Advancewd Debugging tactics needed!

Question

Hi,
I was wondering if any of you more experienced prgrammers can help me with a big problem I'm trying to solve.
I'm working on a large multi-threaded server based app running on Suse Linux and jdk 1.4.1_01.
The code has a lot of database access, and calls a lot of native code for complex calculations and a commercial messaging infrastructure.
The problem is that at a certain point on startup, the application has recently started crashing and I'm having trouble finding out the cause. Its a strange crash because no exceptions are produced, the JVM just dies with no HotSpot reports or anything similar (i.e. the linux process dies). I dont even get a core dump (moving tha app to Solaris saw the same error occurring, this time with a core produced which seemed to say nothing of any significance).
I currently suspect its a bug in the HotSpot compiler, as the program works fine when I used the -Xint VM option (although obviously much slower). The problem is tracking down why this occurs and precisly where in the code.
This leads to my next problem...I have found that by adding logging statements at various points, alters the place that the app crashes (again this points to a hotspot bug as log statements will alter the order that commands are executed by the JVM).
So now I'm stuck....I've got a Schrodingers (sp.) Cat situation where adding logging changes the effect of the bug.
Anyone have any ideas how I should proceed from here?
cheers.

Answer 1

Well there's certainly no throwables/Exceptions occuring in the code. I kinda suspected this would be the case due to the JVM just falling over like it does, but worth a try anyway.
I've got a small sample of code here which seems to be at the root of the problem, maybe someone might spot something. Unfortunately I can't provide a succinct example that can be run on its own, as it seems the problem is caused by the highly concurrent nature of the app. (By the way its running on a dual cpu machine with hyperthreading switched on).
System.out.println("this always gets printed");
synchronized(anObject){
    System.out.println("this sometimes gets printed");
    while(someTest){
        System.out.println("this always gets printed if the one above does");
         callSomeMethod();
         System.out.println("this never gets printed");
}When I say "sometimes gets printed" I mean that either it appears in stdout, or the JVM dies. So in summary, with little or no logging, the JVM dies fairly consistently around the synchronized statement. If I add extra logging, esp. in the callSomeMethod(), then the JVM still dies, but later in the code.
All very strange