Re: isnan(NAN) returns 0!

FROM : Andrew Pinski
DATE : Sun Aug 13 09:01:28 2006

>
> I submitted a bug report to Apple regarding this issue. rdar://
> problem/4678746 .
>
> I'll be quiet now. :)
>
> -Jonathan Grynspan
>
> On 13-Aug-06, at 1:51 AM, Jonathan Grynspan wrote:
>
> > I've narrowed down the cause to "Relax IEEE restrictions" being 
> > enabled on my build. This shouldn't cause anything so drastic, and 
> > yet it does. Is anybody else able to confirm this behaviour?


This is not really a bug if "Relax IEEE restrictions", is really
just a synom for -ffast-math.  -ffast-math turns on -ffinite-math-only
which says: "Assume no NaNs or infinities are generated".

Unless this is x86 which has a known bug and "Relax IEEE restrictions"
is a synom for -funsafe-math-optimizations.  See
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19116

-- Pinski