bug in redhat gcc 2.96

bug in redhat gcc 2.96

[Jim Wright]

We believe we have found a bug in gcc.  We have been trying to track
down why the .../drivers/scsi/sym53c8xx.c driver oopses with a divide
by zero when initializing at line 5265, which reads:

        period = (4 * div_10M[0] + np->clock_khz - 1) / np->clock_khz;

We believe the bug is that gcc is generating incorrect code for this:

                if      (f1 < 55000)            f1 =  40000;
		else                            f1 =  80000;

Here is the test code to demonstrate this:

% cat bug.c
int main (int argc, char *argv[])
{
    unsigned f1;

    f1 = (unsigned)argc;

    if (f1 < 5) {
	f1 = 4;
    } else {
	f1 = 8;
    }
    exit (f1);
}

And here are commands to exhibit the problem.

% for i in 0 1 2 3 4 5 6 ; do ln bug.c bug$i.c ; done
% for i in 0 1 2 3 4 5 6 ; do gcc -save-temps -O$i -o bug$i bug$i.c ; done
% for i in 0 1 2 3 4 5 6 ; do ./bug$i 1 2 ; echo $? ; done
% for i in 0 1 2 3 4 5 6 ; do ./bug$i 1 2 3 4 5 6 7 ; echo $? ; done

The level 0 optimization assembly code appears correct.  For level 1 and
above, the compiler emits a long-subtract-with-borrow statement which
leaves EAX either 0 filled or 1 filled, based on the carry flag.

As this is with Red Hat's version of gcc, I'm not sending
this to the gcc folks.  RPMs of gcc with this problem
include gcc-2.96-69 and gcc-2.96-81.  This has been logged
as http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=39764.
Any suggestions for a way to cope with this?  We have a
customer who's system fails due to this.


-- 
Jim Wright   Software Engineer   Penguin Computing
jwright@penguincomputing.com   v:415-358-2609   f:415-358-2646

Re: bug in redhat gcc 2.96

[Matt Wilson]

This was fixed in 2.96-82, see:

http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=37054

It was a bug in the if conversion optimization.

We're at 2.96-84 in rawhide now.

ftp://ftp.redhat.com/pub/redhat/linux/rawhide/

Cheers,

Matt
msw@redhat.com

On Tue, May 08, 2001 at 08:05:06PM -0700, Jim Wright wrote:
> We believe we have found a bug in gcc.  We have been trying to track
> down why the .../drivers/scsi/sym53c8xx.c driver oopses with a divide
> by zero when initializing at line 5265, which reads:

Re: bug in redhat gcc 2.96

[Alan Cox]

> As this is with Red Hat's version of gcc, I'm not sending
> this to the gcc folks.  RPMs of gcc with this problem

(If you have the time check 3.0 CVS doesnt show the same problem, the RH tree
 diverges from it so may well be unique in having the bug but many bugs are
 shared)

> include gcc-2.96-69 and gcc-2.96-81.  This has been logged
> as http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=39764.

Thanks.

> Any suggestions for a way to cope with this?  We have a
> customer who's system fails due to this.

You can build 2.4 quite sanely with egcs-1.1.2 (aka kgcc)

Re: bug in redhat gcc 2.96

[Tobias Ringstrom]

On Wed, 9 May 2001, Alan Cox wrote:
> > Any suggestions for a way to cope with this?  We have a
> > customer who's system fails due to this.
>
> You can build 2.4 quite sanely with egcs-1.1.2 (aka kgcc)

Since there is no kgcc in RH71, will you be releasing an updated gcc
rpm, or is the best solution to download and compile egcs-1.1.2 from
source?

IMHO, it is best not to revert to an old egcs version, but instead
continue to find bugs in the upcoming 3.0 release.  I'm assuming that
your fixes for your gcc-2.96 are propagated to the pre-3.0 branch.

/Tobias

Re: bug in redhat gcc 2.96

[Jakub Jelinek]

On Wed, May 09, 2001 at 09:56:24AM +0100, Alan Cox wrote:
> > As this is with Red Hat's version of gcc, I'm not sending
> > this to the gcc folks.  RPMs of gcc with this problem
> 
> (If you have the time check 3.0 CVS doesnt show the same problem, the RH tree
>  diverges from it so may well be unique in having the bug but many bugs are
>  shared)

The bug was present in 3.0 and 3.1 as well, but has been fixed the day it
was reported. Use gcc-2.96-82 or above, or gcc-3_0-branch newer than 2001-04-25
(or CVS head newer than that).

