> This way, if we want to reproduce any test, we just start 'make
> testevn', and don't have to figure out which testevn we want.
> Similarly, we always have a DC or two, and a member server running, and
> you just target the tests at the appropriate hosts.
yes, I like this. We create a test network rather than a test server.
The disadvantage is that the network will tend to grow a lot. To test
clustering we will need at least 4 smbd daemons running, as otherwise
all the different inter-node patterns aren't tested. That could lead
to quite a large list of processes running :)