The Cucumber Book

What We Just Learned

When you add asynchronous behavior to a system, you need to make a concerted effort to tame the random effects that it can have on your tests. Build your tests with a knowledge of how the system works and introduce synchronization points where timing issues are likely to arise.

Using sleeps in your steps is not a good way to tackle these timing issues, because it makes your tests slow and doesn’t solve the reliability problem: if the system changes and becomes slower, your sleep may not be long enough and the test will start to break again. The best solution is to listen for events broadcast by the system and pause at the appropriate points in the scenario until those events have been received. That way, you minimize the ...