Bug 173017

Created attachment 27961 [details]
How my desktop looks on first login, just after the crash

Created attachment 27962 [details]
How my desktop looks on second login

That's the bug for which I proposed a patch here: http://lists.kde.org/?l=kwin&m=122416100329812&w=2

*** Bug 174245 has been marked as a duplicate of this bug. ***

*** Bug 174976 has been marked as a duplicate of this bug. ***

It looks like Guillaume posted a patch about a month ago.  (Thanks!)  I don't have the time to set up a full dev environment, but I'm willing to test this on my setup if someone has the time to shoot me a binary.

Unfortunately that patch fixes the symptom, not the cause so further investigation is needed.

*** Bug 174931 has been marked as a duplicate of this bug. ***

*** Bug 163045 has been marked as a duplicate of this bug. ***

Moving the panel to the top of the screen before running nvidia-settings fixes the issue for me - at least what I reported in #174931.

However, a lot of dialogs and right-click menus are appearing on the wrong display.

I have the same problem when changing the screen layout with xrandr or grandr. Moving the primary screen right of the secondary crashes kwin, while moving it left of it doesn't.

The command that led to the crash was:
xrandr --output LVDS --mode 1024x768 --pos 1280x0

KDE version is 4.1.3.

(Probably unusable) backtrace:

Anwendung: KWin (kwin), Signal SIGABRT
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
[Thread debugging using libthread_db enabled]
[New Thread 0xb65846c0 (LWP 5480)]
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
[KCrash handler]
#6  0xb8055430 in __kernel_vsyscall ()
#7  0xb7e41880 in raise () from /lib/tls/i686/cmov/libc.so.6
#8  0xb7e43248 in abort () from /lib/tls/i686/cmov/libc.so.6
#9  0xb79e5795 in qt_message_output () from /usr/lib/libQtCore.so.4
#10 0xb79e5872 in qFatal () from /usr/lib/libQtCore.so.4
#11 0xb79e58cc in qt_assert_x () from /usr/lib/libQtCore.so.4
#12 0xb7fde6c2 in ?? () from /usr/lib/libkdeinit4_kwin.so
#13 0xb7fde9a6 in ?? () from /usr/lib/libkdeinit4_kwin.so
#14 0xb7fdedfb in ?? () from /usr/lib/libkdeinit4_kwin.so
#15 0xb7fdf071 in ?? () from /usr/lib/libkdeinit4_kwin.so
#16 0xb7fe7dbb in ?? () from /usr/lib/libkdeinit4_kwin.so
#17 0xb7fe9269 in ?? () from /usr/lib/libkdeinit4_kwin.so
#18 0xb7fe92d4 in ?? () from /usr/lib/libkdeinit4_kwin.so
#19 0xb7f914f0 in ?? () from /usr/lib/libkdeinit4_kwin.so
#20 0xb7af2a60 in QMetaObject::activate () from /usr/lib/libQtCore.so.4
#21 0xb7af37e2 in QMetaObject::activate () from /usr/lib/libQtCore.so.4
#22 0xb7698963 in QDesktopWidget::resized () from /usr/lib/libQtGui.so.4
#23 0xb711f8af in QApplication::x11ProcessEvent () from /usr/lib/libQtGui.so.4
#24 0xb71497ea in ?? () from /usr/lib/libQtGui.so.4
#25 0xb68626f8 in g_main_context_dispatch () from /usr/lib/libglib-2.0.so.0
#26 0xb6865da3 in ?? () from /usr/lib/libglib-2.0.so.0
#27 0xb6865f61 in g_main_context_iteration () from /usr/lib/libglib-2.0.so.0
#28 0xb7b08478 in QEventDispatcherGlib::processEvents ()
   from /usr/lib/libQtCore.so.4
#29 0xb7148ee5 in ?? () from /usr/lib/libQtGui.so.4
#30 0xb7adc52a in QEventLoop::processEvents () from /usr/lib/libQtCore.so.4
#31 0xb7adc6ea in QEventLoop::exec () from /usr/lib/libQtCore.so.4
#32 0xb7adeda5 in QCoreApplication::exec () from /usr/lib/libQtCore.so.4
#33 0xb70ae767 in QApplication::exec () from /usr/lib/libQtGui.so.4
#34 0xb7fac8ee in kdemain () from /usr/lib/libkdeinit4_kwin.so
#35 0x080485a2 in _start ()
#0  0xb8055430 in __kernel_vsyscall ()

Interestingly I can reproduce the crash of kwin when activating the 2nd head only if I don't run kwin inside gdb.

