Bug 916033 - [abrt] mate-conf-1.4.0-21.fc18: g_settings_constructed: Process /usr/bin/mateconf-gsettings-data-convert was killed by signal 5 (SIGTRAP)
Summary: [abrt] mate-conf-1.4.0-21.fc18: g_settings_constructed: Process /usr/bin/mate...
Keywords:
Status: CLOSED ERRATA
Alias: None
Product: Fedora
Classification: Fedora
Component: mate-corba
Version: 18
Hardware: x86_64
OS: Unspecified
unspecified
unspecified
Target Milestone: ---
Assignee: Dan Mashal
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard: abrt_hash:4c71c4e0b37292e560ebe3e9c4a...
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2013-02-27 04:14 UTC by Jerald Vanderraad
Modified: 2013-03-19 02:17 UTC (History)
8 users (show)

Fixed In Version:
Clone Of:
Environment:
Last Closed: 2013-03-19 02:17:34 UTC
Type: ---
Embargoed:


Attachments (Terms of Use)
File: backtrace (20.62 KB, text/plain)
2013-02-27 04:14 UTC, Jerald Vanderraad
no flags Details
File: cgroup (130 bytes, text/plain)
2013-02-27 04:14 UTC, Jerald Vanderraad
no flags Details
File: core_backtrace (910 bytes, text/plain)
2013-02-27 04:14 UTC, Jerald Vanderraad
no flags Details
File: dso_list (1.69 KB, text/plain)
2013-02-27 04:14 UTC, Jerald Vanderraad
no flags Details
File: environ (1.71 KB, text/plain)
2013-02-27 04:14 UTC, Jerald Vanderraad
no flags Details
File: limits (1.29 KB, text/plain)
2013-02-27 04:14 UTC, Jerald Vanderraad
no flags Details
File: maps (8.73 KB, text/plain)
2013-02-27 04:14 UTC, Jerald Vanderraad
no flags Details
File: open_fds (277 bytes, text/plain)
2013-02-27 04:14 UTC, Jerald Vanderraad
no flags Details
File: proc_pid_status (930 bytes, text/plain)
2013-02-27 04:14 UTC, Jerald Vanderraad
no flags Details

Description Jerald Vanderraad 2013-02-27 04:14:15 UTC
Description of problem:
Crashed upon session login.

Version-Release number of selected component:
mate-conf-1.4.0-21.fc18

Additional info:
backtrace_rating: 4
cmdline:        mateconf-gsettings-data-convert
crash_function: g_settings_constructed
executable:     /usr/bin/mateconf-gsettings-data-convert
kernel:         3.7.9-libre.201.fc18.gnu.x86_64
uid:            1000
var_log_messages: Feb 26 19:25:21 localhost abrt[12334]: Saved core dump of pid 12325 (/usr/bin/mateconf-gsettings-data-convert) to /var/tmp/abrt/ccpp-2013-02-26-19:25:21-12325 (9404416 bytes)

Truncated backtrace:
Thread no. 1 (4 frames)
 #2 g_settings_constructed at gsettings.c:572
 #4 g_object_new_valist at gobject.c:1835
 #6 g_settings_new at gsettings.c:869
 #7 handle_file at mateconf-gsettings-data-convert.c:90

Comment 1 Jerald Vanderraad 2013-02-27 04:14:18 UTC
Created attachment 703252 [details]
File: backtrace

Comment 2 Jerald Vanderraad 2013-02-27 04:14:20 UTC
Created attachment 703253 [details]
File: cgroup

Comment 3 Jerald Vanderraad 2013-02-27 04:14:21 UTC
Created attachment 703254 [details]
File: core_backtrace

Comment 4 Jerald Vanderraad 2013-02-27 04:14:23 UTC
Created attachment 703255 [details]
File: dso_list

Comment 5 Jerald Vanderraad 2013-02-27 04:14:25 UTC
Created attachment 703256 [details]
File: environ

Comment 6 Jerald Vanderraad 2013-02-27 04:14:26 UTC
Created attachment 703257 [details]
File: limits

Comment 7 Jerald Vanderraad 2013-02-27 04:14:28 UTC
Created attachment 703258 [details]
File: maps

Comment 8 Jerald Vanderraad 2013-02-27 04:14:30 UTC
Created attachment 703259 [details]
File: open_fds

Comment 9 Jerald Vanderraad 2013-02-27 04:14:32 UTC
Created attachment 703260 [details]
File: proc_pid_status

Comment 10 Dan Mashal 2013-02-27 04:45:22 UTC
mate-conf is deprecated. You can ignore this.

