On 03/06/2013 03:34 PM, Jason Merrill wrote:
It sounds like the underlying problem is that "decl" isn't a TYPE_DECL. So let's check for that instead.
I see. Thus, also considering that the only other use of check_elaborated_type_specifier happens when the decl is known to be a TYPE_DECL, I thought we could change the caller, go back to something very similar to 4_6-branch + the call when appropriate. Passes testing.

Thanks!
Paolo.

////////////////////
Index: cp/parser.c
===================================================================
--- cp/parser.c (revision 196487)
+++ cp/parser.c (working copy)
@@ -14248,12 +14248,15 @@ cp_parser_elaborated_type_specifier (cp_parser* pa
                                   typename_type,
                                   /*complain=*/tf_error);
       /* If the `typename' keyword is in effect and DECL is not a type
-        decl. Then type is non existant.   */
+        decl, then type is non existent.   */
       else if (tag_type == typename_type && TREE_CODE (decl) != TYPE_DECL)
         type = NULL_TREE; 
-      else 
-       type = check_elaborated_type_specifier (tag_type, decl,
+      else if (TREE_CODE (decl) == TYPE_DECL)
+        type = check_elaborated_type_specifier (tag_type, decl,
                                                /*allow_template_p=*/true);
+      else
+       /* NULL_TREE or error_mark_node (if decl is error_mark_node).  */
+       type = TREE_TYPE (decl); 
     }
 
   if (!type)
Index: testsuite/g++.dg/template/crash115.C
===================================================================
--- testsuite/g++.dg/template/crash115.C        (revision 0)
+++ testsuite/g++.dg/template/crash115.C        (working copy)
@@ -0,0 +1,3 @@
+// PR c++/56534
+
+template < struct template rebind < > // { dg-error "expected" }

Reply via email to