelated, but slightly different: sometimes I don't want to print the field's value at all, I just want to branch on whether it's truthy and emit one string or another. E.g. a `featured` boolean controlling whether the summary shows `★` or nothing, or a `draft` flag controlling whether to show `(draft)` or `(published)`.\n\nToday the only way I can do this is by post-processing the summary outside of the CMS config, which defeats the point of having the filter syntax.\n\nCould the template filter pipeline be extended to cover these two cases? They feel like a natural extension of the existing `upper` / `lower` / `date(...)` filters and would remove a lot of awkwardness around optional / boolean fields in summary strings.\n\nNaming-wise I'd expect something like `default('...')` for case 1 and `ternary('...', '...')` for case 2, to match the existing filter call style."} {"task_id": "format-code-task-000777", "source_id": "format-code-task-000777", "domain": "code", "task_path": "tasks/format-code-task-000777", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e225b251fb8596459111a47ff18c8845317484e93b3c3c11105c24155047c10e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Cannot extend the new role-based model with custom manage roles\n\nWe're rolling out the new aggregation-based role model from `user-authz` to our org and we'd like to compose our own scopes / capabilities on top of it. Two concrete cases:\n\n**Case 1 — a custom scope.** The built-in scopes (`deckhouse`, `kubernetes`, `networking`, ...) don't match how we slice responsibilities internally. I'd like a single \"scope\" that bundles the admin level of the `deckhouse` scope, the admin level of the `kubernetes` scope, and everything from the `user-authn` module.\n\nSo I created a `ClusterRole`, marked it as a manage role, and used a plain Kubernetes `aggregationRule` whose selectors point at the labels of the existing built-in roles, e.g.:\n\n```yaml\napiVersion: rbac.authorization.k8s.io/v1\nkind: ClusterRole\nmetadata:\n name: custom:manage:mycustomscope:admin\n labels:\n rbac.deckhouse.io/kind: manage\n rbac.deckhouse.io/level: scope\n rbac.deckhouse.io/scope: custom\naggregationRule:\n clusterRoleSelectors:\n - matchLabels:\n rbac.deckhouse.io/kind: manage\n rbac.deckhouse.io/aggregate-to-deckhouse-as: admin\n - matchLabels:\n rbac.deckhouse.io/kind: manage\n rbac.deckhouse.io/aggregate-to-kubernetes-as: admin\n - matchLabels:\n rbac.deckhouse.io/kind: manage\n module: user-authn\nrules: []\n```\n\nAggregation itself works (the role ends up with the union of rules as expected). But when I create a `ClusterRoleBinding` against it for a user, **no namespaced `RoleBinding`s show up** — whereas if I bind the same user directly to `d8:manage:deckhouse:admin`, the use-role bindings appear in the deckhouse-managed namespaces as expected. So my custom scope role is effectively ignored by the machinery that materialises the namespaced bindings.\n\n**Case 2 — extending an existing scope with a new capability.** A separate operator team installed a new cluster-scoped CRD (`MySuperResource`). I want users who hold `d8:manage:deckhouse:admin` to also get get/list/watch on this resource. The natural thing is to add a small `ClusterRole` carrying the rules and a label that says \"aggregate me into the deckhouse manage scope\":\n\n```yaml\napiVersion: rbac.authorization.k8s.io/v1\nkind: ClusterRole\nmetadata:\n name: custom:manage:capability:mycustommodule:superresource:view\n labels:\n rbac.deckhouse.io/kind: manage\n rbac.deckhouse.io/aggregate-to-deckhouse-as: admin\nrules:\n - apiGroups: [mygroup.io]\n resources: [mysuperresources]\n verbs: [get, list, watch]\n```\n\nThe rules do get aggregated into `d8:manage:deckhouse:admin`, but if this capability lives in its own namespace and I want a corresponding namespaced use-role binding produced there, I have no way to express that — and nothing happens automatically.\n\n**What I'd expect:**\n- Custom `ClusterRole`s that mark themselves as `rbac.deckhouse.io/kind: manage` and use a standard `aggregationRule` should be first-class citizens of the new role model — they shouldn't need to look like internal Deckhouse roles to be recognised.\n- Binding a subject to such a custom manage role via `ClusterRoleBinding` should automatically produce the same kind of namespaced `RoleBinding`s that binding to a built-in `d8:manage:*` role produces, covering the namespaces of all the roles it (transitively) aggregates.\n- There should be documentation in the user-authz FAQ explaining the contract: which labels a custom role must carry, how to bolt a new capability onto an existing scope, how to introduce a brand-new scope, and how to make the hook create use-role bindings in a namespace that a new capability lives in. Right now operators have to read the hook source to guess at the rules."} {"task_id": "format-code-task-000778", "source_id": "format-code-task-000778", "domain": "code", "task_path": "tasks/format-code-task-000778", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:81f5c196aacaa23a6d2a034cda1628f04e38b94d5166c8eb0ad235fdadd89b39", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## After upgrading to 1.55, the `deckhouse` Service still has stale EndpointSlices and `cm/deckhouse` checks get stuck\n\nAfter updating Deckhouse to 1.55 I noticed something odd on the `deckhouse` Service in `d8-system`:\n\n```\n$ kubectl -n d8-system get endpointslices -l kubernetes.io/service-name=deckhouse\nNAME ADDRESSTYPE PORTS AGE\ndeckhouse IPv4 9650,9651,9652 5m\ndeckhouse-abcde IPv4 9650,9651,9652 30m\n```\n\nThere are **two** EndpointSlices for the same Service:\n\n- `deckhouse` — the one Deckhouse itself manages now (this is the new behavior in 1.55, fine)\n- `deckhouse-` — the old one the built-in endpointslice-controller generated from the Service's selector, left over from before the upgrade\n\nAs a result `cm/deckhouse` checks stay stuck and webhook traffic ends up hitting the wrong endpoint occasionally (the manually-managed slice points at the real deckhouse pod, the controller-managed one points at whatever the selector matches at the moment, and they disagree during rollout).\n\nLooking at the release notes / changelog it seems the intent in 1.55 was that Deckhouse takes over endpoint management itself and the old controller-generated slice gets cleaned up on startup. From the outside it does not look like that cleanup actually happens — I waited a long while, restarted the deckhouse pod, the `deckhouse-` slice keeps coming back / never goes away. Deleting it by hand with `kubectl delete endpointslice` works for a few seconds, then a new `deckhouse-` slice shows up again, presumably re-created by the endpointslice-controller because the Service still has its selector.\n\nExpected: after upgrading, the `deckhouse` Service in `d8-system` should end up with exactly one EndpointSlice — the one Deckhouse manages itself — and the stale controller-generated slice(s) should be gone for good (not re-created on the next reconcile)."} {"task_id": "format-code-task-000779", "source_id": "format-code-task-000779", "domain": "code", "task_path": "tasks/format-code-task-000779", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4615245ec58fb8aa03b553d6d0b3107234ad18704cdbf7e9301b5ebdafc5747a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the cloud-provider validation library to expose `ValidateInstanceClassesEtcdDisk(state *State) Result` as a pure one-shot validator for provider InstanceClass `spec.etcdDisk` placement. The function should read `state.InstanceClassKind`, `state.NodeGroups`, and `state.InstanceClasses`, return a `Result`, and leave the supplied `State` and contained resources unchanged.\n\nWhen `state` is nil, it should return a blocking validation result with code `internal_state_nil`. When an InstanceClass has an empty kind or a kind equal to `state.InstanceClassKind`, and the NodeGroup named exactly `master` references it through `spec.cloudInstances.classReference.kind/name`, that InstanceClass must define `spec.etcdDisk`; otherwise the result should contain an error at `/.spec.etcdDisk` with code `master_etcd_disk_required`. For example, a `State` with `InstanceClassKind: \"DVPInstanceClass\"`, a `master` NodeGroup referencing `master-dvp`, and a `DVPInstanceClass` named `master-dvp` without `spec.etcdDisk` should return that error.\n\nWhen any non-`master` NodeGroup references an InstanceClass through a matching classReference kind and name, that InstanceClass must not define `spec.etcdDisk`; otherwise the result should contain an error at `/.spec.etcdDisk` with code `etcd_disk_forbidden_for_non_master` and the rejected `etcdDisk` value. For example, a `worker` NodeGroup referencing `worker-dvp` and a `DVPInstanceClass` named `worker-dvp` with `spec.etcdDisk: {}` should return this error. The validator should ignore NodeGroups without `cloudInstances.classReference`, references with a different kind or an empty name, and InstanceClasses whose non-empty kind differs from `state.InstanceClassKind`; unattached InstanceClasses may define `spec.etcdDisk` without an error. Calling it repeatedly with equivalent input should produce equivalent validation results and should not perform filesystem, network, or global-state side effects."} {"task_id": "format-code-task-000780", "source_id": "format-code-task-000780", "domain": "code", "task_path": "tasks/format-code-task-000780", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:796ea1af6e74bc9589e6657e2ba8633899ed79ce521103bee29429c77323c235", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## mempool: remove dead `isTreasuryEnabled` / `isAutoRevocationsEnabled` parameters\n\nA bunch of methods on `mempool.TxPool` currently require callers to pass\n`isTreasuryEnabled` and `isAutoRevocationsEnabled` flags, but the methods\nthemselves don't actually do anything with those flags — they're just\nthreaded through to the next call, which also doesn't use them.\n\nFor example, in `server.go` every time we want to evict something from the\npool we end up with code like this:\n\n```go\ntipHash := &s.chain.BestSnapshot().Hash\nisTreasuryEnabled, err := s.chain.IsTreasuryAgendaActive(tipHash)\nif err != nil {\n srvrLog.Errorf(\"Could not obtain treasury agenda status: %v\", err)\n}\n\nisAutoRevocationsEnabled, err :=\n s.chain.IsAutoRevocationsAgendaActive(tipHash)\nif err != nil {\n srvrLog.Errorf(\"Could not obtain automatic ticket revocations agenda \"+\n \"status: %v\", err)\n}\n\nnumEvicted := s.txMemPool.RemoveOrphansByTag(mempool.Tag(sp.ID()),\n isTreasuryEnabled, isAutoRevocationsEnabled)\n```\n\n…just to call into `RemoveOrphansByTag`, which forwards the two booleans to\n`removeOrphan`, which… also doesn't read them. The same pattern shows up\naround `RemoveTransaction`, `RemoveDoubleSpends`, `RemoveOrphan`, and friends\non the public surface of `TxPool`, and across a number of the internal\nhelpers that back them.\n\nThese parameters are effectively dead weight on the API: callers have to go\nfetch agenda state from the chain solely to satisfy a signature, the values\nthen get tunneled through several layers of mempool internals, and nothing\never actually consumes them.\n\nIt would be nice to clean this up — drop the unused agenda flag parameters\nfrom the affected `TxPool` methods and simplify the call sites accordingly,\nso callers no longer have to query agenda status just to feed parameters\nthat don't do anything."} {"task_id": "format-code-task-000781", "source_id": "format-code-task-000781", "domain": "code", "task_path": "tasks/format-code-task-000781", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2629e2dade6cbebef98d4f41976d992f875dd97a31282489f3f1cc9a92bca6e9", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nThe CoSi package is used at a network boundary: callers receive a public-key roster, a message, a participation policy, and signature bytes that may be truncated, oversized, or deliberately malformed. Harden this public surface so malformed inputs are ordinary verification failures, never process panics, while keeping the existing wire format and valid-signature behavior intact. Verify must require a signature containing exactly one encoded commitment, one encoded response, and the roster-sized participation mask; reject masks with bits set beyond the roster; and reject empty, nil-entry, or duplicate-key rosters before attempting cryptographic verification. The default policy remains the complete-roster policy, while a valid threshold policy continues to authorize a signature when enough listed participants contributed. Threshold values that cannot describe a non-empty roster must never authorize a mask. The Mask API must likewise be total over caller-provided indices: negative or out-of-range indices return errors, and rejected SetMask calls (wrong length or out-of-range bits) leave both the mask and aggregate public key unchanged. Existing successful complete and threshold signatures must continue to verify, and all of these failures must be reported through returned errors rather than panics."} {"task_id": "format-code-task-000783", "source_id": "format-code-task-000783", "domain": "code", "task_path": "tasks/format-code-task-000783", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fb84e67cd34235f76b6132b5be531cc836cfc08a491f7eb6a0d857966c1c872e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a public `process_context_with_edu_rerank(processor: EduRerankProcessor, context: str, query: str, temp_dir: str = \"temp\", top_k_per_level: int = 3) -> tuple[str, dict]` helper for one-shot LongBench context compression. It should take raw context text, write it to a temporary `.txt` file under `temp_dir`, call `processor.parse_file()` on that file, delete the temporary file afterward, then call `processor.semantic_filter(parsed_data, query, top_k_per_level=top_k_per_level)` and return the filtered text plus stats.\n\nThe stats dict should always use the keys `orig_tokens`, `new_tokens`, and `ratio`. Token counts should use `tiktoken`'s `cl100k_base` encoding when available, with special tokens disallowed, and otherwise fall back to `len(text) // 4`; successful compression should report `ratio` as `new_tokens / orig_tokens` rounded to 4 decimal places. For an empty context, return `(\"\", {\"orig_tokens\": 0, \"new_tokens\": 0, \"ratio\": 0.0})` without calling the processor. If writing the temp file fails, parsing returns no data, or semantic filtering returns an empty string, return the original context with `new_tokens == orig_tokens` and `ratio == 1.0`.\n\nFor example, with a processor whose `parse_file()` returns parsed data and whose `semantic_filter()` returns `\"Chapter 2: Methods\"`, calling the helper with a non-empty context and query should return `\"Chapter 2: Methods\"` and stats computed from the original and filtered text. With a processor whose `parse_file()` returns `None`, the same non-empty input should return the original context unchanged and a ratio of `1.0`. Running the module directly should also act as a small demonstration: create an `EduRerankProcessor`, process a sample context and query, and print the original token count, filtered token count, compression ratio, and filtered text."} {"task_id": "format-code-task-000784", "source_id": "format-code-task-000784", "domain": "code", "task_path": "tasks/format-code-task-000784", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fee10f54428f5e37f48680276eac2255808920e86f4c03aa551128ff2b49b2b7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI’m converting some SVGs with `svg2rlg`, and the colors defined in a `

Hello!

'\n )\n})\n```\n\nI'd like to help and contribute this feature and have started a possible implementation here: https://github.com/honojs/hono/pull/3685 -> happy for feedback! thanks!\n\n### Related to \n- https://github.com/honojs/hono/pull/1858\n- https://github.com/honojs/hono/pull/2577"} {"task_id": "format-code-task-001517", "source_id": "format-code-task-001517", "domain": "code", "task_path": "tasks/format-code-task-001517", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2f07c8d1a6e423aabf2c0f5d06c713c1513412dd1fc9a601d1c7ef7ad59a4e9f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Attribute `xml` tags in generated structs are missing namespace info\n\nI'm generating Go bindings for a WSDL/XSD that uses XML namespaces (in my case the EPCIS schema, target namespace `urn:epcglobal:xsd:1`). For element-backed struct types, the generated code correctly carries the schema's target namespace via `XMLName`:\n\n```go\ntype Document struct {\n XMLName xml.Name `xml:\"urn:epcglobal:xsd:1 EPCISDocumentType\"`\n ...\n}\n```\n\nBut on the same types, attribute fields are emitted with no namespace info at all:\n\n```go\nSchemaVersion float64 `xml:\"schemaVersion,attr,omitempty\" json:\"schemaVersion,omitempty\"`\nCreationDate soap.XSDDateTime `xml:\"creationDate,attr,omitempty\" json:\"creationDate,omitempty\"`\n```\n\nThis is inconsistent — `XMLName` knows the type lives in `urn:epcglobal:xsd:1`, but the attributes don't. It also breaks round-tripping with real XML where the attributes are namespace-qualified: those attributes don't match the generated fields on unmarshal (they silently get dropped), and on marshal nothing tells `encoding/xml` to emit them under the right namespace either.\n\nThe same problem shows up for nested types whose attributes belong to a different schema. For example, types under the SBDH schema (`http://www.unece.org/cefact/namespaces/StandardBusinessDocumentHeader`) generate attributes with just the local name, even though those attributes really are in that namespace.\n\nIt would be great if attribute tags picked up the target namespace of the schema they were declared in, the same way element `XMLName` already does. You can reproduce this with the `fixtures/epcis` WSDL."} {"task_id": "format-code-task-001518", "source_id": "format-code-task-001518", "domain": "code", "task_path": "tasks/format-code-task-001518", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1d45e31c7605a08a1d460fbbb4ffa9b6f5df8a0b8ff5a1e99e9d7e2e14bfb83c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want PDFPage instances to support page-level text drawing and default text styling. A caller should be able to use a page returned by PDFDocument.addPage() or PDFDocument.getPages(), then call setFont(font: PDFFont): void, setFontSize(fontSize: number): void, setFontColor(fontColor: Color): void, setLineHeight(lineHeight: number): void, and drawText(text: string, options?: PDFPageDrawTextOptions): void.\n\nCalling drawText('Hello', { x: 40, y: 700, size: 18 }) should append PDF content so the saved document renders Hello at that page position in an embedded font; if no font has been selected and no font option is supplied, the page should automatically use StandardFonts.Helvetica. Calling page.moveTo(50, 600); page.setFontSize(12); page.setLineHeight(20); page.drawText('A\\nB') should render A at the current position and B on the next text line using a 20-unit line height. Defaults set with setFont, setFontSize, setFontColor, and setLineHeight should affect later drawText calls on that same page, while explicit drawText options such as font, size, color, lineHeight, x, and y should override the defaults for that call.\n\nThe drawText options should also support opacity, blendMode, rotate, xSkew, ySkew, maxWidth, and wordBreaks. When maxWidth is provided, drawText should split cleaned text into multiple lines using the selected font's width measurements and either the page document's default word breaks or the supplied wordBreaks, then encode each line with the selected font before writing the text operators. Tabs and Unicode line separator characters should be normalized before drawing, backspace and vertical-tab control characters should be removed, and newline/form-feed/carriage-return boundaries should create separate rendered lines. Passing options.font should use that font for the current drawText call without changing the page's existing default font for later calls.\n\nInvalid inputs should be rejected consistently with the rest of the page drawing API: text must be a string, font must be a PDFFont, numeric options such as x, y, size, lineHeight, and maxWidth must be numbers, opacity must be between 0 and 1, wordBreaks must be an array, and blendMode must be one of the public BlendMode values. Separate PDFPage instances should keep independent default text font, size, color, and line-height state."} {"task_id": "format-code-task-001519", "source_id": "format-code-task-001519", "domain": "code", "task_path": "tasks/format-code-task-001519", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d940e96651de11523bd2f94221e9e10707820d3683b86c97f3e66e050ed24811", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI'm trying to manage a MUC from aioxmpp and I need to see which JIDs currently have a given affiliation in the room, like owners or members, instead of only being able to set them. I also want to set the room's vCard/avatar, but `set_vcard` seems to only target my own account right now. Could you add support for those MUC admin use cases, ideally with a small example showing how to do it?\n\n# Expected Outcomes\n\n- MUC affiliation lookup: the MUC service exposes a public async `get_affiliated` API which takes a room JID and an affiliation name, queries that room, and returns the JIDs reported by the room for that affiliation.\n- MUC affiliation lookup errors: authorization or server errors raised while querying affiliations are propagated through the normal aioxmpp IQ error path.\n- Foreign vCard writes: `aioxmpp.vcard.Service.set_vcard` accepts an optional `jid` keyword argument; when supplied, the vCard update targets that entity so callers can update a MUC or other foreign entity vCard.\n- Existing vCard behavior: calling `aioxmpp.vcard.Service.set_vcard(vcard)` without `jid` continues to update the connected account's own vCard, preserving compatibility with existing callers.\n- Examples and documentation: include small examples named `get_muc_affiliations.py` and `set_muc_avatar.py`, and mention the new MUC affiliation lookup and targeted vCard update capabilities in the changelog or equivalent user-facing documentation.\n\n# Implementation Notes\n\n- Follow the existing aioxmpp service conventions for asynchronous APIs, IQ submission, errors, and examples.\n- The exact internal helper structure and validation placement are up to the implementation, as long as the public API behavior above is satisfied.\n- Keep the existing public behavior of unrelated MUC and vCard operations unchanged."} {"task_id": "format-code-task-001520", "source_id": "format-code-task-001520", "domain": "code", "task_path": "tasks/format-code-task-001520", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1cc0652cdf01917267852a4e9e58c034122db5af4c5ff279863d3db70d216950", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Stale gating in `.buildkite/gen-pipeline.sh` should be cleaned up\n\nWhile looking at the Buildkite pipeline generation script, I noticed a couple of conditional checks in there that no longer reflect the current test matrix and the current testing scope.\n\n**1. Dead Python 3 gating around the Elastic tests**\n\nIn `run_gloo_integration`, the Elastic test block is wrapped in a check that only schedules the tests when the test name implies Python 3. Every test configuration we still maintain in the `tests=( ... )` array at the top of the script already targets Python 3, so this guard never evaluates to false anymore — it's just dead code that makes the function harder to read. The Elastic tests should simply run for every gloo configuration unconditionally.\n\n**2. Spark integration tests are skipped on pure-gloo configurations**\n\nIn `run_spark_integration`, several test commands (the Spark Keras Rossmann Run/Estimator, Spark Keras MNIST, and Spark Torch MNIST) are wrapped in a guard that excludes test configurations whose name contains `gloo` unless it also contains `openmpi-gloo`. The original intent seems to have been \"don't run these on gloo-only setups,\" but Spark integration on gloo is something we want to cover going forward. These Spark example runs should execute on the gloo-only configurations as well, just like they already do on the mixed/openmpi setups.\n\nCould we drop both of these stale conditions so the pipeline reflects what we actually want to run? The other surrounding conditions (e.g. the TF1 / `tf2` / `torch0_` / `mpich` / `oneccl` exclusions in `run_spark_integration`, the GPU-queue guard around `test_spark_keras.py` / `test_spark_torch.py`) should stay as they are — only the two gates described above are stale."} {"task_id": "format-code-task-001521", "source_id": "format-code-task-001521", "domain": "code", "task_path": "tasks/format-code-task-001521", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2f706ee9990d887cdbcc06fab518f735e80d55f97755c0dcfa2de1291bdc29d1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want an `EnvironmentLight` class in the package so I can add an HDR environment texture to a Three.js scene and have `RayTracingRenderer` use it as infinite-area lighting. The class should be constructible as `new EnvironmentLight(map, color, intensity)`, extend `THREE.Light`, store the provided `map` on `.map`, and pass `color` and `intensity` through to the base light so `new EnvironmentLight(texture, 0x555555, 0.5)` has that color and intensity. Instances should be addable to a `THREE.Scene` like other lights, and when I later call `renderer.render(scene, camera)` the renderer should recognize valid `EnvironmentLight` instances in that scene.\n\nA valid environment map is a texture with image `width`, `height`, and RGBE or linear-encoded image `data`; invalid environment lights should be skipped with a console warning instead of crashing render setup. When a valid `EnvironmentLight` is present, the renderer should initialize its environment lighting texture from that map's image dimensions and decoded RGB data, scaled by the light's intensity, before adding any ambient or directional light contribution. If multiple valid `EnvironmentLight` instances are in the scene, the first valid one should be the map that seeds the environment lighting texture. Calling `.copy(source)` on an `EnvironmentLight` should preserve normal `THREE.Light` state and copy `source.map` onto the destination light."} {"task_id": "format-code-task-001522", "source_id": "format-code-task-001522", "domain": "code", "task_path": "tasks/format-code-task-001522", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ac1236526da4458b709d999445ab2324b49d1c00ce5764c14a8c3ba5cdec338f", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `RayTracingRenderer(params = {})` instances to support a `maxHardwareUsage` boolean mode for high-throughput rendering. When I create a renderer, set `renderer.maxHardwareUsage = true`, and repeatedly call `renderer.render(scene, camera)`, each render call should trace and present a full-resolution sample for the whole canvas instead of using the normal adaptive tile-by-tile path.