Don't accidentally ignore typeguard validity for staticmethods

[A5rocks]

I was checking through my previous contributions and noticed I messed up the logic here. I added a test case that currently is wrong.

Followup to #14238

I haven't run this on my own machine (it's a one line change....) but presumably CI will pass!

[A5rocks]

Apologies for the commit spam, I was trying out the github.dev editor (which doesn't actually run anything...) as I thought this would be a trivial change xd

[A5rocks]

:/

Why does the line referenced depend on Python version...

[AlexWaygood]

:/

Why does the line referenced depend on Python version...

Possibly because we use typed-ast on 3.7 for parsing Python source files, but the stdlib ast module on 3.8? (Just a hunch, I haven't verified that that's the cause.) It could also be a change in the grammar between the two Python versions.

[A5rocks]

Huh, didn't realize we parsed with different parsers like that. I'm not sure why Python nightly passes though.

[AlexWaygood]

Huh, didn't realize we parsed with different parsers like that.

Yeah, prior to 3.8, type comments weren't parsed by the stdlib ast module, but on 3.8+ they are.

I'm not sure why Python nightly passes though.

We mark the nightly run as a success even if the tests fail ;)

mypy/.github/workflows/test.yml

Line 156 in 764ff6f

- name: Mark as a success

If you click the "details" button, you'll see that many tests currently fail on 3.12. I presume most of these are due to mypyc incompatibilities with changes made to the CPython interpreter that have yet to be released (haven't looked at them in depth/at all).

[hauntsaninja]

Thank you! :-)

[github-actions]

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