Balazs Scheidler <bazsi@balabit.hu> wrote:
On Tue, Oct 15, 2002 at 10:17:04PM +0900, Katsuhiro Kondou wrote:
Any chance to fix dumping core with this version?
not yet, sorry. I once managed to cause a SEGFAULT, but the resulting backtrace was not usable. What seems to be true is that there's some kind of memory corruption going on. It seems to be related to TCP support, but don't I know whether destination or source. Narrowing down the problem to some component (e.g. only occurs with TCP sources or destinations)
I only see segfaults on our central syslog server, not clients, so I'd say source has been my experience.
Does it happen when only a simple, persistent TCP connection exists, or it is somehow related to connection establishment?
Hmm, I'll see if I can pinpoint that. I know nothing is ever written to disk (not a log line, files/directories aren't created, nothing) so it's before that happens.
I'd need to add some more debug info to the output (e.g. dmalloc or something similar).
So there are many questions, and I don't really have enough time now. It also doesn't help that it took several hours to get syslog-ng crash.
I've fixed the most important things (res_init(), the problem with binding sockets, etc.) that's 1.5.22 and libol 0.3.5
Luckily the patch posted to bugtraq applies to several versions :-)