(Possibly) incorrect handling of type comments for non-variadic function with keyword-only parameters

[Strum355]

(Discovered upon using bazel_env here: https://github.com/buildbuddy-io/bazel_env.bzl/blob/main/bazel_env.bzl#L339)

Given the function def bazel_env(*, name, tools = {}, toolchains = {}, **kwargs): with the type comment # type: (string, dict[string, string | Label], dict[string, string | Label]) -> None, starpls treats the * as being of type string, name as dict[string, string | Label] etc.

From the Starlark spec on fuction signatures:

A non-variadic function may also declare keyword-only parameters, by using a bare in place of the args parameter. This form does not declare a parameter but marks the boundary between the earlier parameters and the keyword-only parameters. This form must be followed by at least one optional parameter.

[withered-magic]

Ah, I ended up just copying the behavior from Pyright here, which emits an error diagnostic if the bare * parameter isn't annotated, e.g.

def foo(*, x):
    # type: (str) -> None
    pass

yields

Parameter annotation count mismatch: expected 2 but received 1

while

def foo(*, x):
    # type: (None, str) -> None
    pass

seems to make Pyright happy.

[withered-magic]

I'll take a look through the pep 484 doc again and see if it says anything about annotating * parameters