On 29 April 2016 at 13:06, sebb <[email protected]> wrote: > I assume that the fix works, however I have not been able to reproduce > the original issue. > Maybe it needs a more complicated test. Has anyone else got a reproducer? > > However I wonder if this is the correct solution? > > The bug mentions TidyRMI, but I don't see how that is relevant. > The intention is that the TidyRMI method is be called after the test > is complete, in order to allow JMeter to exit cleanly. Also it only > interrupts the "RMI Reaper" thread, so it should not have an effect on > anything else.
It seems that the stopSoon thread is started under RMI, which would explain the issue. > If it is somehow interfering with threads that are doing work, then > this ought to be fixed if possible. (Not possible) > It also seems wrong to create a non-Daemon thread in a Daemon thread - > that seems very fragile. > I suspect that is the real issue - since we expect the ListenToTest > thread to complete, it should not be a Daemon thread. However it turns out that the testEnded method is not called by the original TestListener thread as may at first appear. It is called by an RMI thread which is a daemon thread. So the fix does now make sense. Sorry for the noise.
