[panel] Indicators missing at Unity startup

I'm a Maverick user and I'm testing Unity Qt from Maverick and I have the same issue, just have the nm-panel (inactiv)

Thanks guys for reporting that important issue.

I have also

updated .xsession-errors file with latest unity from ppa
indicators loaded after some minutes

Yes, it Justamante that the error, though in my case the user change the load several times achievement

I don't have this issue anymore.

(don't know about Charlotte, I don't use it)

Lucazade, thanks for the feedback. Do you still use Maverick and the Unity 2D PPA?

I'm using Natty and Unity 2D PPA.. unfortunately I don't know if this is fixed also for Maverick.

Hi

I'm using Maverick and the Unity-2D PPA and have not had this issue recently.

Hope that helps.

Andy

After much digging, I believe the bug comes from the indicator processes shutting down themselves because the panel is not fast enough to connect to them. This can be workaround with two environment variables:

- INDICATOR_SERVICE_SHUTDOWN_TIMEOUT: the time before the process shutdown itself in milliseconds. Default value is 500 milliseconds, but the system multiplies it by two at startup.

- INDICATOR_ALLOW_NO_WATCHERS: if this environment variable is set the indicator processes won't shut down themselves automatically.

Can you try this?
- Create a file named /etc/X11/Xsession.d/72indicators-no-shutdown, with the following content:
export INDICATOR_ALLOW_NO_WATCHERS=1

- Restart your session and see if you still experience the bug.

created the file /etc/X11/Xsession.d/72indicators-no-shutdown with the contents
export INDICATOR_ALLOW_NO_WATCHERS=1

I did not make a different. Reboot/restarting session still has the long delay before any indicators are shown.

attaching latest .xsession-errors with the ENV variable set

@bill Is this on Maverick or Natty? if it's on Natty it is really weird because DBus names have been changed from org.ayatana to com.canonical in Natty. If it is Maverick, then you may want to grep for "getenv" in libindicator source code, maybe the environment variables changed.

@aurelien
It's actually Lucid with unity + deps backported from Maverick! I will check the code.

It seems that this problem is caused by the autostart of nm-applet.
If you remove /etc/xdg/autostart/nm-applet.desktop out of that directory OR successfully connect to an access point, the indicators seem to be start up when expected.

further debug required here to determine the cause of the block.

not sure if this affects maverick/natty. It is highly reproducible on the charlotte project which is used natty and slightly different networking backend.

sorry, I meant to say charlotte uses Lucid not Natty.

Indicator processes are started by dbus, so their output does not show up in .xsession-errors. I put together a short shell wrapper to get log files. The following procedure should set it up for all indicator processes:

cd /usr/local/bin
sudo wget http://people.canonical.com/~agateau/tmp/log-indicator-service
sudo chmod +x log-indicator-service
for x in /usr/share/dbus-1/services/indicator-*.service ; do sudo sed -i 's,Exec=,Exec=/usr/local/bin/log-indicator-service ,' $x ; done

Then restart your session. You should get log files for all indicator processes in ~/.cache/indicators/

Note that this modifies files handled by the package system, so any update of an indicator-* package will remove the call to log-indicator-service.

To revert the changes:

for x in /usr/share/dbus-1/services/indicator-*.service ; do sudo sed -i 's,Exec=/usr/local/bin/log-indicator-service ,Exec=,' $x ; done
sudo rm /usr/local/bin/log-indicator-service

marking this bug invalid as it can no longer be reproduced. please re-open if it occurs again.