[PATCH] Assume that accept4 is available

[Florian Weimer]

This patch removes conditionals related to accept4 support.  It includes 
a new test case for accept4.

I tested this on x86-64, i386, s390x, s390, aarch64.  I also manually 
compared some object files after cross-compilation; there were no 
changes to accept4.o in the cases I checked (not even on ia64). 
(Comparison is hampered by the fact that the path and executable names 
used by build-many-glibcs.py for binutils are somewhat unpredictable.)

I don't think the minimum kernel version requirement for ia64 is a 
material change.  Too much userspace code unconditionally assumes the 
existence of accept4 already, so ia64 had to be backported to the 3.2 
stable series.

For NaCl and Hurd, my previous comments apply: On NaCl, it is possible 
to emulate accept4 in a race-free manner within glibc (so anyone who 
wants to resurrect the port can implement it).  Hurd needs the Debian 
patches to build, and they contain an implementation of accept4.

Thanks,
Florian
-------------- next part --------------
A non-text attachment was scrubbed...
Name: accept4.patch
Type: text/x-patch
Size: 21261 bytes
Desc: not available
URL: <http://sourceware.org/pipermail/libc-alpha/attachments/20170414/b900cde2/attachment.bin>