[RFA] Further de-specification of gcc.c

[Richard Henderson]

On Wed, May 15, 2002 at 07:51:29AM +0100, Neil Booth wrote:
> I didn't use the new flag_fast_math to determine the settings
> after option processing, because I assume that things like
>   -ffast-math -ftrapping-math
> are expected to be handled in-order, so that trapping-math
> should be set.

I think define __FAST_MATH__ iff the individual components are
as set_fast_math_flags would have set them.  Otherwise the header
isn't going to know which transformations are legal.

Otherwise ok.


r~