[qmtest] passing data from one test to another

[eichin at metacarta.com]

> We have hundreds of tests that simulate telephone calls.  They are
> quite self-contained except for a piece of underlying software that

In tests I have that are similar in concept to this, I've tried two
different approaches - the first is to have scripts that generate the
entire pile of individual tests (with the chosen port numbers) up
front, and then run qmtest on that.  The second is to treat the "set
of calls" as a single test, and just have qmtest run the whole thing
as one test - this scales a lot better, and when the system has passed
the primary smoke tests, these tests are mostly "is this data handled
properly" and so don't need as much detail up the chain.

Another thought is that if your udp port state is related to the SUT
but not to the tests per-se, abstracting it out into a test-runner
that keeps that state and makes the "real" connections, then each
qmtest-level test sends a message to the test-runner, instead of doing
it directly.  That also lets (though this is conflating your two cases
a bit) the "runner" notice the SUT failing and then it can cancel (or
rather, stop accepting requests) for the remainder of the run...