Comment 11 Wolfgang Ulbrich 2013-03-06 16:07:51 UTC
Hi Dan,
abrt redirect me to this report.
I re-open it because we need mate-conf for converting user mateconf data to gsettings.
You had pushed some mate apps with mateconf settings in f17/f18/rawhide, that means we need mateconf-gsettings-data-convert until f20.

Today i run in the same issue, if i try to start Mate-Desktop after mate-netspeed update to mate-netspeed-1.5.1-1.fc18.x86_64.
The issue is definitely caused by latest mate-netspeed, because dowgrading to 1.5.0-1 solve the issue.

Also, the new m-n slow down the start of mate desktop.
Normaly my desktop starts in 1 sec after lightdm, now up to 10 sec.
Exactly the same happens in rawhide, downgrading helps here too.

The break of mateconf indicates that something is wrong with the implementation of mateconf-gsettings-data-convert.
https://github.com/mate-desktop/mate-netspeed/commit/d160715e3b375ad326328e8ae553c676eb333d18

I will fill out a bugreport directly against mate-netspeed.

Comment 12 satellitgo 2013-03-07 14:49:25 UTC
just occurred on netinstall of f18 Mate 1.5.7

https://retrace.fedoraproject.org/faf/reports/64874

Comment 13 Dan Mashal 2013-03-07 18:42:34 UTC
Wolfgang,

mateconf-gsettings-data-convert is for moving from MATE 1.4 -> 1.5

I don't think anyone running Fedora is using MATE 1.4 anymore. 

In addition, mate-conf and mate-corba no longer build on rawhide due to g_mutex deprecations.

If you would like to submit some patches upstream to fix this that would help because releng is going to branch Fedora 19 without them and I am probably going to retire/block those packages.

stefano-k said users running 1.5 do not need mate-conf or mate-corba.

Please correct me if im wrong.

Comment 14 Wolfgang Ulbrich 2013-03-07 19:39:26 UTC
Hi Dan,
obsoleting mate-conf and it's BR mate-corba for f19/20 is a bad idea, because i need it to build compiz :(
I build it with disable-mateconf configure flag, but configure failed without BR mate-conf-devel.
If compiz release stable 1.0.0 which use gsettings than i don't need mate-conf anymore, but i don't think this will happend before end of the year.
But i'am willing to catch mate-conf and mate-corba if you orphoned them.

If f17 gets his EOL the users of my repo will update to fedora mate official.
I don't think that i will update to 1.5 before, because 1.5 breaks to much mate community packages which are in my repo.

So pls give me a favor and don't block them for f19 braching and let me catch them.

Thank you

Comment 15 Wolfgang Ulbrich 2013-03-07 21:42:04 UTC
Dan,
i step back from my proposal, because i did a quick test to build mate-corba.....to much work without help from upstream.
I will change to gconf2 for compiz, hope the gnome guys won't orphoned gconf too.
For me you can oblsolete mate-conf and mate-corba.

Comment 16 Dan Mashal 2013-03-08 04:05:35 UTC
Have you tried patching Compiz 0.8.8 to work with gsettings? Or taking some of the changes upstream and merging them with your compiz package?

Comment 17 Stefano Karapetsas 2013-03-08 11:33:41 UTC
Wolfgang,
in latest mate-control-center, we set gconf and gsettings metacity theme if metacity or compiz are running:

http://git.mate-desktop.org/mate-control-center/commit/?id=088a9625513cc7cb5c8b4d5f18573c4ae42b488f

So, I think you can come back to gconf (that is still supported, I believe) without issues.

