[C++ patch] PR tree-optimization/62053 (VLA arrays)

[Jan Hubicka]

Hi,
this patch triggers assert in tree-inline that verify that array type and its 
main variant have same size
(as pointer).  In this case we have two types that do have same size, but the 
expression is different.
It is created in both cases by layout_type becuase C++ FE calls layout_type 
before linking the variants.
I think it is safe to move the call and avoid the duplication.

Bootstrapped/regtested x86_64-linux, OK?

Honza

        PR tree-optimization/62053
        * g++.dg/torture/pr62053.C: New testcase.
        * tree.c: Avoid ordering issue with layout_type.

Index: testsuite/g++.dg/torture/pr62053.C
===================================================================
--- testsuite/g++.dg/torture/pr62053.C  (revision 0)
+++ testsuite/g++.dg/torture/pr62053.C  (revision 0)
@@ -0,0 +1,7 @@
+// { dg-do compile }
+typedef int GType;
+class YGTable
+{
+  YGTable () { GType a[m_fn1 ()]; }
+  int m_fn1 ();
+};
Index: cp/tree.c
===================================================================
--- cp/tree.c   (revision 216145)
+++ cp/tree.c   (working copy)
@@ -863,12 +863,12 @@ build_cplus_array_type (tree elt_type, t
        {
          t = build_min_array_type (elt_type, index_type);
          set_array_type_canon (t, elt_type, index_type);
-         if (!dependent)
-           layout_type (t);
 
          TYPE_MAIN_VARIANT (t) = m;
          TYPE_NEXT_VARIANT (t) = TYPE_NEXT_VARIANT (m);
          TYPE_NEXT_VARIANT (m) = t;
+         if (!dependent)
+           layout_type (t);
        }
     }