Is variance of List is documented correctly?

[Murtagy]

https://mypy.readthedocs.io/en/stable/generics.html#variance-of-generic-types

Documentation states that lists are invariant, although they act as covariant.

I have reproduced the example of Managers and Employees the docs mentions

class Employee: pass
class Manager(Employee): pass
class Boss(Manager): pass

def salaries(
    staff: list[Manager],
    accountant: Callable[[Manager], int]
) -> list[int]:
    return [accountant(s) for s in staff]

def accountant_M(m: Manager) -> int:
    return 200

And here are the results:

salaries([Boss()], accountant_M)  # valid (most specific)
salaries([Manager()], accountant_M)  # valid (mid specific)
salaries([Employee()], accountant_M)  # invalid (least specific)

If the List was invariant I would expect only middle one to be valid,
Both top and middle are valid, so I guess it acts as covariant

[TeamSpen210]

List is invariant, the reason this works is that type checkers are smart enough to do bidirectional inference to find a solution. What it's doing is casting Boss to Manager (effectively discarding the fact it's a subclass), making the list a list[Manager]. Adding some variable annotations can help demonstrate the behaviour:

bosses: list[Boss] = [Boss()]
salaries(bosses, accountant_M)  # Fails as expected
secret_boss: Manager = Boss()  # Valid cast, it's a subclass.
salaries([secret_boss], accountant_M)  # This works, because it's just an Manager here.

[Murtagy]

Thanks, was trying to settle the terms in my head.

Interesting, wonder wether this is a bug or a feature.
Inline auto casts seem to make sense though for practical purposes.