>>>>> On Wed, 11 Oct 2006 09:56:52 +0200,
>>>>> Marco Schumann said:

>> Oops, sorry, the patch was not really correct. Please use this one
>> (attached below) instead.
>> But the actual behavior of these two patches should be the same...do
>> you mean named exited during initialization by 'refused to start'? Or
>> do you mean it existed just after starting? In any case, if it dumped
>> core, backtrace (of all threads) would be helpful as before.

> Hello,

> again the error:
> 11-Oct-2006 09:40:09.365 general: rbtdb.c:1128: INSIST(((unsigned
> int)((&rbtdb->node_locks[node->locknum].references)->refs))) failed
> 11-Oct-2006 09:40:09.365 general: exiting (due to assertion failure)

> Here the backtrace:

Thanks, and what about the other threads? Whether by
(gdb) thread apply all bt
or the way I showed.

> Next I am going to try your second patch. Would it be helpful to compile
> with debug flags? (I assume that and thus will use --enable-debug
> although not documented)

Not really, because BIND9 is built with debug flags by default. But
building it without compiler optimization flags might be a bit
helpful. I usually do that by

% CFLAGS=-g ./configure (--enable-threads)

but there maybe a better way (at least this won't work with csh).

JINMEI, Tatuya
Communication Platform Lab.
Corporate R&D Center, Toshiba Corp.