-
-
Notifications
You must be signed in to change notification settings - Fork 25.8k
CI Azure: try removing setuptools workaround on windows #22905
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
CI Azure: try removing setuptools workaround on windows #22905
Conversation
I suspect this fails randomly on windows. For example the scheduled wheel builder job sometimes succeeds and sometimes fails on windows: https://github.com/scikit-learn/scikit-learn/actions/workflows/wheels.yml?query=branch%3Amain+event%3Aschedule |
I think it's slightly different. The env var was to tell setuptools to use the legacy behavior of distutils because distutils introduced a breaking change without taking into account the packages that monkeypatch it like numpy. The issue we expected was not random: numpy/numpy#17216. Distutils released a fallback mechanism but in the mean time we needed this env var. If I'm right, it does not mean that the windows build is fixed. It only means that the fallback mechanism works, but we might still be building with the not thread safe CCompiler, (i.e. we still need #22693) |
The PR that introduced this is #22049 which seems to have been to quickly fix the CI because it was failing in every PR, which confirms it being not a random failure |
I think there are two part of this issue:
Edit: This is exactly what you said 😅 |
Now I don't understand why we have deprecation warnings from distutils that we don't have in other jobs... |
This deprecation warning is raised by:
so this is a problem in joblib (not scikit-learn directly). I fixed it in a recently merged PR by backporting |
But if it's not released yet, why don't we trigger that in the other CI jobs ? |
…jeremiedbb/scikit-learn into try-remove-windows-build-setuptools-fix
@thomasjpfan I think we can merge this one |
…#22905) Co-authored-by: Thomas J. Fan <thomasjpfan@gmail.com>
Let's see if we still trigger the issue