Hi Duncan,

The bug appears to be triggered mostly (only?) in multithreaded applications. I guess it might rear its head when a pool in one thread catches an autorelease that originates in an object that was created in another thread. Or maybe it's something completely different. Here's a class that crashes about nine times out of ten when I run it on my machine. Unfortunately, I can't build a proper test case because I'm not familiar with the process of making a multithreaded test.

------------------

import java.lang.ref.WeakReference;

import org.rococoa.cocoa.NSAutoreleasePool; import org.rococoa.cocoa.NSString;

public class NSAutoreleasePoolThreadTest implements Runnable {

static NSString str; NSAutoreleasePool pool;

public void run() { pool = NSAutoreleasePool.new_();

System.out.println(str.substringFromIndex(4));

pool.release(); }

public static void main(String[] args) { NSAutoreleasePool pool_main = NSAutoreleasePool.new_(); str = NSString.CLASS.stringWithString("teststr");

NSAutoreleasePoolThreadTest arptt = new NSAutoreleasePoolThreadTest();

Thread thread = new Thread(arptt); thread.start(); try { thread.join(); } catch (InterruptedException e) { }

WeakReference<Object> reference = new WeakReference<Object>(arptt.pool); arptt.pool = null;

pool_main.release();

while (reference.get() != null) { System.gc(); } }

}

Cheers, Michael

On 27.07.2009, at 17:49, Duncan McGregor wrote: