Crash inside doPoll()

[Sharif Rizal]

Binary package hint: compiz

compiz.real crashed with SIGSEGV

this happens when the screen is locked, and you try to unlock it by putting in your password. it accepts the password, blanks the field, says 'checking' and spins the cursor...and never stops. bumping to high importance, due to the likely amounts of data loss each time this happens.

ProblemType: Crash
Architecture: i386
Date: Fri Aug 10 19:20:26 2007
DistroRelease: Ubuntu 7.10
ExecutablePath: /usr/bin/compiz.real
NonfreeKernelModules: nvidia
Package: compiz-core 1:0.5.2-0ubuntu1
PackageArchitecture: i386
ProcCmdline: /usr/bin/compiz.real --ignore-desktop-hints --indirect-rendering --sm-client-id default0 ccp
ProcCwd: /home/sharif
ProcEnviron:
 PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games
 LANG=en_US.UTF-8
 SHELL=/bin/bash
Signal: 11
SourcePackage: compiz
Stacktrace:
 #0 0x08054b70 in ?? ()
 #1 0x0807b8f8 in ?? ()
 #2 0x00000003 in ?? ()
 #3 0x000000fa in ?? ()
 #4 0x00000000 in ?? ()
StacktraceTop:
 ?? ()
 ?? ()
 ?? ()
 ?? ()
 ?? ()
Title: compiz.real crashed with SIGSEGV
Uname: Linux enigma 2.6.22-9-generic #1 SMP Fri Aug 3 00:50:37 GMT 2007 i686 GNU/Linux
UserGroups: adm admin audio cdrom dialout dip floppy lpadmin netdev plugdev powerdev scanner video

[Apport retracing service (apport)]

StacktraceTop:eventLoop () at ../../src/display.c:1882
main (argc=6, argv=0xbf917c24) at ../../src/main.c:441

[Sarah Kowalik (hobbsee-deactivatedaccount)]

spoken to amaranth, looks like this is really a gnome-screensaver bug, as it's happening under metacity too.

[Steve Langasek (vorlon)]

Not sure how this same bug could be happening under metacity, the segfault is shown to happen in compiz rather than gnome-screensaver? Even if gnome-screensaver is triggering it, that says to me it's a compiz bug.

[Andrea Corbellini (andrea.corbellini)]

I use metacity and this happens to me too so I don't think the cause is compiz.

[Steve Langasek (vorlon)]

Can anyone confirm that this bug is still present with the current version of compiz in hardy, 1:0.6.99+git20080102-0ubuntu3? All of the stack traces are against previous versions of the source, which makes it more challenging to track down a segfault.

[Sarah Kowalik (hobbsee-deactivatedaccount)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Yes, I can still reproduce this.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHhDCJ7/o1b30rzoURAhWDAKCWzYGOFbh49uw87zCO/jXuO3cTlACfQQ/J
hyrAQuOU5fx4WwynfWIvAJs=
=VkRG
-----END PGP SIGNATURE-----

[edschofield (schofield)]

I also get a SIGSEGV on Hardy amd64 when resuming from suspend with an Intel G965 graphics card. I presume it's the same bug. My stack trace is:

Stacktrace:
 #0 0x00007f25ae818f33 in ?? () from /usr/lib/compiz/libworkarounds.so
 #1 0x00007f25ae819dfa in ?? () from /usr/lib/compiz/libworkarounds.so
 #2 0x00007f25ae612a46 in ?? () from /usr/lib/compiz/libezoom.so
 #3 0x00007f25ae40bd45 in ?? () from /usr/lib/compiz/libvpswitch.so
 #4 0x00007f25ae0002ba in ?? () from /usr/lib/compiz/libfade.so
 #5 0x00007f25addf8a04 in ?? () from /usr/lib/compiz/libshift.so
 #6 0x00007f25adbe7c28 in ?? () from /usr/lib/compiz/libwobbly.so
 #7 0x0000000000411b88 in eventLoop ()
 #8 0x000000000040d059 in main ()
StacktraceTop:
 ?? () from /usr/lib/compiz/libworkarounds.so
 ?? () from /usr/lib/compiz/libworkarounds.so
 ?? () from /usr/lib/compiz/libezoom.so
 ?? () from /usr/lib/compiz/libvpswitch.so
 ?? () from /usr/lib/compiz/libfade.so

[Ted Gould (ted)]

I can unlock screensavers using metacity. Have been regularly for several weeks. I'm reasonably certain that this is not a gnome-screensaver bug.

