https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118858
--- Comment #3 from Martin Jambor <jamborm at gcc dot gnu.org> --- The reason why I only added cold was not a question of streaming (I don't think we avoid any this way) but rather me being lazy, in the sense that I really wanted the cold attribute to go in reasonably quickly and so at this time I specifically wanted to avoid discussion which other attributes need to be added and how (what the handlers and conflict hooks should do). At the same time, I did go over the builtin macros builtins.def, rather than just looking at what is in builtin-attrs.def and do think we should be OK, as far as all of builtins defined via DEF_GCC_BUILTIN there are concerned. I have not tried to test if streamed in attributes are dropped because of this. If that is the case, then I believe we likely need more attributes than those in builtin-attrs.def (and so I hope they are not). I think what should be done is that we should add an LTO self-test to check that all builtins added by lto_init_attributes() have the expected attributes and fix any fallout from that as it appears (if it appears).