Comment 18 GeekGirl1 2013-03-08 20:33:22 UTC
I have this error at login since 2013-03-07. If this will help:

Installed Packages
Name        : mate-conf
Arch        : x86_64
Version     : 1.4.0
Release     : 21.fc18

kernel: 3.8.1-201.fc18.x86_64

/var/log/messages contains (I replaced my PC name with 'xxxxx'):

Mar  8 10:09:24 xxxxx kernel: [   39.710702] traps: mateconf-gsetti[1670] trap int3 ip:3c63e4ec67 sp:7fff99a46a60 error:0

Mar  8 10:09:35 xxxxx mate-session[1514]: WARNING: Application 'mateconf-gsettings-data-convert.desktop' failed to register before timeout

Comment 19 Dan Mashal 2013-03-09 07:32:31 UTC
(In reply to comment #17)
> Wolfgang,
> in latest mate-control-center, we set gconf and gsettings metacity theme if
> metacity or compiz are running:
> 
> http://git.mate-desktop.org/mate-control-center/commit/
> ?id=088a9625513cc7cb5c8b4d5f18573c4ae42b488f
> 
> So, I think you can come back to gconf (that is still supported, I believe)
> without issues.

Right but I'm worried about gconf getting deprecated.

Comment 20 Wolfgang Ulbrich 2013-03-09 10:17:46 UTC
mate-netspeed-1.5.1-2.fc18 fix the issue for me
(In reply to comment#17)
> Wolfgang,
> in latest mate-control-center, we set gconf and gsettings metacity theme if
> metacity or compiz are running:
> 
> http://git.mate-desktop.org/mate-control-center/commit/
> ?id=088a9625513cc7cb5c8b4d5f18573c4ae42b488f
> 
> So, I think you can come back to gconf (that is still supported, I believe)
> without issues.
That's not the point because i disabled using the mateconf/gconf settings backend for compiz.
And i can't use marco or metacity for'gwd', because for keybindings compiz use mateconf/gconf xml's files, and both metacity and marco are using gsettings xml files.
I need mate-conf only as BR for building compiz.
I did some tests yesterday to use gconf or patching compiz to build without matconf/gconf. But i wasn't lucky, a lot of plugins crashed in f19.
I need really more time to see why this happens and maybe it is easier to fix mate-corba to build.
Dan,
I didn't try it but i don't think it should work to use gsettings porting patches from compiz-0.9.9, because 0.9.x is ported to c++ and the code is different.
It would be very nice if you don't block mate-corba/mate-conf from branching to f19 for the moment, until i find a solution for compiz. As i said before i can maintain them.

Thanks

Comment 21 Dan Mashal 2013-03-09 11:39:45 UTC
(In reply to comment #20)
> mate-netspeed-1.5.1-2.fc18 fix the issue for me
> (In reply to comment#17)
> > Wolfgang,
> > in latest mate-control-center, we set gconf and gsettings metacity theme if
> > metacity or compiz are running:
> > 
> > http://git.mate-desktop.org/mate-control-center/commit/
> > ?id=088a9625513cc7cb5c8b4d5f18573c4ae42b488f
> > 
> > So, I think you can come back to gconf (that is still supported, I believe)
> > without issues.
> That's not the point because i disabled using the mateconf/gconf settings
> backend for compiz.
> And i can't use marco or metacity for'gwd', because for keybindings compiz
> use mateconf/gconf xml's files, and both metacity and marco are using
> gsettings xml files.
> I need mate-conf only as BR for building compiz.
> I did some tests yesterday to use gconf or patching compiz to build without
> matconf/gconf. But i wasn't lucky, a lot of plugins crashed in f19.
> I need really more time to see why this happens and maybe it is easier to
> fix mate-corba to build.
> Dan,
> I didn't try it but i don't think it should work to use gsettings porting
> patches from compiz-0.9.9, because 0.9.x is ported to c++ and the code is
> different.
> It would be very nice if you don't block mate-corba/mate-conf from branching
> to f19 for the moment, until i find a solution for compiz. As i said before
> i can maintain them.
> 
> Thanks

Stefano, Wolfgang has done an amazing job with compiz on Fedora. Please take some time and fix the glib/g_mutex deprecations for mate conf and mate corba.

