[Bug tree-optimization/27603] [4.1/4.2 Regression] wrong code, apparently due to bad VRP

[rguenth at gcc dot gnu dot org]

------- Comment #4 from rguenth at gcc dot gnu dot org  2006-05-14 18:00 -------
(set_scalar_evolution
  (scalar = j_3)
  (scalar_evolution = {0, +, 1}_1))
)
(instantiate_parameters
  (loop_nb = 1)
  (chrec = {0, +, 1}_1)
  (res = {0, +, 1}_1))
Found new range for j_3: [0, 0]

huh!?  (this is from 4.1 dumps)

<L0>:;
  j_3 = ASSERT_EXPR <j_2, j_2 <= 5>;
  j.0_5 = (unsigned int) j_3;
  D.1285_6 = j.0_5 - 3;
  if (D.1285_6 <= 1) goto <L1>; else goto <L2>;

<L1>:;
  exit (0);

<L2>:;
  D.1286_7 = 6 - j_3;
  D.1287_8 = D.1286_7 * 1000;
  a = D.1287_8;
  j_10 = j_3 + 1;

  # j_2 = PHI <0(0), j_10(3)>;
<L3>:;
  if (j_2 <= 5) goto <L0>; else goto <L4>;

<L4>:;
  abort ();


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27603