Debug GraalPy/uv test failures on Windows

[joerick]

Debugging windows failures in #2734. Perhaps the error only occurs on GP311...

[joerick]

That passed! So it seems the bug is limited to only gp311, when combined with uv/build[uv]. My guess is on some kind of Windows path bug with GraalPy311, does that sound plausible @timfel ? (see #2734 (comment) for the error we've been debugging, I'm curious about the use of D: vs C: drives as a possible trigger).

My inclination is to skip the test, document known issue and move on. This workaround in the tests is ugly though (global state). How long would you like to continue supporting gp311 in cibuildwheel Tim? I remember spring 26 was a possible answer. In that case I'd be happy for this workaround to live in the tests for a few months.

[henryiii]

A test skip & mention in docs is probably fine.

[timfel]

sorry, I have been out sick and had not seen this.

My guess is on some kind of Windows path bug with GraalPy311, does that sound plausible @timfel ?

it does, we've found quite a few windows issues during the upgrade to 3.12

How long would you like to continue supporting gp311 in cibuildwheel Tim?

We released the latest graalpy in September, at some point we said giving people 6 months overlap should be fine, so I wouldn't mind dropping gp311 by the end of March, if that's ok with you?