Run make regen-global-objects

[vstinner]

No description provided.

[erlend-aasland]

When was this regression introduced? There is no linked issue.

[vstinner]

When was this regression introduced? There is no linked issue.

I don't know if it can be qualified as a regression.

It's unclear to me how it is possible to merge a PR if "generated files" are outdated: the GitHub Action job is supposed to fail.

It seems like make regen-global-objects is not deterministic. I don't understand how this issue happened.

[erlend-aasland]

I suspect it was 9205dfe

[vstinner]

I suspect it was 9205dfe

Yep, this change removes _Py_ID(arg) in code generated by Argument Clinic. But the "Check if generated files are up to date" job on the PR succeeded.

The job does:

          # Deepfreeze will usually cause global objects to be added or removed,                    
          # so we run it before regen-global-objects gets rum (in regen-all).                       
          make regen-deepfreeze
          make -j4 regen-all                                                                        
          make regen-stdlib-module-names

where `make regen-alldoes depend onregen-global-objects``. This target is declared as phony: ``.PHONY: regen-global-objects``. So the ``generate_global_objects.py`` script was run.

Maybe depending on the order in which regen-all dependencies are executed, the result is different?

[erlend-aasland]

Maybe depending on the order in which regen-all dependencies are executed, the result is different?

It sounds like a build system bug to me.

[vstinner]

That's the non-deterministic build system bug: #108740