[3.12] gh-105020: Share tp_bases and tp_mro Between Interpreters For All Static Builtin Types (gh-105115)

[miss-islington]

In gh-103912 we added tp_bases and tp_mro to each PyInterpreterState.types.builtins entry. However, doing so ignored the fact that both PyTypeObject fields are public API, and not documented as internal (as opposed to tp_subclasses). We address that here by reverting back to shared objects, making them immortal in the process.
(cherry picked from commit 7be667d)

Co-authored-by: Eric Snow ericsnowcurrently@gmail.com

• Issue: tp_bases of object is NULL: undocumented or unintentional behavior change in 3.12? #105020

[ericsnowcurrently]

@Yhg1s, what do I need to do to fix the ABI check? It's a false positive.

[Yhg1s]

@Yhg1s, what do I need to do to fix the ABI check? It's a false positive.

As the failed check says:

The up to date ABI file should be attached to this build as an artifact.

To learn more about this check: https://devguide.python.org/setup/#regenerate-the-abi-dump

@zooba recently merged support for the CI check to produce the necessary artifact (#105088) but I don't think the devguide has been updated yet. (You may need to update the branch to get the updated check?)

For the record, the failure is not a false positive, it's just a change in the ABI that is okay (because all direct use of the changed struct is internal-only)

[zooba]

The devguide has a mention now, but not step-by-step instructions (which would probably require screenshots of the GitHub UI that would likely go out of date and are probably too big for the doc page).

When the PR check fails, the associated run will have the updated ABI file attached as an artifact. After release manager approval, you can download and add this file into your PR to pass the check.

(Worth pointing out that "the PR check" has already been mentioned in the section, so it's not being newly introduced in the new text.)

Only trick seems to be that the artifacts are listed on the Summary page, not the specific check that published it. Perhaps we can figure out how to infer the URL to the summary page and add a link in that message? In this case, it'd be https://github.com/python/cpython/actions/runs/5127462177?pr=105124#artifacts.

[ericsnowcurrently]

@zooba I was able to use that artifact, which saved me a bunch of work. Thanks!