Don't exit on successful RSpec and Cucumber runs

[sentience]

In our CI setup, we run a single Rake task that invokes multiple test suites, including RSpec and Cucumber. We wish to use Knapsack to parallelise the RSpec and Cucumber suites; however, both RSpecRunner and CucumberRunner end with a call to exit, which exits Rake whether or not the suite ran successfully. This prevented any subsequent Rake tasks from running.

Knapsack’s runners should coexist with other Rake tasks, including any that may depend on the Knapsack tasks running first.

This change only exits from Knapsack when the child process exited with a non-zero status.

[ArturT]

@ciembor Will this change break your rerun failing specs solution? http://blog.lunarlogic.io/2015/randomly-failing-cucumber-scenarios-in-continuous-integration-system/

Related old PR #17

@sentience Thanks for PR. Is that mean when the first test suite failed then chain is stopped. The second test suite won't be executed?
Maybe we should introduce configuration option to enable/disable exit instead of doing it only when non zero exist status.
Your current solution is the default behaviour how regular rake task should work?

[sentience]

@ArturT Currently if you run a set of Rake tasks and the first task fails, the task will normally exit with a non-zero status code. Rake doesn’t catch that; the exit actually exits the Rake process, and the non-zero code indicates to the shell that it failed.

Calling exit inside a Rake task with a status code of 0 is very weird, because it forces the Rake process to exit immediately, but with a status code that tells the shell it was successful.

My fix makes knapsack’s runners behave the way Rake tasks are expected to behave. If there’s a use case where it’s desirable for a task to force Rake to exit with a successful status code, I’d be surprised, but of course very interested! :)

[ArturT]

@sentience Thanks for clarification and PR! I'm going to release new version of the gem today.

[bf4]

or $?.success? or $?.exitstatus.zero? :tmyk:

[sentience]

$?.success? would indeed be cleaner. Happy with that if you want to change it.

[ArturT]

I will do change today so no need for such tiny PR :)