[Bug c++/18084] New: setfill coupled with inline function: incorrect results on Linux x86

[Bug c++/18084] New: setfill coupled with inline function: incorrect results on Linux x86

[knutejunk at austin dot rr dot com]

The problem...

The following program produces incorrect results on a Linux x86 system when 
compiled with gcc v3.4.2 (and 3.4.1) with full optimizations.  The program is 
supposed to print 0xFFFF, but instead produces random results (seems to be a 
stack corruption problem).

The source...

#include <inttypes.h>
#include <iostream>
#include <iomanip>


#define BUG

using namespace std;

class A
{
public:
   
   A () :
      valid_mask_(0x00000000FFFFFFFFull),
      size_(0x000000000000FFFFull)
   {}
      
   uint64_t calc(uint64_t value) const {
#ifdef BUG
      return valid_mask_ & (value | size_);
#else
      // This code will never expose the bug
      uint64_t tmp = valid_mask_ & (value | size_);
      return tmp;
#endif
   }

private:
   uint64_t valid_mask_;
   uint64_t size_;
   
};

uint32_t CheckStuff()
{
   A a;
   uint64_t value = a.calc(0);
#ifdef BUG
   // This code will expose the bug iff BUG is defined for calc()
   cout << "0x" << hex << setfill('0') << value << endl;
#else
   // This code will never expose the bug
   cout << "0x" << hex << value << endl;
#endif
   if(value == 0xffff) {
      return 0;
   }
   else {
      return 1;
   }
}


int main()
{
   return CheckStuff();
}

The build line...

   gcc-3.4.1 -v -W -Wall -O3 -g2 gcc_bug.cpp

The results (it's purely random, but here's an example)...

   0x8048978bfffbb68

What it should print...

   0xFFFF

Observations...

   - Take out the -O3 (-O0) and it works
   - This program works just fine on Solaris no matter how you compile it
   - This program works with gcc 3.3.X, gcc 3.2.X and (shudder) 2.95
   - Take out setfill('0') and it works
   - Keep setfill('0') in the cout statement and return a tmp for the inline 
calc() function and it works

Compiler/system info...

system type:

   Linux x86 (RedHat Enterprise 3) 

gcc -v output:

--enable-shared --with-ld=/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-
linux2.4/bin/ld 

--with-as=/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/bin/as 

--with-nm=/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/bin/nm 

--enable-languages=c,c++,java --disable-nls
Thread model: posix
gcc version 3.4.1
 

/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/libexec/gcc/i686-pc-
linux-gnu/3.4

.1/cc1plus -quiet -v -D_GNU_SOURCE gcc_bug.cpp -quiet -dumpbase gcc_bug.cpp 

-mtune=pentiumpro -auxbase gcc_bug -g2 -O0 -W -Wall -version -o /tmp/cc82jgXc.s
ignoring nonexistent directory 

"/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/lib/gcc/i686-pc-linux-
gnu/3.4.1/

../../../../i686-pc-linux-gnu/include"
#include "..." search starts here:
#include <...> search starts here:
 

/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/lib/gcc/i686-pc-linux-
gnu/3.4.1/.

./../../../include/c++/3.4.1
 

/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/lib/gcc/i686-pc-linux-
gnu/3.4.1/.

./../../../include/c++/3.4.1/i686-pc-linux-gnu
 

/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/lib/gcc/i686-pc-linux-
gnu/3.4.1/.

./../../../include/c++/3.4.1/backward
 /usr/local/include
 /<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/include
 

/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/lib/gcc/i686-pc-linux-
gnu/3.4.1/i

nclude
 /usr/include
End of search list.
GNU C++ version 3.4.1 (i686-pc-linux-gnu)
	compiled by GNU C version 3.4.1.
GGC heuristics: --param ggc-min-expand=100 --param ggc-min-heapsize=131072
 /<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/bin/as -V -Qy -o /tmp/
cc15JK08.o 

/tmp/cc82jgXc.s
GNU assembler version 2.13.2.1 (i686-pc-linux-gnu) using BFD version 2.13.2.1
 

/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/libexec/gcc/i686-pc-
linux-gnu/3.4

.1/collect2 --eh-frame-hdr -m elf_i386 -dynamic-linker /lib/ld-linux.so.2 -rpath 

/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/lib /usr/lib/crt1.o 

/usr/lib/crti.o 

/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/lib/gcc/i686-pc-linux-
gnu/3.4.1/c

rtbegin.o 

-L/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/lib/gcc/i686-pc-
linux-gnu/3.4.1 

-L/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/lib/gcc/i686-pc-
linux-gnu/3.4.1

/../../../../i686-pc-linux-gnu/lib 

-L/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/lib/gcc/i686-pc-
linux-gnu/3.4.1

/../../.. /tmp/cc15JK08.o -lstdc++ -lm -lgcc_s -lgcc -lc -lgcc_s -lgcc 

/<blah>/gnu-gcc-3.4.1-binutils-2.13.2.1/i686-pc-linux2.4/lib/gcc/i686-pc-linux-
gnu/3.4.1/c

rtend.o /usr/lib/crtn.o

-- 
           Summary: setfill coupled with inline function: incorrect results
                    on Linux x86
           Product: gcc
           Version: 3.4.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: knutejunk at austin dot rr dot com
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18084

[Bug rtl-optimization/18084] [3.4 only]setfill coupled with inline function: incorrect results on Linux x86

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-20 17:09 -------
It works on the mainline but not on 3.4.0 and I don't know if this is a regression.

