Posible bug in gcc

Posible bug in gcc

[David]

[-- Attachment #1: Type: text/plain, Size: 858 bytes --]

I hope you will find this information usefull.

I am not in the linux-kernel list so, if posible, I would like to be
personally CC'ed the answers/comments sent to the list in response to
this posting.

I think I heve found a bug in gcc. I have tried both egcs 1.1.2 (gcc
2.91.66) and gcc 2.95.2 versions.

I am attaching you a simplified test program ('bug.c', a really simple
program).

To generate the faulty program from correct code compile as:
  gcc -O2 -o bug bug.c

You can generate good code in two ways:
1. Compiling with:
  gcc -fno-strength-reduce -O2 -o bug bug.c

   So the problem is with the option -fstrength-reduce which is 
   active with the common '-O2' optimization option.

2. Uncomment the printf at line 34. Bugs are surprising.

I have also sent the bug report to the gcc maintainers.

Is it really a bug?

Thank you,

David Llorens.

[-- Attachment #2: bug.c --]
[-- Type: text/plain, Size: 862 bytes --]

/* 
#include <stdio.h>
*/

#define SMALL_N  2
#define NUM_ELEM 4

int main(void)
{
  int listElem[NUM_ELEM]={30,2,10,5};
  int listSmall[SMALL_N];
  int i, j;
  int posGreatest=-1, greatest=-1;

  for (i=0; i<SMALL_N; i++) { 
    listSmall[i] = listElem[i];
    if (listElem[i] > greatest) {
      posGreatest = i;
      greatest = listElem[i];
    }
  }
  
  for (i=SMALL_N; i<NUM_ELEM; i++) { 
    if (listElem[i] < greatest) {
      listSmall[posGreatest] = listElem[i];
      posGreatest = 0;
      greatest = listSmall[0];
      for (j=1; j<SMALL_N; j++) 
	if (listSmall[j] > greatest) {
	  posGreatest = j;
	  greatest = listSmall[j];
	}
      /*
      printf("%d\n", posGreatest);
      */
    }
  }
 
  printf("Correct output: 5 2\n");
  printf("GCC output: ");
  for (i=0; i<SMALL_N; i++) printf(" %.1d", listSmall[i]);
  printf("\n");
  return (1);
}

Re: Posible bug in gcc

[Alan Cox]

> I think I heve found a bug in gcc. I have tried both egcs 1.1.2 (gcc
> 2.91.66) and gcc 2.95.2 versions.
> 
> I am attaching you a simplified test program ('bug.c', a really simple
> program).

Well gcc-bugs would be the better place to send it but this is a known problem
fixed in CVS gcc 2.95.3, CVS gcc 3.0 branch and gcc 2.96 (unofficial, Red Hat)

Re: Posible bug in gcc

[Jakub Jelinek]

On Mon, Feb 26, 2001 at 05:15:28PM +0000, Alan Cox wrote:
> > I think I heve found a bug in gcc. I have tried both egcs 1.1.2 (gcc
> > 2.91.66) and gcc 2.95.2 versions.
> > 
> > I am attaching you a simplified test program ('bug.c', a really simple
> > program).
> 
> Well gcc-bugs would be the better place to send it but this is a known problem
> fixed in CVS gcc 2.95.3, CVS gcc 3.0 branch and gcc 2.96 (unofficial, Red Hat)

I'm not sure if it is known, at least not known to me, but definitely not
fixed in any of gcc 2.95.2, CVS gcc 3.0 branch, CVS gcc 3.1 head, gcc 2.96-RH.

	Jakub

Re: Posible bug in gcc

[Alan Cox]

> > Well gcc-bugs would be the better place to send it but this is a known problem
> > fixed in CVS gcc 2.95.3, CVS gcc 3.0 branch and gcc 2.96 (unofficial, Red Hat)
> 
> I'm not sure if it is known, at least not known to me, but definitely not
> fixed in any of gcc 2.95.2, CVS gcc 3.0 branch, CVS gcc 3.1 head, gcc 2.96-RH.

