issue40192 introduced the use of nanosecond reporting of time.

The new routine called is not available in AIX 5.3:

        xlc_r  -O -I/opt/include -O2 -qmaxmem=-1 -qarch=pwr5 -O    -I../git/python-3.9/Include/internal -IObjects -IInclude -IPython -I. -I../git/python-3.9/Include -I/opt/include   -DPy_BUILD_CORE_BUILTIN  -DPy_BUILD_CORE_BUILTIN -I../git/python-3.9/Include/internal -c ../git/python-3.9/Modules/timemodule.c -o Modules/timemodule.o
"../git/python-3.9/Modules/timemodule.c", line 1358.22: 1506-275 (S) Unexpected text tc encountered.
"../git/python-3.9/Modules/timemodule.c", line 1358.5: 1506-045 (S) Undeclared identifier thread_cputime_t.
"../git/python-3.9/Modules/timemodule.c", line 1359.29: 1506-045 (S) Undeclared identifier tc.
"../git/python-3.9/Modules/timemodule.c", line 1861.19: 1506-196 (W) Initialization between types "void*" and "int(*)(struct _object*)" is not allowed.
make: 1254-004 The error code from the last command is 1.

Does AIX 5.3 officially supported by IBM? As far as I can locate, it is deprecated long time ago.

My source: https://www.ibm.com/support/pages/aix-support-lifecycle-information

No, it is not supported - asin there are no new patches - but there are
organizations that use it.

imho - those organizations are not likely to be using the latest Python
- but I have been packaging Python on AIX 5.3 and AIX 6.1 - as these
packages run unmodified on AIX 7.1 and AIX 7.2 (AIX binary compatibility).

FYI.

On 19/05/2020 13:57, Batuhan Taskaya wrote:
> Batuhan Taskaya <isidentical@gmail.com> added the comment:
>
> Does AIX 5.3 officially supported by IBM? As far as I can locate, it is deprecated long time ago.
>
> ----------
> nosy: +BTaskaya
>
> _______________________________________
> Python tracker <report@bugs.python.org>
> <https://bugs.python.org/issue40680>
> _______________________________________
>

@vstinner what are your thoughts on this? I can propose a patch that introduces a build value to generate this conditionally depending on if the system has it or not. But I am not sure about if our motivation behind this is going to be supporting (very) old versions of AIX.

AIX 5 was released 16 years ago, do you we really care?

According to https://www.ibm.com/support/pages/aix-support-lifecycle-information I suggest to require AIX 7.1 and newer.

There is still a lot of AIX 6.1 out there, and iirc, there may be
"extended" support available.

However, at this point in time the bots run (most of the time) on AIX
7.2, -- occasionally, on AIX 7.1. And when I (personally) test - it is
usually on AIX 6.1 (e.g., to have _uuid support).

In any case, I still package using AIX 6.1 (again, _uuid mod support)
and rely on AIX binary compatibility for installations on AIX 7.1 and
AIX 7.2.

Should you be thinking/considering moving AIX to the "supported" column
- I would "gladly" embrace AIX 7.1 as a requirement.

On 10/06/2020 22:58, STINNER Victor wrote:
> STINNER Victor <vstinner@python.org> added the comment:
>
> According to https://www.ibm.com/support/pages/aix-support-lifecycle-information I suggest to require AIX 7.1 and newer.
>
> ----------
>
> _______________________________________
> Python tracker <report@bugs.python.org>
> <https://bugs.python.org/issue40680>
> _______________________________________
>

> There is still a lot of AIX 6.1 out there, and iirc, there may be "extended" support available.

This issue is about AIX5 support. I'm fine with providing best effort support for AIX 6.1. If supporting AIX 6 becomes too expensive, we can consider to drop support for this version.

Anyway, I close this issue: I don't think that it's worth it to attempt to support AIX 5 in Python 3.10: AIX 5 is no longer supported by its vendor (IBM). Use Python 3.9 which seems to better support AIX 5 (don't need thread_cputime()).

Does this need an update to PEP 11 documenting the drop of support?

Also, I can submit changes to remove pre-AIX 6 dead code paths. Should I open a separate issue for that or reference this one?

> Also, I can submit changes to remove pre-AIX 6 dead code paths. Should I open a separate issue for that or reference this one?

Yes, please do so.

Ok, I've opened https://bugs.python.org/issue42087