[Bug tree-optimization/23960] New: [4.1 regression] ICE in compare_values in VRP

[Bug tree-optimization/23960] New: [4.1 regression] ICE in compare_values in VRP

[steven at gcc dot gnu dot org]

typedef long unsigned int size_t; 
 
size_t __attribute__((noinline)) 
choose_block (char *d, size_t dlen, int col, 
              size_t *wlen) 
{ 
  size_t n, nn; 
 
  n = dlen; 
  for (;;) 
  { 
    ((void) ((d + n > d) ? 0 : (__assert_fail ("d + n > d", "rfc2047.c", 346, 
__PRETTY_FUNCTION__), 0))); 
    if (!nn && (col + *wlen <= 75 + 1 || n <= 1)) 
      break; 
  } 
  return n; 
} 
 
--> compile with -O2 
 
t.c: In function ?choose_block?: 
t.c:6: internal compiler error: in compare_values, at tree-vrp.c:415 
 
 
Breakpoint 1, fancy_abort (file=0xbd5846 "../../mainline/gcc/tree-vrp.c", 
line=415, 
    function=0xbd58b6 "compare_values") at diagnostic.c:590 
590       internal_error ("in %s, at %s:%d", function, trim_filename (file), 
line); 
(gdb) up 
#1  0x000000000094c645 in compare_values (val1=0x2a95a5a000, 
val2=0x2a95a4fe70) at tree-vrp.c:414 
414       gcc_assert (POINTER_TYPE_P (TREE_TYPE (val1)) 
(gdb) l 
409       if (val1 == val2) 
410         return 0; 
411 
412       /* Below we rely on the fact that VAL1 and VAL2 are both pointers or 
413          both integers.  */ 
414       gcc_assert (POINTER_TYPE_P (TREE_TYPE (val1)) 
415                   == POINTER_TYPE_P (TREE_TYPE (val2))); 
416 
417       /* Do some limited symbolic comparisons.  */ 
418       if (!POINTER_TYPE_P (TREE_TYPE (val1))) 
(gdb) p debug_tree_cfg (0) 
choose_block (d, dlen, col, wlen) 
{ 
  static const char __PRETTY_FUNCTION__[13] = "choose_block"; 
  size_t nn; 
  size_t n; 
  size_t D.1627; 
  _Bool D.1626; 
  _Bool D.1625; 
  _Bool D.1624; 
  size_t D.1623; 
  size_t D.1622; 
  size_t D.1621; 
  int iftmp.0; 
 
  # BLOCK 0 
  # PRED: ENTRY (fallthru,exec) 
  n_2 = dlen_1; 
  goto <bb 7> (<L10>); 
  # SUCC: 7 (fallthru,exec) 
 
  # BLOCK 1 
  # PRED: 3 (false,exec) 4 (false,exec) 
<L0>:; 
  # SUCC: 7 [100.0%]  (fallthru,exec) 
 
  # BLOCK 7 
  # PRED: 1 [100.0%]  (fallthru,exec) 0 (fallthru,exec) 
<L10>:; 
  if (dlen_1 == 0B) goto <L1>; else goto <L9>; 
  # SUCC: 2 (true,exec) 6 (false,exec) 
 
  # BLOCK 6 
  # PRED: 7 (false,exec) 
<L9>:; 
  dlen_14 = ASSERT_EXPR <dlen_1, dlen_1 != 0B>; 
  goto <bb 3> (<L3>); 
  # SUCC: 3 [100.0%]  (fallthru,exec) 
 
  # BLOCK 2 
  # PRED: 7 (true,exec) 
<L1>:; 
  __assert_fail (&"d + n > d"[0], &"rfc2047.c"[0], 346, &__PRETTY_FUNCTION__); 
  # SUCC: 3 (fallthru,exec) 
 
  # BLOCK 3 
  # PRED: 2 (fallthru,exec) 6 [100.0%]  (fallthru,exec) 
<L3>:; 
  if (nn_3 == 0) goto <L4>; else goto <L0>; 
  # SUCC: 4 (true,exec) 1 (false,exec) 
 
  # BLOCK 4 
  # PRED: 3 (true,exec) 
<L4>:; 
  D.1621_5 = (size_t) col_4; 
  D.1622_7 = *wlen_6; 
  D.1623_8 = D.1621_5 + D.1622_7; 
  D.1624_9 = D.1623_8 <= 76; 
  D.1625_10 = dlen_1 <= 1; 
  D.1626_11 = D.1624_9 || D.1625_10; 
  if (D.1626_11) goto <L7>; else goto <L0>; 
  # SUCC: 5 (true,exec) 1 (false,exec) 
 
  # BLOCK 5 
  # PRED: 4 (true,exec) 
<L7>:; 
  n_12 = dlen_1; 
  return dlen_1; 
  # SUCC: EXIT 
 
} 
 
 
$1 = void 
(gdb) p debug_generic_expr (val1) 
dlenD.1609_1 
$2 = void 
(gdb) p debug_generic_expr (val2) 
0B 
(gdb) 
 
 
This bug shows up when building mutt.

-- 
           Summary: [4.1 regression] ICE in compare_values in VRP
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P2
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: steven at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23960

[Bug tree-optimization/23960] [4.1 regression] ICE in compare_values in VRP

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-09-19 05:52 -------
Redcued testcase:
void choose_block (char *d, __SIZE_TYPE__ n) 
{ 
  if (!(d + n > d))
    __builtin_abort ();
} 


With the comparision patch from 22368, we get:
t.c:5: error: types mismatch in comparsion
long unsigned intD.4
charD.1 *
nD.1270 == 0B;

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pinskia at gcc dot gnu dot
                   |                            |org
OtherBugsDependingO|                            |22368
              nThis|                            |
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-09-19 05:52:48
               date|                            |
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23960

[Bug middle-end/23960] [4.1 regression] ICE in compare_values in VRP

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-09-19 05:58 -------
The bug is in fold-const.c.
In fold_binary around line 9288 where we don't convert the expressions.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|tree-optimization           |middle-end


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23960

[Bug middle-end/23960] [4.1 regression] ICE in compare_values in VRP

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
OtherBugsDependingO|                            |23968
              nThis|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23960

[Bug middle-end/23960] [4.1 regression] ICE in compare_values in VRP

[kazu at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |kazu at gcc dot gnu dot org
                   |dot org                     |
             Status|NEW                         |ASSIGNED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23960

[Bug middle-end/23960] [4.1 regression] ICE in compare_values in VRP

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-09-26 01:29 -------
Subject: Bug 23960

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	kazu@gcc.gnu.org	2005-09-26 01:28:51

Modified files:
	gcc            : ChangeLog fold-const.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.c-torture/compile: pr23960.c 

Log message:
	gcc/
	PR tree-optimization/23960
	* fold-const.c (fold_binary): Use op0 and op1 instead of arg0
	and arg1 if we are passing them to fold_build2.
	
	gcc/testsuite/
	PR tree-optimization/23960
	* gcc.c-torture/compile/pr23960.c: New.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.10019&r2=2.10020
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/fold-const.c.diff?cvsroot=gcc&r1=1.625&r2=1.626
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.6101&r2=1.6102
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/compile/pr23960.c.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23960

[Bug middle-end/23960] [4.1 regression] ICE in compare_values in VRP

[kazu at gcc dot gnu dot org]

------- Additional Comments From kazu at gcc dot gnu dot org  2005-09-26 01:29 -------
Just checked in a patch.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23960