[NPTH] npth and fork

[NIIBE Yutaka]

On 11/28/2016 06:28 PM, Werner Koch wrote:
> You patch also removed the fix (or well, semi-fix) for AIX.  Was this
> intentional?

It's not intentional for AIX.  My intention is just for macOS.
Nevertheless, this is a good opportunity also addressing AIX.

Executive summary:
I think that the removal is also valid for AIX versions.


I learn the versions of AIX in the Wikipedia:
    https://en.wikipedia.org/wiki/IBM_AIX

I checked IBM Knowledge Center.

    IBM Knowledge Center:
    https://www.ibm.com/support/knowledgecenter/

Current supported versions: 5.3, 6.1, 7.1, 7.2
All has sem_init and it doesn't return ENOSYS.
It seems that it was introduced 5.2.

I found a page saying it was ENOSYS in 4.3.3:

    http://www.verycomputer.com/176_d019c987cbd61c92_1.htm

I don't know about 5.1.

I checked Python bug tracker.  We can see similar issues on AIX
about sem_init.  For example:

    http://bugs.python.org/issue12372


I checked BDW-gc library mailing list:

    http://www.hpl.hp.com/hosted/linux/mail-archives/gc/2011-May/004591.html

It said that it worked on AIX 5.2, 5.3, 6.1 and 7.1.


So, I believe that it (I mean, removal of try_sem_open) works on AIX
5.2, 5.3, 6.1, 7.1 and 7.2.  Note that I can only test on AIX 6.1.


I also find an article:

    Implement POSIX Semaphore APIs using System V Semaphores APIs:
    http://www.ibm.com/developerworks/library/l-semaphore/index.html

It suggests use of System V Semaphores for z/OS.
--