Bug 577512 - gnome-shell doesn't start
Summary: gnome-shell doesn't start
Keywords:
Status: CLOSED ERRATA
Alias: None
Product: Fedora
Classification: Fedora
Component: mesa
Version: 13
Hardware: All
OS: Linux
low
high
Target Milestone: ---
Assignee: Adam Jackson
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard:
: 581323 (view as bug list)
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2010-03-27 18:05 UTC by Vaclav "sHINOBI" Misek
Modified: 2010-05-05 07:25 UTC (History)
11 users (show)

Fixed In Version: mesa-7.8.1-6.fc13
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2010-05-05 07:25:18 UTC
Type: ---
Embargoed:


Attachments (Terms of Use)
perf record requested by walters (203.12 KB, application/octet-stream)
2010-04-29 10:05 UTC, Adam Williamson
no flags Details

Description Vaclav "sHINOBI" Misek 2010-03-27 18:05:51 UTC
Description of problem:
gnome-shell doesn't start.

Version-Release number of selected component (if applicable):
gnome-shell-2.29.1-4.x86_64

How reproducible:

Steps to Reproduce:
1. gnome-shell --replace
2.
3.
  
Actual results:
$gnome-shell --replace
Panel leaving: a new panel shell is starting.

(gnome-panel:12453): EggSMClient-CRITICAL **: egg_sm_client_set_mode: assertion `global_client == NULL || global_client_mode == EGG_SM_CLIENT_MODE_DISABLED' failed
      JS LOG: GNOME Shell started at Sat Mar 27 2010 18:59:13 GMT+0100 (CET)
Shell killed with signal 11


Expected results:
gnome-shell working

Additional info:

clutter-1.2.2-1.fc13
gobject-introspection-0.6.9-3.fc13
gjs-0.6-1.fc13
mutter-2.29.1-1.fc13

It was tested at nouveau and ati drivers with the same result.

Comment 1 Fabian A. Scherschel 2010-04-04 11:08:47 UTC
I get the following on my machine:

[fabsh@serenity ~]$ gnome-shell --replace
Window manager warning: Log level 16: Could not load library [/usr/lib/mutter/plugins/libgnome-shell.so (/usr/lib/mutter/plugins/libgnome-shell.so: undefined symbol: gjs_context_get_context)]
Window manager warning: Log level 8: failed to load plugins
^CShell killed with signal 2
[fabsh@serenity ~]$ Cannot register the panel shell: there is already one running.
^C

Comment 2 Vaclav "sHINOBI" Misek 2010-04-17 09:11:06 UTC
At the current packages there are no errors, but gnome-shell still doesn't run under nouveau and ati drivers. With nvidia proprietary driver, the gnome-shell starts although after a long time, desktop-effects times out and returns to the previous state, but gnome-shell --replace works (also it takes a long time before it's started).

Comment 3 Adam Williamson 2010-04-24 00:35:05 UTC
Fabian, your bug is 577460.

gnome-shell behaves for me exactly as it does for Vaclav on this system. This is a Thinkpad X60, graphics adapter is:

00:02.0 VGA compatible controller [0300]: Intel Corporation Mobile 945GM/GMS, 943/940GML Express Integrated Graphics Controller [8086:27a2] (rev 03)

I have gnome-shell-2.29.1-4.x86_64 . Rest of the system is today's F13 with updates-testing, plus X server 1.8.0-8.fc13.x86_64. Had the same behaviour with -6.

Both from desktop-effects and run directly, I get a screen with only the desktop background; gnome-shell process is running and doesn't crash. At the console I see:

[adamw@x60 ~]$ gnome-shell -r
Panel leaving: a new panel shell is starting.

(gnome-panel:3191): EggSMClient-CRITICAL **: egg_sm_client_set_mode: assertion `global_client == NULL || global_client_mode == EGG_SM_CLIENT_MODE_DISABLED' failed
Terminated

Compiz works fine, as does glxgears - haven't tested anything else. Please advise if I should file a new bug, I'm not sure.



