[Bug c/38364] New: -ftrapv triggers integer miscompilation

[Bug c/38364] New: -ftrapv triggers integer miscompilation

[regehr at cs dot utah dot edu]

Seen on Ubuntu Hardy using r142342.

regehr@john-home:~/volatile/tmp73$ current-gcc -O2 -Wall small.c -o small
regehr@john-home:~/volatile/tmp73$ ./small
0
regehr@john-home:~/volatile/tmp73$ current-gcc -O2 -ftrapv -Wall small.c -o
small
regehr@john-home:~/volatile/tmp73$ ./small
1
regehr@john-home:~/volatile/tmp73$ current-gcc -v
Using built-in specs.
Target: i686-pc-linux-gnu
Configured with: ../configure --program-prefix=current-
--enable-languages=c,c++ --prefix=/home/regehr : (reconfigured) ../configure
--program-prefix=current- --enable-languages=c,c++ --prefix=/home/regehr :
(reconfigured) ../configure --program-prefix=current- --enable-languages=c,c++
--prefix=/home/regehr : (reconfigured) ../configure --program-prefix=current-
--prefix=/home/regehr --enable-languages=c,c++ --no-create --no-recursion
Thread model: posix
gcc version 4.4.0 20081202 (experimental) (GCC) 
regehr@john-home:~/volatile/tmp73$ cat small.c
#include <stdio.h>
#include <stdint.h>

static inline int16_t
safe_mul_int16_t_s_s (int16_t si1, int16_t si2)
{
  if (si1 > 0)
    {
      if (si2 > 0)
        {
          if (si1 > ((32767) / si2))
            {
              return si1;
            }
        }
      else
        {
          if (si2 < ((-32767 - 1) / si1))
            {
              return si1;
            }
        }
    }
  else
    {
      if (si2 > 0)
        {
          if (si1 < ((-32767 - 1) / si2))
            {
              return si1;
            }
        }
      else
        {
          if ((si1 != 0) && (si2 < ((32767) / si1)))
            {
              return si1;
            }
        }
    }
  return si1 * si2;
}

static inline int8_t
safe_sub_int8_t_s_s (int8_t si1, int8_t si2)
{
  if (((si1 ^ si2) &
       (((si1 ^ ((si1 ^ si2) & (1 << (sizeof (int8_t) * 8 - 1)))) -
         si2) ^ si2)) < 0)
    {
      return si1;
    }
  return si1 - si2;
}

uint32_t g_38;

int32_t func_53 (int32_t p_54, uint32_t p_55);
int32_t func_53 (int32_t p_54, uint32_t p_55)
{
  safe_mul_int16_t_s_s (1, 1);
  return 1;
}

int32_t func_97 (uint16_t p_99);
int32_t func_97 (uint16_t p_99)
{
  g_38 = p_99;
  return 1;
}

int32_t func_120 (int32_t p_124);
int32_t func_120 (int32_t p_124)
{
  if (safe_sub_int8_t_s_s (p_124, 1))
    safe_mul_int16_t_s_s (1, func_97 (1));
  return p_124;
}

int32_t func_14 (uint32_t p_16);
int32_t func_14 (uint32_t p_16)
{
  uint32_t l_438 = 0;
  func_97 (p_16 < (1 != func_120 (l_438)));
  return p_16;
}

int main (void)
{
  func_14 (1);
  printf ("%d\n", g_38);
  return 0;
}


-- 
           Summary: -ftrapv triggers integer miscompilation
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: regehr at cs dot utah dot edu
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38364

[Bug c/38364] -ftrapv triggers integer miscompilation

[regehr at cs dot utah dot edu]

------- Comment #1 from regehr at cs dot utah dot edu  2008-12-02 03:46 -------
Bizarrely, deleting the unused func_53() makes the problem go away.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38364

[Bug c/38364] [4.4 Regression] -ftrapv triggers integer miscompilation

[rguenth at gcc dot gnu dot org]

------- Comment #2 from rguenth at gcc dot gnu dot org  2008-12-02 11:15 -------
Confirmed.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu dot
                   |                            |org
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
           Keywords|                            |wrong-code
      Known to fail|                            |4.4.0
      Known to work|                            |4.3.2
   Last reconfirmed|0000-00-00 00:00:00         |2008-12-02 11:15:54
               date|                            |
            Summary|-ftrapv triggers integer    |[4.4 Regression] -ftrapv
                   |miscompilation              |triggers integer
                   |                            |miscompilation
   Target Milestone|---                         |4.4.0
            Version|unknown                     |4.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38364

[Bug c/38364] [4.4 Regression] -ftrapv triggers integer miscompilation

[jakub at gcc dot gnu dot org]

------- Comment #3 from jakub at gcc dot gnu dot org  2008-12-02 17:31 -------
Created an attachment (id=16812)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=16812&action=view)
pr38364.c

Deobfuscated testcase.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38364

[Bug c/38364] [4.4 Regression] -ftrapv triggers integer miscompilation

[jakub at gcc dot gnu dot org]

------- Comment #4 from jakub at gcc dot gnu dot org  2008-12-02 17:46 -------
This is a dup of PR38245, again, DCE removes the only call in the function, but
not the argument push, so that push clobbers slot where caller's %ebx is
preserved.

Either:
--- config/i386/i386.c.i386   2008-11-27 17:08:28.000000000 +0100
+++ config/i386/i386.c        2008-12-02 18:26:24.000000000 +0100
@@ -7566,14 +7566,8 @@ ix86_compute_frame_layout (struct ix86_f

   offset += size;

-  /* Add outgoing arguments area.  Can be skipped if we eliminated
-     all the function calls as dead code.
-     Skipping is however impossible when function calls alloca.  Alloca
-     expander assumes that last crtl->outgoing_args_size
-     of stack frame are unused.  */
-  if (ACCUMULATE_OUTGOING_ARGS
-      && (!current_function_is_leaf || cfun->calls_alloca
-          || ix86_current_function_calls_tls_descriptor))
+  /* Add outgoing arguments area.  */
+  if (ACCUMULATE_OUTGOING_ARGS)
     {
       offset += crtl->outgoing_args_size;
       frame->outgoing_arguments_size = crtl->outgoing_args_size;

fixes this (and other targets would need to be checked for this too), or
DCE needs to be taught to remove also the arguments (if that is at all
possible,
for ACCUMULATE_OUTGOING_ARGS that shouldn't be that hard, but for
-mno-accumulate-outgoing-args it would be much harder).

*** This bug has been marked as a duplicate of 38245 ***


-- 

jakub at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |DUPLICATE


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38364