This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

C++ PATCH for c++/65721 (ICE with invalid using-declaration)


The standard says that the nested-name-specifier in a using-declaration needs to name a base of the class. When we have dependent bases we can't be sure whether a particular type is a base or not, but we can certainly complain if name lookup find the current class rather than a base.

Tested x86_64-pc-linux-gnu, applying to trunk.
commit d9ebf555e29b24e5fa3338471edbc9d922ae0edb
Author: Jason Merrill <jason@redhat.com>
Date:   Sat Apr 11 10:41:22 2015 -0400

    	PR c++/65721
    	* name-lookup.c (do_class_using_decl): Complain about specifying
    	the current class even if there are dependent bases.

diff --git a/gcc/cp/name-lookup.c b/gcc/cp/name-lookup.c
index e3f7cca..9e4e0e3 100644
--- a/gcc/cp/name-lookup.c
+++ b/gcc/cp/name-lookup.c
@@ -3408,7 +3408,7 @@ do_class_using_decl (tree scope, tree name)
 			   tf_warning_or_error);
       if (b_kind < bk_proper_base)
 	{
-	  if (!bases_dependent_p)
+	  if (!bases_dependent_p || b_kind == bk_same_type)
 	    {
 	      error_not_base_type (scope, current_class_type);
 	      return NULL_TREE;
diff --git a/gcc/testsuite/g++.dg/lookup/using55.C b/gcc/testsuite/g++.dg/lookup/using55.C
new file mode 100644
index 0000000..61098b1
--- /dev/null
+++ b/gcc/testsuite/g++.dg/lookup/using55.C
@@ -0,0 +1,19 @@
+// PR c++/65721
+
+template<typename T>
+struct A {
+  typedef T D;
+};
+
+template<typename X>
+class B : public A<X> {
+  using typename B::D;		// { dg-error "not a base" }
+public:
+  D echo(D x) {			// { dg-error "D" }
+    return x;
+  }
+};
+
+int main() {
+  B<int> b;
+}

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]