Bug#867745: libpam-systemd: log out from a TTY and your X input devices get lost!

[Grant Chesy]

On 07/09/2017 10:22 AM, Brian Potkin wrote:
> fvwm here. But I do not think the WM is relevant.

me either, but since three wm reported now, it is more evidence that wm 
isn't relevant.

>
>> I reverted back to sysvinit/uninstalled systemd (and configured Xwrapper to
>> run X as root again, so it would start without the systemd bits). With
>> systemd gone, the problem no longer occurs.
>
> A simple change to PID 1 as init makes no difference to the behaviour
> here. Altering ~/,bash_logout gives a quick fix (workaround?).
>
I don't think it is the pid 1 part of the systemd package that causes 
this issue, but rather logind since that is involved with VTs, and 
logind is invoked when the issue occurs (and logind is removed when 
uninstalling systemd and the issue is resolved); logind is part of the 
main systemd package, so IMO systemd is the correct package to report 
the bug against.

After rebooting with sysvinit-core installed, you have to apt-get remove 
systemd to fix the behavior, or the systemd bits are still there.  And, 
when you do uninstall systemd, you will break X because stretch is setup 
to rely on some systemd bits (?maybe consolekit?) to not need to start X 
as root.  You will need to add the line:
needs_root_rights=yes
to /etc/X11/Xwrapper.config to get X to start again without the systemd 
bits.

I suppose it could be coincidence, and the bug lies elsewhere and some 
systemd component just causes the bug to manifest.  But, so far I have 
seen/ been bitten by *a lot* of buggy behavior in systemd, so biased 
toward suspecting this is yet another systemd bug.