we don't even compile it because it's conditionally including the c file

Yes, but shouldn't we have some coverage for such code then? at least from the CI perspective

I guess optimally we would. This would require a separate test / build target though where we enforce that the fallback implementation is used.

Edit: If that's something we want I can open a separate issue for that. From my perspective we should:

• add a build where we force HAVE_GETADDRINFO and HAVE_GETNAMEINFO to false
• run tests with it (possibly just the socket module tests to save time)

Merged. Well, it happens with code which is not tested by our CIs.

I would say it's always good to add coverage for something that we use as a fallback implementation. However, I don't know how frequent it is to use the fallback implementation. I don't know how hard it will be to write such tests, but we should at least test that the file compiles (namely, we should at least have a build bot that we can test the implementation on).

These changes are to support PlayStation platform: #114918 (comment)

PlayStation is not supported by PEP 11. I don't think that it's worth it to spend more time on this issue, it's now solved.