On 25.05.2013 15:04, Emmanuel Lécharny wrote: > Le 5/25/13 2:01 PM, Stefan Seelmann a écrit : >> On 25.05.2013 12:14, Emmanuel Lécharny wrote: >>> The PasswordPolicy tests are expensive because they need to sleep to >>> check some time condition. At this point, I'm wondering if it would not >>> worth creating a separate module for it. >> Hm, maybe we can avoid using the "real" time (via >> System.currentTimeMillis) and use a "time provider" instead. By default >> it obtails the system time. But for tests the returned time can be >> modified according to the test needs. JodaTime for example has a >> DateTimeUtils class with static accessors for getting and setting the >> time. In other projects we used a TimeProvider interface with different >> implementations for production and tests. > > That would be an extremely useful thing. I think we can benefit from > such an addition for the replication tests, especially those which are > time dependents. We can even simuate some use cases that are time > dependent this way. >
In the apacheds project we have 81 occurrences of Thread.sleep and 347 occurrences of System.currentTimeMillis. I can try to find some time this weekend to find out if and how it can be used, starting with the PP tests. Kind Regards, Stefan
