summaryrefslogtreecommitdiff
path: root/gcc/testsuite/g++.dg/cpp0x/implicit8.C
diff options
context:
space:
mode:
Diffstat (limited to 'gcc/testsuite/g++.dg/cpp0x/implicit8.C')
-rw-r--r--gcc/testsuite/g++.dg/cpp0x/implicit8.C34
1 files changed, 34 insertions, 0 deletions
diff --git a/gcc/testsuite/g++.dg/cpp0x/implicit8.C b/gcc/testsuite/g++.dg/cpp0x/implicit8.C
new file mode 100644
index 000000000..2f3febae8
--- /dev/null
+++ b/gcc/testsuite/g++.dg/cpp0x/implicit8.C
@@ -0,0 +1,34 @@
+// The hack for PR c++/44909 breaks this testcase. We need feedback
+// from the C++ committee to know how to proceed.
+// { dg-options -std=c++0x }
+// { dg-prune-output "implicitly deleted" }
+// { dg-prune-output "cannot bind" }
+// { dg-prune-output "initializing argument" }
+
+struct A
+{
+ A();
+ A(A&);
+};
+
+struct B;
+struct BP
+{
+ BP(const B&);
+};
+
+struct B
+{
+ B();
+ B(B&&);
+ B(const BP&);
+};
+
+// If B(B&&) suppresses the B copy constructor, then copying the B
+// subobject of C should use B(const BP&). But we ignore that constructor
+// in order to break the cycle in 44909. Perhaps the move ctor shouldn't
+// suppress the copy ctor?
+struct C: A, B { };
+
+C c;
+C c2(c); // { dg-bogus "deleted" "" { xfail *-*-* } }