Bug 64671 - Kded uses 100% of the CPU on first Wallet access
Summary: Kded uses 100% of the CPU on first Wallet access
Status: RESOLVED FIXED
Alias: None
Product: kwin
Classification: Plasma
Component: general (show other bugs)
Version: unspecified
Platform: Compiled Sources Linux
: NOR normal
Target Milestone: ---
Assignee: Lubos Lunak
URL:
Keywords:
: 65286 66928 70754 74126 74917 78212 81144 81493 (view as bug list)
Depends on:
Blocks:
 
Reported: 2003-09-21 18:47 UTC by Josh Berry
Modified: 2008-08-20 12:41 UTC (History)
11 users (show)

See Also:
Latest Commit:
Version Fixed In:


Attachments
qt patch (1013 bytes, patch)
2003-10-06 11:32 UTC, Lubos Lunak
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Josh Berry 2003-09-21 18:47:59 UTC
Version:            (using KDE Devel)
Installed from:    Compiled sources
Compiler:          gcc 3.2.x 
OS:          Linux

When first logging in, KDED behaves normally (i.e. sits quietly in the background not using CPU). But when I start Kopete and enter my KWallet password, KDED suddenly starts eating 100% of the CPU.

After killing and restarting kded, the problem seems to go away, even if the wallet has to be reopened.
Comment 1 Maksim Orlovich 2003-09-21 18:57:53 UTC
In case you haven't seen this.. 
 
Comment 2 George Staikos 2003-09-21 19:09:22 UTC
please break in with gdb (gdb kdeinit <pid of kded>) and provide a backtrace 
of where it is.  you can get a backtrace by typing "where". 
 
Thanks 
Comment 3 Josh Berry 2003-09-21 19:27:03 UTC
OK. One minor correction. Steps to reproduce this bug are as follows: 
 
- Login to KDE 
- Start Kopete and enter my Wallet password 
- Start Konqueror and visit a webpage that requires wallet data 
- Close Konqueror 
 
Kded begins eating the CPU. Here's the backtrace: 
 
#0  0x41455912 in select () from /lib/libc.so.6 
#1  0x40f71b1c in __JCR_LIST__ () from /opt/qtcvs/lib/libqt-mt.so.3 
#2  0x40aec1a6 in QEventLoop::enterLoop() () from /opt/qtcvs/lib/libqt-mt.so.3 
#3  0x40aec048 in QEventLoop::exec() () from /opt/qtcvs/lib/libqt-mt.so.3 
#4  0x40ad91a1 in QApplication::exec() () from /opt/qtcvs/lib/libqt-mt.so.3 
#5  0x41b0ec79 in kdemain () from /opt/kdecvs/lib/libkdeinit_kded.so 
#6  0x408538ea in kdeinitmain () from /opt/kdecvs/lib/kde3/kded.so 
#7  0x0804d384 in launch(int, char const*, char const*, char const*, int, char const*, bool, 
char const*, bool, char const*) () 
#8  0x0804faa6 in main () 
#9  0x4139b767 in __libc_start_main () from /lib/libc.so.6 
 
There does not appear to be anything in stdout/err related to the bug. I can send you the 
complete log if you think it will help. 
Comment 4 George Staikos 2003-09-22 07:31:35 UTC
Wow there is lots of corruption in KHTML when this happens.  I still don't see 
the problem in kded but KHTML is indeed not looking good. 
Comment 5 Josh Berry 2003-09-23 06:34:05 UTC
Interestingly enough, this bug (KDED CPU at 100%) also occurs when KNode tries to 
access the wallet. 
Comment 6 George Staikos 2003-09-23 16:58:04 UTC
Subject: Re:  Kded uses 100% of the CPU on first Wallet access

On Tuesday 23 September 2003 00:34, you wrote:
> ------- Interestingly enough, this bug (KDED CPU at 100%) also occurs when
> KNode tries to access the wallet.

  It seems that this might actually be kwin's fault.  Still investigating...  
I can't reproduce it with the old kwin, but people are reporting this with 
the new kwin.

  When did you last update from CVS?

Comment 7 Josh Berry 2003-09-23 17:03:39 UTC
> When did you last update from CVS? 
 
