Well, 3.14 is some way off, so we've not looked at it yet. Feel free to make a PR to add a fix (you can probably add 3.14 to the test matrix as well, marking it experimental like: https://github.com/aio-libs/aiohttp/blob/9482755b40ce69814eea6f65bfbd4c12d1878174/.github/workflows/ci-cd.yml#L158-L161).

Looks like it needs updating to drop 3.8 and add 3.13 as well..

I agree that there is plenty of time. The problem, IIUC, is this:

Because of that setting, DeprecationWarnings are not ignored, which causes the tests to fail.

Yes, which ensures we update them correctly as soon as we add 3.14 to the tests, thus ensuring warnings are not propagated to users. As above, feel free to help up and start those updates.

I've proposed a fix for the warning but I'm not gonna fix the testing/CI because:

• tox config seems to be broken for some time already - the deps in requirements.txt does not contain flake but it's used as a command there
• the dependency on isort is not specified anywhere
• I don't understand the difference between requirements and requirements-dev - they both contain dev dependencies IMHO
• flake8 reports failures that are present there for years apparently

I mean, I can help fix those issues but some decisions have to be made on how you'd like the new CI/testing config to look like. I mean, you have tox, pre_commit, and Makefile and I think this can be simplified.

* tox config seems to be broken for some time already - the deps in requirements.txt does not contain flake but it's used as a command there

I don't think tox has been used since moving to Github Actions. It's all done in the Github runners now.

* I don't understand the difference between requirements and requirements-dev - they both contain dev dependencies IMHO

Yeah, we need to rename them (I'm hoping to work on an org-wide solution in future and make this more consistent across repos). One is for the test job (and needs to work on all Python versions), the other is for the linter job (a superset of dependencies that only needs to work on a specific version of Python).

* flake8 reports failures that are present there for years apparently

I think both isort/flake8 are being run via pre-commit (and the dependencies are specified there). If you get extra errors, that's most likely because you've not used a fresh env and extra plugins installed that we're not using (my biggest annoyance with flake8).

#638 should do it.