Overload is not reported as inconsistend when positional or keyword arguments are missing in implementation

[ImogenBits]

In this overload:

@overload
def f(a: int) -> int: ...
@overload
def f(*args: int) -> int: ...

def f(*args: int) -> int: ...

pyright reports no errors. But the first overload claims the function is callable as f(a=1), which fails at runtime. If you make that overload's argument keyword only, pyright does report the implementation as missing the keyword parameter "a".

I think the keyword or positional variant should also be reported since it leads to runtime errors that aren't caught. From what I understand it, (a: str) -> int should also be a supertype of (*, a: str) -> int, so it seems strange that it raises fewer errors here.

[erictraut]

This is addressed in pyright 1.1.364.