Hello Juerguen,
Juergen Borleis <jbe@...> writes:
> > > Due to the fact it is very hard to reproduce I have no new information on
> > > that issue. It still hasn't hit me. It seems it depends on the userland
> > > workload (for example how many programs allocate the timer...).
I am in the process of testing the latest BSP from GIT as you suggested but
in the meanwhile I found a quite reliable way to reproduce the problem.
Using the following:
- PTXdist 2013.09.0
- mini2440 BSP 2013.10.0
- Kernel 3.11
- Standard ptxdist configuration, plus JamVM and GNU Classpath
Run the following Java program on the target:
===
import java.io.IOException;
public class ThreadTest implements Runnable {
private static int counter = 0;
public void run()
{
while (true)
{
try {
Thread.sleep(100);
}
catch (InterruptedException ex) { System.err.println(ex); }
long time = System.currentTimeMillis();
long ktime = System.nanoTime() / 1000;
if (counter++ % 1000 == 0) {
System.out.println(time + " (" + ktime + ")");
}
}
}
public static void main(String args[]) throws IOException
{
for (int i = 0; i < 10; i++) {
Thread t = new Thread(new ThreadTest());
t.start();
}
}
}
===
This just starts 10 threads, each thread runs an endless loop sleeping
100 ms and waking up. A trace is printed to stdout every 10 seconds approx.
This typically reproduces the problem in a matter of minutes (most of the
time within 1-2 minutes).
Hope this helps,
Stella