Another new crash I see now is that kwin crashes when I try to drag a window from one screen to the other. The crash happens only once the mouse-cursor leaves one screen and should show up on the second screen. I have no desktop effects activated here. I could also trigger this by moving the cursor to the second head (just the cursor), then starting an application via a terminal that sits on the first head. I'm using the "start new apps on the screen with the mouse cursor" setting.
This one is new since maybe 2, at most 3 weeks. Backtrace for that is:
#0  0xb667a566 in raise () from /lib/libc.so.6
#1  0xb667bd88 in abort () from /lib/libc.so.6
#2  0xb759a728 in qt_message_output (msgType=QtFatalMsg,
    buf=0xbfffcab0 "ASSERT failure in QVector<T>::operator[]: \"index out of range\", file /usr/local/include/QtCore/qvector.h, line 321")
    at /home/andreas/src/qt-copy/src/corelib/global/qglobal.cpp:2102
#3  0xb759a78c in qFatal (
    msg=0xb771cb18 "ASSERT failure in %s: \"%s\", file %s, line %d")
    at /home/andreas/src/qt-copy/src/corelib/global/qglobal.cpp:2303
#4  0xb759a7d1 in qt_assert_x (where=0xb7fd2809 "QVector<T>::operator[]",
    what=0xb7fd27e9 "index out of range",
    file=0xb7fd2c78 "/usr/local/include/QtCore/qvector.h", line=321)
    at /home/andreas/src/qt-copy/src/corelib/global/qglobal.cpp:1880
#5  0xb7f9319a in KWin::Workspace::clientArea (this=0x80f89f0,
    opt=KWin::PlacementArea, screen=-1, desktop=1)
    at /usr/local/include/QtCore/qvector.h:321
#6  0xb7f9342f in KWin::Workspace::clientArea (this=0x80f89f0,
    opt=KWin::PlacementArea, p=@0xbfffecc4, desktop=1)
    at /home/andreas/src/kdebase/workspace/kwin/geometry.cpp:279
#7  0xb7f7e134 in KWin::Client::manage (this=0x830ae70, w=52428823,
    isMapped=false) at /home/andreas/src/kdebase/workspace/kwin/manage.cpp:220
#8  0xb7f49060 in KWin::Workspace::createClient (this=0x80f89f0, w=52428823,
    is_mapped=<value optimized out>)
    at /home/andreas/src/kdebase/workspace/kwin/workspace.cpp:497
#9  0xb7f70f97 in KWin::Workspace::workspaceEvent (this=0x80f89f0,
    e=0xbffff280) at /home/andreas/src/kdebase/workspace/kwin/events.cpp:376
#10 0xb7f5d232 in KWin::Application::x11EventFilter (this=0xbffff5b4,
    e=0xbffff280) at /home/andreas/src/kdebase/workspace/kwin/main.cpp:362
#11 0xb6c5f6dc in qt_x11EventFilter (ev=0xbffff280)
    at /home/andreas/src/qt-copy/src/gui/kernel/qapplication_x11.cpp:361
#12 0xb6c622bb in QApplication::x11ProcessEvent (this=0xbffff5b4,
    event=0xbffff280)
    at /home/andreas/src/qt-copy/src/gui/kernel/qapplication_x11.cpp:2918
#13 0xb6c90114 in x11EventSourceDispatch (s=0x80571f0, callback=0,
    user_data=0x0)
    at /home/andreas/src/qt-copy/src/gui/kernel/qguieventdispatcher_glib.cpp:142#14 0xb63a41b8 in g_main_context_dispatch () from /usr/lib/libglib-2.0.so.0
#15 0xb63a7853 in ?? () from /usr/lib/libglib-2.0.so.0
#16 0x08055f10 in ?? ()

This is pretty much a showstopper for the release I think, as it makes kwin unusable with more than 1 screen and XRandR 1.2 (which should be pretty common on current/soon to be released distributions). I basically can't use the second screen because I can't get a window onto that screen.

Andreas, can you try with today's svn? I just committed a fix that should prevent the crash in at least some situations, but I don't know if it's enough to avoid all the clientArea crashes.

This bug has existed since 4.0 so although it may be high priority it is in no way a blocker.

As I indicated in my last comment, there's a new crash making it totally impossible to use more than one monitor (unless you don't want any windows on it). That is a new problem, which is a blocker for me (and I think pretty much for anybody else using Xrandr 1.2). Should I open a new report for that? This bugreport looks like a dumping ground for any xrandr-crashes in kwin, so I just added that instead of opening a new one.

If I'm reading your previous comment correctly you are adding a second monitor while the X server is running. As KWin doesn't support this yet what you are experiencing is not a new bug and therefore belongs here.

