[PATCH] Fix simd-2.c failure on powerpc

[Roger Sayle]

On Thu, 17 Apr 2003, Richard Henderson wrote:
> I see that.  And now re-reading all the threads I see what you
> were after with this change.  But am I not misunderstanding that
> the abort in question is the subject of your rtlanal.c change?

Correct.  I believe that either change alone should fix this
specific failure of simd-2.c.  However, why fix a bug just once
when it can be fixed twice.  The function reg_overlap_mentioned_p
might potentially be called with a zero_extract or a sign_extract
from somewhere else in GCC.  Similarly, if we ever add another RTL
code as a valid SET_DEST then the can_combine_p change will protect
us from surprises in future.

The two changes together make GCC less fragile, for example, should
someone decide to tackle the potential can_combine_p improvements.

Roger
--