[C++ Patch] PR 51621

[Paolo Carlini]

Hi,

this ICE - which leads to worse diagnostics - seems caused by a trivial 
typo: to mean value initialization diagnose_non_constexpr_vec_init is 
passing zoid_zero_node instead of void_type_node to build_vec_init_elt. 
My guess is also confirmed by the beginning of the other caller, 
build_vec_init_expr, which evidently has void_type_node meaning value 
initialization.

The below tested x86_64-linux.

Thanks,
Paolo.

///////////////////
/cp
2011-12-20  Paolo Carlini  <paolo.carl...@oracle.com>

        PR c++/51621
        * tree.c (diagnose_non_constexpr_vec_init): For value initialization
        pass void_type_node, not void_zero_node, to build_vec_init_elt.

/testsuite
2011-12-20  Paolo Carlini  <paolo.carl...@oracle.com>

        PR c++/51621
        * g++.dg/cpp0x/constexpr-ice5.C: New.
Index: testsuite/g++.dg/cpp0x/constexpr-ice5.C
===================================================================
--- testsuite/g++.dg/cpp0x/constexpr-ice5.C     (revision 0)
+++ testsuite/g++.dg/cpp0x/constexpr-ice5.C     (revision 0)
@@ -0,0 +1,13 @@
+// PR c++/51621
+// { dg-options -std=c++0x }
+
+struct A
+{
+  A() {}
+};
+
+struct B
+{
+  A a[1];
+  constexpr B() : a() {} // { dg-error "non-constant|non-constexpr" }
+};
Index: cp/tree.c
===================================================================
--- cp/tree.c   (revision 182525)
+++ cp/tree.c   (working copy)
@@ -564,7 +564,7 @@ diagnose_non_constexpr_vec_init (tree expr)
   tree type = TREE_TYPE (VEC_INIT_EXPR_SLOT (expr));
   tree init, elt_init;
   if (VEC_INIT_EXPR_VALUE_INIT (expr))
-    init = void_zero_node;
+    init = void_type_node;
   else
     init = VEC_INIT_EXPR_INIT (expr);