Hi,
On 10/26/2012 02:44 PM, Jason Merrill wrote:
On 10/25/2012 09:15 PM, Paolo Carlini wrote:
I'm not sure the testcase can't be improved, I'm not very familiar with
dg-final.
You could do an execution test using a user-defined operator new which
initializes the memory to something other than 0. OK with that change.
Also, in case we agree that the fix can be such simple, we
could maybe consider 4_7-branch too...
Sure.
Thanks. The below is what I'm going to apply mainline and 4_7-branch.
Paolo.
////////////////////
Index: testsuite/g++.dg/template/new11.C
===================================================================
--- testsuite/g++.dg/template/new11.C (revision 0)
+++ testsuite/g++.dg/template/new11.C (working copy)
@@ -0,0 +1,28 @@
+// PR c++/54984
+// { dg-do run }
+
+int n = 1;
+
+void* operator new(__SIZE_TYPE__)
+{
+ n = -1;
+ return &n;
+}
+
+template <class T>
+struct Foo
+{
+ Foo()
+ : x(new int)
+ {
+ if (*x != -1)
+ __builtin_abort();
+ }
+
+ int* x;
+};
+
+int main()
+{
+ Foo<float> foo;
+}
Index: cp/init.c
===================================================================
--- cp/init.c (revision 192839)
+++ cp/init.c (working copy)
@@ -2911,7 +2911,8 @@ build_new (VEC(tree,gc) **placement, tree type, tr
orig_placement = make_tree_vector_copy (*placement);
orig_nelts = nelts;
- orig_init = make_tree_vector_copy (*init);
+ if (*init)
+ orig_init = make_tree_vector_copy (*init);
make_args_non_dependent (*placement);
if (nelts)
