[srittau]

Add Python 3.15 and pypy 3.12 to testing matrix

This PR doesn't yet remove 3.9/3.10-specific code branches.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.46%. Comparing base (cfde842) to head (84a63a7).

@@            Coverage Diff             @@
##             main     #816      +/-   ##
==========================================
- Coverage   97.26%   94.46%   -2.80%     
==========================================
  Files           3        3              
  Lines        7859     7859              
==========================================
- Hits         7644     7424     -220     
- Misses        215      435     +220     

Flag	Coverage Δ
3.10	88.85% <ø> (ø)
3.10.20	88.85% <ø> (?)
3.10.4	?
3.11	88.02% <ø> (ø)
3.11.0	?
3.11.15	88.02% <ø> (?)
3.12	87.96% <ø> (ø)
3.12.0	?
3.12.13	87.96% <ø> (?)
3.13	83.39% <ø> (ø)
3.13.0	?
3.13.13	83.39% <ø> (?)
3.14	79.53% <ø> (ø)
3.14.6	79.53% <ø> (?)
3.15	76.70% <ø> (?)
3.15.0	76.70% <ø> (?)
3.9	?
3.9.12	?
pypy3.10	?
pypy3.11	87.88% <ø> (ø)
pypy3.12	87.81% <ø> (?)
pypy3.9	?

Flags with carried forward coverage won't be shown. Click here to find out more.
see 2 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[JelleZijlstra]

3.10 only just went out of support, I think in previous years we waited a little longer?

3.9 can definitely go though.

[srittau]

I'm not sure there's any necessary reason to keep 3.10 support around, it's just a (small) maintenance burden. There aren't any major features planned to be added soon, but also the previous versions of typing-extensions will be around for people who still want to support 3.10.

[AlexWaygood]

I agree that there's not really any harm in dropping support for 3.10 at this point, but it seems like it's also very easy for us to just pin to an older version of Ubuntu rather than using ubuntu-latest. I don't have a particularly strong opinion but I'd lean towards just doing that for now

[srittau]

but it seems like it's also very easy for us to just pin to an older version of Ubuntu rather than using ubuntu-latest. I don't have a particularly strong opinion but I'd lean towards just doing that for now

Meh, that just postpones the problem, and we have a fix ready now. That means someone has to either remember this PR or write it again in a few months. Also, ubuntu-22.04 either is already or will become unmaintained by GitHub at some point, meaning we won't get any new Python versions for it, and I doubt we will notice that, which means that we will silently test only against outdated Python (patch) versions.

[srittau]

@JelleZijlstra @AlexWaygood To unblock CI: I would prefer to drop 3.10 and keep "ubuntu-latest" for the reasons given. If you insist that we should keep 3.10, I will change the PR accordingly, but I'd like to get this in fairly soon.

[AlexWaygood]

I'm definitely not insisting, I don't particularly feel strongly! Changing the ubuntu pin feels like the easier/safer option to me, but I do agree that continuing to support 3.10 doesn't really buy us much at this point.

[AlexWaygood]

I thought the idea with these was that we were deliberately testing the earliest patch version of each minor version, because we ran into issues in the past where 3.13.0 had a bug where 3.13.1 did not, so we had to make sure that our backports worked with both versions

[srittau]

Yes. The .0 versions are unfortunately not available on Ubuntu 26.04, these are the earliest version. On the other hand, the latest versions will not be available on Ubuntu 24.04 going forward, since that won't be updated.