Re: [infinispan-dev] Infinispan smoke tests?
I'd avoid profiles if possible. It is easy for us to remember all these
stupid profiles rules and such but imagine someone contributing code and
now having to understand complex test running, they need to read wiki
instructions... The end results will be more problems down the road.
Keep it simple. Before proceeding into profiles why not give these slow
tests another look to see if they can be somehow sped up? It would be so
cool if we could assign thread pools to test groups in testng :(
On 09-09-04 10:49 AM, Galder Zamarreno wrote:
Hi,
Due to the complexity of tests that have been added to the Infinispan
test suite, the time it takes to run the testsuite has almost doubled.
These tests (i.e. distribution rehashing, non-blocking state transfer)
need to be run, that's for sure.
However, I was thinking whether we could create a brand new group of
tests called "smoke". The aim here is for the tests in that group to run
lightning fast and cover 90% of the testsuite.
I think this would help find most of the regressions that are sometimes
introduced for not running the testsuite locally.
Once hudson issues have been solved, we'll be in a better situation but
I still think having this 'smoke' group could help avoid regressions.
Obviously, the danger here is people always running this profile and
then discovering loads of test fails when the entire testsuite is run
but we have this issue now too.
Thoughts?