A little over 24 hours ago. I'm planning on doing another update tonight (US/Pacific) when I 
get back from school. 
Comment 8 George Staikos 2003-09-23 17:08:03 UTC
Subject: Re:  Kded uses 100% of the CPU on first Wallet access

On Tuesday 23 September 2003 11:03, you wrote:

> > When did you last update from CVS?
>
> A little over 24 hours ago. I'm planning on doing another update tonight
> (US/Pacific) when I get back from school.

  That probably explains things.   I bet it's kwin related now, based on what 
I've heard from other reports.

Comment 9 Josh Berry 2003-09-24 03:51:13 UTC
OK, I think you're right. I just noticed that kded, kwin AND X are all eating CPU at the 
moment. I don't know of anything specific I did to trigger it this time. Perhaps this should be 
sent over to the kwin people as well. 
 
Here's the backtrace from gdb attached to kded: 
 
#0  0x4145b912 in select () from /lib/libc.so.6 
#1  0x40f77b1c in __JCR_LIST__ () from /opt/qtcvs/lib/libqt-mt.so.3 
#2  0x40af21a6 in QEventLoop::enterLoop() () from /opt/qtcvs/lib/libqt-mt.so.3 
#3  0x40af2048 in QEventLoop::exec() () from /opt/qtcvs/lib/libqt-mt.so.3 
#4  0x40adf1a1 in QApplication::exec() () from /opt/qtcvs/lib/libqt-mt.so.3 
#5  0x41b14c79 in kdemain () from /opt/kdecvs/lib/libkdeinit_kded.so 
#6  0x408588ea in kdeinitmain () from /opt/kdecvs/lib/kde3/kded.so 
#7  0x0804d384 in launch(int, char const*, char const*, char const*, int, char const*, bool, 
char const*, bool, char const*) () 
#8  0x0804faa6 in main () 
#9  0x413a1767 in __libc_start_main () from /lib/libc.so.6 
 
Here's the backtrace from kwin: 
 
#0  0x4145b912 in select () from /lib/libc.so.6 
#1  0x40f77b1c in __JCR_LIST__ () from /opt/qtcvs/lib/libqt-mt.so.3 
#2  0x40af21a6 in QEventLoop::enterLoop() () from /opt/qtcvs/lib/libqt-mt.so.3 
#3  0x40af2048 in QEventLoop::exec() () from /opt/qtcvs/lib/libqt-mt.so.3 
#4  0x40adf1a1 in QApplication::exec() () from /opt/qtcvs/lib/libqt-mt.so.3 
#5  0x41b3d6e7 in kdemain () from /opt/kdecvs/lib/libkdeinit_kwin.so 
#6  0x408588da in kdeinitmain () from /opt/kdecvs/lib/kde3/kwin.so 
#7  0x0804d384 in launch(int, char const*, char const*, char const*, int, char const*, bool, 
char const*, bool, char const*) () 
#8  0x0804e3dc in handle_launcher_request(int) () 
#9  0x0804e9c0 in handle_requests(int) () 
#10 0x0804f96f in main () 
#11 0x413a1767 in __libc_start_main () from /lib/libc.so.6 
 
I didn't feel safe getting a backtrace from X... 
 
Interestingly enough, CPU usage went down to 0 when I was attached to kded, but 
remained at 100% when attached to kwin. Don't know if this tidbit helps any or not, but if it's a 
race/deadlock of some sort it might point in the right direction... 
Comment 10 Thomas Vollmer 2003-09-24 08:46:23 UTC
I see this problem also since new kwin. 
 
kded, kwin and X eating 100% of my CPU. 
 
Before CVS from manually killing kded and restarting it with --nofork option 
solved the problem. 
 
I see also a problem with eating up all my memory and swap, when I leave 
the system alone when kded eats my cpu. 
 
Thomas 
Comment 11 Waldo Bastian 2003-09-24 18:34:11 UTC
Joshua: can you try to attach gdb to kded like you did before and then 
let it continue to run with "continue" and break it with ctrl-c, till "bt" gives a 
backtrace that looks significant different from the one you posted above? 
 
 
Comment 12 Lubos Lunak 2003-09-24 18:49:05 UTC
I cannot reproduce the wallet problem with kopete, konqueror or knode. I also doubt it has  
actually anything to do with kwin. Moreover, the backtraces are useless, they just show the  
applications waiting in the eventloop for more events. Could you please try to get some useful  
backtrace from kded, as already explained by Waldo? 
 
