Bug #729230 “gnome-display-properties crashed with SIGSEGV in in...” : Bugs : gnome-control-center package : Ubuntu

Revision history for this message

Eliah Kagan (degeneracypressure) wrote on 2011-03-04:

#1

Dependencies.txt Edit (6.4 KiB, text/plain; charset="utf-8")
Disassembly.txt Edit (597 bytes, text/plain; charset="utf-8")
ProcMaps.txt Edit (25.0 KiB, text/plain; charset="utf-8")
ProcStatus.txt Edit (780 bytes, text/plain; charset="utf-8")
Registers.txt Edit (460 bytes, text/plain; charset="utf-8")
Stacktrace.txt Edit (2.3 KiB, text/plain; charset="utf-8")
ThreadStacktrace.txt Edit (6.4 KiB, text/plain; charset="utf-8")

visibility:

private → public

Revision history for this message

Eliah Kagan (degeneracypressure) wrote on 2011-03-04:

#2

To clarify--this new crash appears to follow exactly the same pattern as the original crash in bug 729065: I had to perform the operation of running gnome-control-center and attempting to change the screen resolution several times before I was able to produce this crash, and I had thought it was exactly the same crash, until Apport informed me differently. Furthermore, as before, to produce the crash I performed the operating multiple times **over a significant time interval**, not just all at once. Like before, I have no idea what causes this crash to occur some of the time, but not all (or, it seems, even most) of the time.

Revision history for this message

Apport retracing service (apport) wrote on 2011-03-04:

#3

StacktraceTop:
bus_get_prop (connection=0x21bc0b98, sender=0x21cf5fb0 ":1.33", path=0x21ce0230 "/com/canonical/menu/380024E", interface=0x21cf0f18 "com.canonical.dbusmenu", property=0x21cf0fc8 "Version", error=0x0, user_data=0x21cc9708) at ../../libdbusmenu-glib/server.c:751
invoke_get_all_properties_in_idle_cb (_data=0x21cf44c8) at /build/buildd/glib2.0-2.28.1/./gio/gdbusconnection.c:4095
g_idle_dispatch (source=0x21ce0570, callback=0x4f3b70 <invoke_get_all_properties_in_idle_cb>, user_data=0x21cf44c8) at /build/buildd/glib2.0-2.28.1/./glib/gmain.c:4537
g_main_dispatch (context=0x21b85540) at /build/buildd/glib2.0-2.28.1/./glib/gmain.c:2440
g_main_context_dispatch (context=0x21b85540) at /build/buildd/glib2.0-2.28.1/./glib/gmain.c:3013

Revision history for this message

Apport retracing service (apport) wrote on 2011-03-04: Stacktrace.txt

#4

Stacktrace.txt Edit (2.8 KiB, text/plain)

Revision history for this message

Apport retracing service (apport) wrote on 2011-03-04: ThreadStacktrace.txt

#5

ThreadStacktrace.txt Edit (6.9 KiB, text/plain)

Changed in gnome-control-center (Ubuntu):
importance:	Undecided → Medium
tags:	removed: need-i386-retrace

Revision history for this message

Michael Terry (mterry) wrote on 2011-03-04:

#6

Hmm... This stacktrace doesn't seem possible to me. I mean, it's a reasonable trace, but I can't see how we get ourselves into this situation [1]. Which, if true, makes me lean toward memory being corrupted by either my patch or something else.

I'd really like to try to reproduce these crashes so I can investigate rather than guess. Let me get some more info from you so I can try myself:

* Are you in unity or classic desktop? If in classic, do you have the appmenu applet running?
* Do you switch resolutions or just try to (I know you just try to on the last time, when it crashes, but other times you open the dialog, you actually switch resolutions?)
* Any pattern to crashes vs. non-crashes that might be a clue?

[1] Looks like a dbus call is coming in after we've destroyed the server object, but one of the things the server does as it dies is make sure that doesn't happen. Though I may be missing something.

