[Bug tree-optimization/64277] [4.9/5 Regression] Incorrect warning "array subscript is above array bounds"

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64277

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2015-01-26
     Ever confirmed|0                           |1

--- Comment #11 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed btw.  I believe we have plenty of dups of this report - it is usually
the last peeled iteration we warn for.

The interesting thing is that we have

  # RANGE [1, 11] NONZERO 15
  i_385 = i_369 + 1;
...
  <bb 52>:
  # RANGE [1, 11] NONZERO 15
  i_461 = ASSERT_EXPR <i_385, i_385 < prephitmp_422>;
...
  <bb 20>:
  # RANGE [0, 10] NONZERO 15
  # i_82 = PHI <i_461(52)>

still we compute

i_461: [14, 32766]  EQUIVALENCES: { i_385 } (1 elements)

so sth is bogus here (the above ranges are simply copied in unrolling
but that's ok - they are conservatively correct).

We should be able to rely on the previous ranges and use them to our advantage,
like with

Index: gcc/tree-vrp.c
===================================================================
--- gcc/tree-vrp.c      (revision 220107)
+++ gcc/tree-vrp.c      (working copy)
@@ -847,6 +847,23 @@ update_value_range (const_tree var, valu
   value_range_t *old_vr;
   bool is_new;

+  /* If there is a value-range on the SSA name from earlier analysis
+     factor that in.  */
+  if (INTEGRAL_TYPE_P (TREE_TYPE (var)))
+    {
+      wide_int min, max;
+      value_range_type rtype = get_range_info (var, &min, &max);
+      if (rtype == VR_RANGE || rtype == VR_ANTI_RANGE)
+       {
+         value_range_d nr;
+         nr.type = rtype;
+         nr.min = wide_int_to_tree (TREE_TYPE (var), min);
+         nr.max = wide_int_to_tree (TREE_TYPE (var), max);
+         nr.equiv = NULL;
+         vrp_intersect_ranges (new_vr, &nr);
+       }
+    }
+
   /* Update the value range, if necessary.  */
   old_vr = get_value_range (var);
   is_new = old_vr->type != new_vr->type

which cuts down the number of warnings to

t.c: In function 'foo':
t.c:18:16: warning: array subscript is above array bounds [-Warray-bounds]
       a[i] = f1[i];
                ^
t.c:18:16: warning: array subscript is above array bounds [-Warray-bounds]
t.c:12:14: warning: 'f1' may be used uninitialized in this function
[-Wmaybe-uninitialized]
    f1[i] = f1[i] + 1;
              ^
from

t.c: In function 'foo':
t.c:18:16: warning: array subscript is above array bounds [-Warray-bounds]
       a[i] = f1[i];
                ^
t.c:18:16: warning: array subscript is above array bounds [-Warray-bounds]
t.c:18:16: warning: array subscript is above array bounds [-Warray-bounds]
t.c:18:16: warning: array subscript is above array bounds [-Warray-bounds]
t.c:18:16: warning: array subscript is above array bounds [-Warray-bounds]
t.c:12:14: warning: 'f1' may be used uninitialized in this function
[-Wmaybe-uninitialized]
    f1[i] = f1[i] + 1;
              ^