🐛 Bug Report

I'm not sure this is the issue for typeshed or mypy, but with attrs, unpacking a dict to initialize a subclassed attr instance generates a weird type error or not depending on the position of the double-starred kwargs, where the runtime behavior is completely fine for both cases.

To Reproduce

Run mypy with the following script:

import attr
from typing import List

@attr.s(auto_attribs=True, slots=True)
class Point:
    x: int
    y: int

@attr.s(auto_attribs=True, slots=True)
class Point3D(Point):
    z: int
    opts: List[str]

p = Point(x=1, y=2)
q = Point3D(
    **attr.asdict(p),
    z=3,
    opts=['a', 'b'],
)

Expected Behavior

There should be no type errors. (Or at least, some different error saying that it cannot validate keys from an arbitrary dict..)

Actual Behavior

When ** comes first:

p = Point(x=1, y=2)
q = Point3D(  # <-- error: "Point3D" gets multiple values for keyword argument "opts"
    **attr.asdict(p),
    z=3,
    opts=['a', 'b'],
)

When ** comes last:

p = Point(x=1, y=2)
q = Point3D(  # no errors!
    z=3,
    opts=['a', 'b'],
    **attr.asdict(p),
)

where the runtime results are same.

Q: Are there other ways to initialize a subclassed attr instance from a super-attr instance with its values, instead of using **attr.asdict()?

Your Environment

• Mypy version used: 0.783
• Mypy command-line flags: python -m mypy test-attr.py
• Mypy configuration options from mypy.ini (and other config files):

[mypy]
ignore_missing_imports = true
namespace_packages = true

• Python version used: 3.8.5
• Operating system and version: Ubuntu 18.04