No error when mutating ParamSpec `**kwargs`

[kmurphy4]

Bug Report

Even though it fails at runtime, mypy doesn't complain if I mutate ParamSpec.kwargs.

To Reproduce

https://mypy-play.net/?mypy=latest&python=3.11&gist=f5cea6637a9ea13c48b8ba75aa8d73dd

If I run this same code in an interpreter, I get

Traceback (most recent call last):
  File "/tmp/repro.py", line 16, in <module>
    print(bar(foo))
  File "/tmp/repro.py", line 13, in bar
    return f(*args, **kwargs)
TypeError: foo() got an unexpected keyword argument 'ka'

Expected Behavior

I would expect this to trigger an error on line 12.

[AlexWaygood]

We could possibly change ParamSpec.kwargs objects so that their upper bound is Mapping[str, object] rather than dict[str, object]. However, note that prior to #12668, previously we received complaints that people couldn't mutate ParamSpec kwargs: #12386. I don't know that there's a foolproof way of catching all unsafe code here without emitting false positives on "safe" transformations of ParamSpec.kwargs objects.

[kmurphy4]

Yeah that makes sense -- the ::pop use-case is pretty idiomatic.

It would be nice if an operation like kwargs["ka"] were able to return something with Concatenate[..., P], but that seems tricky as well: we'd have to return a new object rather than mutate in-place (plus, I'm not sure it's even possible to Concatenate a kwarg...?).

In any case, I think I can work around this on my end. Thanks for the response!

[AlexWaygood]

No worries! Not sure there's anything actionable here without changing the spec, so I'm closing this for now