|
|
|
|
Words of wisdom
|
|
|
|
|
|
|
|
|
|
1. Make sure all changed work on both KDE and Gnome
|
|
|
|
|
|
|
|
|
|
2. The System tray protocol specification can be found at
|
|
|
|
|
http://freedesktop.org/Standards/systemtray-spec/systemtray-spec-0.1.html
|
|
|
|
|
|
|
|
|
|
3. The KDE System tray Docking Protocol can be found at
|
|
|
|
|
http://developer.kde.org/documentation/library/kdetqt/kde3arch/protocols-docking.html
|
|
|
|
|
|
|
|
|
|
4. The site i use for Xlib reference
|
|
|
|
|
http://tronche.com/gui/x/xlib/function-index.html
|
|
|
|
|
|
|
|
|
|
5. Qt Documentation at http://doc.trolltech.com
|
|
|
|
|
|
|
|
|
|
6. There's do or no do. There's no try.
|
|
|
|
|
|
|
|
|
|
KDocker Window Docking Strategy and a little bit of history
|
|
|
|
|
-----------------------------------------------------------
|
|
|
|
|
Couple of things that will help you if you are looking through the code.
|
|
|
|
|
I wanted a single instance of KDocker to be running all the time. So,
|
|
|
|
|
whenever a new instance of KDocker was created it would ask the previous
|
|
|
|
|
instance take over.
|
|
|
|
|
|
|
|
|
|
Two things to know about
|
|
|
|
|
1. Mapping from the PID (process id) to WID (window id)
|
|
|
|
|
2. How to let the previous instance know about app_to_launch
|
|
|
|
|
|
|
|
|
|
Lets deal with them one by one.
|
|
|
|
|
1. PID to WID mapping is easy if people followed rules. Just read _NET_WM_PID
|
|
|
|
|
for a window. If it matches the one app's PID we just created, that window is
|
|
|
|
|
the one we want to dock. Its not as simple as that. Programs have "splash"
|
|
|
|
|
screens. A nice thing is that these "splash" screen have the _WM_TRANSIENT_FOR
|
|
|
|
|
property set. So, we can skip those windows. There are these bunch of programs
|
|
|
|
|
(like gaim, xmms) that create windows and dont put this property on them. Arrgh.
|
|
|
|
|
What we do to counter these applications is to look for interesting windows
|
|
|
|
|
possibly created by the process until the process dies. Have a look at the code
|
|
|
|
|
for a better understanding.
|
|
|
|
|
|
|
|
|
|
2. Sending ClientMessages to previous instances with the program name and
|
|
|
|
|
arguments is painful and makes the code more complex. Before 1.0, I used to dump
|
|
|
|
|
all the program arguments in ~/kdocker.<pid> and send the pid in a
|
|
|
|
|
ClientMessage. The file is then deleted after processing. As of 1.0, I set a
|
|
|
|
|
property _KDOCKER_<pid> on the KDocker initial instance's window which holds
|
|
|
|
|
the program arguments. (The correct way to transfer information is using
|
|
|
|
|
XSelections)
|
|
|
|
|
|
|
|
|
|
TRACING
|
|
|
|
|
-------
|
|
|
|
|
Add the following line in kdocker.pro (in fact just uncomment it)
|
|
|
|
|
DEFINES += ENABLE_TRACING
|
|
|
|
|
|
|
|
|
|
Now do a "make clean" and "then make".
|
|
|
|
|
|
|
|
|
|
Launch KDocker. Try recreating the problem.
|
|
|
|
|
- Right Click on the tray icon.
|
|
|
|
|
- Choose "About KDocker"
|
|
|
|
|
- You should see a button called "Show Trace". Please note that this button
|
|
|
|
|
will appear only when compiled with tracing
|
|
|
|
|
- Try debugging yourself or send it to me
|
|
|
|
|
- If you send SIGUSR1 to the KDocker process, it will dump all the trace to
|
|
|
|
|
~/kdocker.trace
|
|
|
|
|
|