\n\nIn this mode, if the camera changes between render calls, the renderer should reset the accumulated sample count, clear the current accumulated lighting, update the camera state, render one full-resolution sample, reproject/present it to the canvas, and leave `renderer.getTotalSamplesRendered()` at `0` for that first changed-camera frame. If the camera is unchanged on the next `render(scene, camera)` call, it should add another full-canvas sample to the accumulation, present the reprojected and tone-mapped result, and `renderer.getTotalSamplesRendered()` should increase by one. The mode should still do nothing until the renderer is ready to draw, and it should use the same public `render(scene, camera)` lifecycle as the default mode rather than requiring a separate method."} {"task_id": "format-code-task-001523", "source_id": "format-code-task-001523", "domain": "code", "task_path": "tasks/format-code-task-001523", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:090c62530aafca48a8c67345dc39bb68c47d84247594d015d71cf45000962660", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a stream-backed raw Hprose reader API for code that needs to copy encoded values without decoding them. Users should be able to construct it as `new hprose.RawReader(stream)` where `stream` is a `hprose.BytesIO`, then call `readRaw() -> Uint8Array` to receive the exact bytes for the next complete Hprose value while advancing the stream past that value.\n\nFor a stream containing the serialized string `s5\"Hello\"`, `readRaw()` should return bytes whose string form is exactly `s5\"Hello\"`; a second `readRaw()` on a stream containing `i123;s3\"abc\"` should first return `i123;` and then return `s3\"abc\"`. For nested values, `readRaw()` should include the whole encoded structure, such as returning `a2{12}` for a list of two digits and returning both a `c...{...}` class definition and the following object bytes when the next value starts with a class tag. It should also preserve raw error payloads, so a value beginning with the Hprose error tag returns that tag plus the following encoded string.\n\n`hprose.Reader` should expose the same `readRaw() -> Uint8Array` method while still using its current stream position, because RPC callers use it to return `Serialized` results and raw batched response pieces. If the next byte is not a valid Hprose serialization tag, `readRaw()` should raise the same unexpected-tag `Error` style used by the normal reader. Calling `readRaw()` should not decode objects, create class instances, update reference tables, or mutate anything except the stream cursor."} {"task_id": "format-code-task-001524", "source_id": "format-code-task-001524", "domain": "code", "task_path": "tasks/format-code-task-001524", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:003dd14f857bf27d0bccd99401f0e02fb2f52b9707e9b6773831e402bb21d088", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `generate-wrapper.py` to support the optional header-discovery controls `--include-dir`, `--omit-prefix`, `--ignore-headers`, and `--ignore-other` while generating wrapper files from C headers. The invocation remains the normal CLI shape, for example `./generate-wrapper.py --include public.h --sys-include '\"public.h\"' --soname libdemo.so --init-name demo --output-header demo-wrap.h --output-implementation demo-wrap.c`, with any of these discovery options added as needed.\n\n`--include-dir DIR` may be supplied more than once, and each directory should be added to the C preprocessor include search path used for parsing the `--include` headers. For example, if `public.h` contains `#include \"dep.h\"` and `dep.h` is only in `extra/`, then adding `--include-dir extra` should let the command parse successfully, exit 0, and include eligible functions declared through that header graph in the generated wrapper output.\n\n`--omit-prefix PREFIX` may be supplied more than once, and any discovered function whose name starts with one of those prefixes should be excluded from both generated files. For a header declaring `int public_api(void); int _private_api(void);`, running with `--omit-prefix _` should still generate the `public_api` macro, extern pointer declaration, and `dlsym` assignment, but `_private_api` should not appear in the generated header or implementation.\n\n`--ignore-headers TEXT` may be supplied more than once, and functions whose declarations come from a header path containing one of those text values should be excluded. If `public.h` includes `dep.h`, `public_api` is declared in `public.h`, and `dep_api` is declared in `dep.h`, then `--ignore-headers dep.h` should generate wrappers for `public_api` but not for `dep_api`.\n\n`--ignore-other` should restrict discovery to declarations that come from the files explicitly passed with `--include`, so transitive system or dependency header declarations do not leak into the wrapper. The paired `--no-ignore-other` form should also be accepted and should leave transitive declarations eligible. Supplying one of these options that requires a value, such as `--omit-prefix` or `--include-dir`, without its value should be a CLI usage error on stderr with a non-zero exit and should not create successful wrapper output."} {"task_id": "format-code-task-001525", "source_id": "format-code-task-001525", "domain": "code", "task_path": "tasks/format-code-task-001525", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3768833b0e2332cd817e34f919d7fe024ea7a5ce7b79d6910026a954c29612e9", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want generated public wrapper headers to be safe to include from C++ source files. When I run `./generate-wrapper.py --include sample.h --sys-include '\"sample.h\"' --soname libsample.so --init-name sample --output-header sample-wrap.h --output-implementation sample-wrap.c`, the generated `sample-wrap.h` should wrap the exported function-pointer declarations and the `initialize_sample(int verbose);` declaration in an `extern \"C\"` block guarded by `#ifdef __cplusplus` and closed before the final include guard `#endif`. For a header that discovers `int sample_add(int, int);`, the output should contain `#ifdef __cplusplus`, then `extern \"C\" {`, then the wrapper declarations such as `extern int (*sample_add_dylibloader_wrapper_sample)(int, int);` and `int initialize_sample(int verbose);`, followed by `}` and `#endif` for the C++ guard. Plain C consumers should still see the same declarations because the block is only active when `__cplusplus` is defined."} {"task_id": "format-code-task-001526", "source_id": "format-code-task-001526", "domain": "code", "task_path": "tasks/format-code-task-001526", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:eee2cad4e44a1a7a547afc1e7995fc706d50cdaef0d46e04384e37013cc64491", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `SelectBuilder` to support row-locking clauses through these fluent methods: `ForUpdate(tables ...string) *SelectBuilder`, `ForShare(tables ...string) *SelectBuilder`, `SkipLocked() *SelectBuilder`, and `NoWait() *SelectBuilder`. In a builder session such as `sb := NewSelectBuilder(); sb.Select(\"*\").From(\"user\").Where(sb.Equal(\"id\", 1234)).ForUpdate(); sql, args := sb.Build()`, the SQL should be `SELECT * FROM user WHERE id = ? FOR UPDATE` and the args should be `[1234]`.\n\n`ForUpdate(\"u\", \"p\")` should append `FOR UPDATE OF u, p`, and `ForShare(\"accounts\")` should append `FOR SHARE OF accounts`; with no table names they should append just `FOR UPDATE` or `FOR SHARE`. `SkipLocked()` and `NoWait()` should add `SKIP LOCKED` or `NOWAIT` after whichever locking mode was selected, for example `Select(\"*\").From(\"jobs\").ForUpdate().SkipLocked().Build()` should produce `SELECT * FROM jobs FOR UPDATE SKIP LOCKED`, while `Select(\"*\").From(\"jobs\").ForShare().NoWait().Build()` should produce `SELECT * FROM jobs FOR SHARE NOWAIT`. If both wait modifiers are called, the later call should win so the output never contains both `SKIP LOCKED` and `NOWAIT`.\n\nCalling `SkipLocked()` or `NoWait()` without first selecting `ForUpdate` or `ForShare` should not add a stray trailing clause. The locking clause should be emitted after pagination clauses, and `SQL(...)` called after a locking method should inject raw SQL after the locking clause, consistent with the rest of the builder's clause-position injection behavior."} {"task_id": "format-code-task-001527", "source_id": "format-code-task-001527", "domain": "code", "task_path": "tasks/format-code-task-001527", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:06c010dc88ac320d2228add8e4a34a449f71d7d963e043b6f73f4384bb39422d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add the InceptionNeXt model family\n\nWe'd like `timm` to ship the **InceptionNeXt** architecture\n(*\"InceptionNeXt: When Inception Meets ConvNeXt\"*, https://arxiv.org/abs/2303.16900).\n\nInceptionNeXt is a ConvNeXt-style hierarchical convolutional backbone whose token\nmixer is an *Inception depthwise convolution*: the channels of each block are split\ninto groups that are processed by a small square depthwise conv, a 1×k band conv, a\nk×1 band conv, and an untouched identity branch, then concatenated. Blocks follow the\nMetaFormer recipe (token mixer → norm → channel MLP, residual, with a layer-scale\nparameter), arranged into four sequential stages with a patchify stem.\n\nPlease implement it and register it as standard `timm` models so it works through the\nusual public API (`timm.create_model`, `timm.list_models`, feature extraction, etc.).\n\n## Required variants\n\nRegister at least these three named models, with the published per-stage\nconfigurations:\n\n| model name | stage depths | stage channel widths |\n|------------------------|--------------------|-----------------------------|\n| `inception_next_tiny` | (3, 3, 9, 3) | (96, 192, 384, 768) |\n| `inception_next_small` | (3, 3, 27, 3) | (96, 192, 384, 768) |\n| `inception_next_base` | (3, 3, 27, 3) | (128, 256, 512, 1024) |\n\n## Observable contract\n\nEach registered variant must behave like a normal `timm` classification model:\n\n- **Construction & listing.** It is discoverable via `timm.list_models('inception_next*')`\n and buildable with `timm.create_model(name, pretrained=False)`.\n- **Classification forward.** Default `num_classes` is `1000`; a forward pass on an\n `[B, 3, H, W]` input returns a `[B, num_classes]` tensor (free of NaNs). Passing\n `num_classes=N` changes the output width to `N`, and `get_classifier()` returns the\n final classifier layer whose `out_features == N`.\n- **Input channels.** `in_chans` is configurable (e.g. `in_chans=1` accepts single-channel input).\n- **Stride / downsampling.** The backbone has four stages and a patchify stem giving a\n total spatial reduction of 32× (so a 224×224 input yields a 7×7 final feature map).\n `forward_features(x)` returns the *unpooled* 4-D `NCHW` feature map whose channel\n dimension equals `model.num_features`, which in turn equals the last stage's width\n (768 for tiny/small, 1024 for base).\n- **Head / pooling control.**\n - `reset_classifier(0)` removes the classifier; a subsequent forward returns the\n globally-pooled 2-D tensor `[B, num_features]`.\n - `reset_classifier(0, '')` additionally disables global pooling; a forward then\n returns the unpooled 4-D feature map `[B, num_features, h, w]`.\n - Building with `num_classes=0, global_pool=''` produces the same unpooled 4-D output\n directly.\n- **Feature extraction.** Building with `features_only=True` yields a model that returns\n the four stage outputs as a list. `feature_info.channels()` equals the stage widths\n from the table above and `feature_info.reduction()` equals `[4, 8, 16, 32]`; each\n returned feature map has the matching channel count and spatial reduction.\n- **Default config.** `model.default_cfg` reports `num_classes == 1000` and\n `input_size == (3, 224, 224)`, and its `first_conv` and `classifier` entries name\n parameters that actually exist in the model's `state_dict`.\n- **Trainability.** A backward pass through any variant produces gradients for every\n parameter.\n"} {"task_id": "format-code-task-001529", "source_id": "format-code-task-001529", "domain": "code", "task_path": "tasks/format-code-task-001529", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:631239f7eaddb7b4cf1358f2c5fc431b1c433fccca074e5478447a37b35e3435", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## TF OpenAI GPT model fails under mixed precision (AMP) and XLA\n\nI'd like to train/run `TFOpenAIGPTModel` (and `TFOpenAIGPTForSequenceClassification`) with mixed precision and/or XLA, both of which work fine on most other TF models in this repo. With OpenAI GPT specifically, neither works.\n\n### AMP (mixed precision)\n\nMinimal repro:\n\n```python\nimport tensorflow as tf\nfrom transformers import TFOpenAIGPTModel, OpenAIGPTConfig\n\ntf.keras.mixed_precision.set_global_policy(\"mixed_float16\")\n\nconfig = OpenAIGPTConfig()\nmodel = TFOpenAIGPTModel(config)\n\ninput_ids = tf.constant([[1, 2, 3, 4, 5]])\nout = model(input_ids) # blows up with a dtype mismatch\n```\n\nOther TF models in `transformers` run end-to-end under the `mixed_float16` policy, so I'd expect this one to as well.\n\n### XLA\n\nSame story when I try to compile the sequence classification head with XLA:\n\n```python\nimport tensorflow as tf\nfrom transformers import TFOpenAIGPTForSequenceClassification, OpenAIGPTConfig\n\nconfig = OpenAIGPTConfig(pad_token_id=0, num_labels=2)\nmodel = TFOpenAIGPTForSequenceClassification(config)\n\n@tf.function(jit_compile=True)\ndef run(x):\n return model(x).logits\n\nrun(tf.constant([[1, 2, 3, 0, 0], [4, 5, 6, 7, 0]]))\n```\n\nThis fails to compile under XLA, while eager / non-jit calls work fine.\n\n### Expected\n\nBoth AMP and XLA should work for the TF OpenAI GPT models, the same way they do for the other TF model families in this repo. Could the OpenAI GPT TF implementation be made AMP- and XLA-compliant? Happy to help test."} {"task_id": "format-code-task-001530", "source_id": "format-code-task-001530", "domain": "code", "task_path": "tasks/format-code-task-001530", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:54ba133b38f6328e0b4656e905898931b110f973b3bc0e0494badce5a72e0e25", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Loading a SentencePiece tokenizer with `byte_fallback` as a fast tokenizer hard-fails\n\nAfter updating to a recent `transformers` version, converting a slow SentencePiece tokenizer that was trained with the `byte_fallback` option into its fast counterpart no longer works at all — it crashes outright instead of just notifying me about the limitation.\n\nMinimal repro: take any sentencepiece model trained with `byte_fallback=True` in its trainer spec and try to load it as a fast tokenizer (e.g. `AutoTokenizer.from_pretrained(..., use_fast=True)`, which goes through `convert_slow_tokenizer`). The conversion blows up before it returns a tokenizer.\n\nThe message itself is informative — it explains that the fast tokenizer doesn't fully implement `byte_fallback` and that there can be unknown-token differences vs. the sentencepiece version. That's a real and useful caveat. But making it a hard error means I can't load the tokenizer at all, even when I'm fine with the difference in behavior for my use case (and in earlier versions this exact situation just printed a warning and proceeded).\n\nI'd expect this to surface as a warning, not as an exception that aborts the conversion. Users who care can read the warning and decide; users who don't care still get a working fast tokenizer. Right now there's no way to opt in to \"I know, just give me the fast tokenizer anyway\" short of monkey-patching the conversion code."} {"task_id": "format-code-task-001531", "source_id": "format-code-task-001531", "domain": "code", "task_path": "tasks/format-code-task-001531", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0efd5f8b834f2b1e896cd948b36fe68f812d7d3b45e0044d9677c7f54558f7bd", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nSliding window inconsistency between PyTorch and Flax\n### System Info\n\ntransformers main (ae49b218c), Python 3.10.8\n\n### Who can help?\n\n@ArthurZucker, @sanchit-gandhi\n\n### Reproduction\n\nThe attention `sliding_window` has different interpretation for PyTorch and Flax. Here's are matching examples:\n\n**PyTorch**\n\n```python\nfrom transformers import MistralModel\nimport torch\n\nmodel = MistralModel.from_pretrained(\"hf-internal-testing/tiny-random-MistralModel\", sliding_window=2)\n\ninputs = {\n \"input_ids\": torch.tensor([[10, 20, 30, 40, 50, 60, 70, 80, 0, 0]]),\n \"attention_mask\": torch.tensor([[1, 1, 1, 1, 1, 1, 1, 1, 0, 0]])\n}\n\noutputs = model(**inputs)\n\nprint(outputs.last_hidden_state[:, 1:4, 1:4])\n```\n\n**Flax**\n\n```python\nfrom transformers import FlaxMistralModel\nimport jax.numpy as jnp\n\nmodel = FlaxMistralModel.from_pretrained(\"hf-internal-testing/tiny-random-MistralModel\", sliding_window=2, from_pt=True)\n\ninputs = {\n \"input_ids\": jnp.array([[10, 20, 30, 40, 50, 60, 70, 80, 0, 0]]),\n \"attention_mask\": jnp.array([[1, 1, 1, 1, 1, 1, 1, 1, 0, 0]])\n}\n\noutputs = model(**inputs)\n\nprint(outputs.last_hidden_state[:, 1:4, 1:4])\n```\n\nBoth snippets return different results, however, if we use `sliding_window=3` in the PyTorch version, the results are the same.\n\nIn the Flax implementation, `sliding_window=2` means that a position will attend to self, and two previous position inclusive (which intuitively seems correct to me). It looks like in the PyTorch version it is not inclusive. Which behaviour is expected?\n\n### Expected behavior\n\nThe `sliding_window` meaning to be consistent."} {"task_id": "format-code-task-001532", "source_id": "format-code-task-001532", "domain": "code", "task_path": "tasks/format-code-task-001532", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e19409ffa9c745c627e14b268641792e9d1e41b62bd219610e4abd1299b61b34", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nGeneration and speculative-decoding paths increasingly transform key/value caches instead of treating them as immutable tuples. The public `DynamicCache` and `EncoderDecoderCache` APIs need consistent bookkeeping when those transformations are composed, including models whose cached activations legitimately contain zeros.\n\nPlease harden these cache contracts without breaking the existing legacy/indexing and beam-search compatibility:\n\n- Cropping a `DynamicCache` with a negative argument removes that many tokens from the end. If the request removes more tokens than are retained, the cache is empty rather than retaining a suffix or reporting a negative length; its public sequence-length metadata must agree with the tensors.\n- Recombining batch splits must reject an empty split list with a `ValueError`, while normal split/recombine round trips preserve every layer's key/value tensors and sequence length.\n- `EncoderDecoderCache.get_seq_length()` must report the self-attention token count from cache shape, independent of tensor values, and return a normal Python integer. Cross-attention update flags must likewise reflect that a cache is populated even when all cached values are zero.\n- Encoder-decoder batch split/recombine must also work before cross-attention has been populated: preserve the self-attention cache and leave the cross-attention cache empty instead of indexing nonexistent layers.\n\nKeep behavior outside these bookkeeping and transformation boundaries unchanged; callers should continue to use the documented public cache methods and legacy conversion paths."} {"task_id": "format-code-task-001533", "source_id": "format-code-task-001533", "domain": "code", "task_path": "tasks/format-code-task-001533", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2cb460aa84a5a14c8083fc7ac98e1b1f87b371b531c8b388a24fe79caca912b6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Support cancelling an in-flight `client.post()` call\n\nI'm using `crosspost` from a CLI/service that fans a single message out to several networks at once (Bluesky, Mastodon, Twitter, Discord, LinkedIn). It works great in the happy path, but I can't find any way to **cancel** a post once `client.post()` is running.\n\nTwo cases where this hurts:\n\n1. **Timeouts.** One of the upstream services (any of them — pick your favourite) occasionally takes 30+ seconds to respond. I'd like to give the whole call a budget and bail out if it doesn't finish in time.\n2. **User-initiated cancel.** The app has a \"Post\" button and a \"Cancel\" button next to it. If the user hits Cancel while the requests are still in flight, I want to actually stop the requests, not just ignore the eventual result.\n\nRight now `client.post(message, postOptions)` returns a promise I can't interrupt. I can wrap it in `Promise.race` against a timeout, but that only lets *my* code move on — the underlying HTTP requests keep running in the background until each strategy's `fetch` finishes, which wastes API quota and connections.\n\nThe rest of the Node / Web ecosystem (`fetch`, `undici`, most HTTP clients) accepts an `AbortSignal` for exactly this. It would be great if `client.post()` accepted one too via `postOptions`, and propagated it down so that aborting the controller actually rejects the in-flight network calls across all the strategies — not just `Client`, but each individual strategy when used directly (`BlueskyStrategy`, `MastodonStrategy`, `TwitterStrategy`, etc.), since I sometimes use a single strategy on its own.\n\nRough shape of what I'd like to be able to write:\n\n```js\nconst controller = new AbortController();\n\n// somewhere else: controller.abort() when the user clicks Cancel\n// or: setTimeout(() => controller.abort(), 30_000) for a timeout\n\nawait client.post(\"Hello world!\", {\n images: [...],\n // a way to pass controller.signal here\n});\n```\n\nAfter abort, I'd expect the promise from `post()` to reject (per strategy where applicable, since `Client` already aggregates with `Promise.allSettled`) rather than resolve with whatever the server eventually returned."} {"task_id": "format-code-task-001534", "source_id": "format-code-task-001534", "domain": "code", "task_path": "tasks/format-code-task-001534", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:bf2a438657877d3377d0dfc1867a239e6a3eefe38045ab62b461a2388741a12a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `DashboardManager` class to aggregate completed tasks into chart-ready analytics for a dashboard. Calling `calculateMetricsOptimized(completedTasks, startDate, endDate)` should normalize the requested range to local-day boundaries, group tasks by their local `completedAt` day, ignore tasks with missing or invalid `completedAt`, and return an object with `xpData`, `completedTasksData`, `periodXP`, and `metrics`.\n\nFor example, if the range is January 1 through January 3 and the completed tasks are `{ completedAt: '2024-01-01T10:00:00Z', experience: 100, earlyBonus: 20, overduePenalty: -10 }`, `{ completedAt: '2024-01-02T10:00:00Z', experience: 50 }`, and `{ completedAt: 'not-a-date', experience: 999 }`, the XP dataset should contain daily values `[110, 50, 0]`, the completed-tasks dataset should contain `[1, 1, 0]`, and `periodXP` should be `160`. The XP dataset should use the label `XP Gained`, a line chart shape with `fill: false`, the manager's primary color for the line and point background, and the manager's background color for the dataset background; the task dataset should use the label `Tasks Completed`, daily task counts, the manager's background color, and the manager's primary border color.\n\nWhen `calculateMetricsOptimized([], startDate, endDate)` is called, it should return empty chart data for both XP and completed tasks, `periodXP: 0`, and zero-activity metrics for the number of days in the range. Repeating the same aggregation call on the same `DashboardManager` instance should be served from that instance's cache until the cache expires or `clearCache()` is called.\n\nI also need the supporting public methods to work directly: `processTasksBatched(tasks, startDate, endDate)` returns a `Map` keyed by each local day timestamp with the tasks completed on that day; `calculateXPDataFromBatches(tasksByDate, periodDays)` turns those groups into the XP chart dataset using task experience plus early bonus plus overdue penalty with no negative XP totals; `calculateTasksDataFromBatches(tasksByDate, periodDays)` turns them into daily completed-task counts; `calculatePeriodXPFromBatches(tasksByDate)` sums all XP across the grouped tasks; and `createEmptyChartData()` returns an empty `XP Gained` dataset using the manager's chart colors."} {"task_id": "format-code-task-001535", "source_id": "format-code-task-001535", "domain": "code", "task_path": "tasks/format-code-task-001535", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2659a905c092dd2b94d6ee79e1438c20ef78ef35bf23066036371c99ba27b6b6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Battery status from FindMy advertisements is not exposed\n\nWhen I receive a FindMy advertisement and run it through `ParseData` in\n`lib/findmy`, I get back the status byte, but there's no way in this\nlibrary to interpret what that byte means. Looking at what's available,\nthere's only a single `DefaultStatus` constant — everything else I see\nin real advertisements (different values depending on whether the tag's\nbattery is full, getting low, almost dead, etc.) is just an opaque byte\nto me.\n\nApple's FindMy protocol encodes the battery level into that status byte,\nso a tag broadcasting \"battery critical\" looks different from one\nbroadcasting \"battery full\". Right now I'd have to hardcode the magic\nnumbers in my own app to figure out which is which, and every consumer\nof this library would end up duplicating the same table.\n\nIt would be great if `lib/findmy` exposed the battery levels as part of\nits public API so I can tell, from a parsed advertisement, roughly how\nmuch battery a tag has left. Same goes for the other direction —\n`NewData` currently stamps `DefaultStatus` into outgoing advertisements,\nand it'd be nice if the meaning of that default were clearer (a freshly\nprovisioned tag should presumably advertise as having a healthy battery,\nnot just some unnamed default).\n\nI'd expect something like exported `StatusBatteryFull` / `StatusBatteryMedium` / `StatusBatteryLow` / `StatusBatteryCritical` constants, plus a `BatteryStatus(b byte) string` helper that returns names like `\"full\"` / `\"medium\"` / `\"low\"` / `\"critical\"` (and `\"unknown\"` for unrecognized values)."} {"task_id": "format-code-task-001536", "source_id": "format-code-task-001536", "domain": "code", "task_path": "tasks/format-code-task-001536", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3b220197e877e97c541eba4a17cbf0fb66fba73665ceb7a613f1d79e20bfc191", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Missing Go bindings for several OpenCV core array functions\n\nI'm using gocv to do some image/matrix processing in Go, and there are a number of standard OpenCV core array functions that I'd expect to be available but currently aren't exposed in the Go API.