[edschofield (schofield)]

Ted, have you really marked a segfault bug with 22 duplicates as "invalid" because you personally have not experienced it on your particular hardware?

[Roy Jamison (mcisbackuk-deactivatedaccount)]

Why was this marked invalid?? Surely over 100 users are NOT wrong about this bug which we are still experiencing, can someone change this back please, one users installation vs 100 is not right.

[Brian Curtis (bcurtiswx)]

i still experience this error, that is why i've switched it back to confirmed

[Brian Curtis (bcurtiswx)]

I may have moved the wrong one to confirmed, this seems (by the responses) to be a compiz bug. Please correct me if i have confirmed the wrong one, and I can go ahead and fix it.

[Kyubutsu (jf)]

i have experienced this bug on both x86 and x86_64 8.04 hardy versions on random events while running a number of programs.
even though 32bit version seems to be more unstable about it.
resuming from screensaver (password protect on) does not seem to be exclusively affected in my case. i have refrained from submitting duplicate bugs reports on this issue, but i will try to provide further information as it occurs next time.

[Ted Gould (ted)]

On Mon, 2008-02-04 at 23:29 +0000, edschofield wrote:
> Ted, have you really marked a segfault bug with 22 duplicates as
> "invalid" because you personally have not experienced it on your
> particular hardware?

No, I marked it invalid because it doesn't make sense that there is
something gnome-screensaver could do would crash compiz. Even if there
was, that's a bug with compiz not gnome-screensaver. They're at the
wrong places in the stack to affect each other in that way. I'm not
saying the bug is invalid, I'm saying that assigning it to
gnome-screensaver is invalid and that is all that I changed. It might
be an X bug, but it seems most likely to be a compiz bug to me.

[edschofield (schofield)]

On Mon, 2008-02-04 at 06:59 +0000, Ted Gould wrote:

> No, I marked it invalid because it doesn't make sense that there is
> something gnome-screensaver could do would crash compiz. Even if there
> was, that's a bug with compiz not gnome-screensaver. They're at the
> wrong places in the stack to affect each other in that way. I'm not
> saying the bug is invalid, I'm saying that assigning it to
> gnome-screensaver is invalid and that is all that I changed. It might
> be an X bug, but it seems most likely to be a compiz bug to me.

Ted: Okay, thanks for the clarification!

[edschofield (schofield)]

As Ted Gould points out, this seems to be a bug with compiz or X, not gnome-screensaver.

[edschofield (schofield)]

The bug was, it seems, accidentally marked "invalid" for compiz, when it should only be "invalid" for gnome-screensaver.

[Travis Watkins (amaranth)]

Is anyone still having this problem? I thought we had it fixed.

[Travis Watkins (amaranth)]

Forgot to mention, if you are still having this problem, can you produce it while running compiz under valgrind and attach the log?

[Sarah Kowalik (hobbsee-deactivatedaccount)]

I haven't had this in a while - i had it once a month or so ago, but not beyond that. Will be interesting to see what others come up with.

[Sven Almgren (sven-tras)]

Not sure, This but is a result from suspending to ram, and then waking
it up, but I guess I could try, what options do you want from valgrind?
i've only used it for memleaks and a callgraph :P

/Sven

Travis Watkins wrote:
> Forgot to mention, if you are still having this problem, can you produce
> it while running compiz under valgrind and attach the log?
>
>

[Travis Watkins (amaranth)]

No special options needed.

[Steve Langasek (vorlon)]

It certainly appears that this bug is still present, given that new duplicates are still being filed against the current version via apport. Marking as 'confirmed' again, lest the bug be incorrectly expired out.