Revision history for this message

Eliah Kagan (degeneracypressure) wrote on 2011-03-04:

#7

Zim.vmx (virtual hardware specification for this machine) Edit (1.9 KiB, text/plain)

Download full text (5.6 KiB)

"Are you in unity or classic desktop? If in classic, do you have the appmenu applet running?"

This Natty system is a VMware virtual machine without graphics acceleration, so I believe it's a classic desktop (Xorg is the windowing server, and I run applications by selecting them from a menu which is a GNOME panel applet, and I don't have the dock-like bar on the left side of the screen that characterizes Unity), but I haven't actually *selected* classic desktop. In some previous daily builds a while back, I used to get a message telling me to select classic desktop because I didn't have graphics acceleration. That is no longer the case. In GDM the session type selection is "Ubuntu Desktop Edition". Since this is currently my only Natty system, I don't know for sure exactly what the interface is *supposed* to look like, so just in case there's something obvious that I'm missing, I'll attach two screenshots. One is what the desktop looks like during the first login after boot-up. The other is what the desktop looks like during all subsequent logins. This is to say that if I log out (of that first login) and log back in, it's differently themed. I've been meaning to report that as a bug--I hope to get around to doing so later today. The theme that appears the first time I log on after rebooting is the same theme as was in use when I initially booted the live CD, messed around with changing the screen resolution a number of times, and installed Ubuntu.

In case it's relevant, this is a 32-bit uniprocessor virtual machine with 1 GB of RAM being virtualized by VMware Workstation 7.1.3 build-324285. VMware Tools version 8.4.5-324285 (the guest drivers) is installed in the virtual machine, but of course wasn't installed in the live CD system where I first experienced bug 729065. However, some of the drivers for VMware guest systems are included with Ubuntu--this is to say that presumably xserver-xorg-video-vmware 1:11.0.3-1ubuntu2 and xserver-xorg-input-vmmouse 1:12.6.99.901-1ubuntu2, which are installed on and running in this Natty system (i.e. this virtual machine), were also installed and running on the (of course also virtualized) live CD system. In case it's useful, I've attached Zim.vmx, which is the file that contains the virtual hardware specifications for this virtual machine. The *host* system running VMware is itself a Maverick amd64 system with about 4GB of RAM (slightly less for the onboard Intel GMA video) and a 2.2 GHz Intel Core 2 Duo processor; this host system runs kernel version 2.6.35-28-generic.

"Do you switch resolutions or just try to (I know you just try to on the last time, when it crashes, but other times you open the dialog, you actually switch resolutions?)"

Both crashes (i.e. bug 729065 and this bug) occur when I expand the drop-down menu to see the list of resolutions. But when it doesn't crash, I have gone ahead and actually changed the resolution.

I have quit gnome-display-properties after changing the resolution; thus, when I change it (or try to change it) the next time, I reopen gnome-display-properties.

"Any pattern to crashes vs. non-crashes that might be a clue?"

I think the correct answer to this is: no, I ...

"Are you in unity or classic desktop? If in classic, do you have the appmenu applet running?"

This Natty system is a VMware virtual machine without graphics acceleration, so I believe it's a classic desktop (Xorg is the windowing server, and I run applications by selecting them from a menu which is a GNOME panel applet, and I don't have the dock-like bar on the left side of the screen that characterizes Unity), but I haven't actually *selected* classic desktop. In some previous daily builds a while back, I used to get a message telling me to select classic desktop because I didn't have graphics acceleration. That is no longer the case. In GDM the session type selection is "Ubuntu Desktop Edition". Since this is currently my only Natty system, I don't know for sure exactly what the interface is *supposed* to look like, so just in case there's something obvious that I'm missing, I'll attach two screenshots. One is what the desktop looks like during the first login after boot-up. The other is what the desktop looks like during all subsequent logins. This is to say that if I log out (of that first login) and log back in, it's differently themed. I've been meaning to report that as a bug--I hope to get around to doing so later today. The theme that appears the first time I log on after rebooting is the same theme as was in use when I initially booted the live CD, messed around with changing the screen resolution a number of times, and installed Ubuntu.