But I should note that -O2 is enough and -O2 -fno-gcse works.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|c++                         |rtl-optimization
           Keywords|                            |wrong-code
      Known to fail|                            |3.4.0
      Known to work|                            |4.0.0
            Summary|setfill coupled with inline |[3.4 only]setfill coupled
                   |function: incorrect results |with inline function:
                   |on Linux x86                |incorrect results on Linux
                   |                            |x86
   Target Milestone|---                         |3.4.3


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18084

[Bug rtl-optimization/18084] [3.4 only]setfill coupled with inline function: incorrect results on Linux x86

[bangerth at dealii dot org]

------- Additional Comments From bangerth at dealii dot org  2004-10-20 18:00 -------
Here's something slightly smaller: 
------------------ 
#include <inttypes.h> 
#include <iostream> 
#include <iomanip> 
 
uint64_t mask1, mask2; 
 
uint64_t calc() { 
  return mask1 & mask2; 
} 
 
int main() 
{ 
  mask1 = 0x00000000FFFFFFFFull; 
  mask2 = 0x000000000000FFFFull; 
  uint64_t value = calc(); 
 
  std::cout << std::setfill('0'); 
 
  if(value != calc()) 
    abort (); 
} 
------------------------------ 
g/x> /home/bangerth/bin/gcc-3.4.3-pre/bin/c++ -O3 x.cc ; ./a.out  
Aborted 
 
I'll keep working on reducing this. 
 
W. 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18084

[Bug rtl-optimization/18084] [3.4 only]setfill coupled with inline function: incorrect results on Linux x86

[bangerth at dealii dot org]

------- Additional Comments From bangerth at dealii dot org  2004-10-25 14:24 -------
Here is a short testcase: 
---------------------- 
struct X { 
    bool init; 
    void foo() { if (!init) init = true; } 
    void bar() { foo();                  } 
 
}; 
 
typedef unsigned long long int uint64_t; 
uint64_t mask1, mask2; 
 
uint64_t calc() { 
  return mask1 & mask2; 
} 
 
int main() 
{ 
  mask1 = 0x00000000FFFFFFFFull; 
  mask2 = 0x000000000000FFFFull; 
  uint64_t value = calc(); 
 
  X().bar(); 
 
  if(value != calc()) 
    abort (); 
} 
------------------------------ 
g/x> /home/bangerth/bin/gcc-3.4.3-pre/bin/c++ y.cc && ./a.out 
 
g/x> /home/bangerth/bin/gcc-3.4.3-pre/bin/c++ -O3 y.cc && ./a.out 
Aborted 
 
g/x> /home/bangerth/bin/gcc-3.3.4-pre/bin/c++ -O3 y.cc && ./a.out 
 
g/x> /home/bangerth/bin/gcc-4.0-pre/bin/c++ -O3 y.cc && ./a.out 
 
Thus, the problem is confined to the 3.4 branch (which does not mean, 
of course, that the bug isn't latent on mainline as well). 
 
W. 

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2004-10-25 14:24:46
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18084

[Bug rtl-optimization/18084] [3.4 regression]setfill coupled with inline function: incorrect results on Linux x86

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-25 14:56 -------
It also works in 3.3.3.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|4.0.0                       |4.0.0 3.3.3
            Summary|[3.4 only]setfill coupled   |[3.4 regression]setfill
                   |with inline function:       |coupled with inline
                   |incorrect results on Linux  |function: incorrect results
                   |x86                         |on Linux x86


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18084

[Bug rtl-optimization/18084] [3.4 regression]setfill coupled with inline function: incorrect results on Linux x86

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-10-30 22:58 -------
Subject: Bug 18084

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	sayle@gcc.gnu.org	2004-10-30 22:58:49

Modified files:
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/g++.dg/opt: pr18084-1.C 

Log message:
	PR rtl-optimization/18084
	* g++.dg/opt/pr18084-1.C: New testcase.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.4525&r2=1.4526
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/opt/pr18084-1.C.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18084

[Bug rtl-optimization/18084] [3.4 regression]setfill coupled with inline function: incorrect results on Linux x86

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-30 23:06 -------


*** This bug has been marked as a duplicate of 17581 ***

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18084

[Bug rtl-optimization/18084] [3.4 regression]setfill coupled with inline function: incorrect results on Linux x86

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-10-31 01:57 -------
Subject: Bug 18084

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-3_4-branch
Changes by:	sayle@gcc.gnu.org	2004-10-31 01:57:39

Modified files:
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/g++.dg/opt: pr18084-1.C 

Log message:
	PR rtl-optimization/18084
	* g++.dg/opt/pr18084-1.C: New testcase.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.3389.2.301&r2=1.3389.2.302
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/opt/pr18084-1.C.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=NONE&r2=1.1.2.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18084