Subject: Re: crashing kdeinit, anyone?
To: Hubert Feyrer <netbsd-users@NetBSD.org, feyrer@cs.stevens.edu>
From: Hubert Feyrer <hubertf@gmx.de>
List: netbsd-users
Date: 11/19/2005 06:04:52
In article <Pine.LNX.4.61.0511120055310.13897@m24s24.vlinux.de> you wrote:
>         (gdb) bt
>         #0  0x4926ffeb in kill () from /usr/lib/libc.so.12
>         #1  0x492e4a47 in abort () from /usr/lib/libc.so.12
>         #2  0x4871a4fb in kDebugBackend(unsigned short, unsigned, char const*) ()
>            from /usr/pkg/lib/libkdecore.so.4
>         #3  0x4871b032 in kdbgstream::flush() () from /usr/pkg/lib/libkdecore.so.4

After gooing through all the pain of adding support for 
PKG_OPTIONS_DEFAULT "debug" in qt3-libs, kdelibs3 and kdebase3 (and esp. 
rebuilding in my $HOME), I now know that the problem is not the core dump 
(as kDebugBackend() calls abort()), but that some communication problem to 
DCOP exists:

	(gdb) bt
	#0  0x494cafeb in kill () from /usr/lib/libc.so.12 
	#1  0x4953fa63 in abort () from /usr/lib/libc.so.12
	#2  0x4871b5db in kDebugBackend (nLevel=3, nArea=0,
	    data=0x8056c00 "FATAL: DCOP communication problem!\n") at kdebug.cpp:313
	#3  0x4871c112 in kdbgstream::flush() (this=0xbfbfefb0) at qmemarray.h:64

The error string passed to kDebugBackend() here says it.
Now I just have to find the cause of this. After some minor digging (as 
this happens only every 5 days or so), I found that the dcop server 
process was still running, I have no idea (yet) what part of the 
communication went wrong.


 - Hubert