This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[PATCH]: Fix PR tree-optimization/21407


> It would seem rather dangerous to me to implement this optimization.
> 

Again, it's been implemented and turned on for a couple months now, it's
not a proposal.  :)

However, here is a patch to fully disable it and leave a note in it's
place for people who think of the same idea in the future.

I've added 21407 testcase to g++.dg/tree-ssa

Bootstrapped and regtested on i686-pc-linux-gnu
okay for mainline?

2005-05-09  Daniel Berlin  <dberlin@dberlin.org>
	
	Fix PR tree-optimization/21407
	* tree-ssa-operands.c (note_addressable):  Can't
	use COMPONENT_REF target.

Index: tree-ssa-operands.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/tree-ssa-operands.c,v
retrieving revision 2.82
diff -u -p -r2.82 tree-ssa-operands.c
--- tree-ssa-operands.c	6 May 2005 14:04:11 -0000	2.82
+++ tree-ssa-operands.c	10 May 2005 18:42:42 -0000
@@ -1886,35 +1886,14 @@ add_stmt_operand (tree *var_p, stmt_ann_
 static void
 note_addressable (tree var, stmt_ann_t s_ann)
 {
-  tree ref;
   subvar_t svars;
-  HOST_WIDE_INT offset;
-  HOST_WIDE_INT size;
 
   if (!s_ann)
     return;
   
-  /* If this is a COMPONENT_REF, and we know exactly what it touches, we only
-     take the address of the subvariables it will touch.
-     Otherwise, we take the address of all the subvariables, plus the real
-     ones.  */
-
-  if (var && TREE_CODE (var) == COMPONENT_REF 
-      && (ref = okay_component_ref_for_subvars (var, &offset, &size)))
-    {
-      subvar_t sv;
-      svars = get_subvars_for_var (ref);
-      
-      if (s_ann->addresses_taken == NULL)
-	s_ann->addresses_taken = BITMAP_GGC_ALLOC ();      
-      
-      for (sv = svars; sv; sv = sv->next)
-	{
-	  if (overlap_subvar (offset, size, sv, NULL))
-	    bitmap_set_bit (s_ann->addresses_taken, var_ann (sv->var)->uid);
-	}
-      return;
-    }
+  /* Note that it is *NOT OKAY* to use the target of a COMPONENT_REF
+     as the only thing we take the address of.
+     See PR 21407 and the ensuing mailing list discussion.  */
   
   var = get_base_address (var);
   if (var && SSA_VAR_P (var))
Index: testsuite/g++.dg/tree-ssa/pr21407.C
===================================================================
RCS file: testsuite/g++.dg/tree-ssa/pr21407.C
diff -N testsuite/g++.dg/tree-ssa/pr21407.C
--- /dev/null	1 Jan 1970 00:00:00 -0000
+++ testsuite/g++.dg/tree-ssa/pr21407.C	10 May 2005 18:42:42 -0000
@@ -0,0 +1,8 @@
+/* { dg-do run } */
+/* { dg-options "-O2" } */
+extern "C" void abort(void);
+struct T1 {int a, b; virtual void f(){}};
+struct T : T1 { struct T1 w;  int b; };
+void foo (struct T1 *p) { struct T *q = dynamic_cast<T*>(p); if (q->b != 2) abort (); }
+/* We shouldn't kill the store to c.b, because foo uses it.  */
+int main () { struct T c; c.b = 2; foo (&c); return 0; }

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]