Victor Stinner

> Py_SETREF and Py_XSETREF are similar to Py_CLEAR. If change the latter, why not change the formers? You're right. I completed my PR to fix also Py_SETREF() and Py_XSETREF() macros....

> I would document it as a change in 3.12. Ok, I documented the Py_CLEAR() change with a "versionchanged". But Py_SETREF() and Py_XSETREF() are not documented. Maybe it's no purpose,...

> I would consider adding a What's New item. Ok, I added a What's New in Python 3.12 entry. I copied the same text everywhere. Would you mind to review...

I renamed Py_SETREF() and Py_XSETREF() arguments, I documented Py_SETREF() and Py_XSETREF(). I fixed a typo in the documentation. @erlend-aasland: Would you mind to review the updated PR? Sorry, I abused...

> Py_SETREF() and Py_XSETREF() are intentionally not documented because they are not in the public API. If I include ````, I get these macros. What do you mean by "not...

Oh, on these 16 projects, 4 projects only define Py_SETREF() and Py_XSETREF() in a copy of the pythoncapi_capi.h header file: * bitarray-2.6.0 * mypy-0.971 * python-snappy-0.6.1 * zstd-1.5.2.5 They don't...

> Doc/c-api/refcounting.rst contains two times an error: If the argument has side effects, there are no longer duplicated. Further down it is correct with these are no longer duplicated. Ah,...

I created a 3.11 backport without documentation changes: https://github.com/python/cpython/pull/99288

> But that will invalidate the versionchanged in the docs...? I fixed the bug silently, but I cannot document that it's the default behavior in Python 3.11.0 since the fix...

Hum. I read again the discussion and now I'm confused. @serhiy-storchaka @erlend-aasland: Should I backport the fix to 3.10 and 3.11 or not? I just backported the change to 3.11...