[mike-lmctl]

Closes #14619.

This turns a TypeError during collection into a normal parametrize collection error when tuple-style argnames receive a scalar parameter set.

For example, @pytest.mark.parametrize("x,", [None]) now points at the offending test and reports that the parameter set should be a sequence, instead of showing an internal len(None) crash.

I added a regression test for the None case and a changelog entry.

I used AI assistance while preparing this change, reviewed the patch locally, and take responsibility for the contribution.

Checks run locally:

PYTHONPATH=src python3 -m pytest testing/test_mark.py::test_parametrized_collect_with_non_sequence_values testing/test_mark.py::test_parametrized_collect_with_wrong_args -q
PYTHONPATH=src python3 -m pytest testing/test_mark.py -q
git diff --check -- src/_pytest/mark/structures.py testing/test_mark.py changelog/14619.bugfix.rst AUTHORS
python3 -m py_compile src/_pytest/mark/structures.py testing/test_mark.py

testing/python/metafunc.py was not run locally because hypothesis is not installed in this environment.

Checklist:

• New regression test added.
• Maintainers may push to and squash this branch when merging.
• Closes Spurious comma in parametrize causes obfuscated crash #14619.
• AI assistance is credited in the commit trailer.
• Changelog fragment added in changelog/14619.bugfix.rst.
• Added myself to AUTHORS.

[DTiming24]

Good improvement turning len(None) into a collection error. Since the repro in the body uses @pytest.mark.parametrize("x,", [None]), I think the error message should probably mention the single-value tuple syntax explicitly (('value',)), because that's the quickest fix users can make.

[user01010111]

This was closed today without a merge. Is #14619 available for someone else to take over, or do you plan to continue it? I would like to improve the parameter-value diagnostics while preserving the existing tuple-style behaviour.

[RonnyPfannschmidt]

Actually please reopen

This slipped out of my review window

It's Actually good to go

[user01010111]

The source fork is currently unavailable: mike-lmctl/pytest returns 404 from the GitHub API and Git reports “Repository not found”; it is also absent from the author’s public repository list. GitHub still retains the approved head 7cab31585b04c4b83727bd337340a22dd0d3c414 under this PR.

@RonnyPfannschmidt, if the author cannot restore the source and reopen this PR, would you prefer the approved patch carried into a fresh PR with Mike Ma’s authorship and credit preserved? I will hold off on a replacement pending agreement.

[RonnyPfannschmidt]

I just had Claude do a preservation

Ill wait a bit to see if comments from @mike-lmctl come in and merge mid week unless a complaint comes up