[testsuite] fix gcc.dg/torture/builtin-modf-1.c

Jack Howarth howarth@bromo.msbb.uc.edu
Thu Dec 20 01:24:00 GMT 2007


Janis,
   Apparently not. I find that both...

-funsafe-math-optimizations -fno-trapping-math -freciprocal-math 

and just

-fno-trapping-math -freciprocal-math

...produce undefined symbols on powerpc-apple-darwin9. We definitely
need...

-funsafe-math-optimizations -fsigned-zeros -fno-associative-math

                       Jack

On Wed, Dec 19, 2007 at 03:55:12PM -0800, Janis Johnson wrote:
> On Wed, 2007-12-19 at 17:04 -0500, Jack Howarth wrote:
> > Janis,
> >     It appears that on powerpc-apple-darwin9, we need...
> >
> > -funsafe-math-optimizations -fsigned-zeros -fno-associative-math
> 
> That's the same as "-fno-trapping-math -freciprocal-math", right?
> Both of those work for powerpc64-linux with -m32/-m64, so if those
> are needed for darwin in general we could probably use them for
> powerpc*-*-* again.
> 
> >     I am able to compile and execute the test case without errors
> > at -Os, -O0, -O1, -O2 and -O3 with those flags.    
> >                   Jack
> 
> I'll be on vacation for two weeks and not keeping up with email.
> Changing the test to use either of these sets of options for
> powerpc-*-darwin or powerpc*-*-* is fine with me.
> 
> Janis



More information about the Gcc-patches mailing list