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
Dropping support for python 3.7 #11934
Conversation
Wooooh! Walrus operator!
|
Re: walrus operator -- Apparently not everyone was excited about it. See #9606 |
The walrus will just come in slowly, I suspect, as people make commits that use it. Not quite like f-strings, where it is (relatively) easy to search for relevant places (and which were a really nice addition indeed!). |
.circleci/config.yml
Outdated
@@ -6,7 +6,7 @@ jobs: | |||
|
|||
image-tests-mpl302: | |||
docker: | |||
- image: astropy/image-tests-py37-mpl302:1.10 | |||
- image: astropy/image-tests-py38-mpl302:1.10 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
maybe switch directly to py39 instead of 38?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@astrofrog pointed to https://docs.astropy.org/en/latest/development/testguide.html#running-image-tests but someone with server access still needs to generate and upload the new image(s).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
For the record, CircleCI stuff has been resolved over at #12032 . 😸
Just had a look through the changes, and they all look good to me - indeed, rather little is affected. To me, it would seem to make sense to have the matplotlib images be built on 3.9 rather than 3.8, to avoid churn when the minimum version becomes 3.9 and since we test all dependencies with 3.8 elsewhere anyway. |
1973ade
to
7c0fa60
Compare
Weirdly the oldest dependency tests run into issues. I suppose some of the reasons is that old mpl is not providing wheels for python3.8, and something goes somewhat wrong during the build process during CI (the same version seems to be happy with a newer numpy). On the other hand, following the logic of APE18, NEP29, SPEC0 and alike we should really update the mpl min version requirement, and drop mpl 3.0.x, too. The direction that other packages are going is to synchronise and follow SPEC0 above, which has the 2-year cycle. We haven't included any policy in APE18 beyond NumPy and CPython, however, I see no reason to have a more flexible policy for optional dependencies than for our mandatory ones. How does everybody feel about it? cc @astrofrog @Cadair @larrybradley as visualization maintainers |
(btw, locally I couldn't reproduce the issue, so the mpl version may be a red herring) |
Yeah, I don't know what happened there. I saw the same error in CircleCI when I tried to pin both matplotlib and numpy. The only way I know how to fix it is unpin numpy, as you said. I don't mind bumping the minversion but I don't feel like updating the CircleCI truths so soon again... 😬 Or maybe we just don't test with matplotlib in oldest deps Action CI, since CircleCI already does it. |
.github/workflows/ci_cron_weekly.yml
Outdated
|
||
shell: /bin/bash | ||
|
||
install: | | ||
echo "deb http://deb.debian.org/debian buster-backports main" >> /etc/apt/sources.list | ||
echo "deb http://deb.debian.org/debian main" >> /etc/apt/sources.list |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This resulted in malformed sources.list . Do we even need this command now? Not sure.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is there a favourite post-action you use for debugging? E.g. I don't see why line7 on that file should have any issue, etc.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I didn't set this up, so maybe @astrofrog can advise...
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
you're right, based on 0ff2f4f it can be nuked.
@bsipocz I still didn't update wcslib in Debian; it is still 7.4 everywhere (we had a freeze until last week, and I still didn't went through all updates yet). I will do this in the next days (weekend or early next week), if that is OK. My previous experience with wcslib portability is excellent; so I expect to have it everywhere within one step. |
@bsipocz The newest wcslib, 7.7 is now uploaded to Debian (already migrated to testing today). astropy 4.3.1 seems happy with it 👓 |
Thanks for the update! I restarted the affected jobs... 🤞 |
Did I restart the job too early? I still see this:
|
Bullseye comes with 7.4, and will remain so. You need to switch to testing (or bookworm) instead. |
@olebole , what we can use is limited by what is provided at https://github.com/uraimo/run-on-arch-action#supported-platforms . I don't see "testing" or "bookworm". Any advise? |
Hmm, I could backport the new wcslibs to bullseye. This would require that these platforms include the bullseye-backports repository, and one needs to specify the version in apt. |
Yep. that probably would work (with |
@olebole , any update on the status of backport? Thanks! |
@pllim Thank you for the reminder. I uploaded it, and it now needs approval by our ftp-masters. |
@olebole , any timeline on when your patch would be accepted and deployed? Thanks! I am thinking of just merging this PR by the end of today and let the cron job fail until Debian can provide this backport. This is because this PR has been open long enough and I don't want it to have merge conflicts. |
That is hard to predict. However, as their backports queue currently contains entries that are only one day old, I guess it will happen in the next days. Sorry, but the ftp-master require some patience, even from us Debian Developers :-) |
Okay, let's just merge then. I'll open a follow-up issue (#12144). Thanks for the update! 😄 |
EDIT: This had to be broke up into 2 separate PRs to be able to add two changelog entries. See #11935 for dropping numpy
This is in agreement with APE18 and the email thread: https://groups.google.com/g/astropy-dev/c/Z9TPHm8DSmo/m/TgGT6BwFBAAJ?pli=1
Note: the changelog entry is not right, we will need two separate lines, both having the PR number at the end of the line when rendered.
I also expect some of the CI builds to fail, as I may not get the image names right at the first try.
Checklist for package maintainer(s)
This checklist is meant to remind the package maintainer(s) who will review this pull request of some common things to look for. This list is not exhaustive.
Extra CI
label.no-changelog-entry-needed
label.astropy-bot
check might be missing; do not let the green checkmark fool you.backport-X.Y.x
label(s) before merge.