\n\nFor comparison, gocv already wraps things like `ExtractChannel`, `DFT`, `InRange`, `MeanStdDev`, `Multiply`, `Normalize`, etc. — but a bunch of equally common ones from the same OpenCV module (`core_array`) are missing. Right now I either have to do the work in another language and shuttle data across, or write my own cgo wrapper for each one, which kind of defeats the purpose of using the binding.\n\nThe ones I keep running into and would love to see added:\n\n- `FindNonZero` — get the locations of all non-zero pixels in a Mat (super useful after thresholding/masking)\n- `Flip` — flip a 2D array horizontally / vertically / both\n- `Gemm` — generalized matrix multiplication\n- `Hconcat` — horizontal concatenation of two matrices\n- `Idct` — inverse discrete cosine transform (the `DCT` wrapper is already there, but no inverse)\n- `Idft` — inverse discrete Fourier transform (same situation — `DFT` is there but no inverse)\n- `InsertChannel` — insert a single channel into a multi-channel Mat (the inverse of the existing `ExtractChannel`)\n- `Invert` — matrix inverse / pseudo-inverse\n- `Log` — element-wise natural log (the `Exp` wrapper is there, but no `Log`)\n- `Magnitude` — magnitude of 2D vectors from x/y components\n- `Max` — per-element max of two arrays\n\nThese are all standard `cv::` functions in the OpenCV C++ API and have direct equivalents in the Python bindings, so they should be straightforward to wrap following the same pattern as the existing `core` functions in gocv. Could these be added to the `core` package?"} {"task_id": "format-code-task-001537", "source_id": "format-code-task-001537", "domain": "code", "task_path": "tasks/format-code-task-001537", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c42fbe4776d4642ee65a51a3f5ec5f518e5b274cd72496c9741c3578b8e4e9a4", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Missing wrapper for `cv::minEnclosingCircle`\n\nI'm using gocv to analyse contours from `FindContours` and I need to compute the smallest circle that encloses each contour (OpenCV's `cv::minEnclosingCircle`). gocv already exposes `MinAreaRect` for the rotated bounding rectangle, but I can't find an equivalent for the minimum enclosing circle anywhere in the package.\n\nFor my use case (detecting roughly circular objects and measuring their radius), the rotated rectangle isn't a good fit — I really want the circle that the OpenCV function returns, including its center and radius.\n\nReference to the OpenCV function:\nhttps://docs.opencv.org/3.4/d3/dc0/group__imgproc__shape.html#ga8ce13c24081bbc7151e9326f412190f1\n\nCould a binding for `minEnclosingCircle` be added so it can be called on a `[]image.Point` like `MinAreaRect` is?"} {"task_id": "format-code-task-001539", "source_id": "format-code-task-001539", "domain": "code", "task_path": "tasks/format-code-task-001539", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:36becd9eef4f98db18733e5cf9803e3fd5c3aa081a4e38174483241bf3430d09", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a `cachedAsyncBuffer(file: AsyncBuffer, options?: { minSize?: number }) -> AsyncBuffer` helper for wrapping any existing async byte buffer with range caching. The returned object should preserve `file.byteLength` and expose `slice(start: number, end?: number): ArrayBuffer | Promise`.\n\nFor a small file, such as `file.byteLength === 100` with `{ minSize: 200 }`, the wrapper should read the whole source once with `file.slice(0, 100)`, then serve later calls like `slice(0, 10)` and `slice(50, 60)` from that cached whole buffer without calling the source again. For a larger file, such as `file.byteLength === 1000` with `{ minSize: 100 }`, two calls to `slice(0, 10)` should share the same underlying `file.slice(0, 10)` promise, while `slice(10, 20)` should make a separate source call.\n\nThe cache should normalize equivalent ranges: `slice(100)` on a 1000-byte file should cache the same range as `slice(100, 1000)`, and `slice(-10)` should cache the same range as `slice(990, 1000)`. Invalid range requests should fail before calling the source: `slice(-10, 20)` should throw `Error('invalid suffix range [-10, 20]')`, and `slice(20, 10)` should throw `Error('invalid empty range [20, 10]')`. Separate wrappers around the same source should keep independent caches."} {"task_id": "format-code-task-001540", "source_id": "format-code-task-001540", "domain": "code", "task_path": "tasks/format-code-task-001540", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:8d6c3b809d1f9837f8d202f2dd5b6aeb092c24f1febda76fd545064834c9e83a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Lander drops EVM transactions when the node responds with \"nonce too low\"\n\nWe're running the lander against an EVM chain and noticed that messages occasionally vanish even though the underlying transaction actually made it on-chain. Digging into the logs, the pattern is always the same: the provider returns a `nonce too low` error during submission, the inclusion stage treats that as a fatal submission failure, and the transaction gets dropped from the pool right after.\n\nThe thing is — `nonce too low` from an EVM node almost always means the nonce we're trying to use has already been consumed, i.e. some earlier submission of this same transaction (or its replacement) is already in the mempool or already mined. Dropping the transaction in that situation is the worst possible reaction:\n\n- If it was the very first submission attempt and the node rejected us with `nonce too low` because of a stale local nonce, we still want a chance to recover.\n- If it was a re-submission, we already have prior tx hashes recorded for this transaction; one of them is very likely on-chain. Dropping the lander-side `Transaction` means we throw away those hashes and lose track of a payload that actually went through.\n\nToday the EVM adapter's `submit` path bubbles the raw provider error up, the inclusion stage sees a generic submission error, and the tx gets dropped together with all its payloads. That's how we end up with \"successfully delivered on chain, but lander reports it as dropped\".\n\n### Expected behavior\n\nThe lander should recognize `nonce too low` from the EVM provider as a signal that the transaction may already exist on chain rather than as a hard submission failure. Instead of dropping the transaction, the inclusion stage should keep it around and let the normal status-checking path determine what actually happened (mempool / included / finalized / genuinely missing) before deciding whether to drop.\n\nIn other words: a `nonce too low` response on submit must not, by itself, cause the transaction to be removed from the inclusion pool."} {"task_id": "format-code-task-001541", "source_id": "format-code-task-001541", "domain": "code", "task_path": "tasks/format-code-task-001541", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0608844e1c5f40da080bc07610ac0ec8e1ed49462d60c5ad3d8099174ed097e9", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Rename \"transient store\" to \"protocol state store\"\n\nThe framework currently exposes a storage provider called `TransientStorageProvider` (e.g. on the framework `Provider`/`Aries` context, with a matching `WithTransientStoreProvider(...)` option). The name is misleading: the data that lives in this store isn't actually transient — it holds protocol state (connection records, per-state snapshots, namespace/threadID mappings, event data used to resume `AcceptExchangeRequest`, etc.) that protocols rely on across messages and across restarts. Several places in the code even already note this, e.g. the `SaveEvent` comment:\n\n```go\n// SaveEvent saves event related data for given connection ID\n// TODO connection event data shouldn't be transient [Issues #1029]\n```\n\nand there's already a tracking TODO on the field itself:\n\n```go\ntype Aries struct {\n storeProvider storage.Provider\n // TODO Rename transient store to protocol state store\n transientStoreProvider storage.Provider\n ...\n}\n```\n\nCalling this thing \"transient\" gives users (and us) the wrong mental model — people reasonably assume they can back it with an in-memory/ephemeral provider and lose nothing important, but in practice protocols like did-exchange, mediator, message pickup, and out-of-band depend on the records stored here to keep working.\n\nWe should rename this concept across the codebase to something that reflects what it actually is — a **protocol state** store. That includes the provider interface methods all the protocol services and controller commands depend on, the framework option used to inject it, the field/getter on the framework context, and the corresponding mocks/test helpers. Internal helpers and error messages that currently talk about a \"transient store\" should be updated to match so the terminology is consistent end-to-end.\n\nNo behavioral change is intended — this is purely a naming/terminology fix to stop calling persistent protocol state \"transient\"."} {"task_id": "format-code-task-001542", "source_id": "format-code-task-001542", "domain": "code", "task_path": "tasks/format-code-task-001542", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:be10e29641ce37743e8e06e389674d959a66db5e5d89ce761348d8a3d583ffbb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Badger transactions to provide an exact-key historical iterator with the public Go method `func (txn *Txn) NewKeyIterator(key []byte, opt badger.IteratorOptions) *badger.Iterator`. I should be able to create it inside a transaction, call the returned iterator's existing cursor methods such as `Rewind`, `Seek`, `Next`, `Valid`, `Item`, and `Close`, and use it as a stateful cursor scoped to one key.\n\nFor a database that has committed two versions of key `acct` with values `v1` then `v2`, and also has a separate key `acct:meta`, `txn.NewKeyIterator([]byte(\"acct\"), badger.DefaultIteratorOptions)` should produce only items whose `Item().Key()` is exactly `acct`; it must not continue into `acct:meta` just because that key shares the same byte prefix. The iterator should force all versions for that key to be visible, so scanning it from a current read transaction should expose both `acct` versions instead of only the newest version.\n\nIf the caller passes an `IteratorOptions` value with `Prefix` already set, `NewKeyIterator` should panic with the message `opt.Prefix should be nil for NewKeyIterator.` because this constructor owns the exact-key prefix internally. Other iterator options such as value prefetching, reverse traversal, and version cutoffs should be forwarded into the returned iterator while preserving the exact-key behavior."} {"task_id": "format-code-task-001543", "source_id": "format-code-task-001543", "domain": "code", "task_path": "tasks/format-code-task-001543", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a7c25545d1273c8d2134dfa263dea2f7a00f6bd29f96f48d416958928e3854c5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Inconsistent ROI support across signal/model methods\n\nI use ROIs all the time when working with EELS spectra and TEM/holography images in a Jupyter notebook — typically I draw a `SpanROI` or `RectangularROI` interactively on a plot and then want to feed it into whatever processing method I need next (background removal, ZLP alignment, integration, image alignment, picking a fit range in a model, etc.).\n\nThe problem is that the support for passing a ROI as an argument is really inconsistent across the API.\n\nSome methods take a ROI happily:\n\n```python\ns = hs.datasets.example_signals.EDS_TEM_Spectrum()\nroi = hs.roi.SpanROI(left=5, right=15)\ns.remove_background(signal_range=roi, background_type=\"Polynomial\") # works\n```\n\nBut many other methods that conceptually take \"the same kind of thing\" (a range, or a set of coordinates) only accept a plain tuple and reject a ROI:\n\n```python\n# Signal1D / EELS\ns_ll.align_zero_loss_peak(signal_range=hs.roi.SpanROI(-10., 10.)) # I want this to work\n\n# Model1D — I'd love to reuse the same SpanROI I drew on the plot\nm = s.create_model()\nm.set_signal_range(roi)\nm.add_signal_range(roi)\nm.remove_signal_range(roi)\nm.fit_component(g1, signal_range=roi)\n\n# Signal2D\nim = hs.signals.Signal2D(np.random.random((10, 30, 30)))\nrect = hs.roi.RectangularROI(left=2, right=10, top=0, bottom=5)\nim.align2D(roi=rect)\nim.estimate_shift2D(roi=rect)\n```\n\nRight now, for the methods that don't natively accept a ROI, I have to manually pull the attributes off the ROI and rebuild a tuple every time, e.g. `(rect.left, rect.right, rect.top, rect.bottom)` for `align2D`, `(span.left, span.right)` for the model methods, and so on — different shape per ROI type, easy to get wrong, and pretty ugly when I'm just trying to reuse the same ROI object I already have on screen.\n\nIt would be really nice if a ROI could be used in place of the corresponding coordinates/range tuple **anywhere** in the API — so that the same `SpanROI` or `RectangularROI` I drew on the plot can be passed directly to all of these methods uniformly, without me having to remember which methods support ROIs and which don't.\n\nRelatedly, in user code it would be very natural to be able to treat a ROI like the tuple it conceptually represents — e.g. unpack it directly:\n\n```python\nroi = hs.roi.RectangularROI(left=0, right=10, top=20, bottom=20.5)\nleft, right, top, bottom = roi # currently fails\n```\n\nThis would cover the common cases (`SpanROI`, `RectangularROI`, `Point1DROI`, `Point2DROI`, `CircleROI`, `Line2DROI`) and make ROIs feel like first-class arguments throughout hyperspy."} {"task_id": "format-code-task-001544", "source_id": "format-code-task-001544", "domain": "code", "task_path": "tasks/format-code-task-001544", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0cd8543e33c6979eda2190cd64a0b4109bb34da57c599d625d7aed2aa85c9488", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Create-user API rejects users that authenticate through a third-party identity provider\n\nI'm integrating with the create-user API and trying to provision accounts for users who sign in via an external identity provider (SSO / OAuth). These users don't necessarily have an email address on file with us — the identity provider is their source of truth, and I pass that along in the `identities` array.\n\nA request body like:\n\n```json\n{\n \"authority\": \"example.com\",\n \"username\": \"jdoe\",\n \"identities\": [\n {\"provider\": \"acme-sso\", \"provider_unique_id\": \"user-12345\"}\n ]\n}\n```\n\ngets rejected by `CreateUserAPISchema` because `email` is listed as required. But for this class of user I genuinely don't have an email to supply — the whole point is that the third-party provider is handling identity for them.\n\nI think `email` should only be required when the caller hasn't supplied an `identities` entry. If the request already pins the new account to one or more external identities, that should be a sufficient way to identify the user and `email` should be allowed to be omitted.\n\nOne related thing while looking at this: an empty `identities: []` shouldn't count as \"the user has external identities\" — if a caller goes the identities route, they should actually have to provide at least one. Otherwise you could omit `email` *and* pass an empty array and end up with a user that has neither, which doesn't make sense.\n\nCould the schema be relaxed so that a request is accepted as long as it has `authority` + `username` and *either* an `email` *or* a non-empty `identities` array? Existing callers that pass `email` should keep working unchanged.\n\nThis affects both `h/schemas/api/user.py` (the runtime schema) and the published `new-user-schema.json` in the API reference docs."} {"task_id": "format-code-task-001546", "source_id": "format-code-task-001546", "domain": "code", "task_path": "tasks/format-code-task-001546", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4ce2e58c67d40bfe52b2491c90de97bd434253440759945cd35a8d22f52d5f4b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: ability to remove an already-applied voucher from the cart\n\nWe're building a checkout flow on top of flamingo-commerce. Customers can enter a coupon code and we call the existing `ApplyVoucher` flow (`/api/cart/applyvoucher`) to attach it to the cart — that works fine, and the applied codes show up under `cart.AppliedCouponCodes` so we can render them in the cart summary.\n\nThe problem is the inverse operation. A pretty standard piece of cart UX is \"here are the vouchers you've applied, click the X to remove this one\" — e.g. the customer pasted the wrong code, or got handed a better promo and wants to swap. Today there's no way to do that from the cart API: I can keep adding codes, but I can't take one back off without throwing the whole cart away.\n\nI'd expect removing a voucher to be a first-class cart mutation, on the same level as applying one — you say \"drop this code from the cart\", and afterwards the cart no longer lists it and any discount tied to it is gone. It should be reachable via the cart API so the storefront/JS layer can call it directly, and it should flow through the same modify-behaviour layering that `ApplyVoucher` uses (so the in-memory adapter we use in tests/dev keeps working, and other adapters can implement it the same way).\n\nWould it be possible to add this? Naming-wise I'd expect the new operation to mirror `ApplyVoucher` (something like `RemoveVoucher` on the cart service / modify-behaviour)."} {"task_id": "format-code-task-001547", "source_id": "format-code-task-001547", "domain": "code", "task_path": "tasks/format-code-task-001547", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:66dfd8ef835ae5215c55195c3adcedd5a2a715c6daf09fd482bf6fcdc75dafe8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nSupport react-i18next \"defaults\" prop\n#### Version\n- i18next: 11.2.2\n- i18next-scanner: 2.4.6\n- react-i18next: 7.7.0\n\n#### Configuration\n```js\noptions: {\n\t// use strings as keys\n\tnsSeparator: false,\n\tkeySeparator: false,\n\t// settings\n\tdefaultNs: \"frontend\",\n\tlngs: [\"en\"],\n\tresource: {\n\t\tloadPath: \"{{lng}}/{{ns}}.json\",\n\t\tsavePath: \"{{lng}}/{{ns}}.json\",\n\t},\n\tfunc: false,\n\ttrans: false,\n\tdefaultValue: (language, namespace, key) => key,\n},\n```\n\nreact-i18n v7.7.0 introduced the `defaults` prop (see i18next/react-i18next#439). I expect the following code:\n\n```jsx\n{val}!\"\n tOptions={{val: \"World\"}}\n components=[foo]\n/>\n```\n\nto lead to the following output:\n\n```json\n{\n \"Hello <0>{val}!\": \"Hello <0>{val}!\"\n}\n```"} {"task_id": "format-code-task-001548", "source_id": "format-code-task-001548", "domain": "code", "task_path": "tasks/format-code-task-001548", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:88f833dea372e5b0a1c05eb5a6cd607a88659aae07bcfa29679104f415e91cc9", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want initialized i18next instances to support context-specific translation lookup through `t(key, options)`. After I create and initialize an instance with resources, calling `i18n.t('friend', { context: 'male' })` should look for `friend_male` before the base `friend` key, so resources like `{ friend: 'friend', friend_male: 'boyfriend', friend_female: 'girlfriend' }` return `'boyfriend'` for the male context and `'girlfriend'` for the female context. Numeric contexts should work the same way: with `{ step: 'step', step_1: 'first step' }`, `i18n.t('step', { context: 1 })` should return `'first step'`.\n\nThe context separator should be configurable at init time with `contextSeparator`. For example, if an instance is initialized with `contextSeparator: '|'` and resources `{ friend: 'friend', 'friend|formal': 'colleague' }`, then `i18n.t('friend', { context: 'formal' })` should return `'colleague'`.\n\nContext lookup should also combine with plural lookup. With resources `{ item: 'item', item_male_one: 'his item', item_male_other: 'his items' }`, `i18n.t('item', { context: 'male', count: 1 })` should return `'his item'`, and `i18n.t('item', { context: 'male', count: 2 })` should return `'his items'`. If `returnDetails: true` is passed, the returned details should report the actual suffixed key that was used, such as `exactUsedKey: 'friend_male'` for the male friend example. Empty-string, null, or missing `context` should not trigger context suffix lookup.\n\nThe public TypeScript options should expose this API too: `TOptions` should accept a `context` option, and initialization options/type defaults should accept `contextSeparator`."} {"task_id": "format-code-task-001549", "source_id": "format-code-task-001549", "domain": "code", "task_path": "tasks/format-code-task-001549", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a2bcdc96df388377d4a1a3c98fbbcbf803af3c704bd6185babac2e10b5f90b07", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Can't format the timezone name (e.g. `EST`, `Eastern Standard Time`) after `.tz()`\n\nI'm building a meeting/schedule UI with `dayjs` + the `timezone` plugin + the `advancedFormat` plugin, and I'd like to render times together with the human-readable name of the zone, e.g.\n\n```\n2020-04-22 09:00 EST\n```\n\nor with the long form:\n\n```\n2020-04-22 09:00 Eastern Standard Time\n```\n\nIn moment.js this is what `z` / `zzz` give you in the format string, and since `advancedFormat` is the plugin that fills in moment-compatible tokens, I expected the same to work here:\n\n```js\nimport dayjs from 'dayjs'\nimport utc from 'dayjs/plugin/utc'\nimport timezone from 'dayjs/plugin/timezone'\nimport advancedFormat from 'dayjs/plugin/advancedFormat'\n\ndayjs.extend(utc)\ndayjs.extend(timezone)\ndayjs.extend(advancedFormat)\n\nconst t = dayjs.utc('2020-04-22 13:00').tz('America/New_York')\n\nconsole.log(t.format('YYYY-MM-DD HH:mm z')) // wanted: 2020-04-22 09:00 EST\nconsole.log(t.format('YYYY-MM-DD HH:mm zzz')) // wanted: 2020-04-22 09:00 Eastern Standard Time\n```\n\nWhat I actually get is the literal token characters left in the output — they don't get replaced at all, so I end up with something like `2020-04-22 09:00 z` / `2020-04-22 09:00 zzz` in my UI.\n\nI also looked through the `timezone` plugin to see if there was some other way to retrieve the zone's display name (so I could append it manually), but I couldn't find anything — once you call `.tz('America/New_York')`, the resulting instance doesn't seem to expose the zone name in any human-readable form, only the numeric offset.\n\nA couple of things that I think matter:\n\n- The output **must** reflect the zone passed to `.tz(...)`, not the machine's local zone. Otherwise users in different regions would see different labels for the same scheduled meeting, which defeats the point of using `timezone` in the first place.\n- It should also work with DST — if I take a moment in summer in `America/New_York`, I'd expect `EDT` / `Eastern Daylight Time` instead of `EST` / `Eastern Standard Time`, since that's how the zone actually behaves at that instant.\n- Both a short form (`EST`, `PDT`, `CET`, ...) and a long form (`Eastern Standard Time`, ...) would be great to have, the same way moment exposes them.\n\nCould `advancedFormat` (in combination with `timezone`) be extended so that the format string supports rendering the timezone name of the bound zone? Happy to help test if there's a tentative implementation.\n\nIt would also be useful to have a direct accessor on the dayjs instance (something like `.offsetName('short')` / `.offsetName('long')`) so the zone label can be obtained programmatically without going through `format`."} {"task_id": "format-code-task-001550", "source_id": "format-code-task-001550", "domain": "code", "task_path": "tasks/format-code-task-001550", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:7b18ff4deb1c57403038e4e5cfa337d2ecee7991a554bccd1e7211d77083f0bb", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## slate-hyperscript: can't build intentionally-invalid values, and bare `` produces extra text nodes\n\nI'm using `slate-hyperscript` to write test fixtures and ran into two things that look wrong.\n\n### 1. `` always normalizes — can't test validation logic\n\nI'm writing tests for some schema / validation rules. The natural way is to hand-build a value that violates a rule and then assert that my validation code catches/fixes it. But when I build that value with hyperscript, hyperscript itself silently fixes it before my code even runs, so the test can't actually exercise the validation path.\n\nFor test fixtures I'd really like a way to say \"give me back exactly the tree I described, even if it's not strictly valid\" — otherwise hyperscript hides the very bugs I'm trying to test.\n\n### 2. Bare `` / `` adds a phantom empty text node\n\nSeparately, I noticed that if I don't wrap things in `` and just build a document directly, the resulting `document.nodes` doesn't match what I wrote. There's an extra empty text node tacked on at the end that I didn't put there.\n\nRoughly:\n\n```js\nconst doc = (\n \n hello\n world\n \n)\n\n// doc.nodes ends with an unexpected empty text node after the two paragraphs\n```\n\nI didn't catch this earlier because I usually wrap in ``, and apparently the normalization that runs there cleans the stray node up. As soon as I stopped going through `` it showed up.\n\nThese two issues actually interact — fixing #1 (so I can opt out of normalization on ``) makes #2 visible there too, so both should be addressed.\n\nFor the opt-out in #1, I'd expect something like `...` — i.e. a `normalize` attribute on the `` tag."} {"task_id": "format-code-task-001551", "source_id": "format-code-task-001551", "domain": "code", "task_path": "tasks/format-code-task-001551", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:cf2767809453303cc06c81abc0249f3b6979e44529f8876a6e178a9a78836f79", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `createEditor(): Editor` to return an editor whose `editor.apply(operation: Operation): void` can apply `set_selection` operations to the editor's retained selection state. Starting from `const editor = createEditor()` with document text `hello` and `editor.selection === null`, calling `editor.apply({ type: 'set_selection', properties: null, newProperties: { anchor: { path: [0, 0], offset: 0 }, focus: { path: [0, 0], offset: 5 } } })` should set `editor.selection` to that complete range. If the editor already has `{ anchor: { path: [0, 0], offset: 0 }, focus: { path: [0, 0], offset: 5 }, note: 'draft' }`, then applying `{ type: 'set_selection', properties: { focus: { path: [0, 0], offset: 5 } }, newProperties: { focus: { path: [0, 0], offset: 2 } } }` should update only `focus` and preserve the existing `anchor` and custom `note` property. Applying `{ type: 'set_selection', properties: , newProperties: null }` should clear the selection back to `null`. A `set_selection` operation should also clear any pending `editor.marks` so cursor formatting does not survive a selection change. If there is no current selection, applying a partial `set_selection` without both `anchor` and `focus` should throw an `Error`; applying a change that removes `anchor` or `focus` from an existing selection should also throw an `Error`."