unity-2d-shell crashed with SIGSEGV

[Arek]

the panel hungs up, does not hide scroll

ProblemType: Crash
DistroRelease: Ubuntu 12.04
Package: unity-2d-shell 5.4.0-0ubuntu1
ProcVersionSignature: Ubuntu 3.2.0-17.26-generic 3.2.6
Uname: Linux 3.2.0-17-generic x86_64
NonfreeKernelModules: fglrx
ApportVersion: 1.91-0ubuntu1
Architecture: amd64
Date: Sun Feb 19 23:04:58 2012
ExecutablePath: /usr/bin/unity-2d-shell
InstallationMedia: Ubuntu 11.04 "Natty Narwhal" - Release amd64 (20110427.1)
ProcCmdline: unity-2d-shell
ProcEnviron:
 LANGUAGE=en_US:en
 PATH=(custom, user)
 LANG=en_US.UTF-8
 SHELL=/bin/bash
SegvAnalysis:
 Segfault happened at: 0x7f132300cdb2: testb $0x4,(%rbx)
 PC (0x7f132300cdb2) ok
 source "$0x4" ok
 destination "(%rbx)" (0x00000021) not located in a known VMA region (needed writable region)!
SegvReason: writing NULL VMA
Signal: 11
SourcePackage: unity-2d
StacktraceTop:
 ?? () from /usr/lib/x86_64-linux-gnu/libQtDeclarative.so.4
 ?? () from /usr/lib/x86_64-linux-gnu/libQtDeclarative.so.4
 ?? () from /usr/lib/x86_64-linux-gnu/libQtDeclarative.so.4
 ?? () from /usr/lib/x86_64-linux-gnu/libQtDeclarative.so.4
 ?? () from /usr/lib/x86_64-linux-gnu/libQtScript.so.4
Title: unity-2d-shell crashed with SIGSEGV
UpgradeStatus: Upgraded to precise on 2012-02-19 (0 days ago)
UserGroups: adm admin cdrom dialout lpadmin plugdev sambashare vboxusers

[Apport retracing service (apport)]

StacktraceTop:
 QDeclarativeObjectScriptClass::newQObject (this=0xb83630, object=0x243dfc0, type=136) at qml/qdeclarativeobjectscriptclass.cpp:125
 QDeclarativeEnginePrivate::scriptValueFromVariant (this=0xb53330, val=...) at qml/qdeclarativeengine.cpp:2093
 QDeclarativeObjectScriptClass::property (this=<optimized out>, obj=0x2d0e870, name=<optimized out>) at qml/qdeclarativeobjectscriptclass.cpp:320
 QDeclarativeContextScriptClass::property (this=0xb82e00, object=<optimized out>, name=<optimized out>) at qml/qdeclarativecontextscriptclass.cpp:289
 QScript::DeclarativeObjectDelegate::getOwnPropertySlot (this=0x16dcd20, object=0x7f12b7cc5440, exec=0x7f1311b66048, propertyName=..., slot=...) at bridge/qscriptdeclarativeobject.cpp:76

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Andreas Hasenack (ahasenack)]

I'm not on amd64 and it crashed too, my bug was #955958

[Sascha (skbierm-deactivatedaccount)]

For me it crashed on an i386, with Ubuntu 12.04 LTS Beta 1 (nightly build) and unity 5.6, while clicking on a shaking archivemanager on the lauchner.

[Gerry Boland (gerboland)]

Hi guys
if anyone else gets this crash, please describe what you were doing when the crash occurred. We're not managing to reproduce this. Any & all info helpful
Thanks
-G

[Sascha (skbierm-deactivatedaccount)]

I was using my netbook under a high load, so, having open firefox with a couple of tabs, posting to tumblr some pics, while compressing in the background 310 MB to a 7z-file. The same time xchat, empathy, gwibber, thunderbird where open and in use - so a netbook with 2 GB ram under a load > 1.0
When the archive manger symbol stared shaking (because of messaging me something), and I clicked on it, the crash happened.

[Sasa Paporovic (melchiaros)]

I have installed Ubuntu12.04 in kvm(on Ubuntu11.10) and also installed the home media center "freevo".

freevo is heavy conflicting with unity, so that a simple login in the unity 2d environment throw this error.

[Albert Astals Cid (aacid)]

Quick question for those having this problem, did you all have a large number of items in the launcher (let's say more than 20)?