I am grouping all XRandR bugs here so when KWin officially gains XRandR support we can close this one to make it easier to identify all post-supported XRandR bugs from the pre-supported ones.

?? So KWin lost xrandr support between KDE3 and KDE4? Why wasn't that ported? KWin from KDE3 works just fine here with Xrandr 1.2. Or maybe KDE3/KWin just uses the Xinerama Compatibility stuff thats available even if configuring multiple screens via Xrandr 1.2.

Or maybe its because the external monitor is activated while kwin runs? I'll try it out with the external monitor configured via xorg.conf (as is done with my KDE3 system, where everything works just fine).

It looks like I've been mistaken about KWin and XRandR all this time, it's been supported for quite a while. Sorry for the mix up.

Please try the latest trunk to see if Guillaume Pothier's commit did anything.

Well, I'm not so sure you're mistaken. After all there's XRandR and XRandR :)

In particular, up to and including XRandR 1.1 the usual way to setup multiple monitors was having a Xinerama xorg.conf (I'll leave out the custom protocols supported by ATI and NVidia here). Them came XRandR 1.2 and this can now all be done "on the fly", however I think (don't know for sure) the API to query for multiple monitors and the Xinerama-notification-stuff (if that existed) still works as before. That is, Xorg has a compatibility layer for Xinerama-Applications.

Of course, the definitive answer is kwin's sources... I'll test an updated trunk on tuesday, as I won't have the time today.

Had an opportunity to reboot into linux and did the check with up-to-date kwin+kephal. From a short test it seems that kwin still crashes when I add a second monitor during runtime via xrandr command. Or if I start up X11 in "clone" mode and then decide to make one of the monitors a separate screen (i.e. xrandr --output VGA --leftof LVDS).

However what does work is setting up an xorg.conf that creates a 2-monitor X11 screen right from the start. kwin doesn't crash and starts up just fine in that case.

*** Bug 176147 has been marked as a duplicate of this bug. ***

*** Bug 176682 has been marked as a duplicate of this bug. ***

*** Bug 176775 has been marked as a duplicate of this bug. ***

*** Bug 176834 has been marked as a duplicate of this bug. ***

Just wanted to add that I'm seeing this crash on an nVidia card, using "nvidia-settings". I am not sure what that uses internally... maybe it's xrandr 1.2? At any rate, I'm not using an xrandr command directly, but dynamically adding a monitor via nvidia-settings crashes kwin every time.

Is there any way to help attack this problem? I'd really like to help get both kwin and plasma to stop being unusable with dynamically-added monitors.

There are quite a few different KDE versions involved in this report... From those who experience this on current trunk, what might help is:
- output of xrandr --verbose
- output of kded4 right around the time kwin crashes

:-)

(In reply to comment #26)
> Just wanted to add that I'm seeing this crash on an nVidia card, using
> "nvidia-settings". I am not sure what that uses internally... maybe it's xrandr
> 1.2? At any rate, I'm not using an xrandr command directly, but dynamically
> adding a monitor via nvidia-settings crashes kwin every time.
From my experience with KDE 4.2 aka trunk: changing screens in nvidia-settings does not crash KWin any more.

*** Bug 176933 has been marked as a duplicate of this bug. ***

Just moved my comment from  Bug 176933 here:
I have KDE4.2 beta packages from Kubuntu running. When having a second monitor
attached the X-server right after start. Alternatively if I have the monitor
not attached, then later attach it, the X-server crashes then immediately. I am
not sure, if it is correct to report this against KDE.
There was a similar bug in KDE4.1.3/Kubuntu intrepid. The X-server crashed
after 10s independently if the second monitor is attached. This could be
resolved by disabling XRANDR. In this case there is not XRANDR, but KDED which
seems to have to do something with it. If I kill KDED before attaching the
second monitor, the crash does not happen. So this is the reason why I report
it against KDE. If I can provide any more usefull information, please let me
know.

Created attachment 29208 [details]
kwin-crash.txt

I have experienced kwin crash today when issuing the following xrandr command:

  xrandr --output VGA-0 --left-of LVDS

kwin then restarted and I got proper output on both display (switched from cloned to separte display).

This is on HP nw8440 with radeon driver (from git).
I am using up-to-date KDE 4.2 from SVN.

