Running tests on a burrito-wrapped CLI app

[marcelotto]

First of all, thank you for this wonderful project.

I'm writing a Burrito-wrapped CLI and am wondering how it is supposed to run tests. When defining an application entry point with the start/2 function doing the Burrito.Util.Args.get_arguments() processing, this start/2 function is also used when calling mix test with the arguments of the mix call, which doesn't work, of course.

My current solution is to set a compile flag :in_burrito in the prod environment config to only add the Burrito start logic when it is actually running in a Burrito release. This has the nice side effect that it allows the use as an escript during development and not having to do the expensive building of a Burrito release all the time.

defmodule MyEntryModule do  
  if Application.compile_env(:my_app, :in_burrito) do  
    def start(_type, _args) do  
      Burrito.Util.Args.get_arguments()  
      |> main()  
      |> System.halt()  
    end  
  else    
    def start(_, _) do  
      {:ok, self()}  
    end  
  end
end

Am I missing something, or is there a better solution? In any case, some guidance in the README on running tests would be helpful.