JVM crash on IE app on PC's , Error file left on desktop

Question

I have a web based application with Struts and an activeX plug in in and extended desktop with two screens in Kiosk mode. One app with three frames. One frame for each screen and one frame with a small Java pause applet used to augment Javascript functions that need a pause.<br><br>
The application runs fine then randomly and occasionally (about once every day or two) the browser disapears and the following error file from the JVM is left on the desktop.
An unexpected exception has been detected in native code outside the VM.
Unexpected Signal : EXCEPTION_ACCESS_VIOLATION occurred at PC=0x6D07014F
Function=Java_sun_awt_windows_WColor_getDefaultColor+0x28D4
Library=C:\Program Files\Java\j2re1.4.1_02\bin\awt.dll
Current Java thread:
     at sun.awt.windows.WToolkit.eventLoop(Native Method)
     at sun.awt.windows.WToolkit.run(Unknown Source)
     at java.lang.Thread.run(Unknown Source)
Dynamic libraries:
0x00400000 - 0x00419000      C:\Program Files\Internet Explorer\IEXPLORE.EXE
Local Time = Thu Feb 15 18:37:23 2007
Elapsed Time = 667
# The exception above was detected in native code outside the VM
# Java VM: Java HotSpot(TM) Client VM (1.4.1_02-b06 mixed mode)
Any ideas what might be causing this failure. Its a production application so this is a serious problem for us.
Thanks
---John Putnam

Answer 1

At first glance this looks like either a bad oop map being generated from the HotSpot server compiler or bad JNI code storing an invalid object into a field via the JNI interface.
If you run with -client rather than -server, does the crash go away? If so, that would confirm the first hypothesis.
Have you tried running with 5.0u5? If you can't (or don't want to) run with that JDK, you could drop the appropriate jvm.dll from the 5.0u5 JDK into your 1.4.2 JDK and that should run. If the crash still occurs with that configuration then the hs_err log should at least be more useful than the one from 1.4.2. I would also recommend running with -Xcheck:jni on 5.0u5 to see if any bad JNI calls are detected in ColdFusion's native code.