elementary shutdown kills opened applications

[PerfectCarl]

elementary *shuts down so hard* that the applications don't have time to gracefully exit (Firefox proposes to restore sessions when it's reopened and gnome settings are not saved - for elementary apps and others) ?

Same issue with Chrome and other apps.

Note: the reboot dialog doesn't even have time to properly animate out

[Timo Reimerdes (timorei)]

It gets anoying with libreoffice if you get the macro system popping up telling you about something failing during restore.

[PerfectCarl (name-is-carl)]

According to preliminary investigations :
  - the issue is certainly related to gnome-session [1] and is said to be fixed in ubuntu

Looking at gnome-session changelog [2], it seems that the following reports *have nothing to do* with the issue at hand:
   - Fix a crash on session end (#709221) [3]
  - gsm-manager: Don't abort when all clients exit (#710480) [4]

[1] : https://launchpad.net/gnome-session
[2] : http://bazaar.launchpad.net/~vcs-imports/gnome-session/master/view/head:/NEWS
[3] : https://bugzilla.gnome.org/show_bug.cgi?id=709221
[4] : https://bugzilla.gnome.org/show_bug.cgi?id=710480

[ProsjektX (prosjektx)]

This is a very serious bug and should definitely be looked into before beta2. This can cause users data-loss and other issues. If users have unsaved files open they will we force close and any changes lost.

This bug does not only apply when using shutdown from the menu, but also e.g. reboot, and also if using the commandline shutdown.

Doesn't everybody see this issue..?!

[Darrell Swafford (darelster)]

Also annoying in Chrome.

[PerfectCarl (name-is-carl)]

I tried to build gnome-session from trunk

bzr branch lp:gnome-session
sudo apt-get install libgnome-desktop-3-dev
./autogen.sh

and it failed with :

Requested 'gnome-desktop-3.0 >= 3.9.91' but version of gnome-desktop-3.0 is 3.8.4

[PerfectCarl (name-is-carl)]

After some digging it seems that gnome-session that we ship is the same as the one that is shipped with trusty and vivid.
Same base version, same debian patches.

I'm starting to think that this may not be the cause of the issue.
It occurred to me that we might want to write a simple cli app that would log the signal it receives when the session is closed.
If things are going well, it should receive SIGTERM signal and close itself.

If the cli app receives the SIGKILL signal then something is going wrong.

That experience could be redone with a gtk app.

Tom pointed out that it may be related to a systemd service.
After all
On freya, ps -aux | grep systemd returns :

root 349 0.0 0.0 51728 1976 ? Ss 09:47 0:00 /lib/systemd/systemd-udevd --daemon
root 863 0.0 0.0 43452 1852 ? Ss 09:47 0:00 /lib/systemd/systemd-logind
cran 13392 0.0 0.0 14284 932 pts/7 S+ 15:59 0:00 grep --color=auto systemd

[Nolan Shah (nolanshah212)]

Tom is right, gnome-session does uses systemd-logon's poweroff.

If you want to see the chain of events in gnome-session:
1. http://bazaar.launchpad.net/~ubuntu-branches/ubuntu/utopic/gnome-session/utopic/view/head:/gnome-session/gsm-systemd.c at line 446
2. http://bazaar.launchpad.net/~ubuntu-branches/ubuntu/utopic/systemd/utopic/view/head:/src/login/logind-dbus.c at line 2132

[PerfectCarl (name-is-carl)]

Just taking some notes about login here

apt-cache policy libsystemd-login0 returns

libsystemd-login0:
  Installed: 204-5ubuntu20.9
  Candidate: 204-5ubuntu20.9
  Version table:
 *** 204-5ubuntu20.9 0
        500 http://fr.archive.ubuntu.com/ubuntu/ trusty-updates/main amd64 Packages
        100 /var/lib/dpkg/status
     204-5ubuntu20 0
        500 http://fr.archive.ubuntu.com/ubuntu/ trusty/main amd64 Packages

Sources for it are here: https://launchpad.net/ubuntu/+source/systemd/204-5ubuntu20.9

Note: source for vivid is located here:
https://launchpad.net/ubuntu/+source/systemd/208-8ubuntu8.2

[PerfectCarl (name-is-carl)]

This might be relevant https://bugs.freedesktop.org/show_bug.cgi?id=77415

[PerfectCarl (name-is-carl)]

The list of patches that landed in vivid version is long.

Here are for reference the ones with logind in the title:

Launch-logind-via-a-shell-wrapper.patch
logind-add-forgotten-call-to-user_send_changed.patch
logind-add-forgotten-return-statement.patch
logind-allow-unprivileged-session-device-access.patch
logind-also-escape-external-data-when-saving-to-run.patch
logind-bring-polkit-policy-for-hibernate-in-line-wit.patch
logind-fix-bus-introspection-data-for-TakeControl.patch
logind-fix-policykit-checks.patch
logind-fix-reference-to-systemd-user-sessions.servic.patch
logind-given-that-we-can-now-relatively-safely-shutd.patch
logind-if-a-user-is-sitting-in-front-of-the-computer.patch
logind-ignore-failing-close-on-session-devices.patch
logind-initialize-timer_fd.patch
logind-order-all-scopes-after-both-systemd-logind.se.patch
logind-pass-pointer-to-User-object-to-user_save.patch
logind-remove-dead-variable.patch
logind-rework-session-shutdown-logic.patch
logind-save-session-after-setting-the-stopping-flag.patch
logind-save-user-state-after-stopping-the-session.patch
logind-uninitialized-timer_fd-is-set-to-1.patch
logind-use-correct-who-enum-values-with-KillUnit.patch
Make-logind-hostnamed-localed-timedated-D-Bus-activa.patch
Make-sd_login_monitor_new-work-for-logind-without-sy.patch
Start-logind-on-demand-via-libpam-systemd.patch
units-systemd-logind-fails-hard-without-dbus.patch

[PerfectCarl (name-is-carl)]

Another suspect: X11 being killed.

See
https://mail.gnome.org/archives/desktop-devel-list/2014-November/msg00018.html

and the replies
https://mail.gnome.org/archives/desktop-devel-list/2015-January/msg00041.html

[Maxim Taranov (png2378)]

https://www.bountysource.com/issues/3733139-elementary-shutdown-kills-opened-applications

[Ciaran McAleer (ciaran39)]

I removed the $30 you can re add it if you want but I think your problem is solved

[Cameron (cameronp)]

I'm still getting this in loki... was it supposedly fixed?