Another halting Thread problem

Question

Hello,
I want to make a benchmark software using Java, and I want to add capability to stop the test if such setting was selected that the test would take forever. I know that I can do this with using such code like:
public class MyBenchThread extends Thread {
boolean toldToStop = false;
public void run() {
  while (!benchmarkEndConditionHasMet && !toldToStop) {
   // Run heavy benchmark here
public void haltTest() {
  toldToStop = true;
}and calling haltTest() from another Thread.
However, this has two problems:
1) To have the benchmark at its purest level, I would like
to avoid extra computation as much as possible. Is there a way to give the same effect of killing the thread without constantly checking a condition variable?
2) If the benchmark involved some I/O operation (not necessarily over a socket to which you can set a timeout) that takes a while to complete, I would not be able to halt it using the above code or interrupt() function in Thread class. I know that stop() is likely to cause problem... So is there any way to force interrupt execution of the Thread including I/O, do whatever cleanup is needed, and gracefully exit?
Thanks in advance

Answer 1

Sorry for late response... Had been away for weekend.
Well, interrupt() does not seem to work on I/O operation. It only works on a Thread that is actually sleeping or waiting on a monitor. If it was doing some heavy number crunching or I/O, I think it will only set its interrupted status to true, and does not actually suspend its operation. (This is my assumption; correct me if I am wrong)
Hmm... so I guess there is not really a way to get around this. I was having problem with using a proprietary library that established a CORBA session inside it, and it did not provide any way of halting the operation externally. Unless I Ctrl-C whole VM, the library call just kept on going for about 30 secs till it finally gave up.
I don't know... Is force trying to stop a problematic thread that dangerous thing in general to do in multiprocess programming, even if you knew what you were doing?