I expect you'll want to backport this to 3.12?

A different fix is to always implement Py_INCREF() and Py_DECREF() as opaque function calls when Py_DEBUG macro is defined. It may be safer in terms of stable ABI, no?

A different fix is to always implement Py_INCREF() and Py_DECREF() as opaque function calls when Py_DEBUG macro is defined. It may be safer in terms of stable ABI, no?

I created PR #104763 to implement this idea.

Thanks @ericsnowcurrently for the PR 🌮🎉.. I'm working now to backport this PR to: 3.12.
🐍🍒⛏🤖

Sorry, @ericsnowcurrently, I could not cleanly backport this to 3.12 due to a conflict.
Please backport using cherry_picker on command line.
cherry_picker 26e7bbf66e93ee7c94b6e007ec7b2d769c2ced92 3.12

Hmm, I had not meant to merge this before checking with @vstinner about gh-104763. I'm not sure how that happened accidentally. 😟

That said, I was already leaning toward merging this and then letting gh-104763 roll it back if applicable. The most important thing for me is that we get the per-interpreter reftotal working for 3.12+. So I'll go ahead with the backport at this point.

GH-105123 is a backport of this pull request to the 3.12 branch.

I'm not convinced that _DO_NOT_USE_THIS suffix is really needed. We have tons of private functions with _Py prefix that must not be called. If people use them, they are already on their own. For me, this long _DO_NOT_USE_THIS suffix is just a burden, for example if you have to manually type the full name in a debugger or a profiler to look for them.

I'm fine with changing the name, but I'd rather gh-104763 make it irrelevant. 😄