[bind10-dev] python Segmentation fault in notify_out_test.py

[Jelte Jansen]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 10/31/2012 10:21 AM, Shane Kerr wrote:
> Jeremy,
> 
> On Tuesday, 2012-10-30 15:41:38 -0500, "Jeremy C. Reed" 
> <jreed at isc.org> wrote:
>> Some other tests (in other components) also failed and backtraces
>> also showed log4cplus.  I looked at the failure history and it
>> started 2012-10-11 which was same day that I upgraded from
>> log4cplus-1.0.4 to 1.1.0.
>> 
>> I downgraded just now and the segmentation faults are now gone.
>> 
>> Something about our python wrappers for log4cplus and log4cplus 
>> 1.1.0 is causing python to crash on this NetBSD 6.0 i386 system.
> 
> Do we use log4cplus 1.1.0 on any other systems?
> 

I think we do, but not sure where

> I'm just wondering if this is a log4cplus 1.1.0 + NetBSD problem or
> simply a log4cplus 1.1.0 problem. :)
> 

or a problem in our code which only happens to be triggered by 1.1.0
(on netbsd) ;)

Yes I have been able to reproduce the error now. (there are some
related problems I found building log4cplus and bind10 with a
nonstandard log4cplus lib on netbsd, btw, either libtool is has an
issue or we don't link correctly) but that is probably something we
need to discuss separately).

Have not found the real cause yet, it may be that log4cplus needs some
special magic to work in a threaded environment (which is the test
that segfaults)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAlCREQ8ACgkQ4nZCKsdOncWzOwCePVOkX8h+7xBvZ6CvzsU/m+fM
M2YAni7UVI5LpgivBSRYzhEyd/MxxRl5
=lnHT
-----END PGP SIGNATURE-----