FS#35310 - [gnome] 3.8.2 and Catalyst black screen at boot
Attached to Project:
Arch Linux
Opened by Britt Yazel (brittyazel) - Wednesday, 15 May 2013, 18:31 GMT
Last edited by Jan Alexander Steffens (heftig) - Tuesday, 04 June 2013, 03:20 GMT
Opened by Britt Yazel (brittyazel) - Wednesday, 15 May 2013, 18:31 GMT
Last edited by Jan Alexander Steffens (heftig) - Tuesday, 04 June 2013, 03:20 GMT
|
Details
Description:
With yesterdays updates to the Gnome desktop to 3.8.2 many people, including myself, are being greeted at first boot with a solid black screen due to GDM no loading correctly. More details on this issue can be read here: https://bbs.archlinux.org/viewtopic.php?id=163331 To correct this issue a downgrade must be made to mutter and gnome-shell to their 3.8.1 previously working versions. Also, myself and the other members of that forum thread seem to be having the issue pertaining to the Catalyst drivers, though I do not know if that is a coorelation or not. |
This task depends upon
Closed by Jan Alexander Steffens (heftig)
Tuesday, 04 June 2013, 03:20 GMT
Reason for closing: Fixed
Additional comments about closing: Problem disappeared after rebuilding with -Wl,--as-needed.
Tuesday, 04 June 2013, 03:20 GMT
Reason for closing: Fixed
Additional comments about closing: Problem disappeared after rebuilding with -Wl,--as-needed.
I built gnome-shell 3.8.2 locally but it didn't help.
Clutter-CRITICAL **: Unable to initialize Clutter: The OpenGL version could not be determined
% uname -srm
Linux 3.9.2-1-ARCH x86_64
Gnome doesn't start with startx too, this is not gdm trouble
So :
gnome-shell-3.8.2 : black screen
gnome-shell-3.8.1 builded by myself : black screen
old gnome-shell-3.8.1 of extra : works
Maybe gnome-shell 3.8.2 is not directly guilty.
Haven't had time enough to use the system much but it seems to work.
I have to recompile cogl with --enable-cogl-gles2=no, clutter with --enable-wayland-{backend,compositor}=no and without --enable-egl-backend.
Then, I recompile gnome-shell 3.8.2 and it works (it's similar for totem 3.8.2).
Should I add all those packages (including totem) into catalyst repositories until proper fix will arrive?
Sys:
Linux 3.9.2-1-ARCH x86_64
Catalyst-13.1-6
Xorg 1.12
It's work if downgrade gnome-shell (=3.8.1-1) and totem (=3.8.0-1)
[gnome-catalyst]
Server = http://catalyst.wirephire.com/repo/gnome-catalyst/$arch
Is this repo for legacy graphic-card also (HD 4xxx)?
The repo patches gnome so it will work with catalyst. I'd assume it would work with legacy cards, but you may want to confirm with @Vi0L0 first.
Just be sure to patch xorg-server 1.12 against GNOME 3.8.x or simply use packages from xorg112 repo
When I run some applications I get seg-faults, an example of the error logs from totem is:
(totem:8316): GLib-CRITICAL **: g_strsplit: assertion `string != NULL' failed
(totem:8316): Clutter-CRITICAL **: Unable to initialize Clutter: The OpenGL version could not be determined
(totem:8316): Totem-WARNING **: gtk-clutter failed to initialise, expect problems from here on.
(totem:8316): GLib-CRITICAL **: g_strsplit: assertion `string != NULL' failed
(totem:8316): Clutter-CRITICAL **: Unable to initialize Clutter: The OpenGL version could not be determined
Segmentation fault (core dumped)
I am using catalyst from:
[xorg113]
Server = http://catalyst.wirephire.com/repo/xorg113/$arch
[gnome-catalyst]
Server = http://catalyst.wirephire.com/repo/gnome-catalyst/$arch
@Nerzhul I didn't have gnome-shell installed and I still had problems. There is definitely a bigger problem, I think it is more related to gtk and catalyst.
https://bugzilla.gnome.org/show_bug.cgi?id=701445
Any comments that you could make there to get this issue resolved would be appreciated
Can somebody confirm this? If so I will take down the temp repo :)
Any difference in "ldd /usr/bin/gnome-shell" between the problematic and the fixed official packages?
Temporarily turn off ASLR (/proc/sys/kernel/randomize_va_space) to get reproducible output that can be diffed.
Now all works fine. Well done! thanks to Jan Steffens and Vi0L0.
after disabling aslr it's still hard to read diffs output (see attachment)