Silence modules in site-packages and typeshed

[emmatyping]

Fixes #5260 .

[emmatyping]

Huh, this is strange:

from typing import Sequence
from os import fspath
s: Sequence[str]

s = tuple(map(fspath, s))
reveal_type(s)  # error: Revealed type is '<nothing>'

I suppose the map messes things up.

[ilevkivskyi]

@ethanhs Is this a regression or it was always like this? (There is a chance this is related to a recent overload overhaul.)

[emmatyping]

Is this a regression or it was always like this?

Not sure, I created the repro with https://mypy-play.net, which uses 0.610, I haven't tested it on older versions.

[ilevkivskyi]

No, it is a meet.is_overlapping_types problem which is still quite simplistic, it considers typing.Sequence[builtins.str] and builtins.tuple[builtins.str] non-overlapping (surprise).

The fix should be straightforward, but if you don't have time, just open an issue about this.

[ilevkivskyi]

The code looks good, but I think we should add a test that installs a PEP 561 with various errors (both import related and non-import-related) to be sure they are all silenced.

[ilevkivskyi]

Why is this necessary here? Is broken installed as a PEP 561 package?

[emmatyping]

Since opening this, I also implemented silencing typeshed, which broken is simulated as being part of via alt_lib_path.

[ilevkivskyi]

Do we need the general algorithm here? As I understand we essentially only need something like is_sub_path(p1, p2). Not that I think it is important, but maybe it will allow to simplify code?

[emmatyping]

I was originally hoping to eventually use os.path.commonpath once we drop support for 3.4, but I think you are right it would be better to do this ourselves. I will change this when I have time later.

[emmatyping]

I think we should add a test that installs a PEP 561 with various errors

I will add some errors to typedpkg to test the functionality.

[ilevkivskyi]

I will add some errors to typedpkg to test the functionality.

OK, thanks! Just note that I would like to cut the release branch tomorrow morning. So it would be better to have the tests by that time.

[emmatyping]

Just note that I would like to cut the release branch tomorrow morning. So it would be better to have the tests by that time.

Sure thing, I will do that momentarily.

[ilevkivskyi]

OK, looks good now!