} {"task_id": "format-code-task-001552", "source_id": "format-code-task-001552", "domain": "code", "task_path": "tasks/format-code-task-001552", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:212a7bc4fb4bdeba7255c0bf50166b30b91b025a7a0d6459bfd7d91981252326", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nbug: relocate not working as expected\n### What happened?\n\nsetup code -- load the penguins table on Ibis installed on master:\n\n```python\n[ins] In [1]: import ibis\n[ins] In [2]: import ibis.selectors as s\n[ins] In [3]: ibis.options.interactive = True\n[ins] In [4]: ibis.options.repr.interactive.max_rows = 3\n[ins] In [5]: t = ibis.examples.penguins.fetch()\n\n[ins] In [6]: t\nOut[6]:\n┏━━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━┓\n┃ species ┃ island ┃ bill_length_mm ┃ bill_depth_mm ┃ flipper_length_mm ┃ body_mass_g ┃ sex ┃ year ┃\n┡━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━┩\n│ string │ string │ float64 │ float64 │ int64 │ int64 │ string │ int64 │\n├─────────┼───────────┼────────────────┼───────────────┼───────────────────┼─────────────┼────────┼───────┤\n│ Adelie │ Torgersen │ 39.1 │ 18.7 │ 181 │ 3750 │ male │ 2007 │\n│ Adelie │ Torgersen │ 39.5 │ 17.4 │ 186 │ 3800 │ female │ 2007 │\n│ Adelie │ Torgersen │ 40.3 │ 18.0 │ 195 │ 3250 │ female │ 2007 │\n│ … │ … │ … │ … │ … │ … │ … │ … │\n└─────────┴───────────┴────────────────┴───────────────┴───────────────────┴─────────────┴────────┴───────┘\n```\n\nI would expect this to put \"year\", followed by numeric columns -- instead, numeric columns are first:\n\n```python\n[ins] In [7]: t.relocate(\"year\", s.numeric())\nOut[7]:\n┏━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━┳━━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━┓\n┃ bill_length_mm ┃ bill_depth_mm ┃ flipper_length_mm ┃ body_mass_g ┃ year ┃ species ┃ island ┃ sex ┃\n┡━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━╇━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━┩\n│ float64 │ float64 │ int64 │ int64 │ int64 │ string │ string │ string │\n├────────────────┼───────────────┼───────────────────┼─────────────┼───────┼─────────┼───────────┼────────┤\n│ 39.1 │ 18.7 │ 181 │ 3750 │ 2007 │ Adelie │ Torgersen │ male │\n│ 39.5 │ 17.4 │ 186 │ 3800 │ 2007 │ Adelie │ Torgersen │ female │\n│ 40.3 │ 18.0 │ 195 │ 3250 │ 2007 │ Adelie │ Torgersen │ female │\n│ … │ … │ … │ … │ … │ … │ … │ … │\n└────────────────┴───────────────┴───────────────────┴─────────────┴───────┴─────────┴───────────┴────────┘\n```\n\nI would then expect this to do the opposite:\n\n```python\n[nav] In [8]: t.relocate(s.numeric(), \"year\")\nOut[8]:\n┏━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━┳━━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━┓\n┃ bill_length_mm ┃ bill_depth_mm ┃ flipper_length_mm ┃ body_mass_g ┃ year ┃ species ┃ island ┃ sex ┃\n┡━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━╇━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━┩\n│ float64 │ float64 │ int64 │ int64 │ int64 │ string │ string │ string │\n├────────────────┼───────────────┼───────────────────┼─────────────┼───────┼─────────┼───────────┼────────┤\n│ 39.1 │ 18.7 │ 181 │ 3750 │ 2007 │ Adelie │ Torgersen │ male │\n│ 39.5 │ 17.4 │ 186 │ 3800 │ 2007 │ Adelie │ Torgersen │ female │\n│ 40.3 │ 18.0 │ 195 │ 3250 │ 2007 │ Adelie │ Torgersen │ female │\n│ … │ … │ … │ … │ … │ … │ … │ … │\n└────────────────┴───────────────┴───────────────────┴─────────────┴───────┴─────────┴───────────┴────────┘\n```\n\n### What version of ibis are you using?\n\nmaster\n\n### What back"} {"task_id": "format-code-task-001553", "source_id": "format-code-task-001553", "domain": "code", "task_path": "tasks/format-code-task-001553", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d416d28188ba9dfce0a73f56074a0226e442bd04df4cf82d19a8b018464b6dfe", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a package-level `run_inference_retrieval(model: torch.nn.Module | None = None, preprocess: Callable | None = None, tokenizer: Callable | None = None, model_name: str = \"Llama3-MS-CLIP-Base\", pretrained: bool = True, ckpt_path: str | None = None, image_path: str | list[str] | None = None, queries: list[str] | None = None, queries_file: str | None = None, top_k: int = 5, save_path: str | None = None, device: str | None = None, verbose: bool = True) -> pandas.DataFrame` function exported from `msclip.inference` for text-to-image retrieval over custom image sets.\n\nThe function should accept either an already constructed model/preprocess/tokenizer trio or load them from `model_name`, `pretrained`, and `ckpt_path`; it should accept images either as a list of paths or through the normal image path input, encode all images and all query strings, L2-normalize both embedding sets, compute image-query dot-product similarities, and return the top `top_k` image basenames per query. The returned DataFrame should have a `MultiIndex` of `Query` and `Rank`, with columns `Image` and `Similarity`; ranks are 1-based and sorted from highest similarity to lowest for each query.\n\nFor a deterministic injected model that produces unit image embeddings `field.tif -> [1, 0]` and `snow.tif -> [0, 1]`, and a tokenizer/text encoder that maps `queries=[\"green fields\"]` to `[1, 0]`, calling `run_inference_retrieval(model=model, preprocess=preprocess, tokenizer=tokenizer, image_path=[\"/tmp/field.tif\", \"/tmp/snow.tif\"], queries=[\"green fields\"], top_k=1, device=\"cpu\", verbose=False)` should return one row indexed by `(\"green fields\", 1)` with `Image == \"field.tif\"` and `Similarity == 1.0`. With the same image embeddings and two query embeddings `queries=[\"snow cover\", \"green fields\"]` mapped to `[[0, 1], [1, 0]]`, `top_k=2` should return two ranked rows for each query: `snow.tif` then `field.tif` for `\"snow cover\"`, and `field.tif` then `snow.tif` for `\"green fields\"`, with similarities `1.0` then `0.0` in both groups.\n\nIf neither `queries` nor `queries_file` is provided, the function should raise `ValueError` explaining that query text or a query file is required. `queries_file` should support `.txt` files by reading non-empty stripped lines and `.yaml`/`.yml` files by reading the top-level `queries` list; unsupported file extensions should raise `ValueError`. When `save_path` is provided, the function should write the same ranked table as a CSV at that path with a `.csv` suffix, creating parent directories as needed; when `verbose=True`, it should print a retrieval heading and the result table, plus a saved-file message when CSV output is requested. For the same injected model, tokenizer, preprocess, images, query strings, `top_k`, and device, repeated calls should return equal DataFrames and should not mutate the input `queries` or `image_path` lists."} {"task_id": "format-code-task-001554", "source_id": "format-code-task-001554", "domain": "code", "task_path": "tasks/format-code-task-001554", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:98459cbc783284c9d7dc89854f5b1ce7393169528ae98da127590e357b52ca9e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Sarama to expose `NewBalanceStrategyRoundRobin() BalanceStrategy` for computing consumer-group partition assignments in round-robin order. The returned strategy's `Name() string` should return `\"roundrobin\"`, and its `Plan(members map[string]ConsumerGroupMemberMetadata, topics map[string][]int32) (BalanceStrategyPlan, error)` should distribute topic partitions alternately among eligible members after ordering members and topic partitions deterministically. For example, with members M1 and M2 both subscribed to T1 and T2, and T1 and T2 each containing partition 0, the plan should assign T1/0 to M1 and T2/0 to M2; with T1 containing partitions 0 and 1 and both members subscribed to T1, it should assign T1/0 to M1 and T1/1 to M2. An empty member map or empty topic map should return a non-nil error and a nil plan. `AssignmentData(memberID string, topics map[string][]int32, generationID int32) ([]byte, error)` should return `(nil, nil)`. Repeating a call with the same inputs should produce the same plan, the input maps and slices must not be mutated, and the calculation must not perform filesystem, network, or other hidden global side effects."} {"task_id": "format-code-task-001555", "source_id": "format-code-task-001555", "domain": "code", "task_path": "tasks/format-code-task-001555", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4f680d08dd0abe286e717cb184bfbcd03eef9564fcf58c102f0bd060830e0ef7", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want IITC to expose a stateful COMM message filtering API at IITC.comm.declarativeMessageFilter so plugins can register declarative rules and have matching COMM rows hidden when the chat is rendered. The API should support addRule(id: string, rule: {conditions: Array}), removeRule(id: string), getRuleById(id: string), getAllRules(), and filterMessage(message: object): boolean, with rule state retained across calls on the singleton.\n\nA rule matches only when every condition matches. A condition can read a nested message field using dot paths and bracket indexes such as player.name or markup[5][1].plain, then compare either an exact value via value or a regular expression via pattern; if invert is true, that condition result is negated. For example, after addRule('directComparison', {conditions: [{field: 'player.name', value: '43Bad'}]}), filterMessage({player: {name: '43Bad'}}) should return true, while filterMessage({player: {name: 'Spedd'}}) should return false. After addRule('regexPattern', {conditions: [{field: 'markup[0][1].plain', pattern: /UK\\)$/i}]}), a message whose first markup payload plain text is 'Bridge (UK)' should return true, and one with 'Bridge (US)' should return false.\n\nRules should be manageable by ID: getRuleById(id) returns the registered rule or null, getAllRules() returns the current rule map, removeRule(id) removes an existing rule, adding an existing ID overwrites it and logs a warning, and removing an unknown ID logs an error. Conditions with missing fields or without either value or pattern should not match. When the COMM renderer processes stored messages, it should call this filter on the parsed message object and omit rows whose filterMessage result is true, while keeping all non-matching messages in the normal rendered chat output."} {"task_id": "format-code-task-001556", "source_id": "format-code-task-001556", "domain": "code", "task_path": "tasks/format-code-task-001556", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fcf2e278e770734730065adebcafefd5dda4e1fa21f4765580bf47f162e033fa", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\n我这边用 HAP-python 起了个配件，display_name 里如果前后有空格、横杠，或者像 `💡 Reading` 这种名字，Home app 经常搜不到；把名字改成很普通的英文之后又能发现，感觉像是 mDNS 广播出来的名字/hostname 有问题。\n\n# Expected outcomes\n\n- mDNS advertising should remain on the HomeKit service type `_hap._tcp.local.`, and display names that are unsafe for mDNS or hostnames, including the kinds of names described above, should still produce discoverable advertised service information.\n- The advertised mDNS service instance name should be derived from the accessory display name in a cleaned form, while preserving the existing uniqueness suffix.\n- The advertised server hostname for the accessory should be a valid `.local.` hostname derived from the display name and the existing uniqueness suffix.\n- Package metadata should require `zeroconf>=0.32.0`.\n- The README installation guidance should state that, as of version 3.5.1, HAP-python no longer supports Python versions older than 3.6.\n\n# Implementation notes\n\nThe exact normalization helpers, data structures, and where validation is applied are up to the implementation. Preserve the existing public behavior of accessory registration except for making the advertised mDNS service information valid for these display-name edge cases and updating the documented dependency/support metadata above."} {"task_id": "format-code-task-001558", "source_id": "format-code-task-001558", "domain": "code", "task_path": "tasks/format-code-task-001558", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:eea1109f65d02db237443821c5e0d90370e760d436623e84446cdeb76fef9978", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nDeprecation in pillow causing problems with imageio gif writing\nHi there, thanks for all the great work on imageio! 🙂 \n\nI had a bug report at napari/napari-animation#174 which boils down to a deprecation in pillow causing a problem with the imageio gif writing, reporting here \n\n\n```python\nimport numpy as np\nimport imageio\n\nwriter = imageio.get_writer(\n 'test.gif',\n fps=20,\n)\nfor i in range(5):\n writer.append_data(np.random.random((32, 32)).astype(np.uint8))\n```\n\n```txt\n File \"/Users/alisterburt/micromamba/envs/napari-animation/lib/python3.9/site-packages/imageio/v2.py\", line 215, in append_data\n return self.instance.write(im, **self.write_args)\n File \"/Users/alisterburt/micromamba/envs/napari-animation/lib/python3.9/site-packages/imageio/plugins/pillow.py\", line 354, in write\n raise TypeError(\nTypeError: The keyword `fps` is no longer supported. Use `duration`(in ms) instead, e.g. `fps=50` == `duration=20` (1000 * 1/50).\n```"} {"task_id": "format-code-task-001559", "source_id": "format-code-task-001559", "domain": "code", "task_path": "tasks/format-code-task-001559", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c51dd707cb485f091e35d10313942506db1d5836bf0178c8a754ee01aac0f46a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nBytesWarning thrown in imread in 2.11.1\nThere seems to be a small regression when calling `imread` and enabling `BytesWarning`s ([-bb](https://docs.python.org/3/using/cmdline.html#cmdoption-b)):\n\n```py\nfrom pathlib import Path\n\nimport imageio\n\nimage_bytes = Path(\"image.png\").read_bytes()\nimage = imageio.imread(image_bytes)\n```\n\nRun with:\n\n```\npython -bb test.py\nTraceback (most recent call last):\n File \"test.py\", line 6, in \n image = imageio.imread(image_bytes)\n File \"venv/lib/python3.9/site-packages/imageio/core/functions.py\", line 159, in imread\n with imopen(uri, \"ri\", plugin=format) as file:\n File \"venv/lib/python3.9/site-packages/imageio/core/imopen.py\", line 221, in imopen\n plugin_instance = config.plugin_class(request, **kwargs)\n File \"venv/lib/python3.9/site-packages/imageio/config/plugins.py\", line 108, in partial_legacy_plugin\n return LegacyPlugin(request, legacy_plugin)\n File \"venv/lib/python3.9/site-packages/imageio/core/legacy_plugin_wrapper.py\", line 64, in __init__\n f\"`{self._format.name}`\" f\" can not read `{self._request.raw_uri}`.\"\nBytesWarning: str() on a bytes instance\n```\n\nEnabling `BytesWarning`s helps users catch `str`/`bytes` confusion in their own code.\n\n### Environment\n\nPython 3.9.8\n\n```\npip list\nPackage Version\n---------- -------\nclick 8.0.1\nimageio 2.11.1\nnumpy 1.21.4\npep517 0.10.0\nPillow 8.4.0\npip 21.3.1\npip-tools 6.4.0\nsetuptools 57.0.0\ntoml 0.10.2\nwheel 0.36.2\n```"} {"task_id": "format-code-task-001560", "source_id": "format-code-task-001560", "domain": "code", "task_path": "tasks/format-code-task-001560", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f3a632cba3b3edf1abb83da6d45fee3e11027487ec4201fce0ab78bbca87bd35", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the public `compile(code: string, options?: object) -> object` API to compile Imba style `@keyframes` declarations into CSS in the returned `css.code` output.\n\nFor `compile(\"global css @keyframes blink\\n\\t0% o:1\\n\\t100% o:0\", {sourcePath:\"/tmp/example.imba\", sourceId:\"example\"})`, `css.code` should contain `@keyframes blink` with `0%` and `100%` frames, and those frames should render the Imba `o` style shorthand as `opacity: 1;` and `opacity: 0;`.\n\nFor `compile(\"css .test\\n\\t@keyframes blink\\n\\t\\tfrom o:0\\n\\t\\tto o:1\\n\\t.item\\n\\t\\tanimation-name: blink\", {sourcePath:\"/tmp/example.imba\", sourceId:\"example\"})`, `css.code` should contain a scoped animation such as `@keyframes blink-test`, the `from` and `to` frames should render as CSS keyframe blocks, and the owning `.test` rule should expose `--animation-blink: blink-test;` so nested `animation-name: blink` resolves through `var(--animation-blink,blink)`.\n\nThe same `code` and `options` should produce the same `css.code` each time, and this compilation path should not write files, perform network calls, or mutate the input string."} {"task_id": "format-code-task-001561", "source_id": "format-code-task-001561", "domain": "code", "task_path": "tasks/format-code-task-001561", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:898bc5cd352cf0e960cf20fc3ef8af96fd60d41f10c2eb253c4e804a3a23f948", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Fix the request argument-parsing helpers\n\nOur web layer parses incoming request data with a small set of decorators built\non top of webargs + marshmallow. They live in the web \"args\" module and are the\nsingle entry point every request handler uses to read and validate query\nstrings, form data and JSON bodies:\n\n- `use_args` / `use_kwargs`\n- `use_rh_args` / `use_rh_kwargs`\n\nThese helpers no longer work with the versions of webargs and marshmallow the\nproject now depends on — importing the module already fails. Reimplement them so\nthey work again, keeping the public behavior described below. Each decorator\naccepts either a marshmallow `Schema` subclass **or** an \"argmap\" (a plain dict\nmapping field names to marshmallow fields), plus keyword arguments, and wraps a\nhandler function. `use_args`/`use_rh_args` inject the parsed data as a single\npositional argument; `use_kwargs`/`use_rh_kwargs` spread it into keyword\narguments.\n\nThe observable contract:\n\n- **Whitespace stripping.** Surrounding whitespace is stripped from every parsed\n string value, including strings nested inside lists, regardless of whether the\n data came from the query string, form data or a JSON body. Non-string values\n are untouched, and list/collection structure is preserved.\n\n- **Default location.** When no explicit location is given, arguments are read\n from the request body (form data or JSON), *not* from the query string. A\n handler with no location specified must therefore ignore query-string\n parameters.\n\n- **Unknown fields are ignored.** Input fields that the schema does not declare\n are silently dropped instead of producing a validation error.\n\n- **Validation errors.** A failed validation aborts with the framework's\n standard `422 Unprocessable Entity`. The error payload's `messages` must be a\n flat mapping of `field name -> list of errors`; it must **not** be wrapped in\n an extra layer keyed by the input location.\n\n- **Forwarding options.** Standard webargs parsing options passed as keyword\n arguments — the target `location`, an explicit request object, how unknown\n fields are treated, custom validation, error status/headers — are forwarded to\n the underlying parser. All of webargs' input locations (`query`, `form`,\n `json`, `view_args`, `headers`, `cookies`) work.\n\n- **Schema kwargs / partial.** Keyword arguments meant for the schema\n constructor (e.g. `partial=True` for PATCH-style endpoints) are passed through\n to the schema.\n\n- **Schema context.** A `context` mapping may be supplied and is made available\n to the schema's fields. For `use_rh_args`/`use_rh_kwargs`, the context is\n additionally populated from attributes of the current request handler\n (`flask.g.rh`): for a schema class the attribute names come from its\n `Meta.rh_context`, and for an argmap they come from a required `rh_context`\n keyword argument.\n\n- **Misuse errors.**\n - Passing a schema *instance* (rather than a class or an argmap) to any of the\n four decorators raises `TypeError` whose message starts with\n `Pass a schema or an argmap`.\n - Passing the `rh_context` keyword argument together with a schema *class*\n (instead of an argmap) to `use_rh_args`/`use_rh_kwargs` raises `TypeError`\n with the message `The \\`rh_context\\` kwarg is only supported when passing an argmap`.\n"} {"task_id": "format-code-task-001562", "source_id": "format-code-task-001562", "domain": "code", "task_path": "tasks/format-code-task-001562", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:302e98092d16ef60ddd6763ec54ab9d68e5f91cb9a170207492c1293c40a5f1d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nRight now there are four ways to construct an `IPPrefix`:\n\n1. Parsing an string with an IP represented in CIDR notation: `netaddr.ParseIPPrefix(\"127.0.0.1/8\")`\n2. Providing the significant bits to an `IP` struct: `netaddr.IPv4(127, 0, 0, 1).Prefix(8)`\n3. Working with `IPRange`: `netaddr.ParseRange(\"127.0.0.0-127.0.0.255\").Prefix()` (ignoring error returns)\n4. Working with `IPSet`: `ipset.Prefixes()`\n\nI'm working with a kernel API that returns the IP mask itself[^1]. I can turn this into an IPPrefix by first using `net.IPMask` and calling `Size`.\n\n```go\nfunc main() {\n\tvar mask [4]byte\n\tnetmask := make(net.IPMask, net.IPv4len)\n\n\tmask = getKernelMask()\n\tcopy(netmask, mask[:])\n\n\tbits, _ := netmask.Size()\n\tip, _ := netaddr.IPv4(127, 0, 0, 1).Prefix(uint8(bits))\n\n\tfmt.Printf(\"IPPrefix: %s\\n\", ip)\n}\n```\n\nWould there be any interest in allowing an `IPPrefix` to be constructed either directly from a netmask representation, or alternatively from a `net.IPMask`?\n\n[^1]: The kernel API returns the netmask for IPv4. For IPv6 they switched to returning the significant bits instead.\n\nI'd expect something like a new `Netmask` method on `IP` taking the raw mask bytes, e.g. `ip.Netmask(maskBytes)`."} {"task_id": "format-code-task-001563", "source_id": "format-code-task-001563", "domain": "code", "task_path": "tasks/format-code-task-001563", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6fcbd03b0858045f4fd6d5adf62565d2bb2d29027b29469310459feac73b48dc", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Problem Statement\n\nI’m seeing odd results from InfluxQL when I run arithmetic between two fields, like `SELECT a + b` or `a * b`: at timestamps where only one field has a value, the expression still returns a number, as if the missing field was treated as `0`. It happens with float fields, integer fields, and float/integer combinations, so I’m not sure if I’m misunderstanding how null field values are handled in binary expressions.\n\n# Expected outcomes\n\n- Float field arithmetic: for InfluxQL binary arithmetic expressions over two float fields, including addition, subtraction, multiplication, and division, any timestamp where either operand is missing/null should produce a null/missing expression result rather than a numeric result computed with an implicit zero.\n- Integer field arithmetic: for InfluxQL binary arithmetic expressions over two integer fields, including addition, subtraction, multiplication, and division, any timestamp where either operand is missing/null should produce a null/missing expression result rather than a numeric result computed with an implicit zero.\n- Mixed numeric field arithmetic: for InfluxQL binary arithmetic expressions combining float and integer fields in either operand order, including addition, subtraction, multiplication, and division, any timestamp where either operand is missing/null should produce a null/missing expression result rather than a numeric result computed with an implicit zero.\n- When both operands are present at the same timestamp, existing arithmetic semantics should continue to apply for the supported numeric field type combination and operator.\n\n# Implementation notes\n\n- The fix should be expressed in terms of the observable InfluxQL query behavior above; the specific iterator structure, helper functions, data structures, and validation location are implementation choices.\n- Preserve existing behavior outside binary arithmetic expressions over numeric field operands unless it is directly necessary to make missing/null operands propagate correctly."} {"task_id": "format-code-task-001564", "source_id": "format-code-task-001564", "domain": "code", "task_path": "tasks/format-code-task-001564", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:03892a4974f54bb10f0e1316ccff360811ec162be8558949c15aa898be8ad2ea", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nwhen influxdb server shutdown, client still send metric, the background goroutine \"w.writeProc()\" will crash at rand.Intn() after try 52 times\n### Specifications\n\n* Client Version: v2.12.0\n* InfluxDB Version: v2.5.0\n* Platform: Ubuntu 20.04.5 LTS\n\n\n### Steps to reproduce\n\n1. influxdb server is not start, that is shutdown\n2. write points using lib to send metrics to server which in fact is not started\n```\n dbClient := influxdb2.NewClient(serverUrl, token)\n\tdbClient.Options().WriteOptions().SetPrecision(time.Millisecond)\n\tdefer dbClient.Close()\n dbWriteAPI := dbClient.WriteAPI(\"myorg\", \"mybucket\")\n for {\n\t\treq, ok := <-gRecvChannel\n\t\tvar metricName string\n\t\tif !ok {\n\t\t\tbreak\n\t\t}\n //........code about set metricName and tags\n p := influxdb2.NewPoint(metricName, tags,\n\t\t\t\tmap[string]interface{}{\"value\": value},\n\t\t\t\ttime.UnixMilli(timestamp))\n\t\tdbWriteAPI.WritePoint(p)\n }\n```\n3. at first, only print log like \"dial tcp 127.0.0.1:8086: connect: connection refused, batch kept for retrying\", but finally crash\n\n\n### Expected behavior\n\nat normal case, will print the follow log\n\n2022/11/11 10:34:39 influxdb2client E! Write error: Post \"http://127.0.0.1:8086/api/v2/write?bucket=mybucket&org=myorg&precision=ms\": dial tcp 127.0.0.1:8086: connect: connection refused, batch kept for retrying\n2022/11/11 10:34:39 influxdb2client E! Error flushing batch from retry queue: %!w(*url.Error=&{Post http://127.0.0.1:8086/api/v2/write?bucket=mybucket&org=myorg&precision=ms 0xc00022ceb0})\n......\n2022/11/11 10:47:09 influxdb2client E! Error flushing batch from retry queue: %!w(*url.Error=&{Post http://127.0.0.1:8086/api/v2/write?bucket=mybucket&org=myorg&precision=ms 0xc00022c1e0})\n2022/11/11 10:47:24 influxdb2client E! Write error: Post \"http://127.0.0.1:8086/api/v2/write?bucket=mybucket&org=myorg&precision=ms\": dial tcp 127.0.0.1:8086: connect: connection refused, batch kept for retrying\n\n### Actual behavior\n\nbut after about 13 minutes, panic crash, the call stack is as follow\n```\nruntime.fatalpanic (/usr/local/go/src/runtime/panic.go:1143)\nruntime.gopanic (/usr/local/go/src/runtime/panic.go:987)\nmath/rand.(*Rand).Intn (/usr/local/go/src/math/rand/rand.go:168)\nmath/rand.Intn (/usr/local/go/src/math/rand/rand.go:337)\ngithub.com/influxdata/influxdb-client-go/v2/internal/write.(*Service).computeRetryDelay (pkg/mod/github.com/influxdata/influxdb-client-go/v2@v2.12.0/internal/write/service.go:258)\ngithub.com/influxdata/influxdb-client-go/v2/internal/write.(*Service).HandleWrite (pkg/mod/github.com/influxdata/influxdb-client-go/v2@v2.12.0/internal/write/service.go:175)\ngithub.com/influxdata/influxdb-client-go/v2/api.