bpo-43224: Implement PEP 646 changes to genericaliasobject.c

[mrahtz]

These are the parts of the old PR (#30398) relevant to starring tuples. This allows us to do things like *args: *tuple[int, *Ts] (where Ts is a TypeVarTuple).

@gvanrossum Pradeep and I were wondering what the best way of representing a starred tuple would be, given that we'll also need a representation of a starred typing.Tuple for backport purposes. Should we try and make both *tuple[int] and *Tuple[int] refer to the same thing (I guess something we'd add to typing.py), or is it ok to have them refer to different things (as this PR currently assumes, creating a new native type for *tuple)?

Until the PR with the grammar changes has been merged (#31018), the tests for this look a little awkward. Happy to wait until that PR has been merged to be 100% sure this is correct.

@Fidget-Spinner Would you be happy to review this?

(Also, could someone add the 'skip news' label? The news item is being added in the grammar PR.) Jelle suggests creating a separate news item for each PR, so I've added an item for this one.

https://bugs.python.org/issue43224

[JelleZijlstra]

Thanks for splitting up the PR, I think this will be easier to review.

I think you should add a separate news entry for each PR, something like "Allow unpacking GenericAlias objects. Part of PEP 646."

[mrahtz]

@pablogsal Might be good to have your eyes on this too, to make sure we get GC correct :)

[mrahtz]

Apologies for the slow reply - it's been a busy couple of weeks! Taking a look at this feedback now.

[mrahtz]

@Fidget-Spinner Done.

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @JelleZijlstra for commit 1102d27 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

[JelleZijlstra]

Now it's failing tests because of #31801, sorry for that! Once that's merged you'll have to merge main into your branch again.

[JelleZijlstra]

Main is unbroken now, but I'll see if the current refleak builds say anything; maybe they still tell us something useful even if test_asyncio fails.

[mrahtz]

That was fast! I've re-merged.

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @JelleZijlstra for commit eda60f9 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

[JelleZijlstra]

The refleak buildbots are still unhappy

[JelleZijlstra]

Seems like there's a refleak on main. I'll try to find it.

[Fidget-Spinner]

@mrahtz I just merged the fix for the refleak into main. Can I trouble you to merge main into this PR again please?

[mrahtz]

@Fidget-Spinner Wow, that was also fast :) Done.

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @AlexWaygood for commit 5fe058e 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

[AlexWaygood]

Looks like the buildbots are finally happy! 🎉

[Fidget-Spinner]

I hereby use the non-existent powers vested in me to press the big green button. Thank you Mathew. This was admittedly very tricky.

I will leave this for at least 24 hours for anyone else to give their feedback before I merge.

[Fidget-Spinner]

Congrats @mrahtz !

[mrahtz]

Fantastic - thanks @Fidget-Spinner!