[Andreas Hasenack (ahasenack)]

I didn't, but I noticed that it mostly happened when there was something "going on" in the launcher/panel, like an icon about to disappear because I closed the app or window. Sometimes also when maximizing or restoring a window that was maximized.

[Sascha (skbierm-deactivatedaccount)]

No, never. I have only the standard icons fix in the launcher - and running maybe up to 4 programms with icons at the same time. But like Andreas said, anytime that happens, there is something going on in the launcher, like a shaking notification of a icon, or a bar on the icon (like on UbuntuOne) ...
Moving the mouse into the direction of that icon in activity is on my side normally the trigger for this bug here.

[revolverxd (revolverxd)]

When I have this crash I usualy run about 5 applications,
And I download using transmission, won't be surprised if its somehow connected to my network card,
Having problems with it since day one.

[Albert Astals Cid (aacid)]

If you guys can reproduce the crash easily, it would be great if you could install unity-2d-dbg, libqt4-dbg and libqt4-script-dbg packages and confirm the stacktrace you get is the one posted by the Apport retracing service in comments #2 #3

Also if you feel really adventurous (and only if oyu can reproduce the crash easily) install the packages from the unity-staging ppa at https://launchpad.net/~unity-team/+archive/staging and confirm the crash is still there, since we are really failing to get it to misbehave :-/

[Sascha (skbierm-deactivatedaccount)]

Albert, okay, I just provoked it. It took me 30 minutes - my load reached 12, 1300 of 2000 MB in use. The crash did occur, when I said, okay, I don't get it, closed one program after the next one - and crash.
But - because I did days ago the check to "don't report this bug again" - I could now only start ubuntu-bug - and have the tracings etc. in another firefox window.
How can I now link it to this report here?
Just add a new bug report, and tell you here the new number/link to that bug report?

[Andreas Hasenack (ahasenack)]

I installed the debug packages and am waiting for a crash now.

[Albert Astals Cid (aacid)]

Yes, open a new bug, that's fine

[Sascha (skbierm-deactivatedaccount)]

Albert, https://bugs.launchpad.net/ubuntu/+source/unity-2d/+bug/967164 is the one I generated. If that is not enough, is there a way, how I can set back, that the bugreporter shows up again on this crash?

[Govinda Narayana Bhat (govinda-narayana)]

There were only the default items in the launcher.
On Mar 28, 2012 6:21 PM, "Albert Astals Cid" <email address hidden>
wrote:

> Quick question for those having this problem, did you all have a large
> number of items in the launcher (let's say more than 20)?
>
> --
> You received this bug notification because you are subscribed to a
> duplicate bug report (958318).
> https://bugs.launchpad.net/bugs/936560
>
> Title:
> unity-2d-shell crashed with SIGSEGV
>
> Status in Unity 2D:
> New
> Status in “unity-2d” package in Ubuntu:
> Confirmed
>
> Bug description:
> the panel hungs up, does not hide scroll
>
> ProblemType: Crash
> DistroRelease: Ubuntu 12.04
> Package: unity-2d-shell 5.4.0-0ubuntu1
> ProcVersionSignature: Ubuntu 3.2.0-17.26-generic 3.2.6
> Uname: Linux 3.2.0-17-generic x86_64
> NonfreeKernelModules: fglrx
> ApportVersion: 1.91-0ubuntu1
> Architecture: amd64
> Date: Sun Feb 19 23:04:58 2012
> ExecutablePath: /usr/bin/unity-2d-shell
> InstallationMedia: Ubuntu 11.04 "Natty Narwhal" - Release amd64
> (20110427.1)
> ProcCmdline: unity-2d-shell
> ProcEnviron:
> LANGUAGE=en_US:en
> PATH=(custom, user)
> LANG=en_US.UTF-8
> SHELL=/bin/bash
> SegvAnalysis:
> Segfault happened at: 0x7f132300cdb2: testb $0x4,(%rbx)
> PC (0x7f132300cdb2) ok
> source "$0x4" ok
> destination "(%rbx)" (0x00000021) not located in a known VMA region
> (needed writable region)!
> SegvReason: writing NULL VMA
> Signal: 11
> SourcePackage: unity-2d
> StacktraceTop:
> ?? () from /usr/lib/x86_64-linux-gnu/libQtDeclarative.so.4
> ?? () from /usr/lib/x86_64-linux-gnu/libQtDeclarative.so.4
> ?? () from /usr/lib/x86_64-linux-gnu/libQtDeclarative.so.4
> ?? () from /usr/lib/x86_64-linux-gnu/libQtDeclarative.so.4
> ?? () from /usr/lib/x86_64-linux-gnu/libQtScript.so.4
> Title: unity-2d-shell crashed with SIGSEGV
> UpgradeStatus: Upgraded to precise on 2012-02-19 (0 days ago)
> UserGroups: adm admin cdrom dialout lpadmin plugdev sambashare vboxusers
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/unity-2d/+bug/936560/+subscriptions
>

