On Apr 17, 2018, Jason Merrill <ja...@redhat.com> wrote: > I was thinking maybe_get_node
'k > I suppose better would be for > + if (obj->list_p () && obj->get_decl_maybe ()) > + tree_list_freelist ().free (obj->get_node ()); > to be e.g. > obj-> maybe_free_list(); Any objections to making dec_refcount_use a friend of tinst_level's? Otherwise, I'd rather add a free() member function (or maybe static member function) to free both the TREE_LIST and the tinst_level objects. Otherwise, maybe_free_list would leave the object in an inconsistent state, pointing to a freed TREE_LIST, or it could undo to_list, though it's not supposed to be reversible in the lifetime of the object. free()ing the entire object along with the TREE_LIST addresses that concern. But then, dec_refcount_use does that and then some... > Or put the list handling in a destructor for tinst_level, and have > freelist use placement new and explicit destruction. A destructor would enable finalization in GGC, I don't think we want to do that. -- Alexandre Oliva, freedom fighter http://FSFLA.org/~lxoliva/ You must be the change you wish to see in the world. -- Gandhi Be Free! -- http://FSFLA.org/ FSF Latin America board member Free Software Evangelist|Red Hat Brasil GNU Toolchain Engineer
