[PATCH] Allow -msoft-float -mlong-double-128 without a warning

[Mark Mitchell]

Jakub Jelinek wrote:
					\
> +  if (TARGET_SOFT_FLOAT && TARGET_LONG_DOUBLE_128			\
> +      && rs6000_explicit_options.long_double)				\
> +    warning (0, "-msoft-float and -mlong-double-128 not supported");	\

Why not an error?  And, why does it matter whether or not the long
double option was explicit?  What about something like:

  if (TARGET_SOFT_FLOAT && TARGET_LONG_DOUBLE_128)
    error ("-msoft-float requires -mlong-double-64");

It seems scary to have TARGET_LONG_DOUBLE_128 set if it doesn't work
with -msoft-float; that means that any part of the PowerPC back end that
checks TARGET_LONG_DOUBLE_128 must do (TARGET_LONG_DOUBLE_128 &
!TARGET_SOFT_FLOAT).

Of course, I'm not an expert on these things, so it's entirely possible
I'm spouting nonsense.

-- 
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713