Skip to content
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

array-api-strict v2.1.2 breaks CI #9750

Closed
kmuehlbauer opened this issue Nov 8, 2024 · 9 comments · Fixed by #9762
Closed

array-api-strict v2.1.2 breaks CI #9750

kmuehlbauer opened this issue Nov 8, 2024 · 9 comments · Fixed by #9762
Labels
bug CI Continuous Integration tools upstream issue

Comments

@kmuehlbauer
Copy link
Contributor

kmuehlbauer commented Nov 8, 2024

What is your issue?

array-api-strict v2.1.2 (released some hours ago) needs numpy >=2.1, but the conda-forge package is missing that pin:

see:

Some of xarray's dependencies are not available yet for numpy >=2.1, the current environments resolve to numpy 2.0.2. This breaks the tests where array-api-strict is imported.

@kmuehlbauer kmuehlbauer added needs triage Issue that has not been reviewed by xarray team member bug CI Continuous Integration tools and removed needs triage Issue that has not been reviewed by xarray team member labels Nov 8, 2024
@kmuehlbauer
Copy link
Contributor Author

The offending package is numba which has pinned numpy <2.1 for the current version 0.60.

numba 0.61 will be numpy 2.1 compatible, but isn't yet released.

One short-term solution to unlock CI would be to pin numpy <2.1 in all CI with environments containing numba.

@keewis
Copy link
Collaborator

keewis commented Nov 8, 2024

you mean, pin array-api-strict for now?

@kmuehlbauer
Copy link
Contributor Author

That would be one solution, or to pin numpy>=2,<2.1. Not sure what's more reasonable.

@keewis
Copy link
Collaborator

keewis commented Nov 8, 2024

if I understand correctly, the issue is that the new version of array-api-strict requires numpy>=2.1 but the conda-forge package metadata does not reflect that. So I don't think requiring an older version of numpy would help, since we'd still get the new (broken) version of array-api-strict.

@rgommers
Copy link

rgommers commented Nov 8, 2024

The conda-forge array-api-strict metadata is fixed now.

If the current choice is to pin array-api-strict or numpy, I'd constrain the former though.

@keewis
Copy link
Collaborator

keewis commented Nov 8, 2024

numpy is indirectly constrained by numba, so if the metadata has been fixed there might be nothing we need to do here?

Edit: actually, I misunderstood what you said. I guess it would be best to pin array-api-strict (or explicitly forbid the broken versions) either way.

@kmuehlbauer
Copy link
Contributor Author

Thanks @rgommers, much appreciated.

@keewis The all-but-numba check fails even with numpy=2.1.3 and array-api-strict=2.1.2. See https://github.com/pydata/xarray/actions/runs/11739187309/job/32703227103?pr=9751

@keewis
Copy link
Collaborator

keewis commented Nov 8, 2024

I'll look into it, but this might take some time.

@kmuehlbauer
Copy link
Contributor Author

@keewis I've pinned array-api-strict<=2.1 in all relevant environments in #9751. Would be good to get this in to unlock CI again. Would also buy you some time to investigate the issues wrt the failing array-api-strict related tests.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug CI Continuous Integration tools upstream issue
Projects
None yet
Development

Successfully merging a pull request may close this issue.

3 participants