Fix fenv.h fallback returns (bug 14027)

[Andreas Jaeger]

On 04/27/2012 03:23 PM, Joseph S. Myers wrote:
> C99(+TC1), and C11, explicitly say that feclearexcept, feraiseexcept
> and fesetexceptflag return zero when their "excepts" operand is zero,
> because nothing needs to be done in that case.
>
> I propose this patch to fix the fallback implementations - the ones
> for when no exceptions are supported, when no target-specific
> implementation is used - to do so; to conform to the C standard
> requirements, these operations should always be considered supported
> and successful when applied to the empty set of exceptions, even when
> no exceptions are supported.  Sanity checked with a soft-float MIPS
> build.
>
> I submitted this last year
> <http://sourceware.org/ml/libc-alpha/2011-01/msg00005.html>  but it was
> rejected based on an apparent confusion between ISO C and POSIX error
> handling rules.  The notion of implementations defining additional
> error conditions is a POSIX one; here, the semantics are fully defined
> by ISO C ("The feclearexcept function returns zero if the excepts
> argument is zero or if all the specified exceptions were successfully
> cleared.", for example) and it's not valid for the functions to fail
> with a zero argument any more than it's valid for memcpy to fail with
> valid arguments for some implementation-defined reason.
>
> 2012-04-27  Joseph Myers<joseph@codesourcery.com>
>
> 	[BZ #14027]
> 	* math/fclrexcpt.c (__feclearexcept): Return zero if nothing needs
> 	to be done.
> 	* math/fraiseexcpt.c (__feraiseexcept): Likewise.
> 	* math/fsetexcptflg.c (__fesetexceptflag): Likewise.

Thanks, this is fine.

Andreas
-- 
  Andreas Jaeger aj@{suse.com,opensuse.org} Twitter/Identica: jaegerandi
   SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
    GF: Jeff Hawn,Jennifer Guild,Felix Imendörffer,HRB16746 (AG Nürnberg)
     GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126