In case it's relevant, this is a 32-bit uniprocessor virtual machine with 1 GB of RAM being virtualized by VMware Workstation 7.1.3 build-324285. VMware Tools version 8.4.5-324285 (the guest drivers) is installed in the virtual machine, but of course wasn't installed in the live CD system where I first experienced bug 729065. However, some of the drivers for VMware guest systems are included with Ubuntu--this is to say that presumably xserver-xorg-video-vmware 1:11.0.3-1ubuntu2 and xserver-xorg-input-vmmouse 1:12.6.99.901-1ubuntu2, which are installed on and running in this Natty system (i.e. this virtual machine), were also installed and running on the (of course also virtualized) live CD system. In case it's useful, I've attached Zim.vmx, which is the file that contains the virtual hardware specifications for this virtual machine. The *host* system running VMware is itself a Maverick amd64 system with about 4GB of RAM (slightly less for the onboard Intel GMA video) and a 2.2 GHz Intel Core 2 Duo processor; this host system runs kernel version 2.6.35-28-generic.

"Do you switch resolutions or just try to (I know you just try to on the last time, when it crashes, but other times you open the dialog, you actually switch resolutions?)"

Both crashes (i.e. bug 729065 and this bug) occur when I expand the drop-down menu to see the list of resolutions. But when it doesn't crash, I have gone ahead and actually changed the resolution.

I have quit gnome-display-properties after changing the resolution; thus, when I change it (or try to change it) the next time, I reopen gnome-display-properties.

"Any pattern to crashes vs. non-crashes that might be a clue?"

I think the correct answer to this is: no, I have not noticed such a pattern. However, it has occurred to me that most of the crashes on the live CD system were when trying to switch the resolution from 800x600 *back* to 1024x768, whereas most of the crashes on the installed system were when trying to switch the resolution from 1024x768 *back* to 800x600. (That is, the live CD system had decided 1024x768 should be its default resolution, and I didn't change the default; on the installed system, one of the first things I did after installation was to specify 800x600 as the default resolution; I did this in gnome-display-properties itself.) I am not really confident that this description (of which direction caused more crashes) is even accurate, and even if it is, it may well be explainable by pure chance--I doubt I've done this enough times to have a statistically significant sampling. I'm really reaching, with this suggestion.

Furthermore, with the new version of appmenu-gtk (from your PPA), I have only actually produced this crash once. If it would be useful, I can try to produce it some more. If that is useful, please let me know if it would also be useful to file separate duplicate bug reports if/when I succeed. (If memory corruption is happening, perhaps there will be some variability which might be diagnostically useful to know about? I don't know, maybe I just like filing bug reports... ;-) )

Since the crash in bug 729230 was produced on a live CD system, presumably this one can be produced on one too. If it would help, I could try that. Please let me know.

Since this is a virtual machine, I could compress it and send it, if you can specify a way you'd like to receive a really big file. Since I just installed Natty today and have not customized it or installed much in the way of additional applications, it shouldn't be extraordinarily huge--perhaps 2.5 GB compressed.

But since bug 729230 happened when running from the live CD, perhaps I could just send a virtual machine identical to this one before installation. Such a system would easily and quickly compress down to a size where it could be easily attached here on Launchpad.

Of course, if you don't have a suitable VMware virtualization product, then sending a VMware virtual machine on which this bug occurs would probably not be helpful. (You might be able to convert the virtual machine to run on VirtualBox, but this bug wouldn't necessarily persist with the new virtual hardware.) VMware Player is free-as-in-beer-ware for noncommercial use, but I think commercial use might require a paid license.

Revision history for this message

Eliah Kagan (degeneracypressure) wrote on 2011-03-04:

#8

what the first login session after boot-up looks like Edit (356.2 KiB, image/png)

Revision history for this message

Eliah Kagan (degeneracypressure) wrote on 2011-03-04:

#9

what subsequent login sessions look like Edit (355.0 KiB, image/png)

Revision history for this message

Eliah Kagan (degeneracypressure) wrote on 2011-03-04:

#10

"If in classic, do you have the appmenu applet running?"