[Iain Lane (laney)]

Apport has marked my Compiz bug as a dupe of this one. I get it intermittently when switching virtual workspaces (ctrl+alt+arrows). Unfortunately I cannot reliably reproduce it.

[drifter (civilwar1865-gmail)]

sometimes when I reboot X server copiz crashes. I dont think it ever crashes after a full restart.

[Backpacking_Nerd (1backpacker)]

I got this problem when I tried to wake the screen (running on a Presario V6000, nVidia GeForce Go chipset) I restarted the X server (loosing some data) and found that compiz crashed when I logged back in. I can reproduce this reliably if it'll be a help to anyone...just tell me what you need. (my original bug report got filed as #218027).

[Brian Murray (brian-murray)]

Bug 191365 seems to be a duplicate of this one and it has another 18 duplicates.

[Kjell Braden (afflux)]

Sounds like bug 145360 and another 129 duplicates are duplicates too. Opinions?

[Timur (timka.org)]

I was wandering why my lock screen shortcut (Ctrl+Alt+Break) stoped working in Intrepid. I had the System -> Preferences -> Keyboard Shortcuts dialogue active when the crash notification poped up. I'm not sure if I actually pressed my key combination. BTW, I had to change the shortcut to Ctrl+Alt+ScrLk to make locking combination work.

[Rick Spencer (rick-rickspencer3)]

ken - can you please take a look?

[Ken VanDine (ken-vandine)]

I just tried to reproduce this with jaunty on my i386 box with a Intel G965 graphics card, didn't crash. I also tested on my jaunty amd64 laptop with Intell 4500 graphics, didn't crash.

Is anyone able to reproduce this with jaunty?

[Martin Pitt (pitti)]

For the record, I can't reproduce it either.

[Robert Ancell (robert-ancell)]

The stack trace shows the code iterating over a list of file descriptor structures and an iteration returning invalid memory (as shown in the many duplicates). Each file descriptor has an arbitrary callback attached to it. I'd bet a lot of money the bug is the list is being modified inside on of the callbacks.

[Robert Ancell (robert-ancell)]

Bug 145360 and bug 191365 are duplicates of this. They each have 169 and 19 duplicates respectively. I don't seem to be able to mark them as duplicates without manually modifying all their own duplicates first.

[Fernando Miguel (fernandomiguel)]

For all of those that are trying to unsubcrive from this bug and cant (due to a Launchpad timeout bug on AJAX), please user the following URL:
https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/131679/+subscribe

[Kyle Jones (mutiny32)]

Robert, how about not messing with duplicates' status until a solution can be found for the insane amount of notification emails you are causing to be sent out. I've literally been getting hundreds of notification emails for about the past day and so have many, many other people.

[Travis Watkins (amaranth)]

Just so people know (and sorry for the spam) every single duplicate of this bug is probably actually a completely different problem. Crashing in doPoll just means something went weird in compiz that didn't trip it up until the next time it went through this function. I'm not sure how all of these bugs ended up as duplicates (I know many/most were done automatically by apport) but we should probably at least have a separate bug for each action someone was performing when this happened.

Also, unless someone finds the right magic to harden compiz to such problems there is probably no good way to debug this problem since the corruption of compiz state almost certainly happens at some early point that could be quite some time ago.

[mykool (mykool-0)]

How does one un-subscribe?  As you probably know, last week or so ago I (I'm sure I'm not the only one) had my e-mail box bombed with bug reports, it was absurd. This is not the kind of bug bomb I would hope for.(LOL).  Anyway I finally started marking it all as spam.  This is the first one to make it through since the bomb, and yet again - no 'take my name off the list' option. What gives? How do I stop getting mail from you guys?

--- On Thu, 6/25/09, Travis Watkins <email address hidden> wrote:

From: Travis Watkins <email address hidden>
Subject: [Bug 131679] Re: Crash inside doPoll()
To: <email address hidden>
Date: Thursday, June 25, 2009, 2:41 AM

Just so people know (and sorry for the spam) every single duplicate of
this bug is probably actually a completely different problem. Crashing
in doPoll just means something went weird in compiz that didn't trip it
up until the next time it went through this function. I'm not sure how
all of these bugs ended up as duplicates (I know many/most were done
automatically by apport) but we should probably at least have a separate
bug for each action someone was performing when this happened.