[Albert Astals Cid (aacid)]

Sascha, that wasn't enough, i'll send you a private mail (hope that is ok) regarding how to get more info.

[Gerry Boland (gerboland)]

A more comprehensive BT:
http://paste.kde.org/~tsdgeos/448028/

[Sascha (skbierm-deactivatedaccount)]

Albert Astals Cid, I installed halfe an hour ago the staging ppa - and just finished the test, we did via IRC, this time with the files from staging ppa, like unity-2d: Installiert: 5.8.0+bzr1019ubuntu0+inline.
Shell I run again the GNU debugger - and pastebin the backtraces, or don't you need them?
I can provoke that bug again with this version

[Albert Astals Cid (aacid)]

Attaching the paste.kde.org backtrace since it'll die soon

[Albert Astals Cid (aacid)]

Sascha, if you can reproduce the bug with the new version that's ok, i don't need the backtrace again, just wanted to know if it had been fixed by some of the other changes we had made since then or not since the backtrace is probably very similar/the same as the one you got me before and I attahed here at comment #24

[Andreas Hasenack (ahasenack)]

Just got a crash when switching back from youtube fullscreen to regular view. Attaching the crash file. I had the requested dbg packages installed. It's quite big, 12Mb. How do you want me to make it available?

[Andreas Hasenack (ahasenack)]

Albert, I can upload it to chinstrap if you want.

[Andreas Hasenack (ahasenack)]

This was with unity-2d-shell 5.8.0-0ubuntu1

[Albert Astals Cid (aacid)]

Andreas: 12Mb? What is so big? I'm happy enough with the ThreadStacke info you posted in comment #29

[Albert Astals Cid (aacid)]

Andreas, Sascha (anyone else that can reproduce the crash relatively easily). Could you please run unity-2d-shell inside valgrind and attach the output of valgrind to the bug?
To do it:
 - Make sure you have valgrind installed
      sudo apt-get install valgrind
 - Kill the currently running unity-2d-shell (yes twice)
      killall unity-2d-shell
      killall unity-2d-shell
 - Run unity-2d-shell from valgrind
      valgrind --smc-check=all ./shell/app/unity-2d-shell &> valgrind_log.txt
- Try to make unity-2d-shell crash
- If crashes attach the valgrind_log.txt file to the bug
- If you can't make it crash after what you think it be "normal" time that it had crashed i'm also interested in knowing you tried to make it crash and did not succeed and you should attach the valgrind_log.txt file to the bug too.

[George (gtaskeraus)]

In my case I think there are about 40 odd items in the unity bar.

On Wed, 2012-03-28 at 12:43 +0000, Albert Astals Cid wrote:
> Quick question for those having this problem, did you all have a large
> number of items in the launcher (let's say more than 20)?
>

[Andreas Hasenack (ahasenack)]

The core dump was huge
--
Andreas Hasenack
<email address hidden>

Albert Astals Cid <email address hidden> wrote:

Andreas: 12Mb? What is so big? I'm happy enough with the ThreadStacke
info you posted in comment #29

--
You received this bug notification because you are subscribed to a
duplicate bug report (955958).
https://bugs.launchpad.net/bugs/936560

Title:
unity-2d-shell crashed with SIGSEGV

To manage notifications about this bug go to:
https://bugs.launchpad.net/unity-2d/+bug/936560/+subscriptions

[Sascha (skbierm-deactivatedaccount)]

@comment #33 , Albert Astals Cid, here is the valgrind_log.txt
valgrind stopped (of cause) - when the unity-2d-shell crashed in the end with sigsev.

