All commit authors signed the Contributor License Agreement.

Could this be performance checked against the existingstat? Stat is called a lot (in Python 3.14 on linux once for every file opened which includes imports by default) and my worry is issues likehttps://lwn.net/Articles/944214/ where the different variants ofstat have quite a bit of performance delta.

Another take on this: How about not modifyingos.stat behavior at all but instead introduce anos.statx API that uses the more modern system call when available and exposes its functionality - in particular the mask functionality so that people can call it and get only the specific data they want instead of a large everything-bagel os.stat result.

Theos module is all about providing wrappers around the underlying platform APIs.
(edit: raised this question on the issue)

Another take on this: How about not modifying os.stat behavior at all but instead introduce an os.statx API that uses the more modern system call when available and exposes its functionality - in particular the mask functionality so that people can call it and get only the specific data they want instead of a large everything-bagel os.stat result.

os.statx() has been added. This PR is now outdated, I close it. There is a discussion on using os.statx() in pathlib.

Thanks@ntninja for your contribution even if it was taken as it is. Exposing new syscalls is non trivial, it takes time to decide how to expose it in Python.