[Bug c++/53261] [4.8 Regression] ICE in tree_strip_nop_conversions

["manu at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53261

--- Comment #3 from Manuel LÃpez-IbÃÃez <manu at gcc dot gnu.org> 2012-05-07 16:25:11 UTC ---
(In reply to comment #2)
> In general, build_range_check can return NULL_TREE. Just matter of doing:
> 
>     if (tem && integer_zerop (tem))
>       return;
> 
> like a few lines below?

I think you are right. I have launched a bootstrap+regtest with this patch:

Index: gcc/c-family/c-common.c
===================================================================
--- gcc/c-family/c-common.c     (revision 187257)
+++ gcc/c-family/c-common.c     (working copy)
@@ -1627,11 +1627,11 @@ warn_logical_operator (location_t locati
      should be always false to get a warning.  */
   if (or_op)
     in0_p = !in0_p;

   tem = build_range_check (UNKNOWN_LOCATION, type, lhs, in0_p, low0, high0);
-  if (integer_zerop (tem))
+  if (tem && integer_zerop (tem))
     return;

   rhs = make_range (op_right, &in1_p, &low1, &high1, &strict_overflow_p);
   if (!rhs)
     return;
@@ -1642,11 +1642,11 @@ warn_logical_operator (location_t locati
      should be always false to get a warning.  */
   if (or_op)
     in1_p = !in1_p;

   tem = build_range_check (UNKNOWN_LOCATION, type, rhs, in1_p, low1, high1);
-  if (integer_zerop (tem))
+  if (tem && integer_zerop (tem))
     return;

   /* If both expressions have the same operand, if we can merge the
      ranges, and if the range test is always false, then warn.  */
   if (operand_equal_p (lhs, rhs, 0)



Could you test on hppa?


Actually, I am not sure whether "if (!tem || integer_zerop (tem))" is even
better.