Alexander Belopolsky <alexander.belopol...@gmail.com> added the comment:
I think I finally understand what is going on and ready to make what I believe
is a reasonable proposal.
Setting _DARWIN_C_SOURCE (or _DARWIN_UNLIMITED_GETGROUPS) replaces getgroups
with a Darwin extension, but does not change setgroups:
$ nm tg1 | grep etgroups
U _getgroups
U _setgroups
$ nm tg2 | grep etgroups
U _getgroups$DARWIN_EXTSN
U _setgroups
(here tg2 is compiled with -D_DARWIN_C_SOURCE=1)
The extended getgroups$DARWIN_EXTSN is not getgroups at all and instead is
effectively a call to getgrouplist(getpwuid(..)->pw_name, ..). In other words,
it reads the system database instead of the per-process list. (This also
explains the choice made in Apple patch.)
This means that _DARWIN_C_SOURCE is not appropriate for posixmodule.c.
My recommendation is now to close this bug by reverting _DARWIN_C_SOURCE
setting locally in posixmodule.c or globally in pyconfig.h in favor of more
targeted settings in the modules that need Darwin extensions.
This will leave a feature request to add os.getgrouplist() for users who
actually want to read the system database instead of per-process group list.
The other ideas related to allocation of grouplist buffer can be left as a
"resource usage" issue.
Lack of support for more than 16 groups in OSX get/setgroups() should be left
for Apple to take care of. This limitation can be documented possibly as a
conditional appendix to get/setgroups() docstrings.
----------
_______________________________________
Python tracker <rep...@bugs.python.org>
<http://bugs.python.org/issue7900>
_______________________________________
_______________________________________________
Python-bugs-list mailing list
Unsubscribe:
http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
