[fixfon]

Summary

Closes #29535.

Sphinx-style docstrings often escape the stars as \*args and \*\*kwargs, because a bare * starts emphasis in reST. D417 didn't account for that, so it reported those parameters as undocumented even though they were documented.

It went wrong in a different way for each style. In NumPy-style sections, the name before the : was taken verbatim, so \*\*kwargs never matched **kwargs. In Google-style sections, the regex didn't accept a leading \, so the whole line was skipped.

I added a small unescape_parameter_name helper in sections.rs that strips backslashes from the documented name, and both parsers now use it. The Google regex also accepts up to two escaped stars now. Since a backslash can't be part of a Python parameter name, removing all of them should be safe. This also covers non-raw docstrings that write \\*\\*kwargs.

Test Plan

I added five cases to D417.py: escaped NumPy and Google docstrings with every parameter documented, the same two styles with one parameter actually missing, and a non-raw docstring. Without the fix, all five d417 tests fail and report *args and **kwargs as missing. With the fix, the fully documented functions aren't flagged. In the missing variants, only the parameters that really are missing get reported. The snapshots are updated, and cargo test -p ruff_linter -- pydocstyle and clippy both pass.

I used Claude (via T3 Code) to help with the implementation. I've reviewed the change and can explain it.

[astral-sh-bot]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[ntBre]

Thank you!

Replacing all backslashes means we'll be a little overly permissive. For example, we'd convert something invalid like foo\\bar to foobar, rather than only replacing \*, but I think that's okay.