On Wed, 2018-05-02 at 22:01 +0100, Peter Robinson wrote:
>>> today, I've updated my fedora (from 2018-04-17, with updates-testing).
>>> After reboot, I wasn't able to log in (session just crashed right after
>>> typing my password). The bug is already reported in BZ:
>>>
https://bugzilla.redhat.com/show_bug.cgi?id=1573683
>>>
>>> The reason of crash and actual fix for it is pretty weird:
>>> may 02 13:34:48 fanys-laptop gnome-session[1246]:
>>> gnome-session-binary[1246]: WARNING: Application
>>> 'org.gnome.SettingsDaemon.XSettings.desktop' killed by signal 5
>>> may 02 13:34:48 fanys-laptop gnome-session-binary[1246]: WARNING:
>>> Application 'org.gnome.SettingsDaemon.XSettings.desktop' killed by
signal 5
>>> may 02 13:34:48 fanys-laptop gsd-xsettings[1960]: Settings schema
>>> 'org.cinnamon.desktop.a11y.applications' is not installed
>>> may 02 13:34:48 fanys-laptop systemd-coredump[1941]: Process 1798
>>> (gsd-xsettings) of user 42 dumped core.
>>> (stack trace [1])
>>>
>>>
>>> In order to fix this, I had to install cinnamon-desktop. After this, I was
>>> able to log in, both into Wayland an Xorg session. I never had
>>> cinnamon-desktop (or any other cinnamon package) installed on my PC. This
>>> fix worked at least on one another computer. I am currently unable to
>>> reproduce the issue.
>>
>>
https://bugzilla.redhat.com/show_bug.cgi?id=1573546
>>
>> The update that caused this was
>>
https://bodhi.fedoraproject.org/updates/FEDORA-2018-37c3eca456 . It was
>> in updates-testing for 9 hours or so; just about any upgrade of a
>> Workstation install to F28 with u-t that happened during that time was
>> likely affected. Once Matthew and Kevin diagnosed the problem, I
>> unpushed the update. It never reached stable.
>>
>> Removing caribou, installing cinnamon-settings-daemon, or updating to
>> the fixed caribou from
>>
https://bodhi.fedoraproject.org/updates/FEDORA-2018-4c9e1c937c
>> should all resolve the issue.
>
> Thee caribou update crashed my entire session during the dnf update.
Yes. It did that. We know. (It happened to Matt also).
Since it's now been unpushed and a fixed update pushed, are you asking
for something else to happen? If so, what?
In retrospect, if we find a Really Bad Update in updates-testing, the
person who unpushes it should also go and talk to releng to ask for an
emergency updates-testing push, to make sure it's actually removed from
mirrors. Clicking the unpush button in bodhi web ui just queues it to be
unpushed in the next updates-testing push and doesn't do any immediate
action.
I believe what happened here was that there was no updates-testing push
on the day the bad caribou update was unpushed, so it just kept on
lingering on the mirrors and breaking more and more people's systems.
Kalev