Re: candidate sched.h and sys/sched.h for review

[Eric Blake <eblake at redhat dot com>]

On 03/31/2010 10:31 AM, Ralf Corsepius wrote:
>> sched.h includes sys/types.h (which defines timespec)
>>
> SUSv4 mandates timespec in <time.h> and sys/sched.h to receive it from
> <time.h>.

SUSv4 does not mandate <sys/sched.h>.  Portable apps should not be using
it.  So we can stick whatever we want in it, and <sys/types.h> is more
than welcome to use <sys/sched.h> under the hood.

> 
> The fact that it might be indirectly specified elsewhere is not actually
> of importance.
>> sys/types.h includes sys/sched.h.
> 
> I would not want to call this a proper implementation.

Why not?  Since you can't portable use <sys/sched.h>, you also can't
portably tell whether newlib used it under the hood.

The only portable interface that you should care about is <sched.h>, and
whether it includes everything required by SUSv4, and that it does not
pollute the namespace with anything not listed as optional, when going
for strict _POSIX_C_SOURCE compliance.

-- 
Eric Blake   eblake@redhat.com    +1-801-349-2682
Libvirt virtualization library http://libvirt.org

Attachment: signature.asc
Description: OpenPGP digital signature