-- 
Fedora Bugzappers volunteer triage team
https://fedoraproject.org/wiki/BugZappers

Comment 4 Adam Williamson 2010-04-25 12:27:28 UTC
As drago01 points out on IRC, in fact gnome-shell *does* run: it just takes an awful long time. For me, anyway. If I run gnome-shell --replace from a console and wait about 2 minutes, it actually kicks in.



-- 
Fedora Bugzappers volunteer triage team
https://fedoraproject.org/wiki/BugZappers

Comment 5 Colin Walters 2010-04-25 15:52:06 UTC
Adam, can you attach a perf report in a bug report upstream?  Or just find me on irc and pastebin it.  See 

http://live.gnome.org/GnomeShell/Debugging#Performance_Metrics

Comment 6 Owen Taylor 2010-04-25 18:42:44 UTC
I'm pretty sure this is a problem we've been seeing where Mesa advertises INTEL_swap_event, but it's only actually implemented for the intel drivers, and the events never arrive otherwise.

(http://www.opengl.org/registry/specs/INTEL/swap_event.txt)

You could confirm this by running:

 CLUTTER_VBLANK=none ./gnome-shell --replace

Comment 7 Adam Williamson 2010-04-27 00:12:18 UTC
Owen, that could well be the case for Vaclav, who seems to be on an NVIDIA or ATI card (I'm not entirely sure!), but not me: I'm using Intel. And indeed, your suggested command makes no difference.



-- 
Fedora Bugzappers volunteer triage team
https://fedoraproject.org/wiki/BugZappers

Comment 8 Vaclav "sHINOBI" Misek 2010-04-27 06:15:46 UTC
I have both nvidia and ati graphic cards. Under nvidia proprietary driver CLUTTER_VBLANK=none makes no difference for me. Haven't tested ati driver as it doesn't work for me at all (as mentioned in this bugreport).

Comment 9 Adam Williamson 2010-04-29 10:05:10 UTC
Created attachment 410058 [details]
perf record requested by walters

Comment 10 Owen Taylor 2010-04-29 19:12:59 UTC
OK, so there were two problems mixed together here.

 * The problem described above with INTEL_swap_event. There's a partial 
   fix for this built as:

    https://admin.fedoraproject.org/updates/mesa-7.8.1-2.fc13

   It's a partial fix because it fixes the hang, but at the expense of
   making VBlank synchronization not work at all for ATI or NVIDIA cards.
   Getting VBlank synchronization working again apparently requires a
   mechanical change to the DDX drivers.

 * There was a problem where the packaged gnome-shell binary 
   (which has the type unconfined_execmem_t) doesn't get responses
   from ConsoleKit, so hangs until the call times out.

   This has now been filed against selinux-policy as bug 587380

Comment 11 Owen Taylor 2010-04-29 19:13:49 UTC
(In reply to comment #10)

>  * There was a problem where the packaged gnome-shell binary 
>    (which has the type unconfined_execmem_t) doesn't get responses
>    from ConsoleKit, so hangs until the call times out.
> 
>    This has now been filed against selinux-policy as bug 587380    

I should mention that you can test if you are hitting this by starting the shell after running 'setenforce 0' as root.

Comment 12 Owen Taylor 2010-04-29 19:25:52 UTC
(In reply to comment #10)
> OK, so there were two problems mixed together here.
> 
>  * The problem described above with INTEL_swap_event. There's a partial 
>    fix for this built as:
> 
>     https://admin.fedoraproject.org/updates/mesa-7.8.1-2.fc13
> 
>    It's a partial fix because it fixes the hang, but at the expense of
>    making VBlank synchronization not work at all for ATI or NVIDIA cards.
>    Getting VBlank synchronization working again apparently requires a
>    mechanical change to the DDX drivers.

