It is possible to get NullPointerExceptions while backtracking through the
stack frames in Agent.java.
---------------------------------------------------------------------------
I finally got some feedback on this :
> Essentially, when a stack trace is asked for (case CMD_GET_STACK_TRACE), you
> can get a StackFrame sf where sf.className is null. The code then tries to
> out.writeUTF(sf.className), which causes a null ptr exception. I added very
> little code to check for this condition.
>
> Now I can't remember if you could force this condition w/ JDB alone. I KNOW
> we could reproduce it easily w/ ICAT when DosDebug was being used
> because the JVM is halted more frequently (dll loads), and ICAT always unwinds
> the current thread's stack. If the information that I've given you above is
> insufficient to sway JavaSoft, I'll either recreate this *$#@! problem w/ JDB
> OR I'll dig deeper into the code to show them why the null check is a good idea.
Hope this helps,
mick.fleming@Ireland 1998-04-29
stack frames in Agent.java.
---------------------------------------------------------------------------
I finally got some feedback on this :
> Essentially, when a stack trace is asked for (case CMD_GET_STACK_TRACE), you
> can get a StackFrame sf where sf.className is null. The code then tries to
> out.writeUTF(sf.className), which causes a null ptr exception. I added very
> little code to check for this condition.
>
> Now I can't remember if you could force this condition w/ JDB alone. I KNOW
> we could reproduce it easily w/ ICAT when DosDebug was being used
> because the JVM is halted more frequently (dll loads), and ICAT always unwinds
> the current thread's stack. If the information that I've given you above is
> insufficient to sway JavaSoft, I'll either recreate this *$#@! problem w/ JDB
> OR I'll dig deeper into the code to show them why the null check is a good idea.
Hope this helps,
mick.fleming@Ireland 1998-04-29