Unable to add Python 3.12 to CI due to dependency bugs

[supakeen]

For the past few months we've been unable to enable Python 3.12 in our CI due to an outstanding autopep8 formatting bug: hhatto/autopep8#712

We need to find a way forward as we do support Python 3.12 and we're even getting relatively close to Python 3.13. The options we have:

1. Drop autopep8 from the linters.
2. Use a different linter from autopep8.

For the second one I'd prefer not to revisit arguments made during previous discussions assuming they all still stand and which are the reason to use autopep8. So I'd like to go with option 1 instead, unless people want to re-evaluate using ruff for both formatting and linting.

@achilleas-k @ondrejbudai @thozza @mvo5

[mvo5]

The third option would be to fix autopep8, it's strange issues, afaict stock autopep8 is fine, it only acts strongly inside a virtualenv (like the one that tox creates automatically). So we have a fourth option (not a good one, mind you): run autopep8 in a container with python3.12 and without virtualenv.

But if ruff is close enough to the autopep8 output that might be the most pragmatic way forward. I would prefer not to have too many source file re-formated in the process though, I find it makes git blame harder to follow.

[supakeen]

@mvo5 For your second concern, we have a .git-blame-ignore-revs file that we can add any auto formatting commit to which should then be ignored by git blame (though I'm unsure if this is configuration or default behavior).

As far as your first idea, yes it'd be nice to fix up autopep8 however I think the virtual environment behavior is a red herring. The other reports had different versions of Python inside their virtual environments and I have yet to theorize a reason as to why a package would act differently given all versions the same in a virtual environment.

[supakeen]

autopep8 was fixed in #1661.