(*WriteAPIImpl).writeProc (pkg/mod/github.com/influxdata/influxdb-client-go/v2@v2.12.0/api/write.go:192)\ngithub.com/influxdata/influxdb-client-go/v2/api.NewWriteAPI.func2 (pkg/mod/github.com/influxdata/influxdb-client-go/v2@v2.12.0/api/write.go:92)\nruntime.goexit (/usr/local/go/src/runtime/asm_amd64.s:1594)\n```\n\n### Additional info\n\ni modify the code of function \"computeRetryDelay\" at Go_Path/pkg/mod/github.com/influxdata/influxdb-client-go/v2@v2.12.0/internal/write/service.go, print some logs\n```\nfunc (w *Service) computeRetryDelay(attempts uint) uint {\n\tminDelay := int(w.writeOptions.RetryInterval() * pow(w.writeOptions.ExponentialBase(), attempts))\n\tmaxDelay := int(w.writeOptions.RetryInterval() * pow(w.writeOptions.ExponentialBase(), attempts+1))\n\tgTryCnt++ //added by me\n\tfmt.Printf(\"tryCnt=%5d maxDelay=%d minDelay=%d\\n\", gTryCnt, maxDelay, minDelay) //added by me\n\tretryDelay := uint(rand.Intn(maxDelay-minDelay) + minDelay)\n\tif retryDelay > w.writeOptions.MaxRetryInterval() {\n\t\tretryDelay = w.writeOptions.MaxRetryInterval()\n\t}\n\treturn retryDelay\n}\n```\ntryCnt= 1 maxDelay=10000 minDelay=5000\ntryCnt= 2 maxDelay=20000 minDelay=10000\ntryCnt= 3 maxDelay=40000 minDelay=20000\ntryCnt= 4 maxDelay=80000 minDelay=40000\ntryCnt= 5 maxDelay=160000 minDelay=80000\ntryCnt"} {"task_id": "format-code-task-001565", "source_id": "format-code-task-001565", "domain": "code", "task_path": "tasks/format-code-task-001565", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1db143919acfc410f5a313ab7515c8182a67a88ecbf933ca9998cca352feaf81", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nsqlserver input exclude_query config option may be confusing\n## Feature Request\nExcluding what you do not want instead of including what you do want to monitor may get confusing, specially when you just need a few measurements.\n\n### Proposal:\nDeprecate exclude_query configuration option and include a new one to indicate which query metrics one wants to run.\n\n### Current behavior:\nIt is necessary to explicitly state which queries one does not want to be executed when the sqlserver plugin is enabled.\n\n### Desired behavior:\nNone query is executed unless explicitly stated to do so in the config file.\n\n### Use case: \nSupose that I only want to keep track of database IO every 30s and wait stats every 60s. The config file would look something like this:\n\n> [[inputs.sqlserver]]\n interval = \"30s\"\n servers = [\n\t\"Server=server;Port=port;User Id=telegraf;Password=password;app name=telegraf;log=1;\",\n ]\n> \\#\n> \\## Optional parameter, setting this to two will use a new version\n> \\## of the collection queries that break compatibility with the original\n> \\## dashboards. \n> query_version = 2\n> \\## If you are using AzureDB, setting this to True will gather resource utilization metrics\n> \\# azuredb = False\n> \\## If you would like to exclude some of the metrics queries, list them here\n> \\## Possible choices:\n> \\## - PerformanceCounters\n> \\## - WaitStatsCategorized\n> \\## - DatabaseIO\n> \\## - DatabaseProperties\n> \\## - CPUHistory\n> \\## - DatabaseSize\n> \\## - DatabaseStats\n> \\## - MemoryClerk\n> \\## - VolumeSpace\n> \\## - Schedulers\n> \\## - AzureDBResourceStats\n> \\## - AzureDBResourceGovernance\n> \\## - SqlRequests\n> \\## - ServerProperties\n> **exclude_query = [ 'PerformanceCounters', 'WaitStatsCategorized', 'DatabaseProperties', 'CPUHistory', 'DatabaseSize', 'DatabaseStats', 'MemoryClerk', 'VolumeSpace', 'VolumeSpace', 'Schedulers', 'AzureDBResourceStats', 'AzureDBResourceGovernance', 'SqlRequests', 'ServerProperties' ]**\n> [[inputs.sqlserver]]\n interval = \"60s\"\n servers = [\n\t\"Server=server;Port=port;User Id=telegraf;Password=password;app name=telegraf;log=1;\",\n ]\n> \\#\n> \\## Optional parameter, setting this to two will use a new version\n> \\## of the collection queries that break compatibility with the original\n> \\## dashboards. \n> query_version = 2\n> \\## If you are using AzureDB, setting this to True will gather resource utilization metrics\n> \\# azuredb = False\n> \\## If you would like to exclude some of the metrics queries, list them here\n> \\## Possible choices:\n> \\## - PerformanceCounters\n> \\## - WaitStatsCategorized\n> \\## - DatabaseIO\n> \\## - DatabaseProperties\n> \\## - CPUHistory\n> \\## - DatabaseSize\n> \\## - DatabaseStats\n> \\## - MemoryClerk\n> \\## - VolumeSpace\n> \\## - Schedulers\n> \\## - AzureDBResourceStats\n> \\## - AzureDBResourceGovernance\n> \\## - SqlRequests\n> \\## - ServerProperties\n> **exclude_query = [ 'PerformanceCounters', 'DatabaseIO', 'DatabaseProperties', 'CPUHistory', 'DatabaseSize', 'DatabaseStats', 'MemoryClerk', 'VolumeSpace', 'VolumeSpace', 'Schedulers', 'AzureDBResourceStats', 'AzureDBResourceGovernance', 'SqlRequests', 'ServerProperties' ]**\n\nIt is not clear what I wanted to monitor at all. Instead, it would be a lot clearer if the config file read like this: \n\n> [[inputs.sqlserver]]\n interval = \"30s\"\n servers = [\n\t\"Server=server;Port=port;User Id=telegraf;Password=password;app name=telegraf;log=1;\",\n ]\n> \\#\n> \\## Optional parameter, setting this to two will use a new version\n> \\## of the collection queries that break compatibility with the original\n> \\## dashboards. \n> query_version = 2\n> \\## If you are using AzureDB, setting this to True will gather resource utilization metrics\n> \\# azuredb = False\n> \\## Metrics queries\n> \\## Possible choices:\n> \\## - PerformanceCounters\n> \\## - WaitStatsCategorized\n> \\## - DatabaseIO\n> \\## - DatabaseProperties\n> \\## - CPUHistory\n> \\##"} {"task_id": "format-code-task-001566", "source_id": "format-code-task-001566", "domain": "code", "task_path": "tasks/format-code-task-001566", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:02cf97d3a2e98e41c2954902df6afdad5b41516b34801dd1f565a9ab60012232", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nSend metrics from oldest to newest, always\n## Feature Request\n\n### Proposal:\n\nTelegraf should always send metrics point from the older to the newer. During \"normal\" behavior and while catch-up backlog if an output become slow.\n\n### Current behavior:\n\nSince #5287, when buffer is not empty and contains more than one gather cycle, metrics are sent from newest to older.\n\n### Desired behavior:\n\nBatch() should send oldest metrics and metrics in the batch should be in increasing order of age.\n\nWe should still drop the oldest metrics if the buffer become full, so #5194 is still fixed\n\n### Use case:\n\nCurrently metric points are not in consistent order. When buffer is empty, the order is from older to newer, but if buffer start to fill, batch with be in the opposite order (newer to older).\n\nIn practice, the buffer while often contains few metrics from previous gather cycle, so it may happen even if output is not down.\n\nAt the end, that means that the output could not rely on having metrics ordered, which was (mosly) true before #5287.\nHaving metrics ordered allow to easily do some transformation on the fly (e.g. difference between two point to compute the rate) and generally make working with time series easier.\n\nIf this change seems good, I could come with a PR."} {"task_id": "format-code-task-001567", "source_id": "format-code-task-001567", "domain": "code", "task_path": "tasks/format-code-task-001567", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1f61b4ba1c4192618548c35ccbcb8fe3806cccffe0b638879d373a48edab9e67", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want lazy redun task result expressions to support Python item access through `expr[key]` and slicing through `expr[start:stop:step]`. For a task result expression like `make_pair()` where the task eventually returns `(\"Bob\", 20)`, `make_pair()[0]` and `make_pair()[1]` should each produce lazy expressions without executing `make_pair` at workflow construction time, so a workflow can pass them into another task and evaluate to results like `\"Bob is 20\"` once scheduled.\n\nIndexing should work for sequence positions and mapping keys: if a task returns `[\"a\", \"b\", \"c\"]`, using its result as `items()[1]` in a downstream task should evaluate to `\"b\"`; if a task returns `{ \"value\": 3 }`, using `outputs()[\"value\"]` in a downstream task should evaluate to `3`. Slices should expand into a list of lazy item expressions, so `items()[1:3]` behaves like `[items()[1], items()[2]]` and `items()[:2]` behaves like `[items()[0], items()[1]]`.\n\nThe lazy item expressions should have readable representations such as `make_pair()[0]` and `outputs()['value']`, and invalid indexing should surface the underlying Python error during scheduler evaluation, for example an out-of-range list index should raise `IndexError`. Repeating the same indexing operation on the same expression should be deterministic and should not mutate the original expression or perform filesystem, network, or global side effects."} {"task_id": "format-code-task-001568", "source_id": "format-code-task-001568", "domain": "code", "task_path": "tasks/format-code-task-001568", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a0e349a66031dcf5bc2e082c6cd46b84b561c77dec5468951b6dceb5cc6b7f01", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `Scheduler.run(expr, dryrun=True)` to explain why each uncached task would need to run, using the scheduler logger. A typical session is: create a `Scheduler`, run a task expression once to populate provenance/cache state, then call `scheduler.run(...)` again with `dryrun=True` after changing either the task arguments, the task version, or both.\n\nIf only the arguments changed for an existing task, the dry-run log should include a `Miss` line that names the task, says the task is called with new arguments, and shows the task hash plus the old-to-new args hash change. If only the task implementation/version changed while the arguments match an earlier call, the log should include a `Miss` line that names the task, says the task is new while the arguments are reused, and shows the old-to-new task hash change plus the args hash. If both the task and arguments are new, the log should include a `Miss` line that names the task, says the task is called with new arguments, and shows both the task hash and args hash. When no explanation can be determined, the log should still identify the task and say it cannot determine the reason for the cache miss, including the task hash and args hash. If the cache-miss explanation reports an unrecognized reason, the dry-run should fail with `NotImplementedError` so new reason categories do not silently produce misleading output."} {"task_id": "format-code-task-001571", "source_id": "format-code-task-001571", "domain": "code", "task_path": "tasks/format-code-task-001571", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1e121c3ba66978d7a7020f24d2d7fa7fe9115b2a74629c3bf15fe5235baed694", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want LibCST codemods to provide a stateful import cleanup helper through `RemoveImportsVisitor(context: CodemodContext, unused_imports: Sequence[Tuple[str, Optional[str], Optional[str]]] = ())`. A transform should be able to call `RemoveImportsVisitor.remove_unused_import(context: CodemodContext, module: str, obj: Optional[str] = None, asname: Optional[str] = None) -> None` one or more times, then run `RemoveImportsVisitor(context)` over the module and have only matching imports that are no longer referenced removed. For example, if the source is `from typing import Optional, List\\nx: List[int]\\n`, scheduling `RemoveImportsVisitor.remove_unused_import(context, \"typing\", \"Optional\")` should produce `from typing import List\\nx: List[int]\\n`. If the source is `import os, sys\\nprint(sys.version)\\n`, scheduling `RemoveImportsVisitor.remove_unused_import(context, \"os\")` should produce `import sys\\nprint(sys.version)\\n`. If the source is `import os\\nprint(os.getcwd())\\n`, scheduling `RemoveImportsVisitor.remove_unused_import(context, \"os\")` should leave the module unchanged because `os` is still used. The same object should also support direct constructor input, so `RemoveImportsVisitor(context, [(\"typing\", \"Optional\", None)])` applies the same removal behavior without first mutating the context. I also need `RemoveImportsVisitor.remove_unused_import_by_node(context: CodemodContext, node: cst.CSTNode) -> None` to inspect an import node or a subtree that a codemod is deleting and schedule the imported names referenced by that node; in a codemod that deletes `x: Optional[int] = None`, calling it on the original annotation assignment should let the later cleanup remove `from typing import Optional` when there are no other `Optional` references. If the node-based helper is given a relative `ImportFrom` whose absolute module cannot be resolved from the codemod context, it should raise `ValueError`. The cleanup should respect aliases, so scheduling `RemoveImportsVisitor.remove_unused_import(context, \"pkg.mod\", \"Thing\", \"Alias\")` removes `from pkg.mod import Thing as Alias` only when `Alias` is unused and should not remove a differently aliased import. When removing one alias from a multi-name import, the remaining import statement should stay syntactically valid with trailing commas adjusted, and comments attached around parenthesized `from ... import (...)` aliases should be preserved on the remaining import when possible. I also want `CodemodCommand` subclasses to expose convenience methods `remove_unused_import(self, module: str, obj: str | None = None, asname: str | None = None) -> None` and `remove_unused_import_by_node(self, node: cst.CSTNode) -> None` that schedule removals on the command's context. After any `CodemodCommand.transform_module(...)` finishes its main transform, the command should automatically run scheduled import cleanup, so a command body can call `self.remove_unused_import(\"typing\", \"Optional\")` and return the updated tree without manually instantiating `RemoveImportsVisitor`."} {"task_id": "format-code-task-001572", "source_id": "format-code-task-001572", "domain": "code", "task_path": "tasks/format-code-task-001572", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f4a0097404bed4ddfcbf0598d6ccb66dca61cfd4ed3278939950b150df05d948", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Render a list of GitHub deployments as a Slack message\n\nWe want to show a repository's GitHub deployments inside Slack. The data comes from\nGitHub's GraphQL API as an array of deployment nodes, each shaped like:\n\n```js\n{\n creator: { login, url, avatarUrl },\n ref: { name },\n commit: { message, abbreviatedOid, commitUrl },\n description: ,\n task: , // e.g. \"deploy\"\n environment: , // e.g. \"production\"\n state: , // a GitHub deployment state, see below\n createdAt: ,\n latestStatus: { state, description } | null\n}\n```\n\nAdd a new message renderer, available as the default export of\n`lib/messages/deployment-list.js`, that follows the same convention as the other\nrenderers in that directory: it is constructed with the array of deployment nodes and\nexposes a `toJSON()` method that returns the Slack message payload.\n\n`toJSON()` must return an object of the form `{ attachments: [...] }` containing exactly\none attachment per deployment, in the same order as the input array. An empty input array\nyields `{ attachments: [] }`.\n\nEach attachment must contain:\n\n- `fallback`: the plain-text summary `\" triggered a on from \"`,\n where `` is the creator's login and `` is `ref.name`.\n- `color`: a hex color derived from the deployment `state` (see mapping below).\n- `pretext`: the deployment `description` (omit it when the description is null/empty).\n- `author_name`: the creator's login.\n- `author_link`: the creator's `url`.\n- `author_icon`: the creator's `avatarUrl`.\n- `title`: `\" \"`.\n- `title_link`: `commit.commitUrl`.\n- `fields`: a list of short fields (each `{ title, value, short: true }`) in this order:\n `Task` (the task), `Environment` (the environment) and `State` (the state). When the\n deployment has a `latestStatus`, append one more short field titled `Latest Status`\n whose value is `\" \"`. When `latestStatus` is null, no\n such field is present.\n- `footer`: the literal string `\"Created\"`.\n- `footer_icon`: `\"https://assets-cdn.github.com/favicon.ico\"`.\n- `ts`: the `createdAt` timestamp expressed as Unix seconds (i.e. milliseconds since the\n epoch divided by 1000).\n\nColor mapping from deployment state (use the project's existing palette constants where\nthey match these hex values):\n\n| state | color |\n|-------------|-----------|\n| `ABANDONED` | `#24292f` |\n| `ACTIVE` | `#36a64f` |\n| `DESTROYED` | `#cb2431` |\n| `ERROR` | `#cb2431` |\n| `FAILURE` | `#cb2431` |\n| `INACTIVE` | `#24292f` |\n| `PENDING` | `#dbab09` |\n"} {"task_id": "format-code-task-001573", "source_id": "format-code-task-001573", "domain": "code", "task_path": "tasks/format-code-task-001573", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:8d44eff7ec010f4eed1f4e419ff9bf1bcec6fddd3e845f2c513dfafa7815f172", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Confusing error when calling `fit()` without `search()` first in Nano AutoML\n\nI'm trying out BigDL Nano's AutoML/HPO. I defined a model and put a couple of hpo search spaces in it (e.g. a learning-rate range), then just called `model.fit(...)` to train it, the same way I'd train any normal Keras model.\n\nInstead of training, I get a `ValueError` that says something along the lines of:\n\n> study is None. Please call search before calling end_search.\n\nThis is pretty confusing as a user:\n\n- I never called `end_search` myself — it's not in my code at all. The error is pointing at some internal function I don't know about, instead of telling me what I actually did wrong (forgot to run `model.search(...)` before `model.fit(...)`).\n- The message doesn't mention `fit` at all, so it took me a while to realize my mistake was calling `fit` directly when I'd put hpo spaces in the model.\n\nIt'd be much nicer if the error told me, in terms of the public API I actually use, that I need to run `search` before `fit` whenever the model has search spaces in it.\n\nAlso, a related thing I ran into: if I take the search spaces back out of the model (i.e. it's just a plain model with no hpo at all) and call `fit` directly, I'd expect that to just work — there's nothing to search over. Right now the AutoML path still complains about not having done a search even though there's nothing to search.\n\nSo basically two things:\n\n1. When the user *did* define search spaces but forgot to call `search` before `fit`, raise an error that's actually phrased in terms of `search`/`fit` so it's obvious what to do.\n2. When the user defined no search spaces, calling `fit` directly should just train the model without forcing a `search` step."} {"task_id": "format-code-task-001574", "source_id": "format-code-task-001574", "domain": "code", "task_path": "tasks/format-code-task-001574", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:51c4c0c2d9bec418cb82dd50e33c793fae748c8ce11b13c39a1bcf5b8e5e7dda", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Add a Phishing Army blocklist analyzer\n\nWe want IntelOwl to support reputation lookups against the public\n[Phishing Army](https://phishing.army/) blocklist — a plain-text feed where each\nline is a single blocked host (domain or IP), newline-separated.\n\nPlease add a new observable analyzer, exposed under the plugin name\n`PhishingArmy`, that tells the user whether the observable they submitted is\npresent on that blocklist.\n\n## Expected behavior\n\n- The analyzer obtains the blocklist from the Phishing Army feed (a newline-\n separated list of hosts). It should fetch the feed when it doesn't already\n have a local copy of it; a cached copy may be reused.\n- Running the analyzer returns a dictionary that contains a boolean entry under\n the key `found`. `found` is `True` when the observable is present on the\n blocklist and `False` otherwise.\n- For `domain` and `ip` observables, the observable value is matched directly\n against the entries of the list (an entry counts as a match only when it\n equals the value, not when it merely appears as a substring of another entry).\n- For `url` observables, only the URL's **hostname** is matched against the\n list — the scheme, path, query string, etc. must be ignored. A URL whose\n hostname is on the list is reported as found; a URL whose hostname is not on\n the list is reported as not found even if some other part of the URL happens\n to contain a blocked host.\n\nThe analyzer should fit naturally into the existing observable-analyzer\nframework so that it is discoverable like the other analyzers and can be run\nagainst a submitted observable.\n"} {"task_id": "format-code-task-001575", "source_id": "format-code-task-001575", "domain": "code", "task_path": "tasks/format-code-task-001575", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:e7b7729906691a57a6051217c90136eef4ba76182105ddf15749115c8394e6ea", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Meta refresh tag URLs are not being archived correctly\n\nI'm using Zeno to archive some pages that rely on `` refresh tags to redirect users to the actual content, e.g.:\n\n```html\n\n```\n\nI expected Zeno to pick up `https://example.com/real-page` as an asset to archive (the same way it would follow other links/redirects on the page), but that doesn't seem to be happening — the redirect target ends up missing from my archive.\n\nLooking at the archived output, it seems like the current handling of the meta tag's `content` attribute doesn't really understand this format. It looks like it just treats the whole `content` string as a candidate URL if it contains `http`, which means a value like `0; url=https://example.com/real-page` doesn't get processed as a proper URL (it's not a valid URL on its own).\n\nIt would be great if the meta tag handling could recognize the standard `content=\"; url=\"` format used by refresh tags and pull the real URL out of it, so those targets actually make it into the archive.\n\nThe existing behavior for meta tags that just have a plain URL in `content` (e.g. things like ``) should keep working as before."} {"task_id": "format-code-task-001576", "source_id": "format-code-task-001576", "domain": "code", "task_path": "tasks/format-code-task-001576", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:48300394a0b227740d0a1c6d1c0167d7dae9c23ca4b1dc793cae488c182090ec", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n我在把一部分 Firebase web modular database 代码迁到 `@react-native-firebase/database`，现在 `off` 和 `enableLogging` 这两个 import 直接就找不到，TS 也过不了。能不能先让 modular 入口也暴露这两个 API，哪怕 native 这边暂时只是提示不支持，也别让我卡在导入和类型检查这里。\n\nExpected outcomes:\n- `enableLogging` is available as a named import from the modular `@react-native-firebase/database` entrypoint, and TypeScript accepts its Firebase Web modular-compatible call forms.\n- Calling `enableLogging` at runtime does not silently succeed; it reports native non-support by throwing an error whose message is `enableLogging() is not implemented`.\n- `off` is available as a named import from the modular `@react-native-firebase/database` entrypoint, and TypeScript accepts its Firebase Web modular-compatible call forms.\n- Calling `off` at runtime does not silently succeed; it reports native non-support by throwing an error whose message is `off() is not implemented`.\n\nImplementation notes:\n- Keep the behavior aligned with the package’s existing modular API conventions, but the specific file organization and internal implementation details are up to the implementer.\n- These APIs are compatibility stubs for the native implementation, not full native listener/logging implementations."} {"task_id": "format-code-task-001578", "source_id": "format-code-task-001578", "domain": "code", "task_path": "tasks/format-code-task-001578", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ab59a9c19bf5a9fd0262327f4af5a99742972521e039f830edd514c639912263", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## cloud-init fails on AlmaLinux/RHEL: `ValueError: Address default is not a valid ip address`\n\nI'm using cluster-api-provider-proxmox to bring up nodes from an AlmaLinux 9 cloud image template. On Ubuntu 22.04 templates everything is fine, but on RPM-based distros the VM comes up without network — `cloud-init` blows up in the `init-local` stage and never applies the network config.\n\nHere's the relevant tail from `/var/log/cloud-init.log` on the failing AlmaLinux VM:\n\n```\n2023-12-14 13:24:13,681 - stages.py[INFO]: Applying network configuration from ds bringup=False: {'version': 2, 'renderer': 'networkd', 'ethernets': {'eth0': {'match': {'macaddress': '9E:88:BA:6F:CC:1A'}, 'dhcp4': 'no', 'addresses': ['192.168.69.111/24'], 'routes': [{'to': 'default', 'via': '192.168.69.1'}], 'nameservers': {'addresses': ['8.8.8.8', '8.8.4.4']}}}}\n\n2023-12-14 13:24:13,684 - util.py[DEBUG]: failed stage init-local\nTraceback (most recent call last):\n ...\n File \"/usr/lib/python3.9/site-packages/cloudinit/net/network_state.py\", line 1009, in _normalize_net_keys\n raise ValueError(f\"Address {addr} is not a valid ip address\")\nValueError: Address default is not a valid ip address\n```\n\nSo the network-config that capmox generates and feeds into cloud-init contains a route entry that the cloud-init shipped on AlmaLinux/RHEL 9 refuses to accept — it expects the route destination to be a valid IP/CIDR.\n\nThe Ubuntu 22.04 cloud-init happens to accept it, which is why this only shows up once you try to use a non-Debian-family image.\n\nCould the rendered network-config be adjusted so that the default-route entries are expressed in a form that cloud-init accepts on both Ubuntu and RPM-based distros? Needs to cover both IPv4 and IPv6 gateways."} {"task_id": "format-code-task-001579", "source_id": "format-code-task-001579", "domain": "code", "task_path": "tasks/format-code-task-001579", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b05edab99b4bc73513346daac21b3d9514a242459baa26c6d26247a25220a421", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nI'm trying to set up unit tests for the value transfer tangle package and it's a pain that everything wants a real BadgerDB instance to construct things like `tangle.New` and `branchmanager.New`. My tests keep creating temp directories and leaving badger files on disk just to exercise some logic. Could we make these take a more generic store interface instead of requiring a concrete `*badger.DB`, so I could plug in an in-memory backend for tests and dev? Ideally I'd just hand it some in-memory kvstore and not touch the filesystem at all.\n\n## Expected outcomes\n\n- Value transfer tangle construction should no longer require callers to provide a concrete Badger database; callers should be able to construct it with a generic key-value store suitable for in-memory tests.\n- Value transfer branch manager construction should no longer require callers to provide a concrete Badger database; callers should be able to construct it with the same kind of generic key-value store.