Version:            (using KDE KDE 3.2.2)
Installed from:    Compiled From Sources
OS:                Linux

I am writing a screen saver which needs to clean up its resources (open files, child processes, etc.) when the screen saver exits.

I have a SIGTERM handler installed so I get notified when kdesktop_lock ends.  However, there is a race condition with kdesktop_lock: main() may end before my screen saver process cleanly ends.  When main() ends, it destroys the LockProces object which in turn calls the mHackProc destructor, which sends a SIGKILL to my screen saver if it's still running (i.e. I haven't yet finsished cleaning up my resources).

Here's what's happening when the user unlocks the screen:
1) User unlocks screen.
2) This causes an activated() signal on the QSocketNotifier.  The slot is called (sigtermPipeSignal()).
3) sigtermPipeSignla() calls quitSaver()
4) quitSaver() calls stopSaver()
5) stopSaver() calls stopHack()
6) stopHack() sends a SIGTERM to the my screensaver process via kill().
   6a) my screensaver sigterm handler is called initiating cleanup.
7) return back to quitSaver().  quits the KApplicaiton.
8) KApplication event loop ends.  main() ends.
9) Destructors called for stack objects.  Important one being LockProcess
10) LockProcess destroys member objects - Key one being the mHackProc destructor.  This destructor will send a SIGKILL signal to the screensaver process if it's still running.  That's the problem, if I'm still running, it means I'm still cleaning up and it just kills me in the middle of it, potentially leaving corrupt files, orphaned processes, etc.

The Solution:
The solution is to wait for the child process to end before calling kApp->quit().  This can be done either in stopHack() right after the kill() call or the check can be made in the LockProcess destructor

e.g. use the following code:
KProcessController* kpc = KProcessController::theKProcessController;

while ( mHackProc.isRunning() ) {
   kpc->waitForProcessExit(10);
}


There is a potential problem if the mHackProc process never ends (e.g. a bug in the screen saver SIGTERM/cleanup code).  We could modify the the above code so that if the process hasn't exited within a certain amount of time, we just SIGKILL it so the user can get back to their desktop.  In this case, it would be a bug in the running screensaver.

e.g. 
KProcessController* kpc = KProcessController::theKProcessController;

kpc->waitForProcessExit(10);
// We could then issue a mHackProc.kill(SIGKILL) call; 
// - OR -
// If implemented in the destructor, simply let the DTOR run to completion and the 
// KProcess destructor will issue the call for us.