cc. @pablogsal

Re-opening to trigger CI.

LGTM. But to be able to merge this change, please move the news entry to Doc/whatsnew/3.10.rst. I suggest to backport this nice enhancement to Python 3.10.

Ok, moved to 3.10 in commit 6adc9e5. Thanks for reviewing, Pablo & Victor! 🙏🏻

LGTM. But to be able to merge this change, please move the news entry to Doc/whatsnew/3.10.rst. I suggest to backport this nice enhancement to Python 3.10.

Is this considered a bugfix? We are very close to RC 1 so I'm not sure we should risk regressions...

Is this considered a bugfix?

I guess it depends on how you view it:

1. <some-type> used to be a static type before 3.10; now it's a heap type, and has hence lost the option of adding vectorcall support => bug! :)
2. immutable heap types can now implement the vectorcall protocol => (micro) feature! :)

now it's a heap type

But this is something only happens in the stdlib, and those types are not going to change between beta 4 and RC 1. User types are not going to become magically heap types.

I'm missing something?

now it's a heap type

But this is something only happens in the stdlib, and those types are not going to change between beta 4 and RC 1. User types are not going to become magically heap types.

I'm missing something?

No, that's correct. If vectorcall support is not going to be added to any stdlib type in 3.10, there's no reason to backport this.

Sounds like a backport is out of the question, then.
I'll revert the last commit and remove the label.
I've reverted 6adc9e5 and removed the backport label.

@pablogsal:

Is this considered a bugfix? We are very close to RC 1 so I'm not sure we should risk regressions...

Ah, I didn't notice that there are two months between the RC1 and the final versions. Ok, so this change should not be backported.