False negative for restricted self in subclass

This currently passes, but the code is unsafe:

from typing import TypeVar, Generic

T = TypeVar('T')

class Base(Generic[T]):
    def __init__(self, item: T) -> None:
        self.item = item
    def foo(self) -> None:
        pass

class Sub(Base[T]):
    def foo(self: Sub[str]) -> None:
        self.item + 'no'

# This is why the above is unsafe.
a: Base[int] = Sub(0)
a.foo()

On one hand it is a very rare situation, on the other hand in view of making explicit self-types a more "official" feature (see https://github.com/python/mypy/pull/7860) we might want to fix this.

Also it would be great to fix this in a consistent way, so that the fix also works for same issue with multiple inheritance without code duplication (see https://github.com/python/mypy/issues/7724).

Inizia riproducendo l’esempio nell’issue, poi leggi il lavoro esplicito sul self-type in PR 7860 e il caso di ereditarietà multipla nell’issue 7724. La correzione dovrebbe rifiutare coerentemente questo uso non sicuro delle sottoclassi e coprire il comportamento corrispondente dell’ereditarietà multipla senza duplicare la logica.