Bug#915407: libpam-systemd: please add a virtual package "logind" to allow alternatives

[Adam Borowski]

On Mon, Dec 03, 2018 at 04:54:06PM +0100, Adam Borowski wrote:
> On Mon, Dec 03, 2018 at 04:48:51PM +0100, Ansgar Burchardt wrote:
> > On Mon, 2018-12-03 at 16:35 +0100, Adam Borowski wrote:
> > > Use cases, for individual packages: (copied from a mail by smcv):
> > > 
> > >     Packages that need to register their login sessions with logind
> > >     (gdm3, lightdm, openssh-server):
> > >     - remove libpam-systemd dependency
> > >     - add default-logind | logind dependency
> > 
> > Doesn't gdm3 start a gnome-session for the login screen?
> 
> I don't know -- I've tested only with xfce and mate for now.  gdm3 might
> require further work -- but the conversion is opt-in so accepting this patch
> won't break any individual package.
> 
> Sorry for copying this list literally -- gdm3 is not a package I've tested.

And indeed, it does _not_ work.  I tried it with no troubleshooting or
debugging at all -- there was nothing but a text mode cursor on both
monitors, no reaction to attempts to switch to VT1 (I did not try to
recover).  The syslog got spammed with:

Dec  4 23:08:34 umbar gdm3: Child process -2947 was already dead.
Dec  4 23:08:36 umbar gdm3: Child process -2968 was already dead.
Dec  4 23:08:44 umbar gdm3: Child process -3019 was already dead.
Dec  4 23:08:45 umbar gdm3: Child process -3030 was already dead.
Dec  4 23:08:46 umbar gdm3: Child process -3061 was already dead.

So, until someone investigates further, we got an example of testing going
negative.  Ie, the process works, opt-in conversion is required.

Besides the known and worked on[1] policykit-1 reboot/shutdown issue
(buttons appear not grayed out, work until logout but no actual reboot),
combinations I personally tested ok are:
* xfce + lightdm + slick-greeter
* mate + lightdm + slick-greeter
* xfce + slim

But if you apply this patch, the decision whether to convert the dependency
can be done outside src:systemd.  Ie, I'll bother people who are not you.

> > It is a bit different from the standard session and only starts some
> > things (from /usr/share/gnome-session/sessions/gnome-login.session I
> > believe), so I'm not sure about this:
> > 
> > >     Packages that rely on running systemd --user units (dbus-user-
> > > session,
> > >     gnome-session, gnupg):
> > >     - unchanged, elogind is not supported here

No idea about gnome-session, but gnupg?

> > This might imply gdm3 might need `systemd --user` as well; not sure.
> 
> Sounds plausible, yeah.

So it may need systemd only.  Or a pink pony.  I don't _yet_ know.  But we
can test and fix packages one by one, not breaking anything else.


Meow! 

[1]. Existing patches require a separate policykit-1-systemd and
policykit-1-elogind binary; we'd want a runtime check instead.
-- 
⢀⣴⠾⠻⢶⣦⠀ 
⣾⠁⢠⠒⠀⣿⡁ Ivan was a worldly man: born in St. Petersburg, raised in
⢿⡄⠘⠷⠚⠋⠀ Petrograd, lived most of his life in Leningrad, then returned
⠈⠳⣄⠀⠀⠀⠀ to the city of his birth to die.