I'm wrong about that Mesa build fixing it. It's fixing a different problem (with a coincidentally close bug #577142 vs. 577412)

Comment 13 Vaclav "sHINOBI" Misek 2010-04-30 16:14:23 UTC
Yup. You are right Owen. With setenforce 0 nvidia proprietary driver gnome-shell starts quickly as expected.

Ati driver is not working in default config, but when i run export CLUTTER_VBLANK=none and then gnome-shell --replace the gnome-shell appears for a second. X crashes and gdm login screen appears. BTW I'm currently using mesa-drivers-7.8.1-5.fc13.x86_64.

I find in /var/log/gdm/:0.log.1 Xorg: radeon_texture.c:1023: radeon_get_tex_image: Assertion `image->base.Data' failed. message at the end of the file.

Comment 14 Owen Taylor 2010-04-30 16:26:40 UTC
(In reply to comment #13)

> Ati driver is not working in default config, but when i run export
> CLUTTER_VBLANK=none and then gnome-shell --replace the gnome-shell appears for
> a second. X crashes and gdm login screen appears. BTW I'm currently using
> mesa-drivers-7.8.1-5.fc13.x86_64.
> 
> I find in /var/log/gdm/:0.log.1 Xorg: radeon_texture.c:1023:
> radeon_get_tex_image: Assertion `image->base.Data' failed. message at the end
> of the file.    

Can you file this as a separate bug  against Mesa (attach your full Xorg.log to the bug, since that will also have useful information about your particular hardware.) Include in the bug report the version of mesa (as above), and I'd appreciate it if you'd add me to the Cc: of the bug, since I don't see all Mesa bugs by default.

[ It's strange that X is crashing in what appears to be a Mesa function, while it should be using direct rendering, but we can investigate that further in that bug report ]

Thanks!

Comment 15 Vaclav "sHINOBI" Misek 2010-04-30 16:31:51 UTC
OK, I'll do it. In that case we should probably close this bug as it seems to be CLUTTER_VBLANK=none ie. Mesa related if I understand it correctly?

Comment 16 Owen Taylor 2010-04-30 16:57:13 UTC
(In reply to comment #15)
> OK, I'll do it. In that case we should probably close this bug as it seems to
> be CLUTTER_VBLANK=none ie. Mesa related if I understand it correctly?    

I'm taking this bug as the CLUTTER_VBLANK=none bug - which has xserver, Mesa, and Clutter parts to it. I'm working on fixes, but it should be left open for now.

Then bug 587380 is the SELinux issue, so once we have the CLUTTER_VBLANK=none issue fixed we can close this bug.

Comment 17 Vaclav "sHINOBI" Misek 2010-04-30 21:26:32 UTC
I created bug #587793 and added you to the CC.

Comment 18 Fedora Update System 2010-04-30 23:47:44 UTC
clutter-1.2.6-1.fc13 has been submitted as an update for Fedora 13.
http://admin.fedoraproject.org/updates/clutter-1.2.6-1.fc13

Comment 19 Fedora Update System 2010-05-01 00:03:54 UTC
xorg-x11-server-1.8.0-10.fc13 has been submitted as an update for Fedora 13.
http://admin.fedoraproject.org/updates/xorg-x11-server-1.8.0-10.fc13

Comment 20 Fedora Update System 2010-05-01 01:07:41 UTC
mesa-7.8.1-6.fc13 has been submitted as an update for Fedora 13.
http://admin.fedoraproject.org/updates/mesa-7.8.1-6.fc13

Comment 21 Vaclav "sHINOBI" Misek 2010-05-01 09:49:41 UTC
With all above mentioned versions I'm still unable to start gnome-shell with ati drivers without CLUTTER_VBLANK=none.
The error message is changed however.

$ gnome-shell --replace
      JS LOG: GNOME Shell started at Sat May 01 2010 11:46:32 GMT+0200 (CET)
Bug in window manager: Unexpected X error: BadRequest (invalid request code or no such operation) serial 3236 error_code 1 request_code 152 minor_code 179)
Shell killed with signal 6

I'll try nouveau as well.

