Type ignore comment has no effect after argument type comment

[JukkaL]

This program generates an error on the first line even though there is a # type: ignore comment:

def f(x,  # type: x  # type: ignore
      ):
    # type: (...) -> None
    pass

Similarly, this generates an error even though it shouldn't:

def f(x=y,  # type: int  # type: ignore  # Name 'y' not defined
      ):
    # type: (...) -> None
    pass

[emmatyping]

For your first example, here is the AST as parsed by typed_ast:

Module(
    body=[
        FunctionDef(
            lineno=2,
            col_offset=0,
            name='f',
            args=arguments(
                args=[arg(lineno=2, col_offset=6, arg='x', annotation=None, type_comment='x  # type: ignore')],
                vararg=None,
                kwonlyargs=[],
                kw_defaults=[],
                kwarg=None,
                defaults=[],
            ),
            body=[Pass(lineno=5, col_offset=4)],
            decorator_list=[],
            returns=None,
            type_comment='(...) -> None',
        ),
    ],
    type_ignores=[],
)

Note the type_comment='x # type: ignore'.
We generally ignore comments after the first type comment, but it should be easy to make an exception for #\s+type:\s+ignore and add that line to the list of lines to ignore.

[JukkaL]

This happens also with variable type comments:

x = 1 # type: str # type: ignore

[JukkaL]

Increasing priority since it's not at all obvious how to work this around in general.

[emmatyping]

I believe I have hacked together a general implementation in fastparse. I just do a regex search for a # type: ignore when parsing the type comment and keep track of these additional type ignores in the Ast converter class, then add them to the set of ignores when visiting the Module node.

Though this likely should be implemented in typed_ast itself...
Edit: looking over the typed_ast code, it does not seem simple to implement this, maybe putting this in fastparse is ok?