Bug 75944

Version:            (using KDE KDE 3.2.0)
Installed from:    Compiled From Sources
Compiler:          gcc 3.4 prerelease and gcc 3.3.2 
OS:          Linux

Kget does not work on my machine with RedHat 9. It does not crash in the strong sense, but it hangs forever in a call to QThread::wait(). This behavoiur is not deterministic. Sometimes it hangs after the first download, sometimes during the second, but at least before the fifth one it hangs.

This results from a IMHO unappropriate use of a condition variable, QThread::terminate() and QThread::wait(). The patch in the next message solves the problem. The patch was made against the actual CVS KDE_3_2_BRANCH.

Kget hangs because QThread::terminate() is called and the thread does not reach a cancellable state. The patch is a first step to a correct solution. Because it waits on the mutex for new jobs, the thread can be cancelled by QThread::terminate().

Some additional remarks:

- QWaitCondition::wait(int) should not be used.
- QThread::terminate() should not be used at all. A better solution would be to define a TERMINATE Slavecommand. The slave exits Slave::run() when this command is emitted. I could not find out how Slave::slotResult is called and which thread is responsible for what, but I think Slave::slotResult is the appropriate Place to send such a command to the Slave.
- Is there a design document which explains how threads are used in kget? After looking only at slave.h, slave.cpp and transfer.cpp a little bit, it seems that threads are not necessary at all.