Comment 13 Josh Berry 2003-09-25 01:58:17 UTC
I have tried repeatedly to get a different backtrace, to no avail. (Stepping through just plain 
didn't work.) The closest I came was this... and it also looks to be in the Qt event handler: 
 
#0  0x4122a898 in write () from /lib/libpthread.so.0 
#1  0x411feea4 in __JCR_LIST__ () from /usr/X11R6/lib/libX11.so.6 
#2  0x4117093f in _X11TransWrite () from /usr/X11R6/lib/libX11.so.6 
#3  0x411510f2 in _XFlushInt () from /usr/X11R6/lib/libX11.so.6 
#4  0x41150f79 in _XFlush () from /usr/X11R6/lib/libX11.so.6 
#5  0x41151708 in _XEventsQueued () from /usr/X11R6/lib/libX11.so.6 
#6  0x41144a18 in XPending () from /usr/X11R6/lib/libX11.so.6 
#7  0x40a8cd32 in QEventLoop::processEvents(unsigned) () 
   from /opt/qtcvs/lib/libqt-mt.so.3 
#8  0x40af21a6 in QEventLoop::enterLoop() () from /opt/qtcvs/lib/libqt-mt.so.3 
#9  0x40af2048 in QEventLoop::exec() () from /opt/qtcvs/lib/libqt-mt.so.3 
#10 0x40adf1a1 in QApplication::exec() () from /opt/qtcvs/lib/libqt-mt.so.3 
#11 0x41b14c79 in kdemain () from /opt/kdecvs/lib/libkdeinit_kded.so 
#12 0x408588ea in kdeinitmain () from /opt/kdecvs/lib/kde3/kded.so 
#13 0x0804d384 in launch(int, char const*, char const*, char const*, int, char c 
onst*, bool, char const*, bool, char const*) () 
#14 0x0804faa6 in main () 
#15 0x413a1767 in __libc_start_main () from /lib/libc.so.6 
 
I tried an strace as well (with the --nofork option), but couldn't reproduce the bug while 
stracing. 
 
I'll keep trying and see if I can come up with anything else. 
Comment 14 Josh Berry 2003-09-25 02:59:03 UTC
I probably should have mentioned this earlier, though it only just now occurred to me that it 
might be important... when KDED eats CPU, at least 50% of that is spent in kernel-space 
(i.e. system cpu). So it is conceivable that a lot of that time could be spent repeatedly polling 
inside a select()... of course, I could be way off-base here too. 
 
This time kded started eating CPU even while still prompting for the wallet password. Here's 
the backtrace...though I imagine it doesn't show anything unexpected: 
 
