Fix runtime behaviour of PEP 696

[NCPlayz]

In PEP 696 there are two failure modes specified:

DefaultStrT=TypeVar("DefaultStrT", default=str)
DefaultIntT=TypeVar("DefaultIntT", default=int)
DefaultBoolT=TypeVar("DefaultBoolT", default=bool)
T=TypeVar("T")
T2=TypeVar("T2")
# 1. Invalid: non-default TypeVars cannot follow ones with defaultsclassNonDefaultFollowsDefault(Generic[DefaultStrT, T]): ...
classAllTheDefaults(Generic[T1, T2, DefaultStrT, DefaultIntT, DefaultBoolT]): ...
# 2. Invalid: expected 2 arguments to AllTheDefaultsAllTheDefaults[int]

This PR aims to implement the runtime behaviour for this PEP that was missed out in the initial support PR #77.
It also brings support for this to Python 3.11+.

One note I had was that the current error will say Too few parameters for AllTheDefaults; actual 1, expected 5.
For TypeVarLikes with defaults they aren't actually required, so is it worth computing the minimum required then saying expected at least 2 for this case?

Let me know if you have any suggestions!

[ghost]

All commit authors signed the Contributor License Agreement.

[JelleZijlstra]

Thanks! cc @Gobot1234

[cdce8p]

What's the status of this one? With basic / experimental support for PEP 696 coming in the next mypy release this is basically necessary to use TypeVar defaults.

[JelleZijlstra]

It would be good to get more review (if @Gobot1234 or you could review, I'd appreciate it). There's also a CI failure.

[Gobot1234]

I don't have the necessary internal knowledge of how typing.py works anymore really to be able to review this properly

[cdce8p]

It would be good to get more review (if @Gobot1234 or you could review, I'd appreciate it).

👍🏻 I'll look at it tomorrow.

[cdce8p]

Left a few comments. Most of them are just some minor style ones.
Overall this looks good to me! Thanks @NCPlayz 👍🏻

--
@JelleZijlstra I left suggestions everywhere I could. If you like feel free to merge them. Would be awesome to get this released soon.

[NCPlayz]

Thanks for the reviews! Fixed a minor bug but everything seems to be passing.

[JelleZijlstra]

Thanks. We need to be extra careful in the cases where we monkeypatch CPython, so I left some comments there.

I'll look into adding a test to this repo that runs the test_typing suite with the typing_extensions monkeypatches enabled.

[JelleZijlstra]

I'd like to make sure this still passes tests after we merge #353.

[JelleZijlstra]

Ci didn't run

[JelleZijlstra]

Ci didn't run

[JelleZijlstra]

I may have made CI not run any more somehow, will investigate.

[NCPlayz]

Oh.. just noticed that some changes to unit tests might have broken in one of the merge commits -- or I misunderstood whether the merge commits were meant to change those?

[JelleZijlstra]

What specifically is broken? Feel free to send another PR with any fixes.

[NCPlayz]

Seems like these lines got re-added -- but maybe they were breaking for 3.8/3.9?

[JelleZijlstra]

I think so, yes. I see you commented on this in #293 (comment) before; I don't feel too strongly but I think it's best to match CPython's behavior.