On Sat, 15 May 2004 14:29:40 +0200, Pascal Cabaud <pc@eila.jussieu.fr>  
wrote:

> Le Tue, May 11, 2004 at 02:42:17PM +0100, Stephen Borrill disait :
>> I've a client with a server with apache-1.3.29 (plus some other packages
>> such as ap-php) and openssl-0.9.6m. Occasionally, apache is core-dumping
>> with a segmentation fault. GDB reports:
>>
>> (gdb) backtrace
>> #0  0x4898f0dc in ripemd160_md () from /usr/pkg/lib/libcrypto.so.300
>> #1  0x806d8d9 in unique_id_global_init ()
>> #2  0x8073bde in ap_init_modules ()
>> #3  0x807bb9a in standalone_main ()
>> #4  0x807c3f7 in main ()
>> #5  0x80500b0 in ___start ()
>>
>> While ap-ssl is installed, ssl is not configured at all in Apache.
>> It doesn't even LoadModule mod_ssl.
>>
>> unique_id_global_init() is in mod_unique_id. This is a standard module  
>> and
>> cannot be disabled.
>>
>> Any ideas?
>
> I've the same problem, Apache core dump when receiving SIGHUP from
> newsyslog(8). As I rotate logs at fixed time, a crontab(5) "correct"  
> this.

I'm not using newsyslog to rotate the Apache logs, I'm using a home-grown  
script.
But you're right, the timestamp of the core dump is when the script runs  
"apachectl graceful".

-- 
Stephen