The PluginActivated DBus signal is no longer sent

[tari01]

Expected behaviour

When I subscribe to org.mate.SettingsDaemon + /org/mate/SettingsDaemon + PluginActivated via DBus, I expect to receive a notification for each activated plugin.

Actual behaviour

As of version 1.27.1, the daemon no longer sends the PluginActivated signal

Steps to reproduce the behaviour

Launch the mate-settings-daemon and create a signal watcher for PluginActivated.
I have written a small Python test script to demonstrate this (works as expected on 1.27.0, does nothing on 1.28.0):

#!/usr/bin/env python3

from gi.repository import Gio
from gi.repository import GLib
import subprocess

pLoop = None

def onPluginActivated (pConnection, sSender, sPath, sInterface, sSignal, lParameters, pData):

    print (sInterface, sSignal, lParameters[0])

def onNameAppeared (pConnection, sName, sOwner):

    pConnection.signal_subscribe (None, "org.mate.SettingsDaemon", "PluginActivated", "/org/mate/SettingsDaemon", None, Gio.DBusSignalFlags.NONE, onPluginActivated, None)

if __name__ == '__main__':

    Gio.bus_watch_name (Gio.BusType.SESSION, "org.mate.SettingsDaemon", Gio.BusNameWatcherFlags.NONE, onNameAppeared, None)

    pSubprocess = subprocess.Popen (["/usr/lib/mate-settings-daemon/mate-settings-daemon", "--replace"])

    pLoop = GLib.MainLoop.new (None, True)
    pLoop.run ()

m-s-d-test.py.zip

MATE general version

1.28.0

Package version

1.28.0

Linux Distribution

Arch/Debian

[tari01]

The cause of the issue seem to be the changes introduced in b99fca7 by @yetist. Even after going through the whole commit, I am not able to figure out why the daemon stopped sending signals...

[lukefromdc]

I don't have enough understanding of how dbus works to do anything with this one, so a fix will have to wait for one of the other team members.

[joakim-tjernlund]

The cause of the issue seem to be the changes introduced in b99fca7 by @yetist. Even after going through the whole commit, I am not able to figure out why the daemon stopped sending signals...

Can you revert that commit locally and see if that fixes the problem?

[tari01]

The cause of the issue seem to be the changes introduced in b99fca7 by @yetist. Even after going through the whole commit, I am not able to figure out why the daemon stopped sending signals...

Can you revert that commit locally and see if that fixes the problem?

Yes. I've been using my own build of mate-settings-daemon with that commit reverted since I reported this issue. It works properly as before. There are some build changes that also need to to be reverted for the build to succeed.

[joakim-tjernlund]

The cause of the issue seem to be the changes introduced in b99fca7 by @yetist. Even after going through the whole commit, I am not able to figure out why the daemon stopped sending signals...

Can you revert that commit locally and see if that fixes the problem?

Yes. I've been using my own build of mate-settings-daemon with that commit reverted since I reported this issue. It works properly as before. There are some build changes that also need to to be reverted for the build to succeed.

Then perhaps do a MR with that revert, if new stuff breaks and no one fixes it, then it has to go.

[lukefromdc]

The worry is, will distros drop the old code? I don't have enough understanding of dbus to fix this and really can't maintain the whole DE solo, so this may take a while until other maintainers get back

[joakim-tjernlund]

The cause of the issue seem to be the changes introduced in b99fca7 by @yetist. Even after going through the whole commit, I am not able to figure out why the daemon stopped sending signals...

Can you revert that commit locally and see if that fixes the problem?

Yes. I've been using my own build of mate-settings-daemon with that commit reverted since I reported this issue. It works properly as before. There are some build changes that also need to to be reverted for the build to succeed.

I tried the same, revert just this commit(had to add -ldbus-1 too) but this completely lost track of monitor positions in mate-display-properties

I think there are more lost signals here, like when I power off an external display, MATE does not see this properly. If I bring up mate-display-properties and change something, then MATE notices that one monitor is gone.

[sunweaver]

@joakim-tjernlund The step forward, i.e. dropping dbus-glib (deprecated upstream, Debian wants to phase it out), is the right way to go. The way forward here is to fix the GDBus implementation.

@tari01 Maybe in January you can spend some more time on this (@tari01 works in my company...).