Bob, Mary,

Thanks for your quick response. 

I've looked a bit more into stack smashing protection (SSP), and here is
what I found. SSP was a patch in GCC 3.x, but became included in GCC 4.1
release. However, in Debian, SSP code generation is by default still
turned off, and to enable it you must compile with "-fstack-protector"
flag added. Recent Ubuntu releases are the other way around - SSP code
generation is by default on. 

My guess is that if you try recompiling with "-fstack-protector", you'll
be able to reproduce the problem on your Debian box.

By the way, it may seem strange that program works fine without SSP, and
then starts crashing with SSP. However, this is not uncommon. Small
overflows of stack-allocated buffers may not be fatal, depending on what
was overridden, and when. For example, if local variables or function
arguments get overridden past the point of their use, this will not
affect program execution at all. Yet, SSP reorders local variables on
stack such that char[] buffers come at the end just before the special
stack guard variable - so previously hidden bugs suddenly become
visible. 

Also, valgrind and electric fence are unlikely to help here - AFAIK they
only check heap-allocated memory, and don't help at all with
stack-allocated memory. 

I've found a good discussion of SSP here:
http://www.trl.ibm.com/projects/security/ssp/

Please let me know if you were able to reproduce the problem after
recompiling with "-fstack-protector". If not, I'll do my best to debug
the problem on my own box. Also, I'll try to recompile with
"-fno-stack-protector". If my theory above is correct, the problem will
just go away, sending the bug in filter_license back into hiding.  

Best regards,
-Dragoslav

> -----Original Message-----
> From: Bob Gobeille [mailto:[email protected]]
> Sent: Friday, July 31, 2009 1:58 PM
> To: Laser, Mary
> Cc: Mitrinovic Dragoslav-FDM224; [email protected]
> Subject: Re: [FOSSology] stack smashing problem with filter_license
agent
> 
> Great.  The bug contains a test file.  I can't reproduce this problem
> on my debian system.
> 
> Unless Dragoslav can debug for us, I need someone to set this up on
> Ubuntu 9.04.
> 
> Bob
> 
> 
> On Jul 31, 2009, at 12:35 PM, Laser, Mary wrote:
> 
> > FYI - there is a bug filed against this problem with several
> > attachments that may be helpful
> > http://bugs.linux-foundation.org/show_bug.cgi?id=350
> >
> > Mary
> >
> >
> >> -----Original Message-----
> >> On Jul 30, 2009, at 4:29 PM, Mitrinovic Dragoslav-FDM224 wrote:
> >>
> >>> Team,
> >>>
> >>> I am running 1.1.0~rc5 release candidate, which I've rebuilt on my
> >>> Ubuntu 9.04 box using unmodified rc5 debian source package.
> >>>
> >>> I am having problems with Filter_License terminating prematurely.
> >>> Eventually, all activity dies down and analysis remains stuck,
even
> >>> though scheduler is alive, as evidenced by recent "update time" in
> >>> Admin->Scheduler->Status.

_______________________________________________
fossology mailing list
[email protected]
http://fossology.org/mailman/listinfo/fossology

Reply via email to