Calls to a DBus activated service can fail if the sender quits whilst
the service is spawning
(https://gitlab.freedesktop.org/dbus/dbus/-/issues/72) and using
dbus-daemon.
This is a lazy interim fix that just makes these calls block as proper
fixes will require more work.
BUG: 432460
During Plasma 5.18, ksmserver was cleverly split so it hosted the new
new interface registering both services names org.kde.ksmserver and
org.kde.Shutdown.
This way we could do a gradual port, update the libs, and migrate the
final code without any breakages. It was a good plan in theory.
In 5.19 we did the final actual splitting, unfortunately in
libkworkspace on the path where we skip logout confirmation had the
wrong name.
BUG: 423391
Summary:
BUG: 414399
FIXED-IN: 5.18
Test Plan:
1. Lock screen
2. type something in the password field
3. Close lid/suspend system
4. Reopen Lid
Before:
Password text is kept as in step 2
After:
Password text is empty
Reviewers: #plasma, broulik, davidedmundson
Reviewed By: #plasma, davidedmundson
Subscribers: plasma-devel
Tags: #plasma
Differential Revision: https://phabricator.kde.org/D25487
Summary:
The old code path is old and even sometimes uses the ICE protocol to
start a logout from ksmserver. It's quite archaic and unreadable,
KDisplayManager is even worse.
The new code consists of two parts.
A QML friendly front end QObject that we can register in a plugin that
can trigger actions such as prompting a logout or suspending the system
and a singleton that talks to the relevant backend.
Backends are:
* Any logind interface
* Consolekit2
* Consolekit0.4 + UPower
New API loading is also entirely asyncronous.
There are some behavioural changes:
* Creating a logout prompt is called directly instead of going via
ksmserver. The prompt then calls into ksmserver when accepted
* Suspend is called directly instead of going via powerdevil through
some kdelibs4support solid code.
All DBus calls use generated XML files, instead of big QDBusMessage
commands. It's a bit overkill in terms of what we generate, but that's
ready for moving SessionModel in here and killing KDisplayManager.
Patch looks huge because of all the XML files, but is otherwise very
simple.
It is not currently an API break.
Test Plan:
Ran sessionstest on my logind machine
A BSD user tested CK1
Reviewers: #plasma, broulik
Reviewed By: #plasma, broulik
Subscribers: pino, broulik, plasma-devel
Tags: #plasma
Differential Revision: https://phabricator.kde.org/D19389
David Edmundson
Alexander Lohnau
Méven Car