(🐞) [1.6 regression] Can not infer `ParamSpec`

[KotlinIsland]

The function d takes two functions, an original, and a wrapper. The wrapper function needs to take the original function and its parameters.

Mypy is failing to infer the paramspec for the case where the wrapper indicates poisitional parameters. Unfortunately I'm too ignorant of the specifics of ParamSpec to know if this is working correctly or not 😢, but I would think that P would get inferred as [int]

@ilevkivskyi Can you provide any insight for this case?

FYI: This did not show an error in mypy 1.5.1, it inferred P as [i: int].

from collections.abc import Callable
from typing import Concatenate, _P as P

from mypy_extensions import Arg

def o(i: int) -> None:
    print(i)

def f1(c: Callable[[Arg(int, "i")], None], /, i: int) -> None:
    c(i)

def f2(c: Callable[[int], None], /, i: int) -> None:
    c(i)


def d(f: Callable[P, None], fn: Callable[Concatenate[Callable[P, None], P], None]) -> Callable[P, None]:
    def inner(*args: P.args, **kwargs: P.kwargs) -> None:
        fn(f, *args, **kwargs)
    return inner


d(o, f1) # works
d(o, f2)
# main.py:22: error: Cannot infer type argument 1 of "d"  [misc]
# main.py:22: error: Argument 1 to "d" has incompatible type "Callable[[Callable[[int], None], int], None]"; expected "Callable[[Callable[[VarArg(Any), KwArg(Any)], None], VarArg(Any), KwArg(Any)], None]"  [arg-type]

[KotlinIsland]

Actually, I think a simplified version would be:

from collections.abc import Callable
from typing import Concatenate, _P as P

def f(c: Callable[[int], None], /, i: int) -> None:
    c(i)

def d(fn: Callable[Concatenate[Callable[P, None], P], None]) -> None:
    fn

d(f)

mypy 1.5.1: 
no error
mypy 1.6.1: 
main.py:10: error: Cannot infer type argument 1 of "d"  [misc]
main.py:10: error: Argument 1 to "d" has incompatible type "Callable[[Callable[[int], None], int], None]"; expected "Callable[[Callable[[VarArg(Any), KwArg(Any)], None], VarArg(Any), KwArg(Any)], None]"  [arg-type]

[AlexWaygood]

Concentrating on the second example, since it's simpler -- I think this is a valid use of Concatenate (though it took me a little bit of time to get my head around it).

The second error doesn't seem to be reported if you use mypy master, but the first one is still there.

This bisects to 14418bc (cc. @ilevkivskyi).