On the one hand, it's good because it makes the tests 100% deterministic and 
test failures reproducible.

On the other hand, compared to `ThreadLocalRandom.current()` or `new 
Random(seed)` with a _random_ seed, a constant seed may not allow to explore 
all input conditions and discover some failing corner cases.

So ideally the seed should be generated randomly and be a *parameter* of a 
test, recorded in the test name. So when a test fails, it could be run locally 
with the known seed parameter in order to reproduce a test failure.

[ Full content available at: 
https://github.com/apache/incubator-druid/issues/6227 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to