Sorry my error for assuming it was the exsting known strength reduce bug

Re: Posible bug in gcc

[Richard B. Johnson]

On Mon, 26 Feb 2001, Alan Cox wrote:

> > > Well gcc-bugs would be the better place to send it but this is a known problem
> > > fixed in CVS gcc 2.95.3, CVS gcc 3.0 branch and gcc 2.96 (unofficial, Red Hat)
> > 
> > I'm not sure if it is known, at least not known to me, but definitely not
> > fixed in any of gcc 2.95.2, CVS gcc 3.0 branch, CVS gcc 3.1 head, gcc 2.96-RH.
> 
> Sorry my error for assuming it was the exsting known strength reduce bug
> 

Script started on Mon Feb 26 12:54:20 2001
# gcc -o xxx bug.c
# ./xxx
Correct output: 5 2
GCC output:  5 2
# gcc --version
egcs-2.91.66
# gcc -O2 -o xxx bug.c
# ./xxx
Correct output: 5 2
GCC output:  10 5
# exit
exit

Script done on Mon Feb 26 12:55:21 2001

Definitely has something to do with broken optimization. No optimization,
no bug, turn on '-O2' and you have the bug.


Cheers,
Dick Johnson

Penguin : Linux version 2.4.1 on an i686 machine (799.53 BogoMips).

"Memory is like gasoline. You use it up when you are running. Of
course you get it all back when you reboot..."; Actual explanation
obtained from the Micro$oft help desk.

Re: Posible bug in gcc

[Ville Herva]

On Mon, Feb 26, 2001 at 01:02:45PM -0500, you [Richard B. Johnson] claimed:
> 
> Script started on Mon Feb 26 12:54:20 2001
> # gcc -o xxx bug.c
> # ./xxx
> Correct output: 5 2
> GCC output:  5 2
> # gcc --version
> egcs-2.91.66
> # gcc -O2 -o xxx bug.c
> # ./xxx
> Correct output: 5 2
> GCC output:  10 5
> # exit
> exit

Funny:

vherva@babbage:/tmp>/usr/bin/gcc  c.c -o c; ./c ; /usr/bin/gcc --version 
Correct output: 5 2
GCC output:  5 2
2.96
vherva@babbage:/tmp>/usr/bin/gcc -O2 c.c -o c; ./c ; /usr/bin/gcc --version
Correct output: 5 2
GCC output:  10 5
2.96
vherva@babbage:/tmp>/usr/bin/gcc -O6 c.c -o c; ./c ; /usr/bin/gcc --version
Correct output: 5 2
GCC output:  10 5
2.96
vherva@babbage:/tmp>rpm -q gcc
gcc-2.96-74

vherva@babbage:/tmp>kgcc c.c  -o c; ./c ; kgcc --version 
Correct output: 5 2
GCC output:  5 2
egcs-2.91.66
vherva@babbage:/tmp>kgcc c.c -O2 -o c; ./c ; kgcc --version
Correct output: 5 2
GCC output:  10 5
egcs-2.91.66
vherva@babbage:/tmp>kgcc c.c -O6 -o c; ./c ; kgcc --version
Correct output: 5 2
GCC output:  10 5
egcs-2.91.66
vherva@babbage:/tmp>rpm -q kgcc
kgcc-1.1.2-40

vherva@babbage:/tmp>/usr/local/bin/gcc c.c -o c; ./c ;/usr/local/bin/gcc --version    
Correct output: 5 2
GCC output:  5 2
pgcc-2.95.1
vherva@babbage:/tmp>/usr/local/bin/gcc c.c -O2 -o c; ./c ;/usr/local/bin/gcc --version
Correct output: 5 2
GCC output:  5 2
pgcc-2.95.1
vherva@babbage:/tmp>/usr/local/bin/gcc c.c -O6 -o c; ./c ;/usr/local/bin/gcc --version 
Correct output: 5 2
GCC output:  5 2
pgcc-2.95.1

I guess pgcc is not that buggy EVERY time. (Sorry for the off topic post, I
couldn't resist.)