\n- Existing storage-backed test and development paths related to these components should be able to run against in-memory key-value storage without setting up temporary Badger directories or files, while preserving their existing observable behavior.\n\n## Implementation notes\n\n- Keep the storage backend choice abstract at package boundaries; persistent and in-memory backends should be interchangeable from the perspective of callers that only need key-value storage behavior.\n- Do not remove or weaken existing value-transfer, branch-manager, tangle, or message-construction behavior; the change is about decoupling those paths from a concrete filesystem-backed database requirement."} {"task_id": "format-code-task-001580", "source_id": "format-code-task-001580", "domain": "code", "task_path": "tasks/format-code-task-001580", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d8333b50861463e74d6080aa1edd127af1a7e6bf6fce9b333e4cceff905033d3", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Message has no way to indicate which network it belongs to\n\nI'm using `iota.MessageBuilder` to construct messages and ship them to nodes. The current `Message` only carries: version, the two parents, payload and nonce.\n\nThe problem: there's nothing in the message itself that says *which IOTA network* it was produced for. If I have a node running on a testnet and someone replays a binary message that was actually meant for mainnet (or any other separately‑deployed network), the node has no way of telling — the bytes deserialize fine and look like a valid message. Same on the JSON side: a serialized message round‑tripped through `MarshalJSON` / `UnmarshalJSON` carries no network information.\n\nI'd expect a `Message` to carry an identifier of the network it's intended for, set by the producer at build time and preserved across:\n\n- binary `Serialize` / `Deserialize`\n- `MarshalJSON` / `UnmarshalJSON`\n- `MessageBuilder` (so I can set it fluently before `Build()`)\n\nThat way a node receiving a message can reject it up front if it's meant for a different network, instead of accepting arbitrary bytes from foreign networks.\n\nI'd expect the new field on `Message` to be something like `NetworkID` (a `uint64`), with a JSON representation under the key `\"networkId\"` (encoded as a string, consistent with how `nonce` is handled)."} {"task_id": "format-code-task-001581", "source_id": "format-code-task-001581", "domain": "code", "task_path": "tasks/format-code-task-001581", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5370662d6fe228b1578c1b8ae8f1e761479d726f45f4750f5072b4fb56cc5430", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Deleting a state inside a transaction and then reading it back returns the old value\n\nI'm writing some action handler code on top of `stateTX` that, within the same uncommitted transaction, deletes an account state and then reads it back to make sure it's gone. Roughly:\n\n```go\n// inside one stateTX, before Commit()\nif err := ws.DelState(addr); err != nil { ... }\n\nvar acc state.Account\nerr := ws.State(addr, &acc)\n// I expect err to indicate the state no longer exists (state.ErrStateNotExist)\n```\n\nWhat I actually get is `err == nil` and `acc` is filled in with whatever the account looked like before I deleted it. So as far as the caller can tell, the `DelState` had no effect — until the transaction is committed and I re-open the world state, at which point the key really is gone.\n\nThis is pretty surprising, and it's a problem for any protocol logic that wants to do a delete and then verify / branch on the post-delete view of state mid-transaction (e.g. \"delete this entry, then assert it isn't there before re-creating it\"). Reads inside a `stateTX` should reflect the writes (including deletes) that have already happened on that same `stateTX`, not silently fall through to the underlying DB's pre-transaction value.\n\nSame thing happens at the lower level with `KVStoreForTrie`: after `Delete(key)` on a `KVStoreForTrie`, calling `Get(key)` on the same instance (before `Flush`) still returns the value that's sitting in the underlying `KVStore`, instead of telling me the key isn't there.\n\nCould the read path be fixed so that a key deleted earlier in the same uncommitted transaction is reported as not-found, the same way it would be after commit?"} {"task_id": "format-code-task-001582", "source_id": "format-code-task-001582", "domain": "code", "task_path": "tasks/format-code-task-001582", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:2e14d9ac26f18cef38042d9cbd7732e17b7749384a532de91e97755243a0d4e5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI need the cash wallet controller to expose `BaseController.MoveCoins(store weave.KVStore, src weave.Address, dest weave.Address, amount coin.Coin) error` so application handlers and other modules can transfer funds between wallet addresses through an existing controller created with `NewController`.\n\nGiven a store where the source wallet holds `50000 MONY`, calling `MoveCoins(store, source, destination, coin.NewCoin(300, 0, \"MONY\"))` should return nil, subtract `300 MONY` from the source wallet, create the destination wallet if it does not exist, and add exactly `300 MONY` there. Given a store where the source wallet holds `50000 MONY`, calling `MoveCoins(store, source, source, coin.NewCoin(300, 0, \"MONY\"))` should return nil and leave the source wallet with the same total balance.\n\nThe method should reject invalid transfers without partially crediting the destination: a zero amount returns an amount error, a negative amount returns an amount error, a missing source wallet returns an empty-account error, an amount larger than the source balance returns an amount error, and a coin ticker that the source wallet does not hold returns an amount error. If loading or saving either wallet fails, the method should return that error to the caller."} {"task_id": "format-code-task-001583", "source_id": "format-code-task-001583", "domain": "code", "task_path": "tasks/format-code-task-001583", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0213ebb6e66af97cf93365b10847770e4d3028bfe0fe9da75e398114a7282043", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the CLI to support looking up one ASN by passing the ASN as the first argument, for example `ipinfo AS123` or `ipinfo as123`. The command should recognize top-level ASN strings written as `AS` or `as` followed by digits, normalize the ASN sent to IPinfo to uppercase, and perform an authenticated ASN details lookup using `--token/-t` or the saved token.\n\nWith a valid token and an ASN details response for `AS123` whose ASN is `AS123`, name is `Example Network`, and registry is `arin`, running `ipinfo AS123 --field asn,name,registry` should write `asn,name,registry` followed by `AS123,Example Network,arin` on stdout and exit 0. Without `--field`, `ipinfo AS123` should write indented JSON for the ASN details by default; `--yaml/-y` should write YAML instead. `--nocolor` should disable colored output, and `--nocache` should disable cached ASN API responses for this lookup.\n\nRunning `ipinfo AS123 --help` or `ipinfo AS123 -h` should print usage in the form `Usage: ipinfo AS123 []`, document `--token`, `--nocache`, `--field`, `--nocolor`, `--json`, and `--yaml`, exit 0, and not make an API request. If no token is available, the command should fail before making the request, write `err: ASN lookups require a token; login via `ipinfo init`.` to stderr, produce no data output, and follow the CLI's normal error exit behavior. If the ASN details API returns HTTP 401, the command should write `err: Token does not have access to ASN API` to stderr. The top-level help should list `` as a direct command for ASN lookups, and shell completion after an ASN literal should offer the single-ASN lookup flags."} {"task_id": "format-code-task-001584", "source_id": "format-code-task-001584", "domain": "code", "task_path": "tasks/format-code-task-001584", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ec53180e7fede38a071b7506e6cd03025156de91f3d650cbbe905df1d74ec13c", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI'm hitting a netlab validation gap around management addressing: with clab/libvirt topologies, nodes with missing management IPs, management IPs outside `addressing.mgmt`, or an IPv6/IPv4 mgmt address when the matching mgmt prefix isn't defined can get through `netlab create` and then fail later in containerlab/libvirt or end up unreachable. I also noticed `addressing.mgmt.ipv6_pfx` rejects a `/64` with `IPv6 pool prefix cannot be longer than /56`, even though this is the mgmt pool.\n\nExpected outcomes:\n- Management IPv6 pool validation should allow management IPv6 prefixes that are appropriate for node management networks, including `/64`, without weakening the existing IPv6 prefix-length validation for non-management address pools.\n- For clab and libvirt topologies, invalid management-addressing configurations should be rejected during topology validation/conversion, before provider output or runtime execution. The validation should follow the provider’s management-connectivity requirements and cover missing usable management addressing, missing matching management-pool prefixes for configured management address families, and management addresses that do not belong to the configured management subnet.\n- Report these failures as clear user-facing provider-specific errors that let the user identify the node, the relevant address family when applicable, and why the management addressing is invalid.\n\nImplementation notes:\n- The exact validation location, helper structure, data-flow organization, and address-containment mechanism are up to the implementer.\n- Preserve existing behavior for providers and address pools outside the management-addressing cases described above.\n- Error reporting should be actionable, but exact wording, internal helper names, log categories, and call sites are not prescribed."} {"task_id": "format-code-task-001587", "source_id": "format-code-task-001587", "domain": "code", "task_path": "tasks/format-code-task-001587", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:b2be1f452f630da02b12ed08ae451dc1b9332967a6ec326fc67684eb28fcb641", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nLegacy !m query compatibilty for multi-key object classes\nThe `!m` query syntax when querying multi-key RPSL object classes is\nunexpectedly changed between v4 and legacy versions.\n\nLegacy versions expect the key values to be separated with either ` ` or `-`.\nv4 expects the key values to be concatenated with no separator.\n\nFor example, when querying `route` objects:\n\n```\n$ echo -e '!!\\n!v\\n!mroute,41.78.188.0/22AS37271\\n!q' | nc rr.ntt.net 43\nA22\nIRRd -- version 4.1.7\nC\nA346\nroute: 41.78.188.0/22\ndescr: Workonline Communications (Pty) Ltd\norigin: AS37271\nnotify: noc@workonline.co.za\nmnt-by: MAINT-AS37271\nchanged: benmaddison@workonline.co.za 20101201 #15:59:08Z\nsource: RADB\nrpki-ov-state: not_found # No ROAs found, or RPKI validation not enabled for source\nC\n$ echo -e '!!\\n!v\\n!mroute,41.78.188.0/22-AS37271\\n!q' | nc whois.radb.net 43\nA37\n# IRRd -- version 3.0.8 [25Apr2014]\nC\nA233\nroute: 41.78.188.0/22\ndescr: Workonline Communications (Pty) Ltd\norigin: AS37271\nnotify: noc@workonline.co.za\nmnt-by: MAINT-AS37271\nchanged: benmaddison@workonline.co.za 20101201 #15:59:08Z\nsource: RADB\nC\n```\n\nAs a result, it is not possible to construct a query that is acceptable across versions."} {"task_id": "format-code-task-001589", "source_id": "format-code-task-001589", "domain": "code", "task_path": "tasks/format-code-task-001589", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f5c8f4ec11f91faf6c0be0394e7ab1a9ee3eb787fc60407fa9c75c09d5f2c4a6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want ArkUnpacker's AssetBundle resolve mode to support exporting Unity typetrees as JSON. From the command line, `python Main.py -m ab -i -o --typetree` should accept the `--typetree` flag alongside the existing `--image`, `--text`, `--audio`, `--mesh`, `-d`, and `-g` options.\n\nWhen `--typetree` is set, each processed AssetBundle should be loaded, every object that can provide a serialized typetree should be read, and the non-empty typetrees should be written as a JSON file named `TT_.json` in that bundle's output directory. The JSON should be shaped as a top-level object keyed by the bundle name, with nested keys for each object's path id and values containing that object's typetree data; bytes inside typetrees should be serialized using UTF-8 with surrogate escaping, and the configured JSON indentation and export encoding should be respected.\n\nIf I run `python Main.py -m ab -i test/res -o test/upk --typetree -g`, the command should exit 0 and produce files such as `test/upk/client-2.2/chararts-char_002_amiya/TT_char_002_amiya.ab.json` without requiring any image, text, audio, or mesh export flags. When a single object's typetree cannot be read, the command should continue processing the rest of the bundle and log that failure at debug level rather than failing the whole export.\n\nIn the interactive custom resource unpack flow, the resource type prompt should advertise `j=Typetree JSON`; entering `j` should enable the same typetree JSON export, and the confirmation summary should show that Typetree JSON is selected."} {"task_id": "format-code-task-001590", "source_id": "format-code-task-001590", "domain": "code", "task_path": "tasks/format-code-task-001590", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:61adf2685fadb2066045bd90c75f84cadd4ee15f8aa6a2919c78c8caf69ce932", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nHi,\n\nHow to allow Isso to parse pwds correctly?\n`configparser.InterpolationSyntaxError: '%' must be followed by '%' or '(', found: '%!...@c...!Cjq'`"} {"task_id": "format-code-task-001591", "source_id": "format-code-task-001591", "domain": "code", "task_path": "tasks/format-code-task-001591", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:575edbf285ed231bca08d52c3071836afa89b349f200def0c3063885788b8c4b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Envoy fails to start when Envoy Access Log Service is enabled\n\nI'm trying to ship sidecar / gateway access logs to a gRPC Access Log Service (ALS) backend. I followed the same pattern that's already used for `envoyMetricsService` and set the corresponding values for ALS in the istio helm chart, e.g.\n\n```yaml\nglobal:\n proxy:\n envoyAccessLogService:\n enabled: true\n host: accesslog-service.istio-system\n port: 15000\n```\n\nAfter deploying the chart (and likewise when injecting sidecars), the proxy never comes up cleanly:\n\n- For ingress/egress gateways, the `istio-proxy` container in the gateway deployment crash-loops at startup.\n- For injected sidecars, envoy similarly fails to start; the pod's proxy never becomes ready, so the application pod stays in a not-ready state.\n- No access logs ever reach my ALS backend, even after the proxy briefly comes up.\n\nDisabling `envoyAccessLogService` makes everything work again, so the problem is clearly tied to enabling ALS. By contrast, enabling `envoyMetricsService` with the same shape of configuration works fine — that's why I assumed ALS would just work analogously.\n\nI'd expect that enabling `envoyAccessLogService` in the helm values (and/or enabling ALS via mesh config) results in a working envoy that actually sends access logs to the configured backend, the same way `envoyMetricsService` does today."} {"task_id": "format-code-task-001592", "source_id": "format-code-task-001592", "domain": "code", "task_path": "tasks/format-code-task-001592", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fad3814f277fb11aa59cc25ebdd3553b80dcafc5dc3e4f3e47d931205e5d8404", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Problem Statement\n\nWhen I run `istioctl analyze` against my cluster, it doesn't say anything about Service ports that are missing names or named in weird ways — but then Istio silently treats them as TCP and things break in ways that are really hard to debug. Could the analyzer flag Service ports whose names don't follow the `[-]` convention (or have no name at all)? Even just an info-level warning per offending port would save me a ton of time.\n\n## Expected outcomes\n\n- Running `istioctl analyze` or the equivalent Galley configuration analysis over Kubernetes `Service` resources should emit an Info-level diagnostic for every Service port whose name is absent or does not follow Istio's `[-]` port naming convention.\n- Each offending Service port should produce its own diagnostic, so a Service with multiple offending ports should report multiple findings rather than collapsing them into one.\n- The diagnostic should use the public code `IST0118` and should make it clear which offending port was involved, including the port name when present, the numeric service port, and the target port.\n- Service ports whose names follow the convention, such as names with a supported protocol prefix and an optional suffix, should not produce this diagnostic.\n\n## Implementation notes\n\n- The exact analyzer structure, helper functions, and validation location are up to the implementer, as long as the behavior is visible through the normal configuration analysis path.\n- Reuse existing Istio conventions for determining whether a Service port name represents a supported protocol; do not introduce behavior that conflicts with existing supported port naming semantics.\n- The change should integrate cleanly with the existing diagnostic/message infrastructure so that downstream tools can consume the warning like other analysis results."} {"task_id": "format-code-task-001593", "source_id": "format-code-task-001593", "domain": "code", "task_path": "tasks/format-code-task-001593", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4448fc4c236c582176377e14a19b419da24078b37bc0d64e3bc5092fabb8c840", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## EnvoyFilter using deprecated filter names overwrites other EnvoyFilters\n\nWe have two `EnvoyFilter` resources targeting the sidecar listeners. One of them uses a legacy/deprecated filter name (e.g. `envoy.http_connection_manager`) in its `match` block to find the HCM and `MERGE` some config into it. The other uses the canonical name (`envoy.filters.network.http_connection_manager`) to patch a different field.\n\nWhat we observe: after both EnvoyFilters are applied, the patch from the second one disappears — it looks like the two are clobbering each other and only the last one wins. If we rewrite the first EnvoyFilter to use the canonical name instead of the deprecated one, both patches apply correctly and everything works as expected.\n\nSo it seems that mixing deprecated and canonical filter names across EnvoyFilters causes them to step on each other, even though Istio is supposed to accept the deprecated names. From a user perspective, two EnvoyFilters that target the same logical filter (one by old name, one by new name) should compose the same way as if both had used the canonical name — neither should be erased.\n\nRepro is essentially:\n\n1. Apply EnvoyFilter A with `match.listener.filterChain.filter.name: envoy.http_connection_manager` (the deprecated name), `MERGE` some config.\n2. Apply EnvoyFilter B with `match.listener.filterChain.filter.name: envoy.filters.network.http_connection_manager` (the canonical name), `MERGE` some different config.\n3. Inspect the resulting listener config on the sidecar — only one of the two merges is present.\n\nExpected: both merges are applied, regardless of which name spelling each EnvoyFilter uses to refer to the same filter."} {"task_id": "format-code-task-001594", "source_id": "format-code-task-001594", "domain": "code", "task_path": "tasks/format-code-task-001594", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0d2399d5ace050bb8f9ede7c6d3cb719867906ee513d575c75213ba9d22b0113", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: support Apache SkyWalking as a tracing provider\n\nOur observability backend is built around [Apache SkyWalking](https://skywalking.apache.org/) — we already use it to collect traces from our application-side agents (Java, Go, etc.), and we'd like Istio's sidecars to send their distributed traces to the same SkyWalking backend so we have a single place to view end-to-end traces.\n\nLooking at `MeshConfig.extensionProviders`, the available tracing providers right now are Zipkin, Lightstep, Datadog and OpenCensus (and Stackdriver for GCP users). There's no way to point Istio at a SkyWalking backend.\n\nWhat we'd like:\n\n1. Be able to register a SkyWalking backend as a tracing provider under `extensionProviders`, in the same shape as the other tracing providers — i.e. give it a name and tell Istio where the backend lives (a service address + port). Once registered, picking that provider via Telemetry API / mesh defaults should make the sidecar Envoys actually emit traces to that SkyWalking backend.\n\n2. The usual config validation should apply: if someone leaves the service empty, or uses an invalid port, mesh config validation should reject it just like it does today for Zipkin / Datadog / Lightstep / OpenCensus. Behavior here should feel consistent with the existing tracing providers — no surprises.\n\n3. While we're using SkyWalking, it would be very convenient if `istioctl dashboard` had a `skywalking` subcommand that port-forwards to the SkyWalking UI running in the cluster and opens it in the browser, similar to how `istioctl dashboard jaeger` and `istioctl dashboard zipkin` work today. Right now there's no built-in shortcut for SkyWalking and we end up doing the port-forward by hand every time.\n\nHappy to test against a SkyWalking deployment if someone picks this up."} {"task_id": "format-code-task-001595", "source_id": "format-code-task-001595", "domain": "code", "task_path": "tasks/format-code-task-001595", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0eec76c0bcd4ebce08f5b2aea4cdea4144bbe2f264b40d721def979d1aa93bbe", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Bug: revision tag created against an older control plane never gets a CA bundle, so injection fails\n\nI'm trying to use `istioctl x revision tag` to put a stable alias in front of an older Istio control plane (we're running 1.9.x in this cluster and migrating gradually). Something like:\n\n```\nistioctl x revision tag set prod --revision 1-9-5\n```\n\n`istioctl` first warns me that the revision is older than 1.10 and asks \"Continue anyways? (y/N)\". I answer `y` because I genuinely want a tag pointing at this older CP. The command finishes and a `MutatingWebhookConfiguration` for the tag does get created.\n\nThe problem shows up as soon as I try to actually use the tag:\n\n```\nkubectl label ns test-ns istio.io/rev=prod\nkubectl -n test-ns run nginx --image=nginx\n```\n\nThe pod comes up with no sidecar. When I look at the tag's `MutatingWebhookConfiguration`, the `clientConfig.caBundle` field is empty. The webhook for the underlying `1-9-5` revision has a populated `caBundle` just fine — only the tag webhook is missing it, so the API server can't establish TLS to the injector and injection silently doesn't happen.\n\nIf I do the same thing against a 1.10+ revision it works, presumably because the newer istiod fills the bundle in after the fact. But the whole point of revision tags for us is to be able to point them at whichever revision we currently have running, including older ones during an upgrade. Having to upgrade the control plane *first* before tags become usable defeats the purpose.\n\n### What I'd expect\n\n`istioctl x revision tag set` should produce a tag webhook that actually works regardless of the target revision's istiod version, including revisions older than 1.10. I shouldn't have to manually go patch the `caBundle` on the generated webhook to make namespace injection start working.\n\n### Version\nistioctl: 1.10-dev (master)\ncontrol plane: 1.9.5"} {"task_id": "format-code-task-001596", "source_id": "format-code-task-001596", "domain": "code", "task_path": "tasks/format-code-task-001596", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f3c52650b1f3e6d7c97f01b8f080ca731058081ba3b7367be9e365ce45129419", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Bug: ServiceEntries labeled for a different revision are still processed by this istiod\n\nI'm running a canary upgrade with two control planes: `istio.io/rev=stable` and `istio.io/rev=canary`. I expect each istiod to only act on configuration that matches its own revision (which is what `Get`/`List` on the config store already do).\n\nBut for event-driven config (ServiceEntry, VirtualService, etc.), this doesn't seem to hold:\n\n1. If I create a `ServiceEntry` with `istio.io/rev: canary`, the **stable** istiod still picks it up — workloads attached to the stable revision end up with endpoints/clusters from that ServiceEntry, even though the resource is clearly tagged for the other revision.\n\n2. If I take a `ServiceEntry` that previously had `istio.io/rev: stable` (or no rev label, defaulting to stable) and edit its label to `istio.io/rev: canary`, the stable istiod still keeps the old config around. From the stable revision's point of view, that resource is no longer ours, so workloads on the stable revision shouldn't see it anymore — but they do.\n\nListing the configs directly through the store does respect the revision, so this seems specific to the live event path (add / update / delete callbacks coming off the informers, plus whatever runs during the initial sync at startup). Both the steady-state event flow and the bootstrap sync need to honor the revision the same way `List` already does.\n\nRepro is roughly:\n\n```\n# two istiods, revisions \"stable\" and \"canary\"\nkubectl apply -f - < | grep example.com\n# -> shows the cluster, but it shouldn't\n```\n\nAnd for the \"moved between revisions\" case: start with the label set to stable, let things settle, then patch the label to canary. The stable istiod's view of that workload still contains the entry.\n\nExpected: a control plane should completely ignore resources whose `istio.io/rev` doesn't match it, including resources that *used to* match and have since been re-tagged for a different revision."} {"task_id": "format-code-task-001597", "source_id": "format-code-task-001597", "domain": "code", "task_path": "tasks/format-code-task-001597", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f6ae92c0abb8d0d249ec80926b56ff663b6d59c157c805cb00e406b869ed4aa6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### JWT authentication policy with an HTTPS issuer doesn't work\n\nI'm trying to set up end-user (origin) JWT authentication on 0.8 using an `authentication.Policy` that points at a real public OIDC provider (in my case Google, but I'd expect the same for Auth0/Okta/etc.). Something like:\n\n```yaml\napiVersion: authentication.istio.io/v1alpha1\nkind: Policy\nmetadata:\n name: jwt-example\nspec:\n targets:\n - name: my-svc\n origins:\n - jwt:\n issuer: \"https://accounts.google.com\"\n principalBinding: USE_ORIGIN\n```\n\nI'm intentionally not setting `jwks_uri` — I want pilot to discover it via the standard `/.well-known/openid-configuration` endpoint.