https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88766
--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> --- As for the GCC diagnostic pragma, we need it to be a deferred pragma, have no idea how else could we handle that when it is not visible in the token stream. And, if it is visible in the token stream, accepting it everywhere in the exceptions would be quite hard too, e.g. the C FE normally has only 2 tokens lookahead, or how would it play in C++ FE with the tentative parsing? What we could do is accept #pragma GCC diagnostic (and only that?) inside of the braced initializers (so c_parser_braced_init for C and cp_parser_braced_list? for C++), before each init elt and after those, perhaps rejecting it if it appears before a comma. Of course, for gnupg2 they would need to adjust their macros slightly, because the would need to go inside of the compound literal braced init rather than around it. Thoughts on that?
