http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47621
--- Comment #2 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-02-06 23:08:39 UTC --- Note to self: Before the patch the code in the else branch never triggered. The original idea was to preserve MEM[&a] re-writing into SSA form, which didn't happen before the patch. Re-writing the testcase in this PR would also be nice, but I see other problems with this code (bit-field component-refs come to my mind). I have a patch.
