thread deadlocks on solaris but not on windows

[frag]

Hi!

I wrote a little consumer-producer example just to learn how to handle this kind of problem. The application runs perfect on a Windows2000 machine with Java 1.4.2 but if the same application is executed in a solaris environment with Java 1.3.1 the application hangs up (deadlock?).

Why is it like that?

Here are my classes:

import java.util.*;

public class VectorHandler{

  private Vector myVec = new Vector(100,10);
  
  
  public synchronized Object getElement()
  {
  	Object obj;
	while( myVec.isEmpty() )
	{
	 try{
	  wait();
	 }catch( InterruptedException e){
	 }
	}
	obj = myVec.elementAt(0);
	myVec.removeElementAt(0);
	return obj;
  }


  public synchronized Object readElement(int i)
  {
  	Object obj;
  	if( i < myVec.size() )
  	{
  		return myVec.elementAt(i);
  	}
  	else
  	{
  		return null;
  	}
  }


  public synchronized void putElement(Object obj)
  {
  	myVec.add(obj);
  	notify();
  }

}



public class Thread_PopElement extends Thread {

  VectorHandler oh;
  
  public Thread_PopElement(VectorHandler oh)
  {
  	this.oh = oh;
  }

  public void run()
  {
  	int i = 0;
  	while( i < 100 )
  	{
  	 System.out.println(&quot;Thread POP\t, Element: &quot; + (String)oh.getElement());
  	 i++;
  	}
  	System.out.println(&quot;Thread POP died!&quot;);
  }

}


public class Thread_PushElement extends Thread {

  VectorHandler oh;
  
  public Thread_PushElement(VectorHandler oh)
  {
  	this.oh = oh;
  }

  public void run()
  {
  	String str = &quot;&quot;;
  	for( int i = 0; i < 100; i++ )
  	{
  	 str = &quot;&quot; + i;
  	 oh.putElement(str);
  	 System.out.println(&quot;Thread PUSH\t, Element: &quot; + str);
  	}
  	System.out.println(&quot;Thread PUSH died!&quot;);
  }

}



public class Thread_ReadElement extends Thread {

  VectorHandler oh;
  
  public Thread_ReadElement(VectorHandler oh)
  {
  	this.oh = oh;
  }

  public void run()
  {
  	String str = &quot;&quot;;
  	int i = 0;
  	while( oh.readElement(i) != null )
  	{
  	 str = (String)oh.readElement(i);
  	 System.out.println(&quot;Thread READ\t, Element: &quot; + str);
  	 i++;
  	}
  	System.out.println(&quot;Thread READ died!&quot;);
  }

}



public class SyncTest{

  public static void main(String [] arstring)
  {
  	VectorHandler oh = new VectorHandler();
  	Thread_PopElement  pop		= new Thread_PopElement(oh);
  	Thread_PushElement push		= new Thread_PushElement(oh);
  	Thread_ReadElement read		= new Thread_ReadElement(oh);
  	Thread_PushElement push_again	= new Thread_PushElement(oh);
  	
  	push.start();
  	pop.start();
  	
  	while( push.isAlive() || pop.isAlive() ){
  	}// do nothing... just wait

  	push_again.start();

  	while( push_again.isAlive() ){
  	}// do nothing... just wait

  	read.start();

  	while( read.isAlive() ){
  	}// do nothing... just wait
  	
  	System.out.println(&quot;Main thread died!&quot;); 
  }


}

cheers

frag

patrick.metz@epost.de

 

[sedj]

Hi frag,

I've run your code fine on the below :

Win2000 with SDK 1.4.1_02
Win2000 with SDK 1.3.1_07
Linux RedHat 8 with SDK 1.4.1_02
Solaris 9 (OS 5.9) with SDK 1.4.0

And it all seems to run with no problems ... sorry, not much help ...

 

[frag]

Hi sedj,

thank you for your efforts. This is realy strange.
Our Solaris machine has OS version 5.8 (don't know if this is Solaris 8 or 9) and the java version is the build 1.3.1_02-b02.

It starts with pushing 28 elements in the vector and after that one element gets popped from the vector... and that's all, deadlock.

Could this be a bug in the java build or is it possible that the machine has just not enough performance? A lot of people are working on this machine and there is always a high load on the CPU and only a little memory available. This thing is swapping memory to harddisk the hole time.

cheers

frag

patrick.metz@epost.de

 

[sedj]

Well, I would tend to lean to thinking that this is a machine/config problem rather than the JVM/SDK, considering the OS/SDK versions now tested.

Can you add in debug so that it monitors memory usuage -

Runtime.getRuntime().totalMemory();

or maybe start the JVM with more initial memory -
java -Xmx256m ...

 

[frag]

The problem is solved!

I asked our system administrator to install the newest java release (build 1.4.2_02-b03) and now it runs fine.

Really strange! :-/

patrick.metz@epost.de