Hugo van Kemenade

I'm on another Mac right now and they also all work here too... ```console ❯ uv run --python /Library/Frameworks/Python.framework/Versions/3.14/bin/python3 --with markupsafe==3.0.2 1.py 3.14.0b3 (v3.14.0b3:26d485d1225, Jun 17 2025, 12:48:06) [Clang 16.0.0...

Thanks, it was indeed some old cached files. Comparing the problematic machine (left) with the working one (right) shows files were built in April and May respectively: And the `WHEEL`...

Yes, it'd be useful to highlight it. I don't know if we can add colour? If so maybe something like this? Although regular "yellow" is not so good in light...

sphinx-autobuild is very handy, we have a `make htmllive` using it for the devguide, CPython docs and PEPs repos. For example: https://github.com/python/devguide/blob/main/Makefile

For the first one, Python has already evaluated `' long string' == 'long string'` as `False`, and then pytest just does `assert False`. For the second one, it looks like...

Actually, it's not `difflib` that's slow, it's the comprehension to strip newlines. This runs fast: ```diff explanation += [ - line.strip("\n") - for line in ndiff(right.splitlines(keepends), left.splitlines(keepends)) + ndiff(right.splitlines(keepends), left.splitlines(keepends))...

I'd guess the latter is a bit faster as it's doing things in C, but both of those timeout.

Ah yes, I should have made clear the things I posted weren't meant to be correct, just looking for the slow bits :)

The alternative #8869 removed installation of things like `rav1e` and `svt-av1`, are they needed here?

This PR was about x1.8 in April (https://github.com/python-pillow/Pillow/issues/8856#issuecomment-2827278596), and we're getting close to the next release, how is it looking now?