\n\nAfter applying the policy, JWT auth doesn't take effect on the sidecar. In pilot's logs I see it failing to resolve the `jwks_uri` for the issuer; the configured policy never actually gets a usable public key wired in, so requests aren't validated against the JWT as I'd expect.\n\nIf I host the same OIDC metadata + JWKS on a plain HTTP server inside the cluster and point `issuer` / `jwks_uri` at that, things work. The problem only shows up when pilot has to talk to a real HTTPS OIDC provider on the public internet.\n\nThis pretty much blocks using JWT origin authentication with any standard external identity provider, since they're all HTTPS-only. Pilot needs to be able to reach an HTTPS issuer to pull the OpenID discovery doc and the JWKS."} {"task_id": "format-code-task-001598", "source_id": "format-code-task-001598", "domain": "code", "task_path": "tasks/format-code-task-001598", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c93477c3d4a05239197e5054619a042dde79039a5ee739421c5536960e8b28c1", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Stackdriver Context Graph adapter doesn't work with regional GKE clusters\n\nI'm running Istio on a regional GKE cluster (location is something like `us-central1`) and using the Stackdriver Context Graph adapter from `mixer/adapter/stackdriver/contextgraph`. The workload entities reported by the adapter end up with a cluster container path that doesn't resolve to my actual GKE cluster — Context Graph can't tie the workloads back to the cluster resource.\n\nIt works fine when I switch to a zonal cluster (location like `us-central1-a`). Looking at the URLs the adapter generates for the cluster container under `container.googleapis.com`, the path it uses only matches the zonal cluster resource layout. GKE regional clusters live under a different path, so any cluster whose location is a region rather than a zone gets a broken/nonexistent container reference.\n\nCould the contextgraph adapter detect whether `clusterLocation` refers to a region or a zone and emit the right resource path for each? Right now it's effectively zonal-only."} {"task_id": "format-code-task-001599", "source_id": "format-code-task-001599", "domain": "code", "task_path": "tasks/format-code-task-001599", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:fa4febe0c1645e93aec01228b3932316f3f38a5f99fd876ec2c636c11bd9b907", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want `dvc exp show` to let me sort the experiments listed under each baseline by a metric or parameter column. When I run `dvc exp show --sort-by `, DVC should resolve the named column from the metrics and params already shown in the experiment table and order the related experiments by that value, leaving experiments with missing values at the end for ascending order. `--sort-order asc` should be the default behavior, and `--sort-order desc` should reverse that ordering.\n\nThe sort key can be written as just a unique metric or parameter name such as `foo`, or as a qualified path/name form such as `metrics.yaml:foo` or `params.yaml:foo`; names containing colons should still be usable when the path/name split is unambiguous, and a leading colon like `:foo` should mean the column name `foo` without a path qualifier. If the requested sort name does not match any metric or parameter column, `dvc exp show --sort-by missing` should fail with a non-zero exit code and an error that says the sort column is unknown. If the requested sort name matches more than one metric or parameter column, `dvc exp show --sort-by foo` should fail with a non-zero exit code and an error that says the sort column is ambiguous and lists the matching qualified columns."} {"task_id": "format-code-task-001600", "source_id": "format-code-task-001600", "domain": "code", "task_path": "tasks/format-code-task-001600", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:21baffd901d242498e5d3b2ba96e5b0a6e08320407100e20bcf38c01216d03ef", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: customize the name used for each glyph in the generated CSS\n\nI'm using `webfont` to bundle a set of SVG icons into a webfont and emit a CSS stylesheet via the built-in `css` template:\n\n```js\nwebfont({\n files: 'src/svg-icons/**/*.svg',\n fontName: 'my-font',\n template: 'css'\n}).then(/* write result.styles to disk */);\n```\n\nBy default, the class name for each icon in the generated CSS is derived from the SVG file name (e.g. `home.svg` → `.my-font-home`). That's fine as a default, but I'd like to control how those per-icon names look in the final CSS without changing my source files.\n\nConcrete cases I run into:\n\n- The SVGs come from an external icon set / a designer hand-off, so renaming the files just to fit my CSS naming convention isn't practical (it goes out of sync the next time we re-import).\n- Sometimes I want all icon classes to share a common suffix (e.g. `-icon`) so they don't collide with other utility classes in the project.\n- Sometimes I want to massage the name a bit (case, separators, drop a prefix the icon set ships with, etc.).\n\nRight now the only workarounds I can see are:\n\n1. Rename every SVG file — fragile, as above.\n2. Throw away the built-in `css` template and maintain my own template just to change how the name is rendered — way too much overhead for what should be a one-line tweak.\n3. Post-process the generated CSS string with a regex — works but feels wrong, and breaks if the template ever changes.\n\nIt would be really helpful if `webfont` exposed a hook that lets the caller transform each glyph's metadata (its name in particular) before the styles template is rendered, so I can plug in arbitrary logic from my own config. Something I can pass alongside the existing options to `webfont({ ... })` and have it apply to every glyph that goes into the template.\n\nWould you be open to adding that? A new option along the lines of `glyphTransformFn` would work well for me."} {"task_id": "format-code-task-001601", "source_id": "format-code-task-001601", "domain": "code", "task_path": "tasks/format-code-task-001601", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:09ff85a59df22fd679d33fffdaf9fc40fee5671e16fb40fe55c1a5d643f3c874", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Missing `tf.sequence_mask` in the TensorFlow frontend\n\nI'm porting some TensorFlow code over to ivy and ran into a gap. My model\nuses `tf.sequence_mask` to build a boolean mask from a per-row length tensor\n(very common for padding-aware loss / attention).\n\nMinimal example of what I'm trying to do:\n\n```python\nimport ivy.functional.frontends.tensorflow as tf\n\nlengths = tf.constant([1, 3, 2])\nmask = tf.sequence_mask(lengths, maxlen=5)\n# expecting something like:\n# [[ True, False, False, False, False],\n# [ True, True, True, False, False],\n# [ True, True, False, False, False]]\n```\n\nBut `tf.sequence_mask` isn't available under the ivy tensorflow frontend —\nthe attribute doesn't exist, so I can't call it at all. It would be great\nto have it implemented so the rest of the tf API surface lines up with\nupstream TensorFlow (https://www.tensorflow.org/api_docs/python/tf/sequence_mask),\nincluding the optional `maxlen` (inferred from the data when not given) and\n`dtype` arguments."} {"task_id": "format-code-task-001602", "source_id": "format-code-task-001602", "domain": "code", "task_path": "tasks/format-code-task-001602", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:45fde49f4f2861ce7b5b11c8b08a561ae7a52b42763b3f013d69fe913db03c5d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `ivy.hardswish` is missing complex dtype support / `complex_mode` argument\n\nMost of the activation functions in `ivy` (e.g. `relu`, `leaky_relu`, `gelu`, `sigmoid`, `mish`, `softmax`, `log_softmax`, ...) accept a `complex_mode` keyword argument and route complex inputs through `handle_complex_input`, so I can pass complex arrays through them without surprises. `hardswish` is the odd one out.\n\nA small repro:\n\n```python\nimport ivy\n\nivy.set_backend(\"torch\") # same story with jax / numpy / tf\n\nx_real = ivy.array([-3., 0., 3., 5.])\nprint(ivy.hardswish(x_real)) # works fine\n\nx_cplx = ivy.array([1+2j, -1-1j, 0+0j])\nprint(ivy.hardswish(x_cplx)) # blows up depending on backend\n```\n\nA couple of related problems on top of that:\n\n1. The functional API `ivy.hardswish(x)` doesn't take a `complex_mode` kwarg at all, while sibling activations like `ivy.relu` / `ivy.mish` do. So the public surface is inconsistent — I can't write generic code that picks an activation by name and forwards `complex_mode` to it.\n2. The same gap shows up on the `ivy.Array` / `ivy.Container` instance methods and on the stateful `ivy.Hardswish` module — none of them expose `complex_mode` either, so e.g. `Hardswish()` can't be configured for complex inputs the way `ReLU(complex_mode=...)` or `GELU(complex_mode=...)` can.\n\nCould `hardswish` be brought in line with the other activations, both in terms of accepting complex inputs and in terms of exposing `complex_mode` everywhere it's exposed for the others (functional, array method, container method, stateful module)? Ideally the behaviour for purely real inputs should stay exactly as it is today."} {"task_id": "format-code-task-001603", "source_id": "format-code-task-001603", "domain": "code", "task_path": "tasks/format-code-task-001603", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:549017a411c683831560949799dc35aa0d026e3a457937c0886168c946f6c10d", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Feature request: add hardsilu / hardswish activation\n\nI was looking through `ivy`'s experimental activations and noticed quite a lot of them are already there — `silu`, `relu6`, `hardtanh`, `hardshrink`, `selu`, `elu`, `celu`, etc. — but I couldn't find `hardsilu` (a.k.a. `hardswish`).\n\nIt's a pretty standard activation at this point (MobileNetV3 popularized it, and most frameworks ship it natively — PyTorch has `torch.nn.functional.hardswish`, JAX has `jax.nn.hard_silu`, etc.), so it would be nice to have it available through ivy with the same backend-agnostic interface as the others.\n\nConcretely, I'd like to be able to do something like:\n\n```python\nimport ivy\n\nx = ivy.array([1., 2., 3.])\ny = ivy.hardsilu(x)\n```\n\nand also access it as an array / container method the same way the existing activations work:\n\n```python\nx = ivy.array([-0.5, 0.5, 2.])\ny = x.hardsilu()\n\nc = ivy.Container(a=ivy.array([-0.5, -1, 0]), b=ivy.array([0.5, 1., 2.]))\ny = c.hardsilu()\n```\n\nwith consistent results across all the supported backends (tensorflow / numpy / torch / paddle / jax).\n\nWould be great to slot this in next to `silu` / `relu6` in the experimental activations."} {"task_id": "format-code-task-001605", "source_id": "format-code-task-001605", "domain": "code", "task_path": "tasks/format-code-task-001605", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0ccde4df4097ef21c7f817a2bcfa2afabd0640618a54bad05b3cdfdb797ec1a0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want Mockjax to support browser-like follow-through for mocked HTTP redirects in the stateful jQuery workflow where I configure `$.mockjaxSettings`, register handlers with `$.mockjax(options)`, and issue requests with `$.ajax(...)`. The global setting `$.mockjaxSettings.followRedirects` should exist, default to `true`, and `$.mockjax.validateSettings()` should throw a `TypeError` mentioning `followRedirects` when that setting is not a boolean.\n\nWhen a mocked `GET` or `HEAD` request matches a handler whose `status` is `301` or `302` and whose `responseHeaders` contains either `Location` or `location`, Mockjax should start a second Ajax request to that location and deliver the second request's result to the caller. For example, if `/redirect-me` is mocked with `status: 301` and `responseHeaders: { Location: '/final' }`, and `/final` is mocked with `status: 201` and `responseText: 'redirected'`, then `$.ajax('/redirect-me')` should complete with the final mock response: URL `/final`, status `201`, and response text `redirected`. The same should work for a lowercase `location` header and for `302` responses, such as redirecting `/old` to `/new` and returning `/new`'s `202` response.\n\nThe follow-through request should include a `Referer` request header containing the original URL, so a target mock can match `requestHeaders: { Referer: '/redirect-me' }`. If the redirect target is not mocked, Mockjax should fall through exactly as an unmocked Ajax request to the target URL, and retained unmocked call records should show the target URL rather than treating the original redirect response as the final result.\n\nMockjax should not follow redirects when `$.mockjaxSettings.followRedirects` is `false`, when the mocked response has no `Location`/`location` header, or when the original request method is something other than `GET` or `HEAD`. In those cases the caller should receive the original mock response, such as a `POST` to a mocked `301` URL completing with status `301` instead of fetching the location target."} {"task_id": "format-code-task-001606", "source_id": "format-code-task-001606", "domain": "code", "task_path": "tasks/format-code-task-001606", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:60dc729ad5caf33d91be8975be057d9d1c1e89d1f51a0b83f556a6b1c98df409", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI need a `deepdow.experiments.History` object for experiment logging.\n\n`History()` should start empty, and `add_entry(...)` should append one row per call with the fields `model`, `metric`, `value`, `batch`, `epoch`, `dataloader`, `lookback`, `timestamp`, and `current_time`. If I omit `value`, it should default to `NaN`.\n\n`metrics_per_epoch(epoch)` should return a pandas DataFrame for that epoch, and asking for an unknown epoch should raise `KeyError`. `metrics` should return a pandas DataFrame that concatenates every stored row across all epochs.\n\n`pretty_print(epoch=None)` should print mean `value` grouped by `model`, `metric`, `epoch`, and `dataloader`; when `epoch` is `None`, it should use all stored epochs, and when an epoch is provided, it should only use that one.\n\nRepeated `add_entry` calls should accumulate history instead of replacing earlier rows."} {"task_id": "format-code-task-001607", "source_id": "format-code-task-001607", "domain": "code", "task_path": "tasks/format-code-task-001607", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:a973f30b0f025017489028c9ea90bd203e1b146cc305f3ec3014ce958622a6a0", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n`cloup.Group` ignores `command_class` to produce sub-commands\nUsing:\n- Cloup v2.1.1\n- Click v8.1.4\n- Python 3.11.4\n\n#### Bug description\n\n`cloup.Group` ignore the [`command_class` property](https://github.com/pallets/click/blob/d9af5cfa009c927a96d10ed38a3e37979876a12e/src/click/core.py#L1797-L1802) that is [used in `click.Group` to set the default class of the `@group.command()` decorator](https://github.com/pallets/click/blob/d9af5cfa009c927a96d10ed38a3e37979876a12e/src/click/core.py#L1855-L1894).\n\n#### To Reproduce\n\nHere is the minimal CLI, saved in a `cloup_test.py` file, that is reproducing the issue:\n\n```python\nfrom cloup import Command, Group, group\n\n\nclass CustomCommand(Command):\n\n def __init__(self, *args, **kwargs):\n kwargs.setdefault(\"context_settings\", {\"help_option_names\": (\"--help\", \"--my-fancy-help\")})\n super().__init__(*args, **kwargs)\n\n\nclass CustomGroup(Group):\n\n command_class = CustomCommand\n\n\n@group(cls=CustomGroup)\ndef my_cli():\n pass\n\n\n@my_cli.command()\ndef subcommand():\n pass\n\n\nif __name__ == \"__main__\":\n my_cli()\n```\n\nWhen I call the bare CLI at the group level, the `subcommand` is properly registered and appears in the help screen:\n```shell-session\n$ python ./cloup_test.py \nUsage: cloup_test.py [OPTIONS] COMMAND [ARGS]...\n\nOptions:\n --help Show this message and exit.\n\nCommands:\n subcommand\n```\n\nNow when I call the `--help` on the `subcommand` I cannot see any reference to my `--my-fancy-help` custom help option name:\n\n```shell-session\n$ python ./cloup_test.py subcommand --help\nUsage: cloup_test.py subcommand [OPTIONS]\n\nOptions:\n --help Show this message and exit.\n```\n\n#### Expected behavior\n\nInstead of the output above, I expect to get the following results:\n\n```shell-session\n$ python ./cloup_test.py subcommand --help\nUsage: cloup_test.py subcommand [OPTIONS]\n\nOptions:\n --help, --my-fancy-help Show this message and exit.\n```\n\nNotice how `--my-fancy-help` is featured in the help screen, because I expect `@my_cli.command()` to return a decorator of `CustomCommand`, as per the `command_class` property defined on `CustomGroup`.\n\n#### Click behavior\n\nThe same minimal CLI, sourced with Click's primitives, is working as expected.\n\nIn the example above, if you replace:\n```python\nfrom cloup import Command, Group, group\n```\n\nWith:\n```python\nfrom click import Command, Group, group\n```\n\nYou get the expected output:\n```shell-session\n$ python ./cloup_test.py subcommand --help\nUsage: cloup_test.py subcommand [OPTIONS]\n\nOptions:\n --my-fancy-help, --help Show this message and exit.\n```\n\n#### Notes\n\n- This might be related to https://github.com/pallets/click/pull/2417 , which has been fixed in the recent Click 8.1.4\n\n- This example is [inspired by Click's unittest](https://github.com/pallets/click/blob/d9af5cfa009c927a96d10ed38a3e37979876a12e/tests/test_command_decorators.py#L14-L31)"} {"task_id": "format-code-task-001608", "source_id": "format-code-task-001608", "domain": "code", "task_path": "tasks/format-code-task-001608", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:64cc9880060a16e59e4f48c36f4219487b1943d96fcecb0492a69fd1015d64b8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want a `classify_corpus.py` command-line tool that classifies an existing plaintext corpus into a categorized plaintext output corpus. I should be able to run it as `./classify_corpus.py SOURCE_CORPUS TARGET_CORPUS --classifier CLASSIFIER_PICKLE`, where the source corpus is loaded with an NLTK corpus reader and the classifier pickle is loaded from NLTK data.\n\nBy default the tool should use paragraph instances from the source corpus; with `--instances sents` it should classify sentence instances instead. Before classification it should turn each instance into bag-of-words features, lowercasing words, stripping punctuation, and optionally applying `--filter-stopwords LANGUAGE`; `--no-lowercase`, `--punctuation`, and repeated `--ngrams N` flags should control those normalization choices. For each instance, it should call the classifier's probability API, choose the most likely label, and append the original readable text to `TARGET_CORPUS/
` clause for updates and the `DELETE FROM
`\n clause for deletes.\n- **Every subsequent** `.from(...)` call contributes additional tables that are\n joined against. For an update these become a `FROM` clause; for a delete they\n become a `USING` clause. Multiple additional tables are separated by `, ` in\n the order the calls were made.\n\nBehavioral details that must hold:\n\n- When there is only a single `.from(...)` call, no `FROM`/`USING` clause is\n emitted at all (e.g. `update book set ...` or `delete from book`).\n- Clause ordering in the rendered query is:\n - update: `update
set ... from where ... returning ...`\n - delete: `delete from
using where ... returning ...`\n- All existing ways of naming tables in a `.from(...)` call keep working in\n both positions: bare string/template table names, aliased object tables\n (`{ alias: 'table' }`), and the `(values ...)` table produced from an array\n of row objects. Snake-casing of aliases/identifiers and parameter numbering\n are unchanged.\n- The `where`, `returning`, and `set` behavior is otherwise unchanged.\n\nOnly the Postgres builder is affected.\n"} {"task_id": "format-code-task-002730", "source_id": "format-code-task-002730", "domain": "code", "task_path": "tasks/format-code-task-002730", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0925998f1c781ffd3ef9cb08a03de97885886dd4b903ba113af4a1cdab150840", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `String.prototype.interpolate` makes `scheduling.ts` hard to unit test\n\nI was trying to add some unit tests around `scheduling.ts` (specifically `textInterval`, since it has the most branching logic) and hit a wall.\n\nMinimal reproduction — a jest test that just imports `textInterval` and calls it:\n\n```ts\nimport { textInterval } from \"../src/scheduling\";\n\ntest(\"textInterval formats months\", () => {\n expect(textInterval(35, false)).toBeTruthy();\n});\n```\n\nThe test blows up at runtime because `textInterval` ends up calling something like `t(\"MONTHS_STR_IVL\").interpolate({ interval: m })`, and `interpolate` isn't defined on strings in the test environment — the call just isn't a function.\n\nAfter poking around I noticed why: `interpolate` isn't a real utility, it's attached to the global `String.prototype` in `main.ts`. So the only way for the call inside `scheduling.ts` to resolve at runtime is if `main.ts` has already been imported somewhere first, which executes the prototype patch as a side effect. In a unit test that only wants to exercise `scheduling.ts`, that side effect hasn't happened, and there's no clean way to opt into it without dragging the whole plugin entry point into the test.\n\nTwo things feel off about this:\n\n1. From a testing standpoint, a pure function like `textInterval` shouldn't transitively depend on a global side effect from `main.ts` just to format a translated string. I'd like to be able to test scheduling logic in isolation.\n2. More generally, monkey-patching the global `String.prototype` to add an `interpolate` method is the kind of thing that affects every string in the process, not just translated ones. It feels like the wrong layer for this — interpolation is something the translation helper needs, not something every string in the codebase needs.\n\nWould it be possible to rework things so that interpolation lives with the translation layer instead of on the global string type? That would let `scheduling.ts` (and anything else that uses `t(...)` with placeholders) be tested without needing `main.ts` to have run first.\n\nHappy to help add tests for the translation helper itself once the shape of it settles."} {"task_id": "format-code-task-002731", "source_id": "format-code-task-002731", "domain": "code", "task_path": "tasks/format-code-task-002731", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:3c8bf6c80f74fb7bd0d39ae45464f773743c3450eabc4b3d5862127d47f98999", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## Isolate crashes when error handling path itself fails\n\nI'm running an aqueduct-based HTTP server and occasionally seeing the whole isolate die, taking out subsequent requests on it. After narrowing it down, the crashes seem to happen on the framework's error-handling path itself — not on the original handler's exception. I've reproduced two distinct patterns:\n\n### 1. Responding to a request that has already been responded to\n\nIf something in my handler responds to the request and then throws afterwards (cleanup that goes wrong, a follow-up operation that fails, an exception from inside an async continuation, etc.), the framework's catch block tries to send a `500` response on top of the one that already went out. Instead of just losing that one request, the resulting exception isn't caught anywhere and the whole isolate goes down.\n\nSmall repro of the shape:\n\n```dart\nclass MyHandler extends RequestHandler {\n @override\n Future processRequest(Request req) async {\n req.respond(new Response.ok({\"hello\": \"world\"}));\n throw \"post-respond failure\"; // simulate cleanup blowing up\n }\n}\n```\n\nThe first response is fine, but the throw triggers the framework's error path, which tries to `respond` again, and the isolate dies.\n\n### 2. Logging after the response is closed\n\nEven when the framework doesn't try to double-respond, the logging done by the error handler ends up calling something like `req.toDebugString(...)`, which reaches into the request's connection info / remote address. In some circumstances (looks like once the underlying response has been closed), that info isn't available anymore and the debug-string construction itself blows up — again on the catch path, again uncaught, again taking the isolate with it.\n\n### What I'd expect\n\nA single misbehaving request shouldn't kill the isolate and stop the server from handling anything else. Even when the error-handling path itself can't do its job (response already sent, connection info gone, etc.), it should fail quietly for that one request and let the server keep serving."} {"task_id": "format-code-task-002732", "source_id": "format-code-task-002732", "domain": "code", "task_path": "tasks/format-code-task-002732", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:20bceaf0ca70f4c2e8c5e99f89932b6a422a13153f41695d20bc69626a6f511b", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## CVE severity card totals don't add up to the table total\n\nWhen viewing the vulnerabilities page, the total CVE count shown in the table sometimes doesn't match the sum of the per‑severity cards (Critical / Important / Moderate / Low) above it. The mismatch is small but reproducible on clusters that have CVEs whose severity is reported as unknown by the scanner — those CVEs show up in the table row count but aren't reflected in any of the severity cards, so the cards under‑report.\n\nThis is confusing for anyone trying to reconcile the numbers (\"why does the table say 137 but the cards add up to 134?\").\n\nThe GraphQL types backing these cards (`ResourceCountByCVESeverity` / `ResourceCountByFixability`) currently only expose buckets for the four named severities, so the frontend has no way to render or include the leftover CVEs even if it wanted to. The backend needs to start surfacing counts for CVEs that fall outside Critical/Important/Moderate/Low (both total and fixable), so the UI can account for every CVE that the table is showing and the cards stop disagreeing with the table.\n\nThis applies to the same severity-count plumbing used by both image CVE and node CVE views."} {"task_id": "format-code-task-002733", "source_id": "format-code-task-002733", "domain": "code", "task_path": "tasks/format-code-task-002733", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:dc7b831467d124238e6d4ab0ddf0fd740d241ab28075b7d9e968051f03cb3075", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### Cannot override the default image registry on `SecuredCluster` via the operator\n\nWhen deploying RHACS Secured Cluster services through the operator, there is no way to make the secured-cluster components pull their images from a mirror / internal registry instead of the default upstream registry.