#0  0x4145b912 in select () from /lib/libc.so.6 
#1  0x40f77b1c in __JCR_LIST__ () from /opt/qtcvs/lib/libqt-mt.so.3 
#2  0x40af21a6 in QEventLoop::enterLoop() () from /opt/qtcvs/lib/libqt-mt.so.3 
#3  0x40adf201 in QApplication::enter_loop() () 
   from /opt/qtcvs/lib/libqt-mt.so.3 
#4  0x40cb6851 in QDialog::exec() () from /opt/qtcvs/lib/libqt-mt.so.3 
#5  0x4204defe in KWalletD::internalOpen(QString const&, bool, unsigned long) 
    () from /opt/kdecvs/lib/kde3/kded_kwalletd.so 
#6  0x4204ccfa in KWalletD::open(QString const&, unsigned) () 
   from /opt/kdecvs/lib/kde3/kded_kwalletd.so 
#7  0x42055ac6 in KWalletD::process(QCString const&, QMemArray<char> const&, QCS 
tring&, QMemArray<char>&) () from /opt/kdecvs/lib/kde3/kded_kwalletd.so 
#8  0x4083d5c0 in DCOPClient::receive(QCString const&, QCString const&, QCString 
 const&, QMemArray<char> const&, QCString&, QMemArray<char>&) () 
   from /opt/kdecvs/lib/libDCOP.so.4 
#9  0x40836bab in DCOPProcessInternal(DCOPClientPrivate*, int, unsigned long, QM 
emArray<char> const&, bool) () from /opt/kdecvs/lib/libDCOP.so.4 
#10 0x4083655f in DCOPProcessMessage(_IceConn*, void*, int, unsigned long, int, 
IceReplyWaitInfo*, int*) () from /opt/kdecvs/lib/libDCOP.so.4 
#11 0x40848431 in KDE_IceProcessMessages () from /opt/kdecvs/lib/libDCOP.so.4 
#12 0x4083eed4 in DCOPClient::processSocketData(int) () 
   from /opt/kdecvs/lib/libDCOP.so.4 
#13 0x4084049b in DCOPClient::qt_invoke(int, QUObject*) () 
---Type <return> to continue, or q <return> to quit--- 
   from /opt/kdecvs/lib/libDCOP.so.4 
#14 0x40b39b50 in QObject::activate_signal(QConnectionList*, QUObject*) () 
   from /opt/qtcvs/lib/libqt-mt.so.3 
#15 0x40b39cb0 in QObject::activate_signal(int, int) () 
   from /opt/qtcvs/lib/libqt-mt.so.3 
#16 0x40e517a2 in QSocketNotifier::activated(int) () 
   from /opt/qtcvs/lib/libqt-mt.so.3 
#17 0x40b55f30 in QSocketNotifier::event(QEvent*) () 
   from /opt/qtcvs/lib/libqt-mt.so.3 
#18 0x40adef55 in QApplication::internalNotify(QObject*, QEvent*) () 
   from /opt/qtcvs/lib/libqt-mt.so.3 
#19 0x40ade60b in QApplication::notify(QObject*, QEvent*) () 
   from /opt/qtcvs/lib/libqt-mt.so.3 
#20 0x406c4cb5 in KApplication::notify(QObject*, QEvent*) () 
   from /opt/kdecvs/lib/libkdecore.so.4 
#21 0x40aceb5a in QEventLoop::activateSocketNotifiers() () 
   from /opt/qtcvs/lib/libqt-mt.so.3 
#22 0x40a8ca2c in QEventLoop::processEvents(unsigned) () 
   from /opt/qtcvs/lib/libqt-mt.so.3 
#23 0x40af21a6 in QEventLoop::enterLoop() () from /opt/qtcvs/lib/libqt-mt.so.3 
#24 0x40af2048 in QEventLoop::exec() () from /opt/qtcvs/lib/libqt-mt.so.3 
#25 0x40adf1a1 in QApplication::exec() () from /opt/qtcvs/lib/libqt-mt.so.3 
#26 0x41b14c79 in kdemain () from /opt/kdecvs/lib/libkdeinit_kded.so 
---Type <return> to continue, or q <return> to quit--- 
#27 0x408588ea in kdeinitmain () from /opt/kdecvs/lib/kde3/kded.so 
#28 0x0804d384 in launch(int, char const*, char const*, char const*, int, char c 
onst*, bool, char const*, bool, char const*) () 
#29 0x0804faa6 in main () 
#30 0x413a1767 in __libc_start_main () from /lib/libc.so.6 
Comment 15 Lubos Lunak 2003-09-25 10:39:08 UTC
No, sorry, those backtraces also aren't useful. Could you try to find out a way how to 
reproduce the problem with a fresh user, so that I have better chances reproducing it? Also, 
when this happens, could you check in 'top' which applications are active? 
 
Comment 16 George Staikos 2003-09-25 10:40:31 UTC
Subject: Re:  Kded uses 100% of the CPU on first Wallet access

On Thursday 25 September 2003 04:39, you wrote:

> ------- Additional Comments From l.lunak@kde.org  2003-09-25 10:39 -------
> No, sorry, those backtraces also aren't useful. Could you try to find out a
> way how to reproduce the problem with a fresh user, so that I have better
> chances reproducing it? Also, when this happens, could you check in 'top'
> which applications are active?

  Ah but be careful, don't erase your current setup in any way or you may not 
be able to reproduce it again.  Backup files if you need to.

Comment 17 Sami Nieminen 2003-09-28 09:18:59 UTC
I have found out that the only way to reproduce this is to have klaptop running. I disabled 
klaptop for a couple of days and didn't see this problem anymore. Then I enabled it again (just 
to see my notebook battery status) and after a few minutes I had kded, kwin and X eating all 
of my cpu time. 
 
See this post on kde-devel mailing list talking about the same thing: 
http://lists.kde.org/?l=kde-devel&m=106435658526657&w=2 
 
I hope this helps you to reproduce this problem. 
Comment 18 Josh Berry 2003-09-28 09:23:34 UTC
Hmmm... I just noticed that klaptop seems to die whenever I kill kded and it's eating CPU. I 
hadn't noticed that before, but now that you point it out... 
 