This is critical.

Without Wolfgang‘s hard work, we would have never gotten the Fedora 14 experience that so many users long for back. If you don‘t do it for me, please do it for him.

I cant stress enough how urgent it is for compiz and mate to work in harmony on Fedora.

Thanks,
Dan

Comment 22 Stefano Karapetsas 2013-03-10 18:03:13 UTC
Dan, Wolfgang,
I'll be happy to help to fix MateConf/MateCorba deprecations.
Just point me to the fail build logs.

Comment 23 Wolfgang Ulbrich 2013-03-10 20:06:54 UTC
Stefano,
you make my day....
i did some koji scratch builds today. Mate-conf builds well if mate-corba exists in the repo's.
http://koji.fedoraproject.org/koji/taskinfo?taskID=5103047
@ mate-corba
http://koji.fedoraproject.org/koji/taskinfo?taskID=5103065
The configure error i fixed with a patch, renaming configure.in to configure.ac and renaming 'includes' to 'AM_CPPFLAGS'.
https://dl.dropbox.com/u/49862637/Mate-desktop/Bugs/mate-corba.patch
For this i will do a pull request soon.
After applying the patch compile ends up in a glib error.
http://koji.fedoraproject.org/koji/taskinfo?taskID=5103494
This happens with glib2-2.35.8-1.fc19.
Building against glib2-2.34.1-1.fc18 works fine.
It seems that with glib2-2.35.8 compile warnings ends in errors.
I did try to fix them for myself today but i was lost in bits and bytes after some hours :(
I think you have to install glib2-2.35 in a VM to solve the issue.
A interessting note is that building ORBIT2 against glib2-2.35.8 ends up in the same error.
http://koji.fedoraproject.org/koji/taskinfo?taskID=5103728

Thank you for your offer to fix mate-corba

Comment 24 Wolfgang Ulbrich 2013-03-10 20:51:23 UTC
(In reply to comment #23)
> For this i will do a pull request soon.
Dan, opps, i saw you did already the request :)

Comment 25 Dan Mashal 2013-03-11 12:13:11 UTC
mate-conf is okay.
https://koji.fedoraproject.org/koji/buildinfo?buildID=389492

mate-corba (trying to rebuild now): 
http://koji.fedoraproject.org/koji/taskinfo?taskID=5105952

Comment 26 Patrick Monnerat 2013-03-18 09:44:04 UTC
This problem occurs for me at each login (via vnc).
After reading the above thread, I tried to remove package mate-conf, but there are dependencies on it: resign.
If the problem can't be resolved by an update, can you publish a temporary solution here ? Thanks in advance.

Comment 27 Wolfgang Ulbrich 2013-03-18 09:53:59 UTC
Patrick, and all others,
https://bugzilla.redhat.com/show_bug.cgi?id=918641
su -c 'yum update --enablerepo=updates-testing mate-netspeed-1.5.1-2.fc18'
This solves the issue for me.

Comment 28 Stefano Karapetsas 2013-03-18 09:54:26 UTC
A temporary solution is to remove this file:

/usr/share/MateConf/gsettings/mate-netspeed.convert

Comment 29 Patrick Monnerat 2013-03-18 09:59:18 UTC
Thanks Wolfgang, it did the trick !
Thanks also Stefano, although I did not try your solution :-)

Comment 30 Wolfgang Ulbrich 2013-03-18 10:04:17 UTC
Patrick,
if updating to mate-netspeed-1.5.1-2.fc18 works for you, can leave karma at bodhi for obsoleting this bad version of mate-netspeed in fedora stable.

Comment 31 Patrick Monnerat 2013-03-18 10:12:29 UTC
Done. Karma reached and stable push in progress.

Comment 32 Dan Mashal 2013-03-19 02:17:34 UTC
We will discuss what to do Compiz status with upstream (MATE).

CLOSING THIS BUG


Note You need to log in before you can comment on or make changes to this bug.