|
Download tasks/candidate-0847-software-languages/instruction.md from FineEnvs/MiMo-V2.6-RL-harbor-terminal: direct link, hf CLI and curl.
- Browser
- Download file 2.13 kB
-
https://huggingface.co/datasets/FineEnvs/MiMo-V2.6-RL-harbor-terminal/resolve/5708f7d17ec6ffe77bc9ab4585072b4ac0acfb13/tasks/candidate-0847-software-languages/instruction.md
- Command line
-
hf download hf://datasets/FineEnvs/MiMo-V2.6-RL-harbor-terminal@5708f7d17ec6ffe77bc9ab4585072b4ac0acfb13/tasks/candidate-0847-software-languages/instruction.md
-
curl -L -o instruction.md https://huggingface.co/datasets/FineEnvs/MiMo-V2.6-RL-harbor-terminal/resolve/5708f7d17ec6ffe77bc9ab4585072b4ac0acfb13/tasks/candidate-0847-software-languages/instruction.md
2.13 kB
| You are an agent, your current working directory is /app. | |
| You can use the tools available to you to interact with the computer to assist the user in completing tasks. | |
| # Repair independent LALR interactive-parser clones | |
| The frozen Lark tree in `/app/vendor/lark` has a clone-ownership regression. Copying an `InteractiveParser` after partially consuming a source stream creates multiple lexer objects whose references do not agree. A clone may therefore consume the original parser's remaining input, or a superficially independent implementation may leave `parser_state.lexer` disconnected from `lexer_thread`. | |
| Repair the existing implementation in both files: | |
| - `/app/vendor/lark/lark/parsers/lalr_parser_state.py` | |
| - `/app/vendor/lark/lark/parsers/lalr_interactive_parser.py` | |
| ## Required behavior | |
| 1. `ParserState.copy(deepcopy_values=...)` must create an independent lexer thread and independent parser stack while preserving its current shallow-versus-deep value-stack policy. | |
| 2. `InteractiveParser.copy(deepcopy_values=...)` must use one coherent cloned lexer thread: the clone's `parser_state.lexer` must be the clone's `lexer_thread`, and neither may be the original parser's lexer thread. | |
| 3. After partially consuming input, clone and original must both be resumable to the same complete parse tree in either execution order. Advancing either continuation must not change the other's cursor. | |
| 4. Existing manual `feed_token`, `exhaust_lexer`, EOF, and ordinary LALR parsing behavior must remain compatible. | |
| Do not replace the parser with a task-specific implementation, hard-code the sample streams, disable tests, or edit the supplied upstream test asset. | |
| ## Validation and artifact | |
| From `/app/vendor/lark`, run: | |
| ```bash | |
| python3 -m compileall -q lark | |
| python3 tests/test_parser.py TestLalrBasic.test_parser_interactive_parser -q | |
| ``` | |
| Then generate the required artifact: | |
| ```bash | |
| cd /app | |
| python3 tools/run_clone_matrix.py --output /app/repair_report.json | |
| ``` | |
| The artifact must use `output_schema_version` **`lark-clone-repair-v1`** and the driver must report `summary.all_cases_pass: true` after the source repair. | |