http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388
Eric Botcazou <ebotcazou at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |ASSIGNED AssignedTo|unassigned at gcc dot |ebotcazou at gcc dot |gnu.org |gnu.org Target Milestone|--- |4.8.0 --- Comment #10 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-11-22 17:47:38 UTC --- > It's caused by r193599: > http://gcc.gnu.org/ml/gcc-cvs/2012-11/msg00546.html Yep, the problem is that alias.c:nonoverlapping_component_refs_p considers that accesses to bitfields never overlap with accesses to other components... I also think that the size handling in set_mem_attributes_minus_bitpos might need to be further refined, now that bitfields are handled: /* Respect mode size. */ attrs.size_known_p = defattrs->size_known_p; attrs.size = defattrs->size; /* ??? Is this really necessary? We probably should always get the size from the type below. */ I don't see how the final size can be smaller than the mode size.