Since bpo-46541, the global strings are statically allocated so they can now be referenced by deep-frozen modules just like any other singleton. Sharing identifiers with deepfreeze will reduce the duplicated strings hence it would save space.

See https://github.com/faster-cpython/ideas/issues/218
See https://github.com/faster-cpython/ideas/issues/230

I have refactored generate_global_objects.py, and now instead of hard-coding every identifier manually, it now scans *.c files extracts the identifiers used in it and then generate the header file. This has multiple advantages:

- No need to manually add identifiers, as soon as it is used in a c file it is added to the global identifiers struct.
- It simplifies the codegen a lot.
- Remove the need of special casing certain file for checking now it is just a set of identifiers and auto removes unused global strings.

New changeset eb002dbe0da9622245a355db5f0cd5aa2fc70b40 by Kumar Aditya in branch 'main':
bpo-46712: Share global string identifiers in deepfreeze (GH-31261)
https://github.com/python/cpython/commit/eb002dbe0da9622245a355db5f0cd5aa2fc70b40

New changeset 21099fc064c61d59c936a2f6a0db3e07cd5c8de5 by Eric Snow in branch 'main':
bpo-46712: Let generate_global_objects.py Run on Earlier Python Versions (gh-31637)
https://github.com/python/cpython/commit/21099fc064c61d59c936a2f6a0db3e07cd5c8de5

In bpo-47067, there was concern about the addition of the makefile target from PR 31637:

    regen-global-objects: regen-deepfreeze

After a new `&_Py_ID(__orig_class__)` is added to Objects/genericaliasobject.c, running `make regen-global-objects` starts

    gcc -pthread -c [snipped] -DPy_BUILD_CORE -o Objects/genericaliasobject.o Objects/genericaliasobject.c

which fails with a compilation error because that identifier is not yet defined. Is there a good way to convince `make` to regenerate the global objects without this sort of circular dependency? Am I missing a step?

> After a new `&_Py_ID(__orig_class__)` is added to Objects/genericaliasobject.c, running `make regen-global-objects` starts
>
>     gcc -pthread -c [snipped] -DPy_BUILD_CORE -o Objects/genericaliasobject.o Objects/genericaliasobject.c
>
> which fails with a compilation error because that identifier is not yet defined. Is there a good way to convince `make` to regenerate the global objects without this sort of circular dependency? Am I missing a step?

I'm looking into this.  A temporary workaround is to run Tools/scripts/generate-global-objects.py directly.

New changeset febf54bcf3fdc45ad84b4073e24bbaaee0ac8b2a by Eric Snow in branch 'main':
bpo-46712: Do not Regen Deep-Frozen Modules before Generating Global Objects (gh-32061)
https://github.com/python/cpython/commit/febf54bcf3fdc45ad84b4073e24bbaaee0ac8b2a