General goal here is to be able to add tests that behavior isn't unintentionally changed around some common python code patterns, in particular Path("README.rst").read_text().

This started with the code that validated subprocess used vfork, building more general utilities from that core code. The work was motivated in #120755 for how to test that particular change.

TODO

• Resolve how to fix subprocess._USE_VFORK. this PR currently just comments out that test. See: https://discuss.python.org/t/subprocess-use-vfork-escape-hatch-broken-fix-or-remove/56915/2
• Figure out why the print() of markers needs flush=True to not buffer them. Likely add a test for two print statements not being merged using this infrastructure. -- This is the documented behavior of stdout https://docs.python.org/3/library/sys.html#sys.stdout. Because they're files / not interactive it gets block buffered meaning only once a buffer is full does it flush rather than write() when a print() happens.
• Validate at least once CI bot actually runs and passes the test -- Ubuntu / Build and test found that glibc there uses newfstat which caused the test to fail. Updated the test to allow libc to use other fstat variants.
• Where should test for Path().read_bytes() and .read_text() live?
• Update syscall sequence after gh-120754: Reduce system calls in full-file readall case #120755 is merged -- 0606677

• Issue: Speed up open().read() pattern by reducing the number of system calls #120754