Really annoying bug - been seeing it for a while.

yeah, I've noticed this too but just didn't get around to filing it. *last* time we had this bug it was caused by an SELinux denial: did either of you check if it works with enforcing=0?

I just tried enforcing=0 with Fedora-Workstation-Live-x86_64-31-20190909.n.0.iso and it didn't help.

Fedora-KDE-Live-x86_64-31-20190909.n.0.iso seems okay fwiw.

I believe the correct component is spice-vdagent, changing.

I've tested F31 Workstation and F31 KDE Live images, using F30 Workstation host. In both cases, the spice-vdagentd.service is running OK. But the journal output differs.

F31 KDE:
$ journalctl -u spice-vdagentd
-- Logs begin at Wed 2019-09-11 08:03:27 EDT, end at Wed 2019-09-11 12:03:52 EDT. --
Sep 11 08:03:55 localhost-live systemd[1]: Starting Agent daemon for Spice guests...
Sep 11 08:03:55 localhost-live systemd[1]: spice-vdagentd.service: Can't open PID file /run/spice-vdagentd/spice-vdagentd.pi>
Sep 11 08:03:55 localhost-live systemd[1]: Started Agent daemon for Spice guests.
Sep 11 08:03:56 localhost-live spice-vdagentd[1512]: opening vdagent virtio channel
Sep 11 08:03:56 localhost-live spice-vdagentd[1512]: Set max clipboard: 104857600
Sep 11 08:03:56 localhost-live spice-vdagentd[1512]: Set max clipboard: 104857600

F31 Workstation:
$ journalctl -u spice-vdagentd
-- Logs begin at Wed 2019-09-11 07:57:44 EDT, end at Wed 2019-09-11 11:58:05 EDT. --
Sep 11 11:58:01 localhost-live systemd[1]: Starting Agent daemon for Spice guests...
Sep 11 11:58:01 localhost-live systemd[1]: spice-vdagentd.service: Can't open PID file /run/spice-vdagentd/spice-vdagentd.pid>
Sep 11 11:58:01 localhost-live spice-vdagentd[1620]: Error getting session for pid 1614: No data available
Sep 11 11:58:01 localhost-live systemd[1]: Started Agent daemon for Spice guests.

$ ps aux | grep vdagent
liveuser    1614  0.0  0.0  82960  2408 ?        Ssl  07:58   0:00 /usr/bin/spice-vdagent
root        1620  0.0  0.0   7016   256 ?        Ss   07:58   0:00 /usr/sbin/spice-vdagentd
liveuser    2436  0.0  0.0 216116   828 pts/0    S+   08:08   0:00 grep --color=auto vdagent

$ rpm -q spice-vdagent
spice-vdagent-0.19.0-2.fc31.x86_64

Notice the error in F31 Workstation output:
> Error getting session for pid 1614: No data available

I see the same errors on the installed system. Running `spice-vdagent --foreground` manually doesn't discover any useful output.

Nominating as a prioritized bug. This is hurting QA performance quite a bit. It will no doubt annoy our users as well.

Hum. I wonder if this is related to my tty? bug?

https://bugzilla.redhat.com/show_bug.cgi?id=1748681

Given the QA impact, should this get a beta FE?

Also: presumably the testing is with Wayland. Does the problem exist in the X fallback, too?

(In reply to Matthew Miller from comment #8)
> Also: presumably the testing is with Wayland. Does the problem exist in the
> X fallback, too?

Yes, it happens also with X11 (tested with F31 Workstation Live). Also happens with setenforce 0.

Matthew: it's already nominated as Beta FE. The 'blocks' field seems to have been hidden behind a 'show advanced fields' button in Bugzilla recently, which is why you don't see it right away.

I'm +1 for this, as it affects the live environment and that can't entirely be fixed with an update, and it's an annoying issue.

Accepted as a Prioritized Bug in today's meeting

CCing bberg for GNOME session changes, as it seems likely that's related here.

Ugh, sounds like we need a fix in spice-vdagentd. i.e. it tries to look-up the session but fails to do so because it is not technically part of any session (yes, this is the case by design and intentional even tough somewhat crazy).

Mutter/GDM/gnome-session all have workarounds for exactly this. I have not looked at spice-vdagent though. Newer systemd also has support to make these workarounds really simple.

I just checked, and the problem is indeed the call to sd_pid_get_session (in src/vdagentd/systemd-login.c). A fallback will be needed there to find the users graphical (or greeter) session. Similar code is already in e.g. mutter (https://gitlab.gnome.org/GNOME/mutter/blob/master/src/core/main.c#L354). Some of the DBus API has a special "auto" value for the session on the systemd side now, which might also be a way to solve this if we can depend on systemd v243.

A workaround for this is to modify /usr/bin/gome-session to pass --builtin to gnome-session-binary.

I can look into this and cook up a patch, might be a few days though.

Thanks a lot. Benjamin, very appreciated. This problem makes QA life much harder than it used to be :)

(In reply to Adam Williamson from comment #7)
> Hum. I wonder if this is related to my tty? bug?
> 
> https://bugzilla.redhat.com/show_bug.cgi?id=1748681

I think Adam is right, this should be fixed at the Gnome/SystemD level instead.

(In reply to Frediano Ziglio from comment #16)
> … this should be fixed at the Gnome/SystemD level instead.

It is a conceptual issue that occurs because we do not have a separation between the user and session scopes. This lack of separation has existed for years now, the only change here is that the entire desktop is now affect rather than "just" DBus activated applications.

Said differently, a fix is will be prohibitively expensive compared to the small improvements that it may allow us to do.

I created a scratch build https://koji.fedoraproject.org/koji/taskinfo?taskID=37647756

It looks good in my testing (i.e. it gets past the session ID lookup), but I didn't test the actual integration. Could someone please verify before I submit the patch upstream and the MR for the package?

(In reply to Benjamin Berg from comment #18)
> I created a scratch build
> https://koji.fedoraproject.org/koji/taskinfo?taskID=37647756

Benjamin, you're my hero! I installed this in my F31 VM and clipboard integration is working perfectly. I'm looking forward for an official fix.

Thanks, created a pull request for the package: https://src.fedoraproject.org/rpms/spice-vdagent/pull-request/1

FEDORA-2019-0989e01e70 has been submitted as an update to Fedora 31. https://bodhi.fedoraproject.org/updates/FEDORA-2019-0989e01e70

spice-vdagent-0.19.0-3.fc31 has been pushed to the Fedora 31 testing repository. If problems still persist, please make note of it in this bug report.
See https://fedoraproject.org/wiki/QA:Updates_Testing for
instructions on how to install test updates.
You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2019-0989e01e70

The fix works, thanks!

*** Bug 1748123 has been marked as a duplicate of this bug. ***

spice-vdagent-0.19.0-3.fc31 has been pushed to the Fedora 31 stable repository. If problems still persist, please make note of it in this bug report.

I believe Adam reopened this by mistake, closing again.

it was unfortunate timing...thanks.