JVM won't die after shader compile

Jeffg · November 10, 2010, 12:20pm

I have a Windows 7 machine (ATI HD3470), that runs a jogl app and the app exits fine (as far as closing it from the screen). However, the JVM is not terminating in the background. It stays active then jumps to 100% cpu on a core. I’ve isolate the piece of code that’s causing the issue.

  gl.glCompileShaderARB(vertexShader);

But the shaders are working just fine. The glGetInfoLog reports that the shader compiled successfully. But for some reason, when I exit the application, the JVM won’t terminate.

I’ve tried this… but it doesn’t help.

  gl.glUseProgramObjectARB(0);
  gl.glDetachObjectARB(programObject, vertexShader);
  gl.glDeleteObjectARB(vertexShader);
  gl.glDetachObjectARB(programObject, fragmentShader);
  gl.glDeleteObjectARB(fragmentShader);
  gl.glDeleteObjectARB(programObject);

ZbuffeR · November 10, 2010, 12:54pm

Are you using any non-daemon threads ?
After closing the window and having jvm at 100% of one core, run a jstack on its pid, store the result as file, do it a few times, you will have a good idea of the infinite loop which is running …

Jeffg · November 10, 2010, 4:38pm

I haven’t created any additional threads and it works fine on most machines, even works on a VMWare Windows 7 machine fine. But for some reason (and I’ve installed the latest Java) this particular ATI card (running the latest drivers) is giving me an issue. Even a halt() doesn’t kill the JVM.

Runtime.getRuntime().halt(1);

It’s not stuck in a loop. It finishes everything fine and gets to the System.exit() without issue. Something with the glCompileShader is creating a lock that won’t let the JVM close. I can do an exit() right up to the point I compile the shader, any point after that and it will not exit the JVM. I’ll give a look at jstack on Friday.

Jeffg · November 12, 2010, 11:52am

Ok… I got it to die… but perhaps someone could give me some advice on minimizing the potential damage I may be doing, if any. Here is what I had to do.

System.exit(0);

try {
if (OS.startsWith(“Windows”)) {
String pidstr = java.lang.management.ManagementFactory.getRuntimeMXBean().getName();
String pid[] = pidstr.split("@");
Runtime.getRuntime().exec("taskkill /F /PID " + pid[0]).waitFor();
}
}
catch (Exception e) {
}

ZbuffeR · November 12, 2010, 2:04pm

Yeah well to know what damage it would have caused, I would have been nice the see those jstacks … wink wink nudge nudge

Jeffg · November 15, 2010, 7:03am

Ok… ran jstack. It runs fine when the program is running (output is below). If I close the application (JVM still running because it won’t close), jstack gives me “Access Denied”.

Here is the output of the running application:

2010-11-15 10:04:04
Full thread dump Java HotSpot™ Client VM (17.1-b03 mixed mode):

“Direct Clip” daemon prio=6 tid=0x4a414400 nid=0xaf8 in Object.wait() [0x4f75f000]
java.lang.Thread.State: WAITING (on object monitor)
at java.lang.Object.wait(Native Method)
- waiting on <0x19933720> (a java.lang.Object)
at java.lang.Object.wait(Object.java:485)
at com.sun.media.sound.DirectAudioDevice$DirectClip.run(Unknown Source)
- locked <0x19933720> (a java.lang.Object)
at java.lang.Thread.run(Unknown Source)