I get the following chrash-backtrace when I activate my other screen as twinview via nvidia-settings:
Program: KWin (kwin), signal SIGABRT
0x00007fd7de3f25f0 in __nanosleep_nocancel () from /lib/libc.so.6

Thread 1 (Thread 0x7fd7e3645730 (LWP 11117)):
[KCrash Handler]
#5  0x00007fd7de37cfd5 in raise () from /lib/libc.so.6
#6  0x00007fd7de37eb43 in abort () from /lib/libc.so.6
#7  0x00007fd7e179a6b5 in qt_message_output () from /usr/lib/libQtCore.so.4
#8  0x00007fd7e179a7fd in qFatal () from /usr/lib/libQtCore.so.4
#9  0x00007fd7e33cc187 in KWin::Workspace::clientArea (this=0xedefd0, opt=KWin::ScreenArea, screen=-1, desktop=2) at /usr/include/qt4/QtCore/qvector.h:321
#10 0x00007fd7e33cc392 in KWin::Workspace::clientArea (this=0xedefd0, opt=KWin::ScreenArea, p=<value optimized out>, desktop=-1) at /build/buildd/kdebase-workspace-4.1.80/kwin/geometry.cpp:279
#11 0x00007fd7e33cc7a1 in KWin::Workspace::clientArea (this=0xedefd0, opt=KWin::ScreenArea, c=0x1249480) at /build/buildd/kdebase-workspace-4.1.80/kwin/geometry.cpp:284
#12 0x00007fd7e33cca80 in KWin::Client::adjustedClientArea (this=0x1249480, desktopArea=@0x7fffeb8418d0, area=@0x7fffeb8418d0) at /build/buildd/kdebase-workspace-4.1.80/kwin/geometry.cpp:866
#13 0x00007fd7e33d58fb in KWin::Workspace::updateClientArea (this=0xedefd0, force=false) at /build/buildd/kdebase-workspace-4.1.80/kwin/geometry.cpp:115
#14 0x00007fd7e33d6c1b in KWin::Workspace::desktopResized (this=0xedefd0) at /build/buildd/kdebase-workspace-4.1.80/kwin/geometry.cpp:64
#15 0x00007fd7e3386985 in KWin::Workspace::qt_metacall (this=0xedefd0, _c=QMetaObject::InvokeMetaMethod, _id=<value optimized out>, _a=0x7fffeb841af0)
    at /build/buildd/kdebase-workspace-4.1.80/obj-x86_64-linux-gnu/kwin/workspace.moc:431
#16 0x00007fd7e189f134 in QMetaObject::activate () from /usr/lib/libQtCore.so.4
#17 0x00007fd7e2046810 in Kephal::Screens::screenMoved (this=0x2b6d, _t1=0xf1fdf0, _t2={xp = 0, yp = 0}, _t3={xp = 1680, yp = 0})
    at /build/buildd/kdebase-workspace-4.1.80/obj-x86_64-linux-gnu/libs/kephal/libkephal/moc_screens.cxx:149
#18 0x00007fd7e2033e4f in Kephal::DBusScreens::screenMovedSlot (this=0xf1d4b0, id=<value optimized out>) at /build/buildd/kdebase-workspace-4.1.80/libs/kephal/screens/dbus/dbusscreens.cpp:91
#19 0x00007fd7e2046d57 in Kephal::DBusScreens::qt_metacall (this=0xf1d4b0, _c=QMetaObject::InvokeMetaMethod, _id=1, _a=0x7fffeb841d20)
    at /build/buildd/kdebase-workspace-4.1.80/obj-x86_64-linux-gnu/libs/kephal/libkephal/moc_dbusscreens.cxx:71
#20 0x00007fd7e189f134 in QMetaObject::activate () from /usr/lib/libQtCore.so.4
#21 0x00007fd7e20486a1 in OrgKdeKephalScreensInterface::screenMoved (this=0x2b6d, _t1=0) at /build/buildd/kdebase-workspace-4.1.80/obj-x86_64-linux-gnu/libs/kephal/libkephal/screens_interface.moc:116
#22 0x00007fd7e2048f36 in OrgKdeKephalScreensInterface::qt_metacall (this=0xf27c80, _c=QMetaObject::InvokeMetaMethod, _id=<value optimized out>, _a=0x7fffeb8422c0)
    at /build/buildd/kdebase-workspace-4.1.80/obj-x86_64-linux-gnu/libs/kephal/libkephal/screens_interface.moc:84
#23 0x00007fd7e1ddfea3 in ?? () from /usr/lib/libQtDBus.so.4
#24 0x00007fd7e1de6eef in ?? () from /usr/lib/libQtDBus.so.4
#25 0x00007fd7e1899da5 in QObject::event () from /usr/lib/libQtCore.so.4
#26 0x00007fd7e0dacc3d in QApplicationPrivate::notify_helper () from /usr/lib/libQtGui.so.4
#27 0x00007fd7e0db49ba in QApplication::notify () from /usr/lib/libQtGui.so.4
#28 0x00007fd7e2b24c3b in KApplication::notify (this=0x7fffeb842d10, receiver=0xf27c80, event=0x10ee860) at /build/buildd/kde4libs-4.1.80/kdeui/kernel/kapplication.cpp:307
#29 0x00007fd7e188ad61 in QCoreApplication::notifyInternal () from /usr/lib/libQtCore.so.4
#30 0x00007fd7e188b9fa in QCoreApplicationPrivate::sendPostedEvents () from /usr/lib/libQtCore.so.4
#31 0x00007fd7e18b34d3 in ?? () from /usr/lib/libQtCore.so.4
#32 0x00007fd7dcf5bd3b in g_main_context_dispatch () from /usr/lib/libglib-2.0.so.0
#33 0x00007fd7dcf5f50d in ?? () from /usr/lib/libglib-2.0.so.0
#34 0x00007fd7dcf5f6cb in g_main_context_iteration () from /usr/lib/libglib-2.0.so.0
#35 0x00007fd7e18b315f in QEventDispatcherGlib::processEvents () from /usr/lib/libQtCore.so.4
#36 0x00007fd7e0e3ea6f in ?? () from /usr/lib/libQtGui.so.4
#37 0x00007fd7e1889682 in QEventLoop::processEvents () from /usr/lib/libQtCore.so.4
#38 0x00007fd7e188980d in QEventLoop::exec () from /usr/lib/libQtCore.so.4
#39 0x00007fd7e188bcbd in QCoreApplication::exec () from /usr/lib/libQtCore.so.4
#40 0x00007fd7e339ffe1 in kdemain (argc=1, argv=0x7fffeb8431b8) at /build/buildd/kdebase-workspace-4.1.80/kwin/main.cpp:513
#41 0x00007fd7de368466 in __libc_start_main () from /lib/libc.so.6
#42 0x00000000004007f9 in _start ()