Also, unless someone finds the right magic to harden compiz to such
problems there is probably no good way to debug this problem since the
corruption of compiz state almost certainly happens at some early point
that could be quite some time ago.

--
Crash inside doPoll()
https://bugs.launchpad.net/bugs/131679
You received this bug notification because you are a direct subscriber
of a duplicate bug.

Status in Compiz: Confirmed
Status in “compiz” package in Ubuntu: Triaged
Status in “gnome-screensaver” package in Ubuntu: Invalid

Bug description:
Binary package hint: compiz

compiz.real crashed with SIGSEGV

this happens when the screen is locked, and you try to unlock it by putting in your password.  it accepts the password, blanks the field, says 'checking' and spins the cursor...and never stops.  bumping to high importance, due to the likely amounts of data loss each time this happens.

ProblemType: Crash
Architecture: i386
Date: Fri Aug 10 19:20:26 2007
DistroRelease: Ubuntu 7.10
ExecutablePath: /usr/bin/compiz.real
NonfreeKernelModules: nvidia
Package: compiz-core 1:0.5.2-0ubuntu1
PackageArchitecture: i386
ProcCmdline: /usr/bin/compiz.real --ignore-desktop-hints --indirect-rendering --sm-client-id default0 ccp
ProcCwd: /home/sharif
ProcEnviron:
 PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games
 LANG=en_US.UTF-8
 SHELL=/bin/bash
Signal: 11
SourcePackage: compiz
Stacktrace:
 #0  0x08054b70 in ?? ()
 #1  0x0807b8f8 in ?? ()
 #2  0x00000003 in ?? ()
 #3  0x000000fa in ?? ()
 #4  0x00000000 in ?? ()
StacktraceTop:
 ?? ()
 ?? ()
 ?? ()
 ?? ()
 ?? ()
Title: compiz.real crashed with SIGSEGV
Uname: Linux enigma 2.6.22-9-generic #1 SMP Fri Aug 3 00:50:37 GMT 2007 i686 GNU/Linux
UserGroups: adm admin audio cdrom dialout dip floppy lpadmin netdev plugdev powerdev scanner video

[James P. Carter (jpcarter)]

You log into the bugs.launchpad.net and unsubscribe from 'your'
submitted bugs. It's easy but you will have to wait until the
subscribers column populates. Once this happens you hit the '-' (minus
sign).

The problem last week or perhaps 2 weeks ago was that the majority of
folks were responding with 'unsubscribe' and then getting mad and
sending even more messages. I was surprised at the amount of people
that continued to try this method for unsubscribe even after a
launchpad admin informed them they had to log in. The problem is not
the site it is the people who can't seem to read their instructions in
the email.

