[PATCH] Assume that SOCK_CLOEXEC is available and works

[Florian Weimer]

* Mike Frysinger:

> On 13 Oct 2015 14:27, Florian Weimer wrote:
>> This fixes (harmless, I assume) data races when accessing the various
>> __have_sock_cloexec variables.
>> 
>> Compiled and tested on x86_64-redhat-linux-gnu.  According to the
>> previous thread (“O_CLOEXEC and SOCK_CLOEXEC“), there might be some
>> impact of this change on Hurd and NaCl, but it seems manageable.
>
> looks good, but shouldn't we also drop __ASSUME_O_CLOEXEC ?

Thanks.  Yes, __ASSUME_O_CLOEXEC is next, but it seemed it would be a
much larger patch, so I wanted to test the waters first.