Errors for union of callables could be better

Given this definition (playground):

f: Callable[[int, int], None] | Callable[[int, str], None]

this code results in the following errors:

f(5, 42)  # E: Argument 2 has incompatible type "int"; expected "str"  [arg-type]
f(5, '42')  # E: Argument 2 has incompatible type "str"; expected "int"  [arg-type]

Looking at the errors (which are probably emitted in a for-loop) one would miss the big picture, which is that it's a union of two incompatible types.

Perhaps we could "meet" the callables, thus resulting in:

f(5, 42)  # E: Argument 2 has incompatible type "int"; expected <nothing>  [arg-type]
f(5, '42')  # E: Argument 2 has incompatible type "str"; expected <nothing>  [arg-type]

Notably we handle defaults well, i.e. in this code:

class FInt:
    def __call__(self, a: int, b: int=42) -> None:
        return

class FStr:
    def __call__(self, a: int, b: str='42') -> None:
        return

f: FInt | FStr

we get

f(5, 42)  # E: Argument 2 to "__call__" of "FStr" has incompatible type "int"; expected "str"  [arg-type]
f(5, '42')  # E: Argument 2 to "__call__" of "FInt" has incompatible type "str"; expected "int"  [arg-type]

but no error on

f(5)

Comienza con los dos ejemplos enlazados de mypy playground y compara los diagnósticos para uniones de tipos invocables, incluido el caso con argumento predeterminado. Traza la ruta de comprobación de argumentos de los Callables que produce los errores indicados. El trabajo estará terminado cuando los diagnósticos comuniquen que no se pueden satisfacer firmas de Callable incompatibles, preservando al mismo tiempo el comportamiento válido del argumento predeterminado.