-
-
Notifications
You must be signed in to change notification settings - Fork 46
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
Rebuild for python 3.13 #329
Rebuild for python 3.13 #329
Conversation
…nda-forge-pinning 2024.08.29.02.39.48
Hi! This is the friendly automated conda-forge-linting service. I just wanted to let you know that I linted all conda-recipes in your PR ( |
Hi! This is the friendly conda-forge automerge bot! I considered the following status checks when analyzing this PR:
Thus the PR was passing and merged! Have a great day! |
19cfe0d
to
274d302
Compare
@conda-forge/core, we're a bit stuck here, mainly because ustream numpy was only testing with Cython 3.1.0 nightlies, but there's no tag we can use for that (short of building from head). OTOH, the failures here are limited to osx (for reasons unknown), and it's only about ~20 test failures out of almost 50k tests. Do we want to:
|
IMHO we wait. This is an issue in numpy and numpy needs to either fix it or make everyone wait for cython 3.1. That's for them to decide. |
Given how things went with cython 3.0, that's potentially many months if not longer.
Normally I'd agree, though in this case I think a migration without numpy will not get us very far. However, where impact justified it (and the expected fallout was ~0), we have skipped a handful of tests here and there for numpy/scipy/etc. many times already. Skipping 20 tests out of 50000 does not break the package in any meaningful way IMO. And doubly so for RC builds. As maintainer here, I wouldn't hesitate in skipping those tests under current circumstances, my question was more around the fact that it's quite possible that we'll run into the "actually requires cython 3.1.0" with several packages (i.e. everyone who's been working on nogil support must use those nightlies), and how we want to deal with that. That's why I asked for a 3.1.0 alpha, and that's why I kind of favour the second point above, i.e. put a cython 3.1.0 build under |
These dtype failures are scary and really hard to debug. We should submit patches to numpy or cython or wait |
Which dtype failures are you referring to? All but one failure are due to some test module (presumably compiled by cython) running into:
Thanks for the input! |
from the test names, they just seem to have to do with "date time dtypes". I didn't look into it much further but:
|
key part here is (IMO)
IOW, this is not about numpy datetypes not working per se, but about testing interaction with cython. There are many other tests for datetimes (example) that are passing. Looking closer, the tests in that module are all failing because this fails to get transpiled/compiled correctly. Which is admittedly more impactful than I first thought, because IIUC it means you cannot use (parts of) the numpy API through cython. |
I added a failure from the log to the issue description of numpy/numpy#27314 (logs disappear ....). What stands out there is that
That is likely causing a difference between how |
Ah of course, missed the separate That leaves explaining why it's only happening for Python 3.13 on macOS x86-64 and not for Python 3.12. Is |
No, all our libraries are linked against a static library. For some reason, the test adds |
That may be a bug in the |
LIBPYTHON is not a sysconfig variable anymore. https://github.com/mesonbuild/meson/blob/18427adbf21909f66a307a54ba4c47fd91e18fba/mesonbuild/scripts/python_info.py#L74C36-L74C45 |
It should be, since we got that change reverted: python/cpython#122842. That commit made it into 1.13.0rc1 it looks like. |
0303ba5
to
74f5c84
Compare
Thanks for the debugging & fix! I've rebased out the debug commits. From my POV this is OK now, but happy to pick up other fixes around libpython if we want to wait for that. |
This should be fine to merge if CI is green; the |
]3.13.0rc1 was released on Jul 31, but the PR is from Aug 8. |
Yes, you're right. I was looking at the wrong commit. Okay, that explains it then. The problem has been fixed and will go away for the final 3.13.0 release. |
This PR has been triggered in an effort to update python313.
Notes and instructions for merging this PR:
Please note that if you close this PR we presume that the feedstock has been rebuilt, so if you are going to perform the rebuild yourself don't close this PR until the your rebuild has been merged.
This package has the following downstream children:
and potentially more.
If this PR was opened in error or needs to be updated please add the
bot-rerun
label to this PR. The bot will close this PR and schedule another one. If you do not have permissions to add this label, you can use the phrase@conda-forge-admin, please rerun bot
in a PR comment to have theconda-forge-admin
add it for you.This PR was created by the regro-cf-autotick-bot. The regro-cf-autotick-bot is a service to automatically track the dependency graph, migrate packages, and propose package version updates for conda-forge. Feel free to drop us a line if there are any issues! This PR was generated by https://github.com/regro/cf-scripts/actions/runs/10607148608 - please use this URL for debugging.