David Brown wrote:
The main bottleneck we started running into is that the whole test suite was taking longer and longer to run,
That's one of the killers for this sort of thing. Another I've run into is where it either costs money to run the tests (because, for example, you're accessing someone else's for-fee server) or when you're trying to test long timeouts (where, say, you should do something if you don't get an answer within 30 minutes).
I guess you could let the check-in provisionally pass if when you check it in you say which tests you think you affected, and those tests do what they're supposed to, then at lunch or in the evening you run the whole suite of tests again.
This sounds like an entire project itself, doesn't it? Sadly, I tend to wind up working in groups too small to support this kind of thing.
-- Darren New / San Diego, CA, USA (PST) -- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-lpsg
