https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70594
--- Comment #27 from Nathan Sidwell <nathan at gcc dot gnu.org> --- (In reply to Jason Merrill from comment #26) > With this patch gengtype seems to assign gt_ggc_mx(tree&) to cp/constexpr.o, > breaking cc1 link. I'm not going to try to fix that now, though someone > else is welcome to. Yeah, a similar link failure led me to adding all that gt_ggc_mx stuff in yesterday's protopatch. However, I think you're on the right minimal track. When the constexpr_call_table is gc-deleted, the preservation of the fn_copies_table will guarantee we won't change UID allocation. We'll still want the post-parseing hook to kill the copies table. I'll see what I can do with your patch.