Hi, thanks for posting.

Right now, this is a feature that is specific to the NCrunch VS client and it isn't available for the console tool. Could you tell me more about your use case? How will having this feature in the console tool help you?

Is the purpose of your script to force the runner to execute different groups in tests in different orders so that you can tease out the issues?

When run locally on a machine under VS, NCrunch churn mode does cycle the tests into different execution orders and groups. It will also cycle the test processes randomly to try and trigger sequence-dependent issues. Is there a reason you need this to be on your build server and you aren't able to use a client machine?

Thanks for confirming this. This would make sense as a feature request if you'd like to submit it at uservoice.

Unfortunately, I can't right now provide you with a way to get churn mode to work using the console tool - but I can provide some general advise around its use that I hope will help with your situation.

The problem you're facing is very common and needs to be kept in check as test suites expand in size. Intermittent test failures slow down development and destroy developer sanity. There are two approaches I would suggest in bringing this problem under control:

1. Nominate a team member to take responsibility for turning on churn mode when they finish up work for the day. Let it run overnight. In the morning, check the results and examine the list of tests that failed. Construct a theory for each of these failures. If there is not enough information to act on them, add trace diagnostics to get more information so that the next time they fail, you'll have more to assess the situation.

2. For the tests that you've identified as intermittently failing, get NCrunch to target them for churn mode in isolation (just select the test and run churn mode). This will give you a way to stimulate the failure at will so that you can analyse it. Unfortunately it isn't possible to do this with a debugger attached, but as long as you can make the test fail, you can continue to add diagnostics until you've identified the reason for the failure. This won't work for many test-sequence dependent issues, but it works very well for race conditions and resource related problems.

We have some features coming with the V5 release that I expect will help immensely in dealing with this particular problem. I can't share details on this right now but please take this in confidence that we are doing what we can to help with this problem.