Do we do this kind of stuff for every calls? do we support-fpermissive flags? should we do it? I would prefer having an issue (maybe an old one) for other places that could be affected by this.

The internal headers (pycore_xxx.h) are often not C++ compatible, even though they all haveextern "C" { guards (often in the wrong places).

The change itself seems fine, but I'm curious how this came up. What's includingpycore_cell.h from C++?

To answer@picnixz, we generally do this sort of thing when people ask, especially when the changes are tiny, but we don't proactively make the internal header files C++ compatible.

What's including pycore_cell.h from C++?

This came up while I’m adaptingCinderX to FT-Python: C++ codegen path forLOAD_DEREF/STORE_DEREF ends up includingpycore_cell.h and calling the internal helpers as a good starting point (vs. generating the atomic-swap sequence right away). I initially worked around it with wrappers on our side, but since the header already hasextern "C" guards and the change is tiny, I wanted to check whether it’s okay to make this compile cleanly from C++.

Also agreed thatpycore_* headers aren’t guaranteed to be C++-compatible and may change; in CinderX we already adapt to internal changes per version.

Ok, make sense. Can you open an issue like@picnixz asked? Also edit the PR title to associate it with the issue.

Created the issue#144490

@picnixz,@colesbury: Thank you so much for the review.