This repository was archived by the owner on Nov 4, 2025. It is now read-only.
Remove problematic content slicing in test output parsing - #6
Open
ryanhoangt wants to merge 149 commits into
Open
ryanhoangt wants to merge 149 commits into
ryanhoangt wants to merge 149 commits into
Conversation
Some of the repo_setup.sh scripts leave the working tree in a dirty state which can make it difficult to generate a patch that applies cleanly. This change commits any outstanding changges such that any patch generated with `git diff` will cleanly apply to a newly launched container during the evaluation step.
* Simplify installation guidelines for inference submodule * Fixes SWE-bench#368 * Update version
* add docs * Add leaderboard * Remove unused import * Update docs * Update version * Update: docs
* Support multilingual evaluation * CI: Fix documentation building vs deploying * Minor fixes * Remove some redundancy * Update dataset ref --------- Co-authored-by: Kilian Lieret <kilian.lieret@posteo.de> Co-authored-by: John Yang <byjohnyang@gmail.com>
…re are subsequent successful futures. (SWE-bench#370)
…h#358) * fix: preserve all issue references with same keyword in PRs * Modified extract_resolved_issues to use a set instead of list to store references
Co-authored-by: John Yang <byjohnyang@gmail.com>
…t.py SWE-bench#368 (SWE-bench#369) * fix prompt_col from text_inputs to text * update log --------- Co-authored-by: changqingai <changqingai@tencent.com>
Match the documentation for installing additional dependencies with the contents of `pyproject.toml`
…set lacks v5 fields
Updated the GitHub repository URL for cloning SWE-bench, old one wasn't working. Someone should check other search results for `princeton-nlp`: https://github.com/search?q=repo%3ASWE-bench%2FSWE-bench%20princeton-nlp&type=code
* Default images build --tag to latest so make_image_spec does not assert * Run eval containers with seccomp unconfined so browser sandboxes can start * Decode test output with errors=replace so a stray byte cannot abort an instance
parse_log_seaborn indexed line.split()[1] / parts[1] without checking length, so a log line containing only a status word (e.g. a bare 'PASSED' or 'FAILED' from interleaved or captured test output) raised IndexError and crashed grading for mwaskom/seaborn instances. Other pytest parsers in this module already guard against short lines; apply the same guard here. Add a regression test.
Jest only appends a "(N ms)" duration suffix above a small timing threshold, and the optional duration group combined with the non-greedy name capture can leave a trailing space in the captured test name depending on whether that suffix is present. The same test can then map to two different dict keys across runs, which can hide a real fail-to-pass transition.
662 parametrized ids in SWE-bench_Verified are truncated mid-parameter (unbalanced brackets), so exact-match grading counts them as missing -> failed even when they passed. Resolve truncated ids (and only those) by prefix: accept a unique match, or multiple matches that all share one status. Exact ids keep exact semantics; ambiguous conflicts stay failed.
john-b-yang
force-pushed
the
fix-modal-patch-eval
branch
from
August 12, 2026 03:07
aa0f1ed to
74690d7
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
20 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Reference Issues/PRs
Fix SWE-bench#377
What does this implement/fix? Explain your changes.
When running patch eval on Modal, I see that for some instances, the content of test output files being captured are out of order, which causes the test summary to fall outside the
>>>>> Start Test Outputand>>>>> End Test Outputmarkers. I attached a sample log file below.test_output_astropy__astropy-12907.txt
This PR removes the content slicing line and uses the whole file content for parsing.
Any other comments?
🧡 Thanks for contributing!