Sorry, I had meant to answer this, but when I was writing my post, I forgot to. Yes, indicator-applet-appmenu is running. (It does suffer from bug 718184, but I doubt that's relevant to this bug.)

Revision history for this message

Eliah Kagan (degeneracypressure) wrote on 2011-03-07:

#11

/var/log/dpkg.log Edit (952.1 KiB, text/plain)

Provided that I waited a while (I think 15 minutes or more) in between tries, bug 729065 never took more than three or four tries to reproduce. On the other hand, throughout the weekend and a bit this morning (I'm in the eastern United States), I've been trying to reproduce this bug. I haven't been putting enormous effort into this, but I think I've probably tried about 10-15 times, and I've logged out and back in, restarted, and shut down and started back up at various points in the process, while also leaving it on and logged in for over 8 hours before trying. I have still only ever produced this (as you say, impossible-seeming) crash once (when I reported it as this bug).

I have not tried replacing appmenu-gtk with the PPA version in a live CD system; perhaps that would be more effective. Should I try that? Should I send you a virtual machine?

It occurs to me that my description of this crash may imply a pattern that is not present. That is, I had thought this crash followed the same pattern as bug 718184 because it was produced in the same was as that crash, and because I had tried to produce it unsuccessfully several times before finally producing it. However, unlike bug 718184, this crash has not been recurring from time to time (or at all, in spite of my efforts). Perhaps the version of appmenu-gtk in your PPA really *does* completely fix bug 718184, and this crash is due to some obscure and rarely-manifesting dbus bug?

Or perhaps the mechanism that triggered the crash involved some other package which has since been updated. I'm attaching a current /var/log/dpkg.log, in case it's relevant.

(By the way, the issue with the desktop being themed differently between logins appears to have gone away -- now my desktop just always looks like http://launchpadlibrarian.net/65635660/natty-desktop-first-login.png -- which is why I have still not reported that as a bug.)

Revision history for this message

Eliah Kagan (degeneracypressure) wrote on 2011-03-07:

#12

Another thing: Sometimes, when I click on the drop-down menu, it pops up momentarily, and then disappears. When I click on it again, it stays up. Sometimes, it stays up the first time. When i do this a few times in a short interval, re-running gnome-display-properties each time, it starts popping up the first time.

In both the crash in bug bug 729065 and this bug, I believe that what had happened was that I clicked on the menu, it came up momentarily and then disappeared (as I had just described), and then gnome-display-properties crashed.

Sebastien Bacher (seb128) on 2011-03-07

Changed in gnome-control-center (Ubuntu):
assignee:	nobody → Michael Terry (mterry)
importance:	Medium → Low

Revision history for this message

Michael Terry (mterry) wrote on 2011-03-07:

#13

So it sounds like maybe this crash was an oddity or at least a different issue. I'm going to have the maintainers of appmenu-gtk review my branch in bug 729065 and try it out. It may have actually been correct then.

Revision history for this message

Eliah Kagan (degeneracypressure) wrote on 2011-03-07:

#14

Sounds good; and if I do reproduce this bug again, I'll make sure to post about it.

Revision history for this message

Sebastien Bacher (seb128) wrote on 2011-04-20:

#15

there was no update or duplicate since so let's consider it fixed and close it, you can open a new bug if you get an issue

Changed in gnome-control-center (Ubuntu):
status:	New → Fix Released

Ubuntu
gnome-control-center package

gnome-display-properties crashed with SIGSEGV in invoke_get_all_properties_in_idle_cb()

Bug Description

Other bug subscribers

Bug attachments

Remote bug watches

Ubuntugnome-control-center package

gnome-display-properties crashed with SIGSEGV in invoke_get_all_properties_in_idle_cb()

Bug Description

Other bug subscribers

Bug attachments

Remote bug watches

Ubuntu
gnome-control-center package