On Fri, Jul 6, 2012 at 3:26 PM, Isabel Drost <[email protected]> wrote: > On Fri, Jul 6, 2012 at 9:28 AM, Frank Scholten <[email protected]> wrote: >> Yeah, I think that would be useful to have. An alternative is to use >> Maven profiles and use conventions like *IntegrationTest or >> *MapReduceTest as part of the name of the test class. > > Jepp - true. One thing though: I never managed to convince my IDE to > not run tests named *ITest* when right-clicking on the package or > folder containing the tests.
This can be solved by having subpackages like 'integration' and 'mapreduce' in the test source tree. > (And yes, I have to admit that I also> like some other features the random > testing component provides - e.g. > tracking leaking threads, executing tests out of order etc. and needed > an excuse to play with them during my holiday.) > >> Would it make a difference in test speed if we used MRUnit for some of >> the tests? Just curious. > > I haven't looked too deeply into what our tests do - the ones I did > mark @Nightly mostly looked like integration tests that check the > whole workflow instead of the individual units. Which does not mean > that using MRUnit for some of our tests might not actually speed > things up. I think there are also a few tests that use more iterations / data then necessary. > > > Isabel >
