http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48063
--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-03-10 19:09:49 UTC --- Created attachment 23617 --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=23617 gcc46-pr48063.patch This seems to fix it, but no idea whether it is the right fix. It seems all other places that check call_stmt_cannot_inline_p also call tree_can_inline_p though.