stdlib: Add several missing @abstractmethod decorators

[AlexWaygood]

The following methods are all abstract at runtime, but not in the stub:

• typing.ContextManager.__exit__ (and therefore also contextlib.AbstractContextManager.__exit__)
• typing.AsyncContextManager.__aexit__ (and therefore also contextlib.AbstractAsyncContextManager.__aexit__)
• numbers.Complex.__eq__
• numbers.Complex.__abs__
• numbers.Complex.__conjugate__
• os.PathLike.__fspath__

This means that inheriting from any of these classes at runtime without overriding these methods would cause an exception to be raised. Type checkers, however, would not have flagged the possibility of an exception being raised, believing these methods to be concrete.

[AlexWaygood]

This class doesn't seem to be a context manager at runtime, and doesn't seem to have ever been a context manager at runtime.

[AlexWaygood]

Diff from mypy_primer, showing the effect of this PR on open source code:

pandera (https://github.com/pandera-dev/pandera)
- pandera/checks.py:74: error: Argument 1 to "ChainMap" has incompatible type "MappingProxyType[str, Any]"; expected "MutableMapping[str, Callable[..., Any]]"  [arg-type]
- pandera/model.py:362: error: Need type annotation for "attrs" (hint: "attrs: Dict[<type>, <type>] = ...")  [var-annotated]

pydantic (https://github.com/samuelcolvin/pydantic)
- pydantic/class_validators.py:332: error: Need type annotation for "all_attributes"  [var-annotated]
- pydantic/class_validators.py:332: error: Argument 1 to "ChainMap" has incompatible type "*List[MappingProxyType[str, Any]]"; expected "MutableMapping[<nothing>, <nothing>]"  [arg-type]

These are weird -- looks like this crossed the streams with #7423 being merged, maybe?

[JelleZijlstra]

The boostedblob errors are unfortunate. They're in places like https://github.com/hauntsaninja/boostedblob/blob/eb6e95b41ba55c4c595cbe754d05a48f069ce42e/boostedblob/path.py#L328, where request.execute is an async CM. The new annotations make mypy catch on to the fact that async CMs can eat exceptions.

And yeah, not sure what caused the ChainMap changes.

[JelleZijlstra]

Oh wait, I bet mypy special cases -> bool as "may swallow" and assumes that "bool | None" means "won't swallow". So we should probably just change _AsyncGeneratorContextManager.__aexit__ back; sorry for the thrash.

[JelleZijlstra]

And now pyright in CI is complaining about shutil, which you didn't change.

[AlexWaygood]

And now pyright in CI is complaining about shutil, which you didn't change.

Yeah, it's because #7384 made chown unavailable on Windows, but it's still declared in __all__ on Windows. It wasn't picked up by CI running on that PR, because I increased pyright strictness for checking __all__ after that PR was filed.

[JelleZijlstra]

Ah good catch

[github-actions]

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