Half-random NullPointerException in OpenAL

Troubleshooting sound
1

This is pretty suspicious stuff. Most of the time, the program runs fine. But sometimes, we get the following stacktrace (given in context - not that it matters).

22-jun-2007 14:08:30 com.jme.util.lwjgl.LWJGLTimer <init>
INFO: Timer resolution: 1000 ticks per second
22-jun-2007 14:08:30 com.jmex.game.DefaultUncaughtExceptionHandler uncaughtException
SEVERE: Main game loop broken by uncaught exception
java.lang.NullPointerException
        at com.jmex.sound.openAL.SoundSystem.update(Unknown Source)
        at com.jmex.game.StandardGame.update(Unknown Source)
        at com.jmex.game.StandardGame.run(Unknown Source)
        at java.lang.Thread.run(Unknown Source)
22-jun-2007 14:08:30 com.jmex.sound.openAL.scene.SoundNode <init>
INFO: Node created.



On another run, the position might be slightly different (below the SoundNode constructor, for example). So, this sounds a damn lot like a racing condition of some sorts. I started digging at the sources and the only suspicious line I could find is line 202 in com.jmex.sound.openAL.SoundSystem.java

nodes[nodeName].updateWorldData(time);



Proposed patch:

Index: src/com/jmex/sound/openAL/SoundSystem.java
===================================================================
RCS file: /cvs/jme/src/com/jmex/sound/openAL/SoundSystem.java,v
retrieving revision 1.15
diff -u -r1.15 SoundSystem.java
--- src/com/jmex/sound/openAL/SoundSystem.java   6 Mar 2007 15:30:00 -0000   1.15
+++ src/com/jmex/sound/openAL/SoundSystem.java   22 Jun 2007 12:26:00 -0000
@@ -199,7 +199,8 @@
     public static void update(int nodeName, float time){
         if(nodes==null) return;
         if(nodeName<0 || nodeName>=nodes.length) return;
-        nodes[nodeName].updateWorldData(time);
+        if(nodes[nodeName] != null)
+            nodes[nodeName].updateWorldData(time);
         updateListener();
        
     }



Because of the random nature of this crash, I can't tell for sure this solves it. I haven't had any problems in the couple of runs I did with patched sources, I'll tell you if I have. And perhaps the root of the problem is deeper than this. Have fun figuring it out.
2

If you compiled jME with debug option on wouldn't it tell you the exact line saving the guess work?

3

Err… Probably yes. But then, I still have to get that crash again…

4

Well, I got the NullPointerException again. Now I'm officially confused. Researching this tonight.



Hmm. Of course, it was the other update getting called. New patch looks something like

Index: src/com/jmex/sound/openAL/SoundSystem.java
===================================================================
RCS file: /cvs/jme/src/com/jmex/sound/openAL/SoundSystem.java,v
retrieving revision 1.15
diff -u -r1.15 SoundSystem.java
--- src/com/jmex/sound/openAL/SoundSystem.java   6 Mar 2007 15:30:00 -0000   1.15
+++ src/com/jmex/sound/openAL/SoundSystem.java   26 Jun 2007 21:10:42 -0000
@@ -185,7 +185,8 @@
         if(nodes==null) return;
        
         for(int a=0; a<nodes.length; a++){
-            nodes[a].updateWorldData(time);
+           if(nodes[a] != null)
+              nodes[a].updateWorldData(time);
         }
         updateListener();
        
@@ -199,7 +200,8 @@
     public static void update(int nodeName, float time){
         if(nodes==null) return;
         if(nodeName<0 || nodeName>=nodes.length) return;
-        nodes[nodeName].updateWorldData(time);
+        if(nodes[nodeName] != null)
+            nodes[nodeName].updateWorldData(time);
         updateListener();
        
     }



I've noticed a lot of those 'unchecked' array references that might be null if threads decide to intermingle. This is of course not unlikely (hell, it happened to me and I'm Mr. Center-of-the-Gauss-Curve) if StandardGame's update() gets called every few moments while complex objects are being moved around in the createSoundNode code.
5

Good points…  I will check the jmex.audio package for similar problems as jmex.sound will eventually be dropped.

6

i just wanted to add, that i get the same Exeption maybe once ever 10 starts.