I haven't yet had a chance to try to reproduce this. I'll do it tomorrow if I can (it's past 
midnight here already). 
Comment 19 Thomas Vollmer 2003-09-28 10:13:56 UTC
This can be the cause also at my machine. Also a laptop with klaptop and ACPI. 
Comment 20 Josh Berry 2003-09-29 23:35:15 UTC
I discovered a couple new bits on this bug today, just in the normal course of using it. 
 
- KDED spiked today when I had nothing more than a terminal open (well, and the usual 
system tray stuff; KLaptop included), so I don't think it's directly KWallet's fault. KWallet does 
trigger the bug readily, however. 
 
- I can't seem to trigger the bug when KLaptop isn't open. But it's quite possible that I'm just 
not inventive enough. 
 
- KDED, KWin and X continue to eat CPU even after KLaptop is closed. 
 
I still haven't had a chance to test using a vanilla configuration. I'll try to do that today. 
Comment 21 George Staikos 2003-09-29 23:38:35 UTC
Subject: Re:  Kded uses 100% of the CPU on first Wallet access

On Monday 29 September 2003 17:35, you wrote:
> ------- I discovered a couple new bits on this bug today, just in the
> normal course of using it.
>
> - KDED spiked today when I had nothing more than a terminal open (well, and
> the usual system tray stuff; KLaptop included), so I don't think it's
> directly KWallet's fault. KWallet does trigger the bug readily, however.
>
> - I can't seem to trigger the bug when KLaptop isn't open. But it's quite
> possible that I'm just not inventive enough.
>
> - KDED, KWin and X continue to eat CPU even after KLaptop is closed.
>
> I still haven't had a chance to test using a vanilla configuration. I'll
> try to do that today.

  I see this locally on a Dell laptop as well.  I have also seen kdesktop_lock 
using 100% cpu and ram

Comment 22 Josh Berry 2003-09-29 23:59:03 UTC
OK. Steps to consistently reproduce the bug: 
 
1. Login to a fresh account (one that has never seen KDE before) 
2. Accept all the defaults in the personalization wizard 
3. Note that KLaptop is running in the system tray. 
4. Start Konqueror, and visit a website that needs a username/password. Log in to the 
website. 
5. Setup KWallet when prompted, using the basic setup. 
6. Enter your KWallet password when prompted. 
7. Check to see that your CPU usage is normal. Note that the wallet is unlocked. 
8. Close Konqueror and note that the wallet is now locked. 
9. Watch your CPU usage soar. Run 'top' and "enjoy". :) 
 
Steps for NOT reproducing the bug: 
 
Same as above, except for the following: 
 
3. Check to see if KLaptop is running, and if so, close it. 
9. Note that your CPU is continuing to idle. :) 
 
Hope this helps. 
Comment 23 Lubos Lunak 2003-09-30 11:23:01 UTC
Bad luck. I don't have a laptop, and I cannot reproduce the problem :(. 
 
Comment 24 Lubos Lunak 2003-10-01 12:55:14 UTC
*** Bug 65286 has been marked as a duplicate of this bug. ***
Comment 25 Carl Thompson 2003-10-01 18:42:03 UTC
Also note that it currently doesn't seem to matter if the wallet subsystem is disabled in the KDE control 
center.  However, that might be because the wallet subsystem doesn't honor the setting-- I notice the 
wallet icon showing up in my system tray even when I have it disabled. 
 
Comment 26 George Staikos 2003-10-01 22:09:32 UTC
Subject: Re:  Kded uses 100% of the CPU on first Wallet access

On Wednesday 01 October 2003 12:42, you wrote
> ------- Also note that it currently doesn't seem to matter if the wallet
> subsystem is disabled in the KDE control center.  However, that might be
> because the wallet subsystem doesn't honor the setting-- I notice the
> wallet icon showing up in my system tray even when I have it disabled.

   How do you reproduce this?

