bug_report

[Sergei Steshenko]

--- On Wed, 7/8/09, Thomas Weber <thomas.weber.mail at gmail.com> wrote:

> From: Thomas Weber <thomas.weber.mail at gmail.com>
> Subject: Re: bug_report
> To: "John W. Eaton" <jwe at octave.org>
> Cc: bug-octave at bevo.che.wisc.edu, "Ryszard Tanas" <tanas at zon1.physd.amu.edu.pl>
> Date: Wednesday, July 8, 2009, 2:28 PM
> On Wed, Jun 17, 2009 at 12:09:38PM
> -0400, John W. Eaton wrote:
> > On 17-Jun-2009, Ryszard Tanas wrote:
> > 
> > | --------
> > | Bug report for Octave 3.0.5 configured for
> x86_64-pc-linux-gnu
> > | 
> > | Description:
> > | -----------
> > | I get the following error with my Octave running on
> Debian testing, when I try to find eigenvalues of the
> complex 15x15 matrix
> > | 
> > | ** On entry to ZGEEV  parameter number  5
> had an illegal value
> > 
> > This message comes from LAPACK and indicates an
> incorrect call to
> > ZGEEV.
> > 
> > I can duplicate this error with the Octave 3.0.5 in
> Debian using the
> > 1:3.0.5-6 version of the package on a system running
> Debian testing
> > (updated in the last few days).
> > 
> > The problem does not appear for me with a copy of
> Octave that I built
> > from the current development sources on the same
> system.
> > 
> > With 3.2.0 from Debian unstable on the same system, I
> see
> > 
> >   panic: Segmentation fault -- stopping
> myself...
> >   attempting to save variables to
> `octave-core'...
> >   save to `octave-core' complete
> >   Segmentation fault
> 
> Hmm, I built 3.2.0 with debug symbols and -O0 and then
> Octave doesn't
> crash anymore. Building it with -O2 results in a crash.
> 
> I will try to gather more date about this one.
> 
>     Thomas

Depending on optimizations 'gcc' generates functionally different code in
case of, say, integer overflows.

I once was hit by this problem in 'libsndfile' (no, I do not claim to be
the author) and even filed a bug report against 'gcc' - this is how I know.

'gcc' developers do not consider such a behavior to be a bug.

Regards,
  Sergei.