C++ PATCH: Ensure destructor calls have type "void"

[Mark Mitchell <mark at codesourcery dot com>]

In some internal testing, we noticed that
g++.dg/template/dependent-expr1.C was failing on ARM ABI targets.  The
problem turned out to be that the C++ front-end was treating an
expression of the form "x->~X()" as having type "void *", rather than
"void" as the ISO C++ standard requires.  The cause was that, in the
ARM ABI, [cd]tors return "void *" (the "this" pointer).  Because we
don't have a separate C++ lowering phase, we must be careful to
convert to the correct type.

Tested on arm-none-eabi, applied to the mainline.  I shall also apply
shortly to 4.1 and 4.2 branches, since this is a regression.

--
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713

2007-02-19  Mark Mitchell  <mark@codesourcery.com>

	* call.c (build_new_method_call): Ensure that explicit calls of
	destructors have type "void".

Index: call.c
===================================================================
--- call.c	(revision 121831)
+++ call.c	(working copy)
@@ -5520,6 +5520,20 @@ build_new_method_call (tree instance, tr
 		  && TREE_SIDE_EFFECTS (instance_ptr))
 		call = build2 (COMPOUND_EXPR, TREE_TYPE (call),
 			       instance_ptr, call);
+	      else if (call != error_mark_node
+		       && DECL_DESTRUCTOR_P (cand->fn)
+		       && !VOID_TYPE_P (TREE_TYPE (call)))
+		/* An explicit call of the form "x->~X()" has type
+		   "void".  However, on platforms where destructors
+		   return "this" (i.e., those where
+		   targetm.cxx.cdtor_returns_this is true), such calls
+		   will appear to have a return value of pointer type
+		   to the low-level call machinery.  We do not want to
+		   change the low-level machinery, since we want to be
+		   able to optimize "delete f()" on such platforms as
+		   "operator delete(~X(f()))" (rather than generating
+		   "t = f(), ~X(t), operator delete (t)").  */
+		call = build_nop (void_type_node, call);
 	    }
 	}
     }