Comment 27 Carl Thompson 2003-10-02 00:29:18 UTC
Reproduce what?  The wallet manager not respecting the control center setting?  That would be a 
different bug but it can be reproduced by turning off the wallet manager in control center and noticing 
that the wallet icon pops up in the system tray anyway when you go to a site requiring a password. 
Comment 28 George Staikos 2003-10-02 02:28:41 UTC
Subject: Re:  Kded uses 100% of the CPU on first Wallet access

On Wednesday 01 October 2003 18:29, you wrote:
> ------- Reproduce what?  The wallet manager not respecting the control
> center setting?  That would be a different bug but it can be reproduced by
> turning off the wallet manager in control center and noticing that the
> wallet icon pops up in the system tray anyway when you go to a site
> requiring a password.

   You'd better update your code and try again.  I can't reproduce that 
behaviour at all.

Comment 29 Lubos Lunak 2003-10-03 13:26:43 UTC
Could people who can reproduce this bug please read 
http://lists.kde.org/?l=kde-core-devel&m=106511153216993&w=2 and so as it says? The 
rule about sending the result directly to me applies too, just please include the bugreport 
number in the mail. Also, it doesn't matter what kind of computer you have, the thing that 
matters is that you have this problem. 
The command for compiling should be something like g++ a.cpp -o a.out -I/usr/X11R6/include 
-I/opt/kde3/include -I/opt/qt3/include -L/opt/kde3/lib -L/opt/qt3/lib -lkdecore 
 
Comment 30 Lubos Lunak 2003-10-06 11:32:01 UTC
Created attachment 2698 [details]
qt patch

Qt bug, triggered by klaptopdaemon's xautolock code.
Comment 31 Lubos Lunak 2003-10-06 11:32:29 UTC
*** Bug has been marked as fixed ***.
Comment 32 Carl Thompson 2003-10-11 06:37:17 UTC
Building from CVS updated today this does _not_ seem to be fixed. 
 
Comment 33 George Staikos 2003-10-11 06:42:34 UTC
Subject: Re:  Kded uses 100% of the CPU on first Wallet access

On Saturday 11 October 2003 00:37, you wrote:
> ------- Building from CVS updated today this does _not_ seem to be fixed.

   Did you update qt-copy?

Comment 34 Carl Thompson 2003-10-11 08:54:50 UTC
Subject: Re:  Kded uses 100% of the CPU on first Wallet access

Quoting George Staikos <staikos@kde.org>:

> ...

>    Did you update qt-copy?

No, I didn't!  I didn't realize I needed to!  Sorry for the bother...


Comment 35 Reinhold Kainhofer 2003-10-30 12:09:46 UTC
*** Bug 66218 has been marked as a duplicate of this bug. ***
Comment 36 Lubos Lunak 2003-11-03 22:32:21 UTC
*** Bug 66928 has been marked as a duplicate of this bug. ***
Comment 37 Lubos Lunak 2003-12-22 11:43:35 UTC
*** Bug 70754 has been marked as a duplicate of this bug. ***
Comment 38 Lubos Lunak 2004-02-05 14:37:53 UTC
*** Bug 74126 has been marked as a duplicate of this bug. ***
Comment 39 Lubos Lunak 2004-02-13 14:31:26 UTC
*** Bug 74917 has been marked as a duplicate of this bug. ***
Comment 40 Lubos Lunak 2004-03-22 15:24:36 UTC
*** Bug 78212 has been marked as a duplicate of this bug. ***
Comment 41 Lubos Lunak 2004-05-08 17:27:02 UTC
*** Bug 81144 has been marked as a duplicate of this bug. ***
Comment 42 Lubos Lunak 2004-05-13 13:07:00 UTC
*** Bug 81493 has been marked as a duplicate of this bug. ***
Comment 43 Risto H. Kurppa 2008-03-15 06:31:06 UTC
This still happens on Kubuntu Gutsy & Hardy on KDE 3.5.8 and 3.5.9, see https://bugs.launchpad.net/ubuntu/+source/kdelibs/+bug/86168

I'm using Kpowersave, not klaptop.
Comment 44 Jeremy W. Murphy 2008-07-01 08:27:01 UTC
Yeah, unfortunately this or a similar bug is still present.  I don't use KLaptop or such since this computer is not a laptop.  The bug occurs when I log into KDE and seems to be associated with kwalletmanager (triggered by either Kontact or one of many Konqueror windows).  Killing kded there and then seems to leave KDE unstable, so I quit KDE, kill the kded process from the console and then return.  I'm running Gentoo x86_64, KDE 3.5.9, Qt 3.3.8/4.3.3, gcc 4.1.2, glibc 2.6.1.