Daniel-Constantin Mierla writes:

 > One possibility is caused by syslog -- syslog does synchronized writing, 
 > if it was writing when the kill is done will get locked by next log 
 > messages in destroy functions. It is hard to avoid it. 

this must have been the cause of processes not getting killed in my
case.  i thought that it had something to do with presence that i
started testing, but the real reason was that i run the tests at debug
level 7, which causes lots of syslog activity.  at debug level 3 init
restarts works fine.

i would guess that on a busy server there can be enough xlog messages
for the same problem to show up even at debug level 3.

-- juha

_______________________________________________
Devel mailing list
[email protected]
http://openser.org/cgi-bin/mailman/listinfo/devel

Reply via email to