"Check if the ABI has changed" is failing, this is what I was worried about. I'll try to think of a solution that doesn't involve new entries on the struct.

My recollection (admittedly vague) is that the ABI dump doesn't try to distinguish internal ABI, so an ABI check failure isn't a definitive no-go, it just means RM should take a look.

If we are concerned that someone might be building with Py_BUILD_CORE and depending on offsets into internal symtable entry structs, an easy change that would eliminate that concern would be to move the new field to the end of the struct in the 3.12 version. That would still require regenerating the ABI dump to pass CI (and RM should still be consulted) but would mean nobody accessing existing fields would be broken.

@Yhg1s what do you think of merging this into 3.12? It fixes two bugs with name mangling around PEP 695 (#119311, #119395), but makes some changes to the internal C API to do so.

Yes, I think it makes sense to backport this to 3.12. The ABI breakage doesn't matter because the structs aren't visible as Carl said, and the new function is fine because it's private and not used by extension modules, just the core itself.

You do need to regen the ABI dump as part of the PR.

I couldn't get the abi regen tool to run locally (I tried the instructions in https://devguide.python.org/getting-started/setup-building/#regenerate-the-abi-dump but got errors related to mpdecimal). But I was able to edit the dump file manually and the check succeeds now.