-
james

On Thu, Jun 25, 2009 at 5:17 AM, mykool<email address hidden> wrote:
> How does one un-subscribe?  As you probably know, last week or so ago I
> (I'm sure I'm not the only one) had my e-mail box bombed with bug
> reports, it was absurd. This is not the kind of bug bomb I would hope
> for.(LOL).  Anyway I finally started marking it all as spam.  This is
> the first one to make it through since the bomb, and yet again - no
> 'take my name off the list' option. What gives? How do I stop getting
> mail from you guys?
>
> --- On Thu, 6/25/09, Travis Watkins <email address hidden> wrote:
>
> From: Travis Watkins <email address hidden>
> Subject: [Bug 131679] Re: Crash inside doPoll()
> To: <email address hidden>
> Date: Thursday, June 25, 2009, 2:41 AM
>
> Just so people know (and sorry for the spam) every single duplicate of
> this bug is probably actually a completely different problem. Crashing
> in doPoll just means something went weird in compiz that didn't trip it
> up until the next time it went through this function. I'm not sure how
> all of these bugs ended up as duplicates (I know many/most were done
> automatically by apport) but we should probably at least have a separate
> bug for each action someone was performing when this happened.
>
> Also, unless someone finds the right magic to harden compiz to such
> problems there is probably no good way to debug this problem since the
> corruption of compiz state almost certainly happens at some early point
> that could be quite some time ago.
>
> --
> Crash inside doPoll()
> https://bugs.launchpad.net/bugs/131679
> You received this bug notification because you are a direct subscriber
> of a duplicate bug.
>
> Status in Compiz: Confirmed
> Status in “compiz” package in Ubuntu: Triaged
> Status in “gnome-screensaver” package in Ubuntu: Invalid
>
> Bug description:
> Binary package hint: compiz
>
> compiz.real crashed with SIGSEGV
>
> this happens when the screen is locked, and you try to unlock it by
> putting in your password.  it accepts the password, blanks the field,
> says 'checking' and spins the cursor...and never stops.  bumping to high
> importance, due to the likely amounts of data loss each time this
> happens.
>
> ProblemType: Crash
> Architecture: i386
> Date: Fri Aug 10 19:20:26 2007
> DistroRelease: Ubuntu 7.10
> ExecutablePath: /usr/bin/compiz.real
> NonfreeKernelModules: nvidia
> Package: compiz-core 1:0.5.2-0ubuntu1
> PackageArchitecture: i386
> ProcCmdline: /usr/bin/compiz.real --ignore-desktop-hints --indirect-rendering --sm-client-id default0 ccp
> ProcCwd:...

[James P. Carter (jpcarter)]

You log into the bugs.launchpad.net and unsubscribe from 'your'
submitted bugs. It's easy but you will have to wait until the
subscribers column populates. Once this happens you hit the '-' (minus
sign).

The problem last week or perhaps 2 weeks ago was that the majority of
folks were responding with 'unsubscribe' and then getting mad and
sending even more messages. I was surprised at the amount of people
that continued to try this method for unsubscribe even after a
launchpad admin informed them they had to log in. The problem is not
the site it is the people who can't seem to read their instructions in
the email.
--
james

On Thu, Jun 25, 2009 at 5:17 AM, mykool<email address hidden> wrote:
> How does one un-subscribe?  As you probably know, last week or so ago I
> (I'm sure I'm not the only one) had my e-mail box bombed with bug
> reports, it was absurd. This is not the kind of bug bomb I would hope
> for.(LOL).  Anyway I finally started marking it all as spam.  This is
> the first one to make it through since the bomb, and yet again - no
> 'take my name off the list' option. What gives? How do I stop getting
> mail from you guys?
>
> --- On Thu, 6/25/09, Travis Watkins <email address hidden> wrote:
>
> From: Travis Watkins <email address hidden>
> Subject: [Bug 131679] Re: Crash inside doPoll()
> To: <email address hidden>
> Date: Thursday, June 25, 2009, 2:41 AM
>
> Just so people know (and sorry for the spam) every single duplicate of
> this bug is probably actually a completely different problem. Crashing
> in doPoll just means something went weird in compiz that didn't trip it
> up until the next time it went through this function. I'm not sure how
> all of these bugs ended up as duplicates (I know many/most were done
> automatically by apport) but we should probably at least have a separate
> bug for each action someone was performing when this happened.
>
> Also, unless someone finds the right magic to harden compiz to such
> problems there is probably no good way to debug this problem since the
> corruption of compiz state almost certainly happens at some early point
> that could be quite some time ago.
>
> --
> Crash inside doPoll()
> https://bugs.launchpad.net/bugs/131679
> You received this bug notification because you are a direct subscriber
> of a duplicate bug.
>
> Status in Compiz: Confirmed
> Status in “compiz” package in Ubuntu: Triaged
> Status in “gnome-screensaver” package in Ubuntu: Invalid
>
> Bug description:
> Binary package hint: compiz
>
> compiz.real crashed with SIGSEGV
>
> this happens when the screen is locked, and you try to unlock it by
> putting in your password.  it accepts the password, blanks the field,
> says 'checking' and spins the cursor...and never stops.  bumping to high
> importance, due to the likely amounts of data loss each time this
> happens.
>
> ProblemType: Crash
> Architecture: i386
> Date: Fri Aug 10 19:20:26 2007
> DistroRelease: Ubuntu 7.10
> ExecutablePath: /usr/bin/compiz.real
> NonfreeKernelModules: nvidia
> Package: compiz-core 1:0.5.2-0ubuntu1
> PackageArchitecture: i386
> ProcCmdline: /usr/bin/compiz.real --ignore-desktop-hints --indirect-rendering --sm-client-id default0 ccp
> ProcCwd:...