	Jakub

Re: bug in redhat gcc 2.96

[David Woodhouse]

msw@redhat.com said:
> This was fixed in 2.96-82, see:
> http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=37054
> It was a bug in the if conversion optimization. 

So if I run up2date, will I get the fixed version?

--
dwmw2

Re: bug in redhat gcc 2.96

[Stefan Hoffmeister]

: On Wed, 9 May 2001 11:09:14 +0200 (CEST), Tobias Ringstrom wrote:

>On Wed, 9 May 2001, Alan Cox wrote:
>> > Any suggestions for a way to cope with this?  We have a
>> > customer who's system fails due to this.
>>
>> You can build 2.4 quite sanely with egcs-1.1.2 (aka kgcc)
>
>Since there is no kgcc in RH71, 

There is an compat-egcs RPM (on CD2?) that contains kgcc. Took me a while
to find that.

Re: bug in redhat gcc 2.96

[Jeremy Hogan]

This bug is fixed in gcc-2.96-82 and higher, as per
http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=37054. I've posted
gcc-2.96-84.src.rpm at your enterprise ftp folder.

--jeremy


Jim Wright wrote:
> 
> We believe we have found a bug in gcc.  We have been trying to track
> down why the .../drivers/scsi/sym53c8xx.c driver oopses with a divide
> by zero when initializing at line 5265, which reads:
> 
>         period = (4 * div_10M[0] + np->clock_khz - 1) / np->clock_khz;
> 
> We believe the bug is that gcc is generating incorrect code for this:
> 
>                 if      (f1 < 55000)            f1 =  40000;
>                 else                            f1 =  80000;
> 
> Here is the test code to demonstrate this:
> 
> % cat bug.c
> int main (int argc, char *argv[])
> {
>     unsigned f1;
> 
>     f1 = (unsigned)argc;
> 
>     if (f1 < 5) {
>         f1 = 4;
>     } else {
>         f1 = 8;
>     }
>     exit (f1);
> }
> 
> And here are commands to exhibit the problem.
> 
> % for i in 0 1 2 3 4 5 6 ; do ln bug.c bug$i.c ; done
> % for i in 0 1 2 3 4 5 6 ; do gcc -save-temps -O$i -o bug$i bug$i.c ; done
> % for i in 0 1 2 3 4 5 6 ; do ./bug$i 1 2 ; echo $? ; done
> % for i in 0 1 2 3 4 5 6 ; do ./bug$i 1 2 3 4 5 6 7 ; echo $? ; done
> 
> The level 0 optimization assembly code appears correct.  For level 1 and
> above, the compiler emits a long-subtract-with-borrow statement which
> leaves EAX either 0 filled or 1 filled, based on the carry flag.
> 
> As this is with Red Hat's version of gcc, I'm not sending
> this to the gcc folks.  RPMs of gcc with this problem
> include gcc-2.96-69 and gcc-2.96-81.  This has been logged
> as http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=39764.
> Any suggestions for a way to cope with this?  We have a
> customer who's system fails due to this.
> 
> --
> Jim Wright   Software Engineer   Penguin Computing
> jwright@penguincomputing.com   v:415-358-2609   f:415-358-2646

Re: bug in redhat gcc 2.96

[Dan Kegel]

Stefan Hoffmeister wrote:
> >Since there is no kgcc in RH71,
> 
> There is an compat-egcs RPM (on CD2?) that contains kgcc. Took me a while
> to find that.

OH.  I kept looking for a package called 'kgcc'.  Silly me.

Guess it's time for a "How to compile a kernel on Red Hat 7.1" FAQ.
- Dan

Re: bug in redhat gcc 2.96

[Bernhard Rosenkraenzer]

On Wed, 9 May 2001, Dan Kegel wrote:

> OH.  I kept looking for a package called 'kgcc'.  Silly me.
>
> Guess it's time for a "How to compile a kernel on Red Hat 7.1" FAQ.

The answer is that you can safely use gcc 2.96. No need to install kgcc or
any other old compiler.

LLaP
bero

Re: bug in redhat gcc 2.96

[Matt Wilson]

This is because Jakub fixed it in GCC 3.0 CVS at the same time that he
fixed it in 2.96-82, which was on the same day it was reported.  It
was broken in GCC CVS until that moment.

Cheers,

Matt

On Wed, May 09, 2001 at 09:56:24AM +0100, Alan Cox wrote:
> > As this is with Red Hat's version of gcc, I'm not sending
> > this to the gcc folks.  RPMs of gcc with this problem
> 
> (If you have the time check 3.0 CVS doesnt show the same problem, the RH tree
>  diverges from it so may well be unique in having the bug but many bugs are
>  shared)