I’m lately having issues with my desktop-pc (not the newest) and i suspect it might have to do with a recent update. I’m performing updates every week or so but reboot my machine rather seldom so if this is kernel related I’m unsure since which update this may have occurred since there were several performed since my previous reboot.
After booting, the gnome-greeter appears as usual but when I try to log in, the screen stays blank (in case of xorg-gnome; waited for >10 mins) or goes back to the user-selection (in case of default gnome (wayland)).
I can switch to tty3 and log in there;
# systemctl status gdm gives me:
> gdm.service - GNOME Display Manager > > Loaded: loaded (/usr/lib/systemd/system/gdm.service; enabled; vendor preset: enabled) > Active: active (running) since Thu 2019-07-04 16:36:37 CEST; 39min ago > Main PID: 1670 (gdm) > Tasks: 3 (limit: 4915) > Memory: 10.0M > CGroup: /system.slice/gdm.service > └─1670 /usr/sbin/gdm > > Jul 04 16:36:37 f2x systemd: Starting GNOME Display Manager... > Jul 04 16:36:37 f2x systemd: Started GNOME Display Manager. > Jul 04 16:36:38 f2x gdm-launch-environment]: accountsservice: Could not get current seat: No data available > Jul 04 16:38:01 f2x gdm-password]: accountsservice: Could not get current seat: No data available > Jul 04 16:38:07 f2x gdm-password]: gkr-pam: unable to locate daemon control file
Peeking into the journal for something obvious, an entry mentioning ‘gnome-session-binary[…]: Unrecoverable failure in required component org.gnome.Shell.desktop’ but upon closer inspection I found that this error is also present in much earlier log-sections where gnome was running fine.
As far as I can remember, everything worked fine with the old-style-gdm-launch (before the recent prettified loader with the fedora-logo in colour).
Some data about my system:
- Fedora-30, updated from at least F29, maybe 28.
- Running on some old nvidia-card (GeForce 210)
- nouveau-driver used (
# lsmod | grep nouveauyields 8 hits such as video, drm, wmi,…)
- no nvidia-package installed (
# dnf list installed | grep nvidiayields no results)
- gnome-shell-3.32.2-2, mutter-3.32.2-3 currently installed
Some things I’ve tried so far:
- underlying filesystems are all clean (checked offline),
- disabling wayland through uncommenting
WaylandEnable=falsein /etc/gdm/custom.conf + reboot did not bring any change,
- booting an older kernel (5.1.8 instead of 5.1.15) did not bring any change.
As a temporary workaround I’ve installed LXQT which works ok but on this machine I’d prefer gnome and would like to get it back working.
I’m currently considering to
# dnf history rollback <id> or
# dnf history undo <id> to get closer to the possible cause of my issues.
Update: a rollback to a specific ID failed with a python-exception in /usr/lib/python3.7/site-packages/dnf/db/history.py, see below post for details.
Update: undoing certain history-IDs also seems not possible as the underlying packages seem to have been rotated out of the repo-metadata and the respositories. At least i get
Error: no package matched after a series of
No package <name> available.
Any further ideas where else to look into?