http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59226
--- Comment #9 from Jan Hubicka <hubicka at ucw dot cz> --- > Honza, > > I've tested your patch from comment 7 and it doesn't work. > However your suggestion "to simply return NULL when inner_binfo is NULL" > does seem to work fine. I've successfully build Chromium with it. Good to know. I am however not sure if it is safe - this scenario happens in "diamond" shaped virtual inheritance and here we know we visited the base in question on different path. But I wonder if we can encounter different types of thunks. I put more complete attempt to http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58252 Honza