[patch] fix pr 29302

[Roger Sayle]

Hi Eric,

On Fri, 15 Dec 2006, Eric Christopher wrote:
> 2006-12-15  Eric Christopher  <echristo@apple.com>
>
> 	PR target/29302
> 	* real.c (real_maxval): Fix for ibm_extended_format.
> 	(ibm_extended_real_maxval): New function.
> 	* c-cppbuiltin.c (builtin_define_float_constants): Update comment.

I appreciate that the actual code is cut'n'paste from c-cppbuiltin.c,
but have you considered recoding the algorithm to avoid the use of
real_from_string?  It makes sense to construct this floating point
constant as an ASCII string in c-cppbuiltin.c as that's how it'll be
used there, but this seems like unnecessary overhead in real.c where
the hex values you want can be written directly into the REAL_VALUE_TYPE.

p.s. I think the simple test case from your posting seems sufficient.

Thanks in advance,

Roger
--