bpo-44392: Add Py_GenericAlias to C API docs

[Fidget-Spinner]

• Added docs for Py_GenericAlias and Py_GenericAliasType.
• Fixed the type for Py_GenericAliasType so the docs can compile. That's a separate PR now at bpo-44392: Fix exported type of Py_GenericAliasType in stable ABI #26739 which needs to be merged first.

https://bugs.python.org/issue44392

[Fidget-Spinner]

BTW, I also noticed the stable ABI manifest exposes the wrong return type for Py_GenericAliasType. It exposes it as a function when it's a var/type/struct. So I fixed that in the PR too otherwise the docs weren't compiling. But that means the exported symbol for Windows is also modified.

[Fidget-Spinner]

BTW, I also noticed the stable ABI manifest exposes the wrong return type for Py_GenericAliasType. It exposes it as a function when it's a var/type/struct. So I fixed that in the PR too otherwise the docs weren't compiling. But that means the exported symbol for Windows is also modified.

@encukou, is this alright with you and can I backport this? I was wondering if such a change to the stable ABI (for a bugfix) is okay.

[encukou]

Yes, fine with me, as long as it's backported to 3.10. This is a documentation change, except the function is not marked as DATA for the MSVC export. I'm not sure of the exact ABI impact of that, but it was clearly wrong before and we fixed things like this for 3.10.

cc @pablogsal: another ABI-related change for 3.10.

[Fidget-Spinner]

Yes, fine with me, as long as it's backported to 3.10. This is a documentation change, except the function is not marked as DATA for the MSVC export. I'm not sure of the exact ABI impact of that, but it was clearly wrong before and we fixed things like this for 3.10.

cc @pablogsal: another ABI-related change for 3.10.

Great, thanks for your advice. I created a separate PR GH-26739 to fix the stable ABI errors and make backporting this docfix PR easier (it needs to land in 3.9 too). That will need to be merged first before this PR.

[pablogsal]

cc @pablogsal: another ABI-related change for 3.10.

As we are before beta 4, this is still ok to me, but please, backport this before Thursday so it goes into beta 3.

[gvanrossum]

Thanks! Here are some suggestions. What do you think? Otherwise this looks good!

[Fidget-Spinner]

Thanks! Here are some suggestions. What do you think? Otherwise this looks good!

Thanks for the review and suggestions. I think they make sense so they're added.

[miss-islington]

Thanks @Fidget-Spinner for the PR, and @gvanrossum for merging it 🌮🎉.. I'm working now to backport this PR to: 3.10.
🐍🍒⛏🤖 I'm not a witch! I'm not a witch!

[bedevere-bot]

GH-26756 is a backport of this pull request to the 3.10 branch.

[gvanrossum]

I don't know why the Address Sanitizer test is failing but I don't think it was your code, so I've merged regardless. Now on to the backport(s).

[Fidget-Spinner]

I don't know why the Address Sanitizer test is failing but I don't think it was your code, so I've merged regardless. Now on to the backport(s).

Yeah I noticed it passed at first, then failed after I added a blurb entry, which probably means it was something else.

This doesn't affect much, but I initially split off the C API changes into GH-26739 which was supposed to be merged first (along with a separate news entry). However, since you already merged this PR, we don't need that anymore.

Thanks for the merge, and I'll work on the docs only backport to 3.9 :). This needs some manual work.