http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49003
--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> 2011-05-16 11:09:05 UTC --- ah I was thinking of PR 45908 (where I provided that reduced testcase - I knew it looked familiar!) but that was for an ICE which is fixed, so this one should be kept for the change in semantics