Arnaud, just out of interest, what are your new tests that crash and restart the broker for?
Testing recover of a persistent store perhaps? Just in case you did not realize it, there are actually some tests that already do this. At the moment we run them manually, not automated, though. On 27/09/2007, Arnaud Simon <[EMAIL PROTECTED]> wrote: > > On Thu, 2007-09-27 at 13:53 +0100, Martin Ritchie wrote: > > > > > > ok, I agree with that and I am adding the step of starting/stopping a > > > remote broker. I am doing it in the client module for now. I would > > > suggest that we then review those tests and decide together the ones > we > > > want to move into integration Tests. > > > > Before you rush in to doing this might be work taking a look at the > > way the python tests are run against the java broker as part of the > > build just now. There is the RunBrokerWithCommand class that starts a > > broker runs an external command then quits > > > > It would be very easy to duplicate the pom configuration and simply > > provide the command 'mvn test broker=localhost:<port>'. All we need is > > the tests to understand the broker property. Though IIRC this won't > > exactly work as the current code base will not have been installed > > until the test cycle has completed. > > We need a way of running a new broker instance before a test. For code > reuse sake I don't want to change all the existing tests and it looks to > me that the current tests don't do the necessary cleanup as they rely on > a in_VM broker that is started before each test. So, we need to start > and stop the broker before and after running a test. > Moreover, the new tests I am adding are relying on crashing and > restarting the server. So, we need a way of doing that from within the > tests and I don't think we can rely on maven for that. > >
