gh-123681: Check the strftime() behavior at runtime instead of at the compile time

[serhiy-storchaka]

It is needed to support cross-compiling.
Remove macros Py_NORMALIZE_CENTURY and Py_STRFTIME_C99_SUPPORT.

• Issue: test_strftime_y2k fails on embedded Linux #123681

[erlend-aasland]

This PR is not needed anymore. Please see the issue history.

[serhiy-storchaka]

No, it is needed to support cross-compiling. You cannot check the runtime behavior at compile time when cross-compiling.

[erlend-aasland]

No, it is needed to support cross-compiling. You cannot check the runtime behavior at compile time when cross-compiling.

There is no need; we require C99 compliant strftime.

[serhiy-storchaka]

It is not C99 compliant (or at least not Python documentation compliant) on Linux. This is why this workaround was needed at first place.

[erlend-aasland]

It is not C99 compliant (or at least not Python documentation compliant) on Linux. This is why this workaround was needed at first place.

If that was the case, all our Linux build bots and Linux CI would now be broken. However it is not.

[serhiy-storchaka]

I wonder why? I guess because ac_cv_normalize_century=yes if cross-compiling.

[erlend-aasland]

I wonder why?

Because it is C99 compliant, presumably.

I guess because ac_cv_normalize_century=yes if cross-compiling.

ac_cv_normalize_century is not used. It is also set to the empty string, not yes if cross-compiling.

[serhiy-storchaka]

It is used:

cpython/configure.ac

Line 6669 in fa985be

if test "$ac_cv_normalize_century" = yes

And it is set to "yes":

cpython/configure.ac

Line 6668 in fa985be

[ac_cv_normalize_century=yes])])

[erlend-aasland]

Sorry, I misread it as ac_cv_strftime_c99_support.

[picnixz]

What's the status of this PR?

[picnixz]

@serhiy-storchaka apparently, the JIT tests for test_strftime_y2k don't fail anymore; should we do something here?