I'm controlling a creaky old FORTRAN simulator from a VB.NET GUI, using redirected I/O to communicate with the simulator executable. The GUI pops up a "status" window with a progress bar, estimated time, and a "STOP" button (Button_Stop).

Now, I want the Button_Stop to terminate the simulator process immediately. The obvious way to do this is to call Kill() on the Child Process object. This gives an exception if it's done after the process has exited, but I can test whether the process is exited before trying to kill it, right?

OK, so I do the following when the button is clicked:

If Not Child.HasExited Then
    Child.Kill()
    Button_Stop.Enabled = False
End If

However, what if the process happens to exit between the test and the call to Kill()? In that case, I get an exception.

The next thing to occur to me was that I can do Button_Stop.Enabled = False in the Process.Exited event handler, and thus prevent the Child.Kill() call in the Button_Stop.Clicked handler. But since the Process.Exited handler is called on a different thread, that still leaves the following possible interleaving:

1. Child process exits.
2. Process.Exited fires, calls Invoke to schedule the Button_Stop.Enabled = False
3. User clicks on Button_Stop, triggering Child.Kill()
4. Button_Stop.Enabled = False actually happens.

An exception would then be thrown on step 3.

How do I kill the process without any race conditions? Am I thinking about this entirely wrong?