[Andreas Hasenack (ahasenack)]

I probably won't be able to do the valgrind run, because it took a long time to crash yesterday (the whole day basically). I don't have yet a reliable way to get it to crash. I hope the stacktrace with debug I provided earlier, and Sascha's valgrind output, are helpful enough.

[Gerry Boland (gerboland)]

Hi everyone,
this bug is still mystifying us. One possibly related thing I've occasionally noticed is that sometimes icon tiles loose their tooltip - as in when you mouse over a tile, no tooltip appears. Does anybody notice this happening to particular tiles, and is that tile the one which later causes the crash?
Thanks
-G

[LGB [Gábor Lénárt] (lgb)]

@Gerry: I am not sure it's related, but I have the experience since several Ubuntu releases that showing a tooltip then switching workspace (with configured key combination I use) causes that tooltip remains "floating" above the new workspace. It will remain there until I move the mouse (no need to click or so) above it: then it disappears.

[LGB [Gábor Lénárt] (lgb)]

BTW: What I wrote seems to be true for various programs, one which always makes this: Google chrome.

[Gerry Boland (gerboland)]

@Lénárt
thanks for the input, every bit helps! That's a separate bug you've noticed, please report! :) I can reproduce it easily (you're in auto-hide mode).

But even after doing that, the tile still has its tooltip afterwards. I'm suspecting the time when the tooltip vanishes, something more important goes wrong.

It's just happened to me again now, and I've no idea how. But shell still running ok for now, it could crash later. Still digging.
-G

[LGB [Gábor Lénárt] (lgb)]

@Gerry: that's ok, but I have no idea how to report it; exactly what package should I report to have this bug? I notice it with google-chrome usually, but it would be hard to report that, as that software is not provided by Canonical but by Google, and I am not sure that chrome itself causes it, or some other components of the system. Well, I try to reproduce it with other softwares, now I am not sure anymore if it's with chrome only or there are others too really ...

[Gerry Boland (gerboland)]

@Lénárt
tooltip problem is a unity-2d bug. Report and we can investigate more. Thanks!
-G

[Andreas Hasenack (ahasenack)]

Developers,

I don't know what you are not doing. This crash happens daily. Are you running precise by any chance or just trying out things on ec2 and a chroot? :)

[Gerry Boland (gerboland)]

@Andreas,
I assure you we "eat our own dogfood" - we all use the latest trunk of unity2d so we spot problems immediately. But this one is eluding us, we're struggling to reliably reproduce it.
-G

[Andreas Hasenack (ahasenack)]

So if the stacktrace with debug symbols didn't help enough, how would you like to proceed? Maybe with a special build that prints some more information?

[Gerry Boland (gerboland)]

Hi Andreas,
thank you for your help so far, it's been invaluable. Could I ask you to grab another stack trace, but this time running shell with the environment variable QML_DISABLE_OPTIMIZER=1.

So running these commands from terminal

killall unity-2d-shell
killall unity-2d-shell
QML_DISABLE_OPTIMIZER=1 gdb /usr/bin/unity-2d-shell

then use shell until you get crash. Then back in the terminal gdb session, type bt to get the backtrace.
Thanks
-Gerry

[Andreas Hasenack (ahasenack)]

I have it like that for a few hours now, let's sit and wait.

[Andreas Hasenack (ahasenack)]

Aha, got the backtrace. Attached "bt" and "bt full".

[Gerry Boland (gerboland)]

I want to add that we have a preliminary patch that may fix this crash that Andreas is kindly testing for us now.

[Andreas Hasenack (ahasenack)]

I'd say it got much better with 5.10.0, I got very few crashes and I think they were all unrelated to this one in particular. I also don't think I hit that code that the patch added, because I didn't see the log message in my console.

[Sasa Paporovic (melchiaros)]

I am on Ubuntu 12.10 Quantal 64bit for several days now, and could not abserve this bug anymore.

[Ivan Dubrov (wfragg)]

Still crashes couple of times a day. Usually, I get something similar to the following in the dmesg:

[918670.729214] unity-2d-shell[25850]: segfault at 6c ip 00007f3228385062 sp 00007fff06513e00 error 4 in libQtDeclarative.so.4.8.1[7f322812c000+386000]

I think, it somehow related with Qt applications. I have two of them running: skype and P4V (Perforce Visual client).