On Tue, Aug 8, 2017 at 5:01 PM, David Malcolm <dmalc...@redhat.com> wrote: > On Mon, 2017-08-07 at 14:25 -0400, Jason Merrill wrote: > > Thanks for looking at this. > >> On 08/01/2017 04:21 PM, David Malcolm wrote: >> > @@ -27632,6 +27769,9 @@ cp_parser_sizeof_operand (cp_parser* >> > parser, enum rid keyword) >> > { >> > tree type = NULL_TREE; >> > >> > + matching_parens parens; >> > + parens.peek_open (parser); >> >> I was puzzled by this until I found that >> cp_parser_compound_literal_p >> consumes the open paren. Let's remove that in favor of calling >> consume_open here, so we don't need peek_open anymore. > > Done. > >> About passing parser in or not, I'm happy with the current approach; >> adding things to the stack isn't free in a highly recursive program >> like GCC. > > Thanks; I'll keep "parser" out of the new classes then. > > Here's an updated "v3" patch. > > Successfully bootstrapped®rtested on x86_64-pc-linux-gnu in conjunction > with the other patches (1 and 2 of the v2 kit). > > OK for trunk, assuming the other patches are approved? (patch 2 in the kit, > for the C frontend, still needs approval).
OK. Jason