\n\nThe Helm chart for secured cluster services already supports a registry override (the same knob that lets you rewrite e.g. `nginx:latest` → `/library/nginx:latest`), and the `Central` CR exposes equivalent configuration. But the `SecuredCluster` CR (`platform.stackrox.io/v1alpha1`) does not expose anything analogous, so operator users in disconnected / air‑gapped / mirrored-registry environments are stuck.\n\n#### What I tried\n\nI install the operator and apply a `SecuredCluster` resource:\n\n```yaml\napiVersion: platform.stackrox.io/v1alpha1\nkind: SecuredCluster\nmetadata:\n name: stackrox-secured-cluster-services\n namespace: stackrox\nspec:\n clusterName: my-cluster\n # ... no way to point image pulls at our internal mirror here\n```\n\nThere is no field on `SecuredClusterSpec` that gets plumbed through to the Helm `registryOverride` value, so the resulting workloads always try to pull from the default registry. In a disconnected cluster that has no egress to the default registry, the pods just keep failing to pull.\n\n#### What I expected\n\nParity with what the Helm chart and `Central` already allow: I should be able to specify a default-registry override on the `SecuredCluster` CR, and the operator should pass that through to the underlying chart so that all images used by secured cluster components get rewritten to come from my chosen registry.\n\nThis should be a normal optional field (an empty / unset value should keep today's behavior — no rewriting), and it should show up in the operator's CSV/CRD so it's visible to users editing the resource through the OperatorHub UI as an advanced setting, alongside things like `imagePullSecrets`, `customize`, `misc`, etc."} {"task_id": "format-code-task-002734", "source_id": "format-code-task-002734", "domain": "code", "task_path": "tasks/format-code-task-002734", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:696973870063ea9c9c49e0e907ef26f1f5bce9c4669282af9a8dfbf604a86e97", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the `st2 key load` CLI command to bulk load StackStorm datastore key-value pairs from a JSON or YAML file. The command should be invoked as `st2 key load FILE`, with an optional `-c`/`--convert` flag before the file path.\n\nThe file may contain either one key-value object or a list of key-value objects. Each object must provide `name` and `value`, and may also include `scope`, `user`, `encrypted`, `secret`, and `ttl`; for each item, the CLI should create or update that datastore key through the normal key API and print a table containing `name`, `value`, `secret`, `scope`, `user`, and `ttl` unless JSON or YAML output is requested.\n\nIf an item's `value` is already a string, load it as-is. If `value` is a non-string such as an object, array, boolean, integer, or float, `st2 key load --convert FILE` should JSON-encode that value before saving it; without `--convert`, the command should fail with a non-zero exit and explain that non-string values must be converted or written as strings in the file.\n\nFor a concrete example, loading a JSON file containing `{ \"name\": \"kv_name\", \"value\": \"super cool value\", \"scope\": \"system\" }` should exit 0 and send a datastore update for `kv_name` with that string value. Loading a file containing `[ { \"name\": \"first\", \"value\": \"one\" }, { \"name\": \"second\", \"value\": \"two\", \"secret\": true, \"ttl\": 100 } ]` should update both keys and exit 0. An empty YAML file should not error; it should exit 0 and produce the standard empty-result output. Missing files, unsupported file extensions, missing required fields, or non-string values without `--convert` should exit non-zero."} {"task_id": "format-code-task-002735", "source_id": "format-code-task-002735", "domain": "code", "task_path": "tasks/format-code-task-002735", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9a2a751dac60731e0af6f4d2e8e4c4152506a56c80c4db9fbf7d953d3cf7186a", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want the mysql2rgeo adapter to expose a pure Arel expression method `st_distance_sphere(rhs, units = nil)` on spatial Arel expressions, including table attributes such as `Place.arel_table[:lonlat]`, constants created with `Arel.spatial(rgeo_geometry)`, and spatial function nodes returned by other Arel spatial methods.\n\nCalling `Place.arel_table[:lonlat].st_distance_sphere(\"SRID=4326;POINT(-72.099 42.099)\")` should return an `RGeo::ActiveRecord::SpatialNamedFunction` named `ST_Distance_Sphere` with the receiver and RHS as its two expressions, so it can be used in predicates like `.lt(500)` inside ActiveRecord `where` clauses and serialize as a MySQL `ST_Distance_Sphere(...)` call with both geometry arguments handled as spatial values.\n\nCalling `Place.arel_table[:lonlat].st_distance_sphere(other_geom, :meter)` should return the same kind of Arel node with a third expression equal to the string `\"meter\"`; that units argument should be a normal non-spatial argument while the receiver and RHS remain spatial arguments.\n\nThe method should be referentially transparent: repeated calls with the same receiver, RHS, and units produce equivalent Arel nodes, and it must not mutate the receiver, the RHS string or geometry object, database state, filesystem, network, or global ActiveRecord configuration."} {"task_id": "format-code-task-002736", "source_id": "format-code-task-002736", "domain": "code", "task_path": "tasks/format-code-task-002736", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:80b7930a978605a2080e679cb09d51312fb6fd50d0c7e5396432a542a26e55fe", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nENH: Allow recursive estimation for samp size less than window size in RollingOLS\n#### Is your feature request related to a problem? Please describe\nSometimes you want estimates for sample sizes smaller than the full window size. \n\n\n#### Describe the solution you'd like\nAllow recursive estimation before the full sample is reached in RollingOLS so that the window would be\n```\n[1,n1]\n[1,n1+2]\n...\n[1,window]\n[2,window+1]\n....\n```\n\nThis is expanding window until the full window length is reached."} {"task_id": "format-code-task-002737", "source_id": "format-code-task-002737", "domain": "code", "task_path": "tasks/format-code-task-002737", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4382a8669f1148d5ec4db7e2f12dbdc6a4945bcd25536f2c42c4ea522303d547", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Let Archimedean copula parameters be supplied per call\n\nThe single-parameter Archimedean copulas (`ClaytonCopula`, `FrankCopula`,\n`GumbelCopula`) currently bake their dependence parameter into the instance: you\npass `theta` to the constructor and every `cdf`/`pdf`/`logpdf` call uses that\nstored value. That makes them awkward to use for anything that needs to evaluate\nthe same copula family at many different parameter values (e.g. plugging a\ncopula into an optimizer), because you have to keep building new objects.\n\nMake the dependence parameter overridable on a per-call basis, consistently\nacross all three of these copulas.\n\nConcretely:\n\n- `cdf`, `pdf` and `logpdf` accept an optional `args` argument holding the\n copula parameter(s) for that single call. For these one-parameter families\n that means `args=(theta,)`.\n- When `args` is non-empty, the value in `args` is used for that call and\n overrides whatever parameter the instance was constructed with. Evaluating\n `SomeCopula(theta=a).cdf(u, args=(b,))` must give exactly the same result as\n `SomeCopula(theta=b).cdf(u)`, and likewise for `pdf` and `logpdf`.\n- When `args` is empty (`()`, the default) or omitted, the parameter supplied at\n construction is used, exactly as before.\n- It must be possible to construct any of these copulas without supplying a\n parameter at all (i.e. leaving `theta` unset / `None`) without raising. Such an\n instance produces correct results when the parameter is provided through\n `args`.\n\n`pdf` and `logpdf` must stay consistent with each other: for the bivariate case,\n`pdf` equals `exp(logpdf)` regardless of whether the parameter came from the\nconstructor or from `args`.\n\nThe existing behaviour for instances built with a fixed `theta` and called\nwithout `args` must be unchanged.\n"} {"task_id": "format-code-task-002739", "source_id": "format-code-task-002739", "domain": "code", "task_path": "tasks/format-code-task-002739", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:94d1b37e14d3ebd1e8db3589db00539582006ad52285d94f3ac287b26520b0c8", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nDIckey Fuller test breaks on constant values\n#### Describe the bug\n\nDickey-Fuller test gives error when testing with constant values.\n\n#### Code Sample, a copy-pastable example if possible\n\n\n```python\nadfuller(np.full(8, 5.0))\n```\n`RuntimeWarning: divide by zero encountered in log\n llf = -nobs2*np.log(2*np.pi) - nobs2*np.log(ssr / nobs) - nobs2`\n\n#### Expected Output\n\nI have expected that this series is stationary\n\n#### Output of ``import statsmodels.api as sm; sm.show_versions()``\n\n
\n\nINSTALLED VERSIONS\n------------------\nPython: 3.10.6.final.0\nOS: Linux 5.15.0-50-generic #56-Ubuntu SMP Tue Sep 20 13:23:26 UTC 2022 x86_64\nbyteorder: little\nLC_ALL: None\nLANG: en_US.UTF-8\n\nstatsmodels\n===========\n\nInstalled: 0.13.2 (/home/michael/.local/lib/python3.10/site-packages/statsmodels)\n\nRequired Dependencies\n=====================\n\ncython: Not installed\nnumpy: 1.23.3 (/home/michael/.local/lib/python3.10/site-packages/numpy)\nscipy: 1.9.2 (/home/michael/.local/lib/python3.10/site-packages/scipy)\npandas: 1.5.0 (/home/michael/.local/lib/python3.10/site-packages/pandas)\n dateutil: 2.8.2 (/home/michael/.local/lib/python3.10/site-packages/dateutil)\npatsy: 0.5.3 (/home/michael/.local/lib/python3.10/site-packages/patsy)\n\nOptional Dependencies\n=====================\n\nmatplotlib: 3.6.1 (/home/michael/.local/lib/python3.10/site-packages/matplotlib)\n backend: module://matplotlib_inline.backend_inline \ncvxopt: Not installed\njoblib: Not installed\n\nDeveloper Tools\n================\n\nIPython: 8.5.0 (/home/michael/.local/lib/python3.10/site-packages/IPython)\n jinja2: 3.1.2 (/home/michael/.local/lib/python3.10/site-packages/jinja2)\nsphinx: Not installed\n pygments: 2.13.0 (/home/michael/.local/lib/python3.10/site-packages/pygments)\npytest: Not installed\nvirtualenv: 20.16.5 (/home/michael/.local/lib/python3.10/site-packages/virtualenv)\n\n
"} {"task_id": "format-code-task-002740", "source_id": "format-code-task-002740", "domain": "code", "task_path": "tasks/format-code-task-002740", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:820a386f68cea59f3693ae6b6b24011ea075fd7ea8f6f9b83409875f759d2079", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Resolve sync committee seats to validator indices\n\nWhen we process Altair (and Merge) blocks and validator duties, we constantly\nneed to know *which validator* occupies each seat of the current and next sync\ncommittees. The beacon state only stores the committees as lists of public keys\n(`current_sync_committee.pubkeys` / `next_sync_committee.pubkeys`), so every\nconsumer that wants a `ValidatorIndex` ends up rebuilding a pubkey → index\nmapping by scanning the entire validator registry. That scan is wasteful when it\nhappens on a hot path such as block replay.\n\nGive us a single helper that turns a sync-committee-bearing state into the\nvalidator indices behind both committees.\n\n## What to add\n\n- A public `SyncCommitteeCache` type with two fields, `current_sync_committee`\n and `next_sync_committee`, each an `array[SYNC_COMMITTEE_SIZE, ValidatorIndex]`.\n\n- A public `get_sync_committee_cache(state, cache)` that accepts an Altair or\n Merge `BeaconState` together with a mutable `StateCache` and returns a\n `SyncCommitteeCache`.\n\n## Behavior\n\nFor the returned value, at every position `i`:\n\n- `current_sync_committee[i]` is the index of the validator in\n `state.validators` whose public key equals\n `state.current_sync_committee.pubkeys[i]`.\n- `next_sync_committee[i]` is the index of the validator in `state.validators`\n whose public key equals `state.next_sync_committee.pubkeys[i]`.\n\nThe mapping must be exact for both committees across all `SYNC_COMMITTEE_SIZE`\nseats (the committees contain duplicates — the same validator may appear in\nmany seats — which is fine).\n\nBecause recomputing the mapping is expensive, the result should be memoized in\nthe supplied `StateCache`, keyed by the state's sync committee period, and\nreused on subsequent calls. The result must be deterministic and stable: calling\nthe helper repeatedly — whether reusing the same cache or starting from a fresh\none — yields the same indices.\n"} {"task_id": "format-code-task-002741", "source_id": "format-code-task-002741", "domain": "code", "task_path": "tasks/format-code-task-002741", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:559599c9d662ceccd557c93f48057fb7217153dc8e065dc63f60a52717337467", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nYet another rAF/cAF shim in client.js\nHey, many thanks for the nice project. \n\n`react-helmet-async` carries requestAnimationFrame shim on board. This is just the 7th rAF shim in our projects bundle. It's [pretty safe][0] to use the API directly in 2018 (yep, we'd also worred about rAF shiming to support old browsers [in 2014][1]).\n\n[0]: https://caniuse.com/#feat=requestanimationframe\n[1]: https://stackoverflow.com/questions/21177323/"} {"task_id": "format-code-task-002742", "source_id": "format-code-task-002742", "domain": "code", "task_path": "tasks/format-code-task-002742", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:79f6ecf5e67dfb11efa1d7163cdd57621b9059430c995515a6492d1c12ae426e", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `response.data` comes back as a plain dict instead of the typed object\n\nI'm writing unit tests against code that calls the Steamship SDK. After a call returns I'd like to assert on individual fields the natural way — using attribute access on the returned object, e.g.\n\n```python\nresponse = some_steamship_call(...)\nassert response.data.file.blocks[0].tags == [...]\n```\n\nBut this blows up because `response.data` is actually a plain `dict` at that point, not the typed model I was expecting. To get at the fields I have to fall back to string-keyed dict access:\n\n```python\nassert response.data[\"file\"][\"blocks\"][0][\"tags\"] == [...]\n```\n\nThis is painful for a few reasons:\n\n- No IDE autocomplete on the nested fields — easy to typo a key and not catch it until runtime.\n- No static type checking; the dict is `Dict[str, Any]` so everything underneath is opaque.\n- Tests end up coupled to the exact JSON-y key spelling instead of the model field names.\n\nIt looks like the body of the response is being serialized into a dict somewhere on the way back, even when I had (or could have had) the real typed model in hand. I'd like `response.data` to keep the original typed object so attribute access works in tests and downstream code. The dict form is only useful right before going over the wire — by the time it lands in `response.data` on the client side it should be the structured object again."} {"task_id": "format-code-task-002744", "source_id": "format-code-task-002744", "domain": "code", "task_path": "tasks/format-code-task-002744", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:0ae8b101c34947dd98d831f354ce74348910220091fa7e7ad0254a31304e91bd", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nI want an exported `MFAMethod` class that tracks which iCloud MFA channel is currently selected and exposes the request details needed by the authentication flow. It should be used as `const method = new MFAMethod(mfaMethod?: 'device' | 'voice' | 'sms', numberId?: number)`, with no arguments defaulting to the trusted device flow and phone-based flows defaulting to phone number id `1` when no id is provided.\n\nThe instance should be mutable through `update(mfaMethod?: 'device' | 'voice' | 'sms' | string, numberId?: number): void`. After `new MFAMethod()`, `isDevice` is true, `isSMS` and `isVoice` are false, `toString()` returns `'Device'`, `getResendURL()` returns `https://idmsa.apple.com/appleauth/auth/verify/trusteddevice`, `getResendPayload()` returns `undefined`, `resendSuccessful(202)` is true while `resendSuccessful(200)` is false, `getEnterURL()` returns `https://idmsa.apple.com/appleauth/auth/verify/trusteddevice/securitycode`, `getEnterPayload('123456')` returns `{ securityCode: { code: '123456' } }`, and `enterSuccessful(204)` is true while `enterSuccessful(200)` is false.\n\nFor `new MFAMethod('sms', 7)`, `isSMS` should be true, `toString()` should return `'SMS' (Number ID: 7)`, resend should target `https://idmsa.apple.com/appleauth/auth/verify/phone` with payload `{ phoneNumber: { id: 7 }, mode: 'sms' }`, entering code `123456` should target `https://idmsa.apple.com/appleauth/auth/verify/phone/securitycode` with payload `{ securityCode: { code: '123456' }, phoneNumber: { id: 7 }, mode: 'sms' }`, and both resend and enter success checks should accept HTTP 200 and reject the device-only success codes. For `new MFAMethod('voice', 3)`, the same phone endpoints and HTTP 200 success rules apply, but the payload mode should be `'voice'` and `toString()` should return `'Voice' (Number ID: 3)`.\n\nCalling `update()` should replace the instance state used by all later getters and helper methods. For example, starting with `new MFAMethod('sms', 7)` and then calling `update('device')` should make the same instance behave exactly like the default device instance, and calling `update('voice', 4)` afterward should make it behave like a voice method for phone number id `4`. Unknown method strings should fall back to the device flow instead of throwing, and separate `MFAMethod` instances should not share state."} {"task_id": "format-code-task-002745", "source_id": "format-code-task-002745", "domain": "code", "task_path": "tasks/format-code-task-002745", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:1d46bf3ed638ab2ccc5091837414ab74aeafe923284999a685850a395c8779c5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n### `txnbuild.PathPaymentStrictReceive` is declared but unusable\n\nI'm updating some transaction-building code to match the Protocol 12 / [CAP-0024](https://github.com/stellar/stellar-protocol/blob/master/core/cap-0024.md) naming, where the old `path_payment` op was renamed to `path_payment_strict_receive`. I noticed `txnbuild` already exposes a `PathPaymentStrictReceive` type, so I tried to use it directly:\n\n```go\nop := &txnbuild.PathPaymentStrictReceive{\n SendAsset: txnbuild.NativeAsset{},\n SendMax: \"10\",\n Destination: destAddress,\n DestAsset: txnbuild.NativeAsset{},\n DestAmount: \"1\",\n Path: []txnbuild.Asset{ /* ... */ },\n}\n\ntx := txnbuild.Transaction{\n SourceAccount: &sourceAccount,\n Operations: []txnbuild.Operation{op},\n Timebounds: txnbuild.NewInfiniteTimeout(),\n Network: network.TestNetworkPassphrase,\n}\nerr := tx.Build()\n```\n\nThis doesn't compile — `*PathPaymentStrictReceive` is not accepted as a `txnbuild.Operation`, and if I try calling `op.BuildXDR()` / `op.Validate()` directly they aren't found either. The only way I can actually send this operation today is to keep using `txnbuild.PathPayment{...}`, which is the pre-Protocol-12 name.\n\nI'd expect `PathPaymentStrictReceive` to be the usable, first-class way to build this operation now that the protocol has renamed it, while existing code using `PathPayment` still keeps working so nobody has to migrate in a panic.\n\nAlso, since CAP-0024 is about making path payments symmetrical, I'd expect `txnbuild` to additionally expose the new dual operation (something like `PathPaymentStrictSend`) so both directions of the symmetric pair are usable."} {"task_id": "format-code-task-002746", "source_id": "format-code-task-002746", "domain": "code", "task_path": "tasks/format-code-task-002746", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:6883e96dbb3448311c379a3f66e03aab55c4e9731276b76518bcb50998b6fba4", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nNetwork passphrase is easy to accidentally get wrong\n**Describe the bug**\nThe [runtime warning](https://github.com/stellar/js-stellar-base/blob/64785eb2e3dfece3b28e3c3767edaa6b8e09bb06/src/transaction.js#L31-L35) when passing a bad `networkPassphrase` to a `new Transaction` is easy to miss. The [laboratory had incorrect behavior](https://github.com/stellar/laboratory/pull/429/files) for a number of weeks and a stronger warning might have caught it earlier. \n\n**What version are you on?**\n2.1.2\n\n**To Reproduce**\n1. Pass something other than a string to `new Transaction(envelope, garbagePassphrase)`.\n2. See deprecation warning in console at runtime.\n\n**Expected behavior**\nA strong signal that an incorrect argument was passed.\n\n**Additional context**\nAdd any other context about the problem here."} {"task_id": "format-code-task-002747", "source_id": "format-code-task-002747", "domain": "code", "task_path": "tasks/format-code-task-002747", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:4a1320509b13ad9cc24faec66238f07df2f07a62d2e851c81dd09f6c8215fe19", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `disconnect()` is missing on the mock client\n\nI'm using `ioredis-mock` as a drop-in replacement for `ioredis` in my test suite. My application code creates a Redis client, does its work, and then calls `.disconnect()` on it during cleanup. Pretty standard pattern.\n\nAgainst a real `ioredis` client this works fine. Against `ioredis-mock` the test blows up at the disconnect call — the method just isn't there on the mock, so I get a \"not a function\" failure and nothing past that point in the teardown runs.\n\nWould be great if the mock supported `disconnect()` so the same code can run against either client without having to branch on whether we're in a test."} {"task_id": "format-code-task-002748", "source_id": "format-code-task-002748", "domain": "code", "task_path": "tasks/format-code-task-002748", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9f54ee2fce4967b024c280ad18b18d532ff63423daf95b6f474380857bf274a6", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\nkeyPrefix does not seem to work as expected\nThe tests for `keyPrefix` don't seem to be valid, and correcting them results in a failure. In particular, if you have two Redis instances that share the same data, but use a different key prefix, they will interfere with one another.\n\nHere's a test that shows the issue:\n\n```\n\ndescribe('multiple intance use same key with diffence keyPrefix', () => {\n const redis1 = new MockRedis({\n data: {\n foo: 'bar',\n hello: 'world',\n },\n keyPrefix: 'test:',\n });\n\n const redis2 = redis1.createConnectedClient({\n keyPrefix: 'test2:',\n });\n\n it('should return null on keys that do not exist', () =>\n redis2.get('foo').then(result => expect(result).toBe(null)));\n```\n\nHere's another example test that fails:\n\n```\n describe('multiple instance use same key with diffence keyPrefix', () => {\n const redisBase = new MockRedis({\n data: {\n foo: 'bar',\n hello: 'world',\n },\n });\n const redis1 = redisBase.createConnectedClient({\n keyPrefix: 'test:',\n });\n\n const redis2 = redis1.createConnectedClient({\n keyPrefix: 'test2:',\n });\n\n it('should not be able to read something set with one prefix using another prefix', () =>\n redis2\n .set('hello', 'ioredis')\n .then(status => expect(status).toBe('OK'))\n .then(() => redis1.get('hello'))\n .then(result => expect(result).toBe(null)));\n });\n```\n\nThe current test suite uses separate instances of `MockRedis` and tests that they do not interfere. However, each instance of `MockRedis` already will not interfere, that's how `MockRedis` works! To only way to test that a `keyPrefix` results in non-interference between instances is to use `createConnectedClient` as above."} {"task_id": "format-code-task-002749", "source_id": "format-code-task-002749", "domain": "code", "task_path": "tasks/format-code-task-002749", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:8a69d9f8b165532a8b6178955d747f0d2425674df3e2cf2139efe091a6edea14", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n# Bug report\n\n## Describe the bug\n\nComponents created with `styled` doesn't accept styles via `className` prop.\n\nA clear and concise description of what the bug is.\n\n## To Reproduce\n\nhttps://codesandbox.io/s/sleepy-dubinsky-04gio?file=/src/App.js\n\n## System information\n\n- OS: Windows\n- Browser: latest Chrome\n- Version of Stitches: @stitches/react@0.1.6"} {"task_id": "format-code-task-002750", "source_id": "format-code-task-002750", "domain": "code", "task_path": "tasks/format-code-task-002750", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:74bc026f505f8e8e4263c4fc3b244fa0de926719cfd672985725a85bcaf81403", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `Manager.Reset` fails on HPE iLO4 (gen9) when the manager doesn't advertise `ResetType@Redfish.AllowableValues`\n\nI'm using gofish to reset the BMC on an HPE Gen9 server (iLO4). The Actions block returned by the manager looks like this — note there's only a `target`, no `ResetType@Redfish.AllowableValues`:\n\n```json\n\"Actions\": {\n \"#Manager.Reset\": {\n \"target\": \"/redfish/v1/Managers/1/Actions/Manager.Reset/\"\n }\n}\n```\n\nSo `SupportedResetTypes` ends up empty on the parsed `Manager`. When I call `manager.Reset(...)`, the server rejects the request with `Base.0.10.ActionNotSupported` and the BMC is not reset. Sniffing the request shows gofish posts a body of `{\"Action\":\"Manager.Reset\"}`, which iLO4 doesn't accept.\n\nPer the [Manager schema](https://redfish.dmtf.org/schemas/v1/Manager.v1_18_0.yaml), `ResetType` on the reset request body is described as:\n\n> This parameter shall contain the type of reset. The service can accept a request without the parameter and perform an implementation specific default reset. Services should include the @Redfish.AllowableValues annotation for this parameter to ensure compatibility with clients, even when ActionInfo has been implemented.\n\nSo when a manager doesn't advertise any allowable reset types, the spec-compliant thing is to send the request without a `ResetType` and let the service do its default reset. I confirmed on iLO4 that posting an empty body to the reset target works and the BMC reboots as expected.\n\nCould `manager.Reset` handle this case so it works against managers (like HPE Gen9 / iLO4) that don't expose `ResetType@Redfish.AllowableValues`?"} {"task_id": "format-code-task-002751", "source_id": "format-code-task-002751", "domain": "code", "task_path": "tasks/format-code-task-002751", "docker_image": "docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:9a1a6dd1c8cfa0252cdccf1eca24d5b15ce9650eb63a899598276defa77795e5", "step_limit": 500, "agent_timeout_sec": 3600.0, "verifier_timeout_sec": 2100.0, "instruction": "Fix the following issue:\n\n## `no-redundant-story-name` doesn't catch redundant names in CSF2 stories\n\nI have a project with a lot of CSF2-style stories, e.g.:\n\n```js\nexport const PrimaryButton = () =>