Le 28/10/2024 à 14:38, Andrew MacLeod a écrit :

On 10/26/24 15:08, Mikael Morin wrote:
Hello,

Le 24/10/2024 à 14:53, Andrew MacLeod a écrit :
diff --git a/gcc/range-op-ptr.cc b/gcc/range-op-ptr.cc
index dd312a80366..ef2b2cce516 100644
--- a/gcc/range-op-ptr.cc
+++ b/gcc/range-op-ptr.cc
(...)
-void
-pointer_or_operator::wi_fold (irange &r, tree type,
-                  const wide_int &lh_lb,
-                  const wide_int &lh_ub,
-                  const wide_int &rh_lb,
-                  const wide_int &rh_ub) const
+operator_bitwise_or::fold_range (prange &r, tree type,
+                 const prange &op1,
+                 const prange &op2,
+                 relation_trio) const
 {
   // For pointer types, we are really only interested in asserting
   // whether the expression evaluates to non-NULL.
-  if (!wi_includes_zero_p (type, lh_lb, lh_ub)
-      && !wi_includes_zero_p (type, rh_lb, rh_ub))
+  if (!op1.zero_p () || !op2.zero_p ())
     r.set_nonzero (type);

this doesn't feel right.  It checks that the operand range is anything but singleton [0], instead of checking that it does not contain 0.


Hmm, i think you are right.  But I also think the original is wrong? Shouldn't it be nonzero if either operand doesn't contain zero?

It's Monday morning.. Check my logic please... :-)

The original wasn't strictly speaking wrong; I mean it was just producing more VARYING than necessary.
But otherwise, yes, your logic appears correct.


diff --git a/gcc/range-op-ptr.cc b/gcc/range-op-ptr.cc
index 9f500c52d2c..ccce05035b9 100644
--- a/gcc/range-op-ptr.cc
+++ b/gcc/range-op-ptr.cc
@@ -388,7 +388,7 @@ operator_bitwise_or::fold_range (prange &r, tree type,
  {
    // For pointer types, we are really only interested in asserting
    // whether the expression evaluates to non-NULL.
-  if (!op1.zero_p () || !op2.zero_p ())
+  if (!range_includes_zero_p (op1) || !range_includes_zero_p (op2))
      r.set_nonzero (type);
    else if (op1.zero_p () && op2.zero_p ())
      r.set_zero (type);

And this looks good.


Reply via email to