Re: mkstemp declaration fix

[Eric Blake]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

According to Bruce Korb on 3/20/2007 10:00 AM:
>> The reason is that mkstemp is declared in <unistd.h>, not <stdlib.h>, on
>> this platform. But for POSIX compatibility, only gnulib's <stdlib.h> should
>> be sufficient to declare the function. I'm applying this:
> 
> Just a question:  is including "unistd.h" the right approach?

Yes, since the inclusion is only done when needed.  And this is the
approach taken in other situations of broken system headers, such as with
the wchar module.

> Not clear to me that pulling in all of unistd.h is the right way
> since that means ``#include <stdlib.h>'' implies ``#include <unistd.h>''.

MacOS is already broken for having to use <unistd.h> instead of <stdlib.h>
to get mkstemp, so it doesn't break things any worse to have gnulib
automatically pull in an extra header in this case.  And since system
headers are supposed to be idempotent, it shouldn't matter if the gnulib
client does any of these three:

#include <stdlib.h>
/* use mkstemp */

#include <unistd.h>
#include <stdlib.h>
/* use mkstemp */

#include <stdlib.h>
#include <unistd.h>
/* use mkstemp */

You are correct that on MacOS, the namespace is a bit more polluted than
necessary on a strict POSIX system, but in practice, that doesn't seem to
be too much of a problem.  Although the standard often allows you to use
names for your own use if you avoid the headers where they are
standardized, you are only asking for problems by doing that.

- --
Don't work too hard, make some time for fun as well!

Eric Blake             address@hidden
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Cygwin)
Comment: Public key at home.comcast.net/~ericblake/eblake.gpg
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGAHzc84KuGfSFAYARAu6RAJ4iKl/L3XsgAZCx2iwuNYcpmz0WDQCgn8zY
jQbZMDnr5/LADl5s5yoJp4M=
=JT7N
-----END PGP SIGNATURE-----