Funny business debugging code today: The 1.5 JVM hung/exited my program instead of showing stack trace for a NullPointerException. There was no message at all. I salted my program with println's and finally found where there was possibly a NPE, but the java command would not show it. I tried the 1.4 java.exe and immediately got the NPE trace! What is going on??? This is the first time I've seen this kind of thing. Going back several years.
The problem is not always repeatable. After I used JVM 1.4, then 1.5 would get and display the stack trace for the NPE. Since there are 17 source files of about 240K bytes, I haven't tried to isolate the problem because it doesn't always happen. It may be related to the mini-IDE/smart editor I'm using (WinEdit). I think I'll go back to JVM 1.4 to finish developing this project.
Frankly, I'm rather amazed to hear that Tiger supports Windows 98. I don't think you're going to be running any C# programs on that OS!
But in any case, it sounds to me like an issue whereby error output doesn't always get flushed on process exit. If you're saying, as I think you are, that this happens when the output is going into an IDE, rather than to the command window itself -- well, again, I'm bloody amazed this works at all.
If you can't afford to upgrade to a 21st century version of Windows, you might consider Linux.