[Matt Zimmerman (mdz)]

On Thu, Jun 25, 2009 at 12:17:10PM -0000, mykool wrote:
> How does one un-subscribe?  As you probably know, last week or so ago I
> (I'm sure I'm not the only one) had my e-mail box bombed with bug
> reports, it was absurd. This is not the kind of bug bomb I would hope
> for.(LOL).  Anyway I finally started marking it all as spam.  This is
> the first one to make it through since the bomb, and yet again - no
> 'take my name off the list' option. What gives? How do I stop getting
> mail from you guys?

At the bottom of each mail you received, there is a link. Click on that
link, then click on the "Unsubscribe" button.

Or just click here:

https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/131679/+subscribe

--
 - mdz

[Launchpad Janitor (janitor)]

This bug was fixed in the package compiz - 1:0.8.3+git20090914-0ubuntu1

---------------
compiz (1:0.8.3+git20090914-0ubuntu1) karmic; urgency=low

  * Pre-Version of the upcomming 0.8.4 bugfix release
    - fixes crashes LP: #268683, #284362, #346391, #351878, #352037, #380338
    - fixes crashes on 64bit arches due to issue in va_list use
    - fixes at least some of the doPoll/eventLoop crashes reported in
      LP: #131679, #145360, #279820
    - likely fixes crashes LP: #340428 and #360643
    - fixes placing of dialogs (e.g. PolicyKit) behind currently focused
      window (LP: #165161)
    - fixes placing of new windows behind fullscreen window (LP: #153676)
    - fixes wobbly bouncing near panel and screen edges (LP: #153374)
    - screenshot plugin now saves to the correct desktop directory by default
      (LP: #175116)
    - lowering a window now activates the topmost window, unless
      click-to-focus is off (LP: #122507)
    - fixes inconsistent icon sizes in switcher, ring, shift, staticswitcher
      plugins
    - fixes some memory leaks
    - texture compression is now off by default due to texture quality issues
    - ported KDE4 window decorator to KDE 4.3
  * debian/control:
    - add versionized dependency on compiz-wrapper from
      compiz-core

compiz (1:0.8.3+git20090907-0ubuntu1~ppa1) karmic; urgency=low

  * new git snapshot

compiz (1:0.8.3+git20090904-0ubuntu1~ppa2) karmic; urgency=low

  [ Travis Watkins ]
  * debian/compiz-manager:
    - set COMPIZ_PLUGINS to plugins we want to always have enabled

  [ Saïvann Carignan ]
  * debian/compiz-gnome.gconf-defaults: Re-enable desktop switching by
    default as it has been fixed by appropriate GNOME schemas. (LP: #414170)

compiz (1:0.8.3+git20090904-0ubuntu1~ppa1) karmic; urgency=low

  * new git snapshot:
    - fixes segfaults on 64bit arches
    - warning fixes
    - avoid adding core twice

compiz (1:0.8.3+git20090831-0ubuntu1) karmic; urgency=low

  [ Michael Vogt ]
  * new git snapshot with fixes from 0.8 branch:
    - fix stacking below problem with the polkit1 dialog
    - fixes doPoll() crash (LP: #131679)
    - many more bugfixes
  * debian/control:
    - add "Provides: x-window-manager" to the compiz package
      (LP: #364242)

  [ Robert Ancell ]
  * debian/control:
  * debian/rules:
    - Build a compiz-dbg package (LP: #420321)

 -- Michael Vogt <email address hidden> Mon, 14 Sep 2009 09:39:42 +0200