The way cp_parser_lookup_name handles object scope (i.e. the scope on the
RHS of a . or -> expression) is a bit subtle: before the lookup it's in
parser->context->object type, and after the lookup it's in
parser->object_scope.  But a couple of places that elide lookups were
failing to do the same transform.

I'm not aware of this breaking anything currently.

Tested x86_64-pc-linux-gnu, applying to trunk.

gcc/cp/ChangeLog:

        * parser.c (cp_parser_template_name): Move object type.
        (cp_parser_pre_parsed_nested_name_specifier): Likewise.
---
 gcc/cp/parser.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/gcc/cp/parser.c b/gcc/cp/parser.c
index ab1dc81b997..7a0b6234350 100644
--- a/gcc/cp/parser.c
+++ b/gcc/cp/parser.c
@@ -18405,6 +18405,7 @@ cp_parser_template_name (cp_parser* parser,
            {
              /* We're optimizing away the call to cp_parser_lookup_name, but
                 we still need to do this.  */
+             parser->object_scope = parser->context->object_type;
              parser->context->object_type = NULL_TREE;
              return identifier;
            }
@@ -33575,7 +33576,8 @@ cp_parser_pre_parsed_nested_name_specifier (cp_parser 
*parser)
   /* Set the scope from the stored value.  */
   parser->scope = saved_checks_value (check_value);
   parser->qualifying_scope = check_value->qualifying_scope;
-  parser->object_scope = NULL_TREE;
+  parser->object_scope = parser->context->object_type;
+  parser->context->object_type = NULL_TREE;
 }
 
 /* Consume tokens up through a non-nested END token.  Returns TRUE if we

base-commit: bd55fa102715c7442c050b193dadfdb5337e2377
prerequisite-patch-id: 6aead84b2823eea91fa3886a64c032aa4882cec3
-- 
2.27.0

Reply via email to