Using pry-rescue to debug exceptions in Cucumber steps

Question

I've added an Around hook to my Cucumber features that I had hoped would cause pry-rescue to start pry when an exception was thrown:

Around do |scenario, block|
  Pry::rescue do
    block.call
  end
end

The Around hook is definitely being called, however exceptions thrown within steps aren't rescued. E.g. this step:

When(/^I perform the action$/) do
  raise 'hell'
end

... causes the feature to fail, but doesn't drop me into pry at the console.

Is it possible to use pry-rescue with Cucumber? I've raised this as an issue as well, as I suspect it might be a bug.

Updated: as per a suggestion from AdamT in the comments, I've:

• added the @allow-rescue tag to the feature calling the deliberately broken step
• added puts logging to verify that the Around hook is being called

It's still failing to enter pry when the exception is raised, but I can see from the puts statements that it's entering the Around hook.

Answer 1

I wanted to do the same thing - debug when a step fails. Your hook cannot work because a failing step exception is caught already. There seems to be no standard way of doing what you want with cucumber. But if you look at lib/cucumber/ast/step_invocation.rb the invoke(runtime, configuration) method, you will see what I am talking about.

In the method step level exceptions are caught. And the last rescue block is where we want to insert our debugging code. So in latest cucumber 1.3.12, on line 74 I inserted:

        require 'byebug'
        byebug

And now once the transient failure happens I get a prompt:

[71, 80] in /home/remote/akostadi/.rvm/gems/ruby-2.1.1/gems/cucumber-1.3.10/lib/cucumber
/ast/step_invocation.rb
   71:             failed(configuration, e, false)
   72:             status!(:failed)
   73:           rescue Exception => e
   74:             require 'byebug'
   75:             byebug
=> 76:             failed(configuration, e, false)
   77:             status!(:failed)
   78:           end
   79:         end
   80:       end

You can insert other debugging code in there though.

I'm thinking if cucumber project will accept a contribution to have a hook there instead.

UPDATE: here's my latest version. The positives in that version are that you get failure log before falling into a debugger. Also you can reach (at least with pry) to the cucumber World and launch pry inside to play around as if this is your test code. I've opened a discussion in the cuke google group to see if something similar can be implemented upstream. Give your voice and suggestions if you want to have it standard in cucumber. So just put the below code in support/env.rb:

  Cucumber::Ast::StepInvocation.class_eval do
    ## first make sure we don't lose original accept method
    unless self.instance_methods.include?(:orig_accept)
      alias_method :orig_accept, :accept
    end

    ## wrap original accept method to catch errors in executed step
    def accept(visitor)
      orig_accept(visitor)
      if @exception
        unless @exception.class.name.start_with?("Cucumber::")
          # @exception = nil # to continue with following steps
          # cd visitor.runtime/@support_code
          # cd @programming_languages[0].current_world
          # binding.pry
          require 'pry'
          binding.pry
        end
      end
    end
  end

Answer 2

In the version of Cucumber 2.4.0 the #accept method is resided in Cucumber::Formatter::LegacyApi::Ast::StepInvocationso redefine it and proceed the required action inside it:

Cucumber::Formatter::LegacyApi::Ast::StepInvocation.class_eval do
  alias_method :orig_accept, :accept

  def accept formatter
    orig_accept(formatter)
    if status == :failed
      formatter.runtime.support_code.ruby.current_world.instance_eval do
        # do something as if you are inside the cuke test step
        # like: expect(something).to be_something_else
      end
    end
  end
end