-
-
Notifications
You must be signed in to change notification settings - Fork 7.9k
[MNT]: Python 3.14.0a7 test failures #29959
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
Comments
FYI, a NumPy optimization that elides temporaries is broken on 3.14. Probably best to wait until numpy/numpy#28748 (comment) is merged to work on downstream libraries, otherwise you'll be dealing with deeply confusing test failures. |
FYI, I jumped to test matplotlib on numpy 2.2.5 with that PR patched on top - it fixed most of the problems, remaining ones seem to be genuine for matplotlib:
|
It looks like the refcount is coming in one better than expected (I suspect this is due to an optimization in refcounting where it changes using "borrow" bytecodes when it decides it is safe so the refcount inside of |
In NumPy I handled updating tests that depend on refcounts to check the change in the refcount rather than a constant refcount value. That should be the same in all Python versions. There was only one other obscure case in NumPy where that didn’t work because we happened to trigger another new refcount optimization on top of the stackref optimization. |
Summary
I ran tox on the main branch of matplotlib with the fix from #29393 applied on top. I ran the same set of tests which we run during the Fedora matplotlib build, so invocation looked like this:
And the results:
Proposed fix
No response
The text was updated successfully, but these errors were encountered: