https://bugs.kde.org/show_bug.cgi?id=416652
Konrad Materka <mate...@gmail.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Attachment #125539|0 |1
is obsolete| |
Attachment #125540|0 |1
is obsolete| |
--- Comment #16 from Konrad Materka <mate...@gmail.com> ---
Created attachment 125558
--> https://bugs.kde.org/attachment.cgi?id=125558&action=edit
session bus monitor logs (cleared)
My investigation:
There are 3 parties involved:
:1.10 = org.kde.StatusNotifierWatcher
:1.33 = org.freedesktop.Notifications = com.canonical.Unity = plasmashell
:1.763 - Telegram
Tray icon is created twice:
> Sender=:1.33 Destination=org.kde.StatusNotifierWatcher
> Path=/StatusNotifierWatcher Interface=org.kde.StatusNotifierWatcher
> Member=RegisterStatusNotifierItem -> ":1.765"
> Sender=:1.763 Destination=org.kde.StatusNotifierWatcher
> Path=/StatusNotifierWatcher Interface=org.kde.StatusNotifierWatcher
> Member=RegisterStatusNotifierItem -> ":1.766"
The ":1.765" is immediately destroyed. I was able to observe similar behavior
with snap package on my computer to.
There is a race condition, if service is destroyed too soon,
StatusNotifierWatcher can't notice it. From StatusNotifierWatcher cpp:
//check if the service has registered a SystemTray object
org::kde::StatusNotifierItem trayclient(service, path,
QDBusConnection::sessionBus());
if (trayclient.isValid()) { // it exists here
m_registeredServices.append(notifierItemId); // it dies here
m_serviceWatcher->addWatchedService(service); // it is dead by now, no
chance to get notification
emit StatusNotifierItemRegistered(notifierItemId);
}
Race condition is a problem on it's own, but bigger issue is: why plasmashell
is creating SNI? For a brief moment there a SNI object with hidden menu with
only one item: toggled checkbox.
--
You are receiving this mail because:
You are watching all bug changes.
