FS#44587 - [gdm] update to 3.16.1 causes unsable system
Attached to Project:
Arch Linux
Opened by st.binder (nium) - Wednesday, 15 April 2015, 09:34 GMT
Last edited by Jan de Groot (JGC) - Monday, 13 June 2016, 22:22 GMT
Opened by st.binder (nium) - Wednesday, 15 April 2015, 09:34 GMT
Last edited by Jan de Groot (JGC) - Monday, 13 June 2016, 22:22 GMT
|
Details
Description: after updating gdm and libgdm to version 3.16.1
the system is able to boot but shows no login screen, this
happens with or without WaylandEnable set to false in
/etc/gdm/custom.conf. Now using previous versions of gdm and
libgdm as workaround.
Additional info: * gdm 3.16.1 * kernel 3.19.3-3-ARCH x86_64 Steps to reproduce: * update gdm to 3.16.1 |
This task depends upon
Closed by Jan de Groot (JGC)
Monday, 13 June 2016, 22:22 GMT
Reason for closing: Fixed
Additional comments about closing: Should be no issue anymore. If it still is, check your dbus configuration files for changes.
Monday, 13 June 2016, 22:22 GMT
Reason for closing: Fixed
Additional comments about closing: Should be no issue anymore. If it still is, check your dbus configuration files for changes.
The X output http://pastebin.com/dk55YT9p
I use a bumblebee setup (intel+nvidia)
What looks like repeated restarts of gdm also hinders typing in a username and password at the console (ctrl-alt-Fx). For users with this setup, and a long and cryptic password, rebooting in single user mode would be the only solution, as they would not be able to type in the password without the input being interrupted.
Downgrading GDM and LIBGDM to 3.16.0 save my day :)
pacman -U /var/cache/pacman/pkg/*gdm*-3.16.0.1-*
Downgrading gdm only (not libgdm) to 3.16.0.1 solved the problem.
FS#44588https://www.archlinux.org/packages/extra/x86_64/gdm/
Confirmed working now
https://mail.gnome.org/archives/gnome-announce-list/2015-March/msg00071.html
"The login screen is now using Wayland by default."
It will fallback to X backend if fails.
However, there are two problems:
1. Archlinux does not have xorg-server-xwayland as a/an (optional) dependency of gdm, and thus gdm complains about no Xwayland.
2. gdm does not fall back to X backend once wayland fails. Instead, it continues to try wayland for at least three times (which you can see from the log).
*. If you switch to other tty and switch back, gdm will finally start the X backend.
The two problems combined cause no login screen. I installed xorg-server-xwayland and that solved the problem, before 3.16.2 rolling out.
[2015-04-17 08:06] [ALPM] upgraded libgdm (3.16.1-3 -> 3.16.1.1-1)
[2015-04-17 08:06] [ALPM] upgraded gdm (3.16.1-3 -> 3.16.1.1-1)
Both this current and the previous version have a 100% repeatable (I think related) bug which occurs across all my platforms where Arch is installed. I use auto-login and if I log out then I never get an X login window. I also can not get virtual terminals. If I disable auto-login and reboot then GDM does present a login window. This occurs if I log in with GNOME or with GNOME with wayland.
hope these help as well...
Xorg.0.log (1.7 KiB)
lightdm.log (2.5 KiB)
https://bbs.archlinux.org/viewtopic.php?pid=1527421#p1527421
Whenever I boot my machine I see the login screen normally, but I get a "blank screen" after I login. I have to manually kill gnome-shell (at another tty) and then things start to work normally. When I uncomment "#WaylandEnable=false" in custom.cfg, then the machine is unusable (previous method does not work).
Part of the `journalctl -r -u gdm` is attached.
I have an intel graphic card (Lenovo t420s) + mesa, nm-connection-editor installed, intel microcode updates configured.
Not sure if relevant, but my configuration forces edid in KMS to allow using the needed external screen resolution (as shown here https://wiki.archlinux.org/index.php/Kernel_mode_setting#Forcing_modes_and_EDID).
https://bugzilla.gnome.org/show_bug.cgi?id=751308
Also seen in Fedora:
https://bugzilla.redhat.com/show_bug.cgi?id=1232786
And I believe some people were running into other crashes, and it seems 3.16.1.1 fixed them. Not sure why there were still crash reports after 3.16.1.1 (e.g. @bulletmark), but I tend to think those are independent issues.