-- v --

v@iki.fi

Re: Posible bug in gcc

[J . A . Magallon]

On 02.26 Alan Cox wrote:
Also fails in gcc-2.96-0.38mdk (Mandrake Cooker):
rpm -q --changelog gcc
* Sat Feb 17 2001 Chmouel Boudjnah <chmouel@mandrakesoft.com> 2.96-0.38mdk

- exit 0 if [ $1 = 0 ] if we are in %postun (to don't screwd up the
  alternatives).

* Thu Feb 15 2001 David BAUDENS <baudens@mandrakesoft.com> 2.96-0.37mdk

- Fix build on PPC :)

* Thu Feb 15 2001 Chmouel Boudjnah <chmouel@mandrakesoft.com> 2.96-0.36mdk

- Break build on PPC ;).
- Red Hat patches, Jakub Jelinek (rel74) 5 new patches :

-- 
J.A. Magallon                                                      $> cd pub
mailto:jamagallon@able.es                                          $> more beer

Linux werewolf 2.4.2-ac4 #2 SMP Mon Feb 26 00:21:23 CET 2001 i686

Re: Posible bug in gcc

[David Relson]

At 01:02 PM 2/26/01, Alan Cox wrote:
> > > Well gcc-bugs would be the better place to send it but this is a 
> known problem
> > > fixed in CVS gcc 2.95.3, CVS gcc 3.0 branch and gcc 2.96 (unofficial, 
> Red Hat)
> >
> > I'm not sure if it is known, at least not known to me, but definitely not
> > fixed in any of gcc 2.95.2, CVS gcc 3.0 branch, CVS gcc 3.1 head, gcc 
> 2.96-RH.
>
>Sorry my error for assuming it was the exsting known strength reduce bug


It's broken in my copy of gcc.2.95.3 ...

David

Re: Posible bug in gcc

[J . A . Magallon]

On 02.26 David wrote:
> I hope you will find this information usefull.
> 
> I am not in the linux-kernel list so, if posible, I would like to be
> personally CC'ed the answers/comments sent to the list in response to
> this posting.
> 
> I think I heve found a bug in gcc. I have tried both egcs 1.1.2 (gcc
> 2.91.66) and gcc 2.95.2 versions.
> 

gcc2.95.2 is sane in irix6.2, irix6.5 and solaris7sparc.

The optimizer is not in the common front-end ?

-- 
J.A. Magallon                                                      $> cd pub
mailto:jamagallon@able.es                                          $> more beer

Linux werewolf 2.4.2-ac4 #2 SMP Mon Feb 26 00:21:23 CET 2001 i686

Re: Posible bug in gcc

[Erik Mouw]

On Mon, Feb 26, 2001 at 11:30:13PM +0100, J . A . Magallon wrote:
> On 02.26 David wrote:
> > I think I heve found a bug in gcc. I have tried both egcs 1.1.2 (gcc
> > 2.91.66) and gcc 2.95.2 versions.
> 
> gcc2.95.2 is sane in irix6.2, irix6.5 and solaris7sparc.
> 
> The optimizer is not in the common front-end ?

Yes and no. There is some common code in the optimiser, but each target
has its own optimisation tricks for which the code is not shared with
other targets.


Erik

-- 
J.A.K. (Erik) Mouw, Information and Communication Theory Group, Department
of Electrical Engineering, Faculty of Information Technology and Systems,
Delft University of Technology, PO BOX 5031,  2600 GA Delft, The Netherlands
Phone: +31-15-2783635  Fax: +31-15-2781843  Email: J.A.K.Mouw@its.tudelft.nl
WWW: http://www-ict.its.tudelft.nl/~erik/