Testing code with at_exit conflicts with raise_error

[edisonywh]

Hey guys! I currently have a class like this:

module Gem
  def custom_method
    at_exit do
      # code_logic
      raise StandardError
    end
  end
end

The reason I am wrapping my code inside at_exit is because I want to run the execution of the code AFTER my base class has finished defining the methods. For example:

class Foo
  extend Gem
  custom_method
  def method_one; end
end

I want to access method_one from inside my gem, I have two ways to do it:

1. Put my custom_method all the way at the bottom of the file
2. Wrap my custom_method around an END block or an at_exit block, both of which will ensure that my custom_method runs after the file that extended the file.

I've decided to go with the second option because I do not want to put the method call at the end of the file.

It all works fine when I test it manually, however, the issue is as follow when I try to test it with rspec.

I want to test my method like so:

expect do
  Foo = Class.new do
    extend Gem
    custom_method
    def method_one; end
  end
end.to raise_error StandardError

But this won't work, rspec will think that there's no error, I presume this is because my at_exit runs after the rspec raise_error, which means the error is raised after the expectation.

How can I make my test green? Appreciate any pointers!

[edisonywh]

For a more concrete example, I just published the gem in question, so you can dig into the source code too.

https://github.com/edisonywh/behaves

[JonRowe]

👋 This is a tough one, your test is working correctly, because at_exit hooks are installed at... well.. exit, by which time your test suite has already failed. The way to test this is, probably, to assert that at_exit is received with a stub and then check the block raises StandardError when invoked.

As far as I know there is no way to intercept the blocks and check them, and in any case would cause RSpec to exit uncleanly.

[edisonywh]

Cool, thanks for confirming what I suspected!

With that, I think I'm going to break the rules of no testing private method, as I don't think it should be a hard rule but rather a guideline, and one that's reasonable to break in my case.

My fix is over at edisonywh/behaves#16 along with my rationale, if someone is interested.

Thanks for taking the time to answer my question, Jon!

[JonRowe]

You could also test it by shelling out and running your code that way, which is how we test some parts of RSpec.

[edisonywh]

I don't understand what you mean by shelling out, could you explain what's that? Thanks!

[JonRowe]

Running your code as a seperate process on the command line and testing the resulting output.

[JonRowe]

You could also experiment with fork a quick check says at_exit is run, but not rescuable but again you might be able to assert on the output