From the mypy primer output it seems like you need to exclude class definitions in stub files from your check.

I think I fixed the issue. Could someone rerun the workflows?

I'll see if I can set up mypy primer locally to get a better idea of what's going wrong.

I tested the changes locally on yarl and discord.py and it was not throwing the errors anymore.

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

If the mistake is in a third-party library, does it still show an error in my own code? So, for example:

some_library/__init__.py:

from dataclasses import dataclass, field
from typing import Final

@dataclass
class C:
    x: Final[int] = field(init=False)

    def __post_init__(self) -> None:
        self.y = 3  # typo here

my_code.py:

from some_library import C

C().x  # an error here?

I'm a bit confused about the self.y = 3 #typo here line. Shouldn't it be okay to assign a final field in the __post_init__ method?

I'm silly and didn't read the difference between x and y.

I do understand your main point here though, and I'll take a look at how my code handles errors in imported libraries.

It does show errors from third-party libraries, I assume we want to suppress those?

It does show errors from third-party libraries, I assume we want to suppress those?

Nevermind actually, I realized that a lot of other mypy checks have the same problem. So it should be fine!

This PR is going in the right direction, but it feels to me like the implementation is too complicated, especially considering this is a relatively minor edge case. In particular, iterating over the body of a function in the dataclass plugin is not the best way to approach this, and it seems to duplicate functionality in the semantic analyzer which will make things harder to maintain and understand. Also I'd prefer it checkmember.py wouldn't need to be changed.