[Bug tree-optimization/66678] loop counter not accurately described by vrp

[vries at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66678

--- Comment #1 from vries at gcc dot gnu.org ---
I.
(In reply to vries from comment #0)

> AFAIU:
> - the loop iv i_1 has range [0, 4294967294], and 

This bit is incorrect. i_1 can actually be 4294967295. 


II.

With this demonstrator patch:
...
diff --git a/gcc/tree-vrp.c b/gcc/tree-vrp.c
index fdaebe4..3dd2992 100644
--- a/gcc/tree-vrp.c
+++ b/gcc/tree-vrp.c
@@ -1890,7 +1890,8 @@ extract_range_from_assert (value_range_t *vr_p, tree
expr)
                max = fold_build2 (PLUS_EXPR, TREE_TYPE (max), max,
                                   build_int_cst (TREE_TYPE (max), -1));
              else
-               max = fold_build2 (MINUS_EXPR, TREE_TYPE (max), max,
+               max = fold_build2 (MINUS_EXPR, TREE_TYPE (max),
+                                  TYPE_MAXVAL (TREE_TYPE (max)),
                                   build_int_cst (TREE_TYPE (max), 1));
              if (EXPR_P (max))
                TREE_NO_WARNING (max) = 1;
...

we get:
...
  # RANGE [1, 4294967295]
  i_17 = i_1 + 1;
...