*** Bug 177742 has been marked as a duplicate of this bug. ***

With KDE 4.1.85 (4.2 beta 2) on Fedora 10 (intel or ATI graphics), it seems pretty much impossible to add a second display with xrandr after logging into KDE -- the second display is messed up, with a panel in the middle of it, and is pretty much useless. Removing a display "on-the-fly" also doesn't work at all -- KDE still thinks the second display is there and tries to open windows onto it, and maximising windows on the existing display doesn't get anywhere close to fullscreen.

Setting up an xorg.conf so that the displays are configured before starting KDE does seem to work great, so this is only an issue with on-the-fly changing. Is there any ETA on when that's likely to work again?

Created attachment 29684 [details]
KWin crashing immediately upon xrandr changing resolution

Since it seems related: running 4.2 beta 2 from the kubuntu-experimental ppa, kwin crashes every single time I use xrandr to change resolution. Attaching the kcrash file.

*** Bug 179602 has been marked as a duplicate of this bug. ***

*** Bug 164107 has been marked as a duplicate of this bug. ***

*** Bug 180434 has been marked as a duplicate of this bug. ***

SVN commit 910287 by gpothier:

Fixed many Kephal bugs.
BUG: 173017

 M  +5 -1      kwin/geometry.cpp
 M  +2 -0      libs/kephal/CMakeLists.txt
 M  +2 -2      libs/kephal/kded_kephal/kephald.cpp
 M  +2 -2      libs/kephal/kephald/kephald.cpp
 M  +6 -17     libs/kephal/outputs/xrandr/xrandroutputs.cpp
 M  +3 -4      libs/kephal/outputs/xrandr/xrandroutputs.h
 M  +3 -1      libs/kephal/screens/configuration/configurationscreens.cpp
 M  +36 -10    libs/kephal/screens/output/outputscreens.cpp
 M  +6 -0      libs/kephal/screens/output/outputscreens.h
 M  +8 -3      libs/kephal/xrandr12/randrcrtc.cpp
 M  +8 -2      libs/kephal/xrandr12/randrdisplay.cpp
 M  +8 -7      libs/kephal/xrandr12/randroutput.cpp
 M  +57 -21    libs/kephal/xrandr12/randrscreen.cpp
 M  +5 -1      libs/kephal/xrandr12/randrscreen.h

As there is so many different bugs that have been merged with this one if there is still a problem please report a new bug and link to here instead of reopening this one.

holy crap! I hope this is fixed for good now :) the most annoying bug for multi-monitor setup. Time to check it out later this evening if its all good... fingers crossed.

thanks gpothier, and Lucas for posting info

*** Bug 180636 has been marked as a duplicate of this bug. ***

*** Bug 181166 has been marked as a duplicate of this bug. ***