[Bug middle-end/48493] [4.6/4.7 Regression] ice in expand_expr_addr_expr_1 with complex types and mem_ref

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48493

--- Comment #13 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-05-30 
19:08:58 UTC ---
(In reply to comment #12)
> >on arm?  That's pre-approved if it passes bootstrap & regtest 
> 
> I will try it on MIPS since I saw the exact same ICE and the exact same
> "(concat:SC (reg/v:SF 134 [ s ])
>     (reg/v:SF 135 [ s+4 ]))" .

It fixes the bug and I saw no regressions on mips.

Thanks,
Andrew