[PATCH] PR c++/17554, ice in create_tmp_var, cleanup_expr

[Andrew Pinski <pinskia at physics dot uc dot edu>]

On Oct 5, 2004, at 12:19 AM, Andrew Pinski wrote:

> I don't know if this is the right fix but it works for this testcase.
> But should the cleanup point expression have the same type as what is
> containing it or not?  If have it set to the same as the expression
> which contains, we ICE in create_tmp_var as we are creating a temporary
> variable for struct which we cannot do.

Here is a patch which is safer and only changes the type to void if
the type was an aggregate before.

OK? Bootstrapped and tested on powerpc-darwin with no regressions

Thanks,
Andrew Pinski

Testcase:
struct A { int i;  A();  A(const A&); };

void bar()
{
    A a;
    for ( ;; a=A() ) ;
}

ChangeLog:
	* semantics.c (maybe_cleanup_point_expr): Set the type of the
	cleanup point expression to void type if we have an aggregate
	type.

Index: semantics.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/cp/semantics.c,v
retrieving revision 1.441
diff -u -p -r1.441 semantics.c
--- semantics.c	1 Oct 2004 15:11:21 -0000	1.441
+++ semantics.c	5 Oct 2004 14:44:23 -0000
@@ -358,7 +358,12 @@ static tree
 maybe_cleanup_point_expr (tree expr)
 {
   if (!processing_template_decl && stmts_are_full_exprs_p ())
-    expr = build1 (CLEANUP_POINT_EXPR, TREE_TYPE (expr), expr);
+    {
+      tree type = TREE_TYPE (expr);
+      if (AGGREGATE_TYPE_P (type))
+	type = void_type_node;
+      expr = build1 (CLEANUP_POINT_EXPR, type, fold (expr));
+    }
   return expr;
 }