Comment 22 Vaclav "sHINOBI" Misek 2010-05-01 10:28:06 UTC
Nouveau driver:
$ gnome-shell --replace
      JS LOG: GNOME Shell started at Sat May 01 2010 12:24:32 GMT+0200 (CET)
Bug in window manager: Unexpected X error: BadRequest (invalid request code or no such operation) serial 2527 error_code 1 request_code 153 minor_code 179)
Shell killed with signal 6

nouveau with CLUTTER_VBLANK=none :
$ gnome-shell --replace
Panel leaving: a new panel shell is starting.

(gnome-panel:3255): EggSMClient-CRITICAL **: egg_sm_client_set_mode: assertion `global_client == NULL || global_client_mode == EGG_SM_CLIENT_MODE_DISABLED' failed
      JS LOG: GNOME Shell started at Sat May 01 2010 12:24:46 GMT+0200 (CET)

(mutter:3289): Cogl-glx-WARNING **: ./cogl-framebuffer.c:421: Failed to create an OpenGL framebuffer

(mutter:3289): Cogl-glx-WARNING **: ./cogl-framebuffer.c:421: Failed to create an OpenGL framebuffer

(mutter:3289): Cogl-glx-WARNING **: ./cogl-framebuffer.c:421: Failed to create an OpenGL framebuffer

(mutter:3289): Cogl-glx-WARNING **: ./cogl-framebuffer.c:421: Failed to create an OpenGL framebuffer

(mutter:3289): Cogl-glx-WARNING **: ./cogl-framebuffer.c:421: Failed to create an OpenGL framebuffer

(mutter:3289): Cogl-glx-WARNING **: ./cogl-framebuffer.c:421: Failed to create an OpenGL framebuffer
Bug in window manager: Unexpected X error: BadRequest (invalid request code or no such operation) serial 3618 error_code 1 request_code 153 minor_code 179)
Shell killed with signal 6

Comment 23 Owen Taylor 2010-05-01 10:40:42 UTC
*** Bug 581323 has been marked as a duplicate of this bug. ***

Comment 24 Vaclav "sHINOBI" Misek 2010-05-01 13:40:48 UTC
Ccorrection. I closed the bug #587793 due to the nvidia proprietary libraries interference so current results are:

The problems under ati drivers were solved by the updates and gnome-shell runs without problems.

under nouveau drivers there is still problem.

$ gnome-shell --replace
Panel leaving: a new panel shell is starting.

(gnome-panel:3555): EggSMClient-CRITICAL **: egg_sm_client_set_mode: assertion `global_client == NULL || global_client_mode == EGG_SM_CLIENT_MODE_DISABLED' failed
      JS LOG: GNOME Shell started at Sat May 01 2010 15:28:55 GMT+0200 (CET)
Shell killed with signal 11

Comment 25 Fedora Update System 2010-05-04 06:19:16 UTC
mesa-7.8.1-6.fc13 has been pushed to the Fedora 13 testing repository.  If problems still persist, please make note of it in this bug report.
 If you want to test the update, you can install it with 
 su -c 'yum --enablerepo=updates-testing update mesa'.  You can provide feedback for this update here: http://admin.fedoraproject.org/updates/mesa-7.8.1-6.fc13

Comment 26 Fedora Update System 2010-05-04 23:54:21 UTC
xorg-x11-server-1.8.0-12.fc13 has been pushed to the Fedora 13 stable repository.  If problems still persist, please make note of it in this bug report.

Comment 27 Fedora Update System 2010-05-05 07:20:45 UTC
clutter-1.2.6-1.fc13 has been pushed to the Fedora 13 stable repository.  If problems still persist, please make note of it in this bug report.

Comment 28 Fedora Update System 2010-05-05 07:25:12 UTC
mesa-7.8.1-6.fc13 has been pushed to the Fedora 13 stable repository.  If problems still persist, please make note of it in this bug report.


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