Looks like you found a mypy bug: ast.pyi defines its own class called slice, but somehow it's checking @final on builtins.slice?

Looks like you found a mypy bug: ast.pyi defines its own class called slice, but somehow it's checking @final on builtins.slice?

😣

Hmm, the other weird thing is that the mypy tests pass, it's only when running the mypy build in the stubtest invocation does it fail...

hmm hmm hmm, locally:

stubtest --custom-typeshed-dir . ast  # passes
stubtest --custom-typeshed-dir . ast _ast  # passes
stubtest --custom-typeshed-dir . _ast ast  # fails

mypy --python-version 3.10 --no-incremental --custom-typeshed-dir . stdlib/ast.pyi stdlib/_ast.pyi  # passes
mypy --python-version 3.10 --no-incremental --custom-typeshed-dir . stdlib/_ast.pyi stdlib/ast.pyi   # fails

Spent some more time digging through mypy code, it's pretty thorny.

Here's a one character fix:

diff --git a/stdlib/builtins.pyi b/stdlib/builtins.pyi
index b4890072..53575f8f 100644
--- a/stdlib/builtins.pyi
+++ b/stdlib/builtins.pyi
@@ -19,7 +19,7 @@ from _typeshed import (
     SupportsRDivMod,
     SupportsWrite,
 )
-from ast import AST, mod
+from _ast import AST, mod
 from io import BufferedRandom, BufferedReader, BufferedWriter, FileIO, TextIOWrapper
 from types import CodeType, MappingProxyType, TracebackType
 from typing import (

There's really a lot going on here: 1) collision with a name in builtins, 2) builtins depending on ast, 3) ast doing from _ast import *. This removes point 3 from the equation, which helps.

Spent some more time digging through mypy code, it's pretty thorny.

Here's a one character fix:
[...]

Amazing, thanks for looking into it! I'll see if I can edit my patch tomorrow.

Thanks for investigating @hauntsaninja!

Not sure if it will make a difference, but it looks like we can also remove mod from that import. It's only used in a union with AST in an argument type, and _ast.mod is a subclass of _ast.AST.

Huzzah!