Patch fixing 3.3 bug PR 9745 and PR 10021

[Mark Mitchell]

> 	Expanding on Jim's original analysis, we need to find a way for
> the current GCC aliasing infrastructure to correctly deduce and return a
> valid ADDRESS RTL instead of a naked hard register.

Independent of what targets the bug affects, this reinforces my key
point: that someone is going to have to work hard to fix this aliasing
issue, and until that point we have to choose between wrong code and
slow code.

Personally, I'd choose slow code.  Not only is that better for our
users, some of whom use GCC to build very critical systems, but it would
also encourage someone to go solve the optimization problem.

-- 
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com