gh-83714: Fix compiler warning in stat_nanosecond_timestamp()

[vstinner]

Disable the fast path on systems with 32-bit long.

• Issue: Add os.statx() function #83714

[vstinner]

Fix the warnings:

./Modules/posixmodule.c:2638:31: warning: comparison is always true due to limited range of data type [-Wtype-limits]
./Modules/posixmodule.c:2638:45: warning: comparison is always true due to limited range of data type [-Wtype-limits]

cc @cmaloney @jbosboom

[vstinner]

!buildbot x86 Debian

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @vstinner for commit c86068e 🤖

Results will be shown at:

https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F141043%2Fmerge

The command will test the builders whose names match following regular expression: x86 Debian

The builders matched are:

• x86 Debian Installed with X PR
• x86 Debian Non-Debug with X PR

[vstinner]

https://buildbot.python.org/#/builders/1285/builds/530 confirms that the compiler warnings on Modules/posixmodule.c are gone.

[vstinner]

On 32-bit, the fast path is not interesting, it only covers seconds in the range [-3; 2]:

>>> -2**31 // 1_000_000_000
-3
>>> (2**31-1) // 1_000_000_000
2

[serhiy-storchaka]

I do not think this is correct. The comparison is always true, not always false if SIZEOF_LONG < 8.

[vstinner]

Would you mind to elaborate? I'm not sure that I understand your remark. What do you suggest?

[serhiy-storchaka]

Previously, if SIZEOF_LONG < 8, the fast path PyLong_FromLongLong(sec * SEC_TO_NS + nsec) was always used. Now it is never used.

[vstinner]

I wrote a different fix: #141069

[vstinner]

I think that I now understood my mistake. Sorry about that :-( I read LONG_MIN (32-bit) instead of LLONG_MIN (64-bit).