Immortal objects and the C API docs

[JelleZijlstra]

https://docs.python.org/3.12/c-api/none.html#c.Py_None currently says that Py_None "needs to be treated just like any other object with respect to reference counts". That's not really true now that we have PEP-683 immortal objects: there is no need to incref/decref it. There are several similar places, e.g. https://docs.python.org/3.12/c-api/slice.html#ellipsis-object and https://docs.python.org/3.12/c-api/bool.html#c.Py_False.

The docs should be updated to somehow refer to the fact that these objects are immortal. However, extension authors who want to retain support for 3.11 and lower should continue to refcount None etc. correctly, so we shouldn't completely omit discussion of refcounting.

cc @eduardo-elizondo

[eduardo-elizondo]

Oh that's a great point, let me make a note of this to go through the C-API Docs this weekend and update all the relevant singletons.

Debating whether to add a 3.12 note only or if we should change the whole section to reflect immortality and just have a note for < 3.11

[carljm]

Debating whether to add a 3.12 note only or if we should change the whole section to reflect immortality and just have a note for < 3.11

The docs are per-version, so the 3.12 docs should document the state of 3.12. It's good to include a .. versionchanged: 3.12 note, though.

[eduardo-elizondo]

This has now been addressed by @iritkatriel in GH-105195! We can probably close this issue now 🙂

[iritkatriel]

Sorry I didn't notice this issue. My PR is not reviewed/merged yet though.

[eduardo-elizondo]

My PR is not reviewed/merged yet though.

Ah my bad, you are right, I misread. Alright let's see if we can push your PR to resolve this issue

[vstinner]

cc @ZeroIntensity

[ZeroIntensity]

It looks like most of the objects in question have been documented with an immortality note, is there anything else to do here?

[vstinner]

The docs should be updated to somehow refer to the fact that these objects are immortal.

I confirm that the doc has been updated. I close the issue.