[Bug tree-optimization/106112] [10/11/12/13 Regression] wrong code at -Os and above on x86_64-linux-gnu since r10-2711-g3ed01d5408045d80

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106112

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
             Status|NEW                         |ASSIGNED

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
  <bb 2> [local count: 1073741824]:
  e.0_1 = e;
  _2 = e.0_1 | 4294967292;
  e = _2;
  d.1_3 = d;
  _4 = (long unsigned int) d.1_3;
-  # RANGE [2147483647, 6442450942] NONZERO 8589934591
+  # RANGE ~[2147483647, 18446744071562067967]
  _10 = _4 + 4294967295;
  _7 = _10 - _2;
  _8 = 18446744073709551615 % _7;
  _9 = (int) _8;
  c = _9;
  _11 = 18446744073709551614 - _2;
-  _13 = d.1_3 + -1;
-  # RANGE ~[2147483647, 18446744071562067967]
-  _14 = (long unsigned int) _13;
-  _15 = _11 % _14;
+  _15 = _11 % _10;

is the key change, replacing (unsigned long)(intvar + -1) with (unsigned
long)intvar + 4294967295

It looks like

  /* For constants simply extend it.  */
  if (TREE_CODE (op) == INTEGER_CST)
    return wide_int_to_tree (wide_type, wi::to_wide (op));

doesn't sign-extend the -1 from int to unsigned log.  Using wi::to_widest
does and fixes the testcase.