Release a version of CPython’s `threading_helper` on pypi

[ngoldbaum]

CPython has a threading_helper that’s part of the internal test module they use for testing CPython itself.

It has a lot of useful helpers for writing multithreaded tests. If we release a version on PyPI that people can add as a test dependency then projects across the ecosystem can make use of it as well.

Technically projects can use it already becsause test is part of CPython but it is not necessarily shipped with the interpreter in all installs and is also not public API.

It would need to keep the PSF license and optimally the original git history as well.

[Carreau]

That might pull more than you wish, while threading_helper is only 10 commits, it – itself – import support, and support verify that it is from the cpython source tree named test.support.

If you bypass all that, then (on main), it imports annotationlib which is 3.14 only, so it seem like a cat and mouse game to me.

How to:

• recursively copy cpython source into a new dir
• The next command will directly edit all branches and git structure directly it will mess up everythings, really do it in a copy.
• git filter-repo --path Lib/test/support/ --force

The final repo is ~610 commits.

I can go back to 3.13 and do the same and push it to a separate repo if you wish

[ngoldbaum]

I don't think it matters too much which python version you extract the code from, although ideally we would start with the most recent version to get all bugfixes that have been done. Unfortunately since the implementation is tightly coupled to CPython and what is present in any given version, that indicates this might be tricky and just extracting the history like you tried to do might not be the best way.

I'd maybe try to see how annotationlib is being used. Is it actually particularly important? Maybe it can be removed?

Or maybe the original plan in this issue won't actually work, and instead we should try to figure out what testing utilities we want and need and then only try to extract the parts we need or even rewrite them without relying on CPython internals.