Repository navigation
CI: sample the unsplit WASM test suite with the semicolon-lapack IFAIL fix (do not merge) - #12
Closed
agriyakhetarpal wants to merge 6 commits into
Closed
agriyakhetarpal wants to merge 6 commits into
agriyakhetarpal wants to merge 6 commits into
Conversation
…L fix Sampling bed for scipy#25713. Builds SciPy untouched against BLIS 2.1 and semicolon-lapack from the branch behind ilayn/semicolon-lapack#36, runs the whole test suite unsplit in one interpreter (pytest-split and its driver script removed), and un-skips test_wrapper_module_collected (scipygh-26267). The conftest workarounds stay as they are. Only the Emscripten/Pyodide job is kept, as a 16-job matrix per push with no concurrency group. Not for merging.
Owner
Author
|
Conclusion reached! This is the right fix, based on 87 passing runs (all others unrelated). Closing. |
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 join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Sampling bed for the Emscripten/Pyodide job, follow-up to scipy#25713. Not for merging.
The hangs and traps in that job were caused by an out-of-bounds write in semicolon-lapack's Fortran-ABI shim, which converts
IFAILto 1-based indices even whenJOBZ = 'N'while SciPy passes a 0 or 1 element dummy buffer in that case (ilayn/semicolon-lapack#35, fix in ilayn/semicolon-lapack#36). This branch checks whether the mitigations added while that was unknown can go.What is different from
main:fix-fabi-shim-ifail-jobzbranch behind BUG: fabi shim: convertIFAILonly whenJOBZ = 'V'ilayn/semicolon-lapack#36 instead of the v0.01.3-pre tarball; BLIS and SciPy are untouched.python -m pytest -vra --pyargs scipy -m 'not slow', with pytest-split andtools/wheels/cibw_test_command_pyodide.shremoved.test_wrapper_module_collectedis removed. The other TST: linalg: fix Pyodide hang and crash in wrapper GC tests scipy/scipy#26267 skip stays, since that test hits the WASM stack limit, which is unrelated.conftest.pyworkarounds stay as they are: theget_native_idstub and the threadpoolctl filter are unrelated to the bug, and the two GC-related pieces cover an end-of-session collection that still traps on a fixed stack (a separate problem, see the notes in CI: pyodide job can hang scipy/scipy#25713).Every push runs the job as a 16-job matrix with every other workflow removed and no concurrency group, so samples are never cancelled. Read results with
gh run list --workflow emscripten.yml --branch wasm-ci-unsplit-sampling; a hang shows up ascancelledat the 45 minute limit, a trap asfailurewith afatal errorin the log. Push an empty commit to add another 16 samples.🤖 Generated with Claude Code