Not every project has an ::AplicationController

[mhenrixon]

In our project there is no root ApplicationController and hence when you are trying to use it for comparison my anonymous controller tests blow up like the pull request suggests.

[JonRowe]

As a workaround you can specify your own base class:

describe "my controller anonymously" do
  controller(MyController) do
  end
end

or ask RSpec to infer the controller from the described_class e.g.

RSpec.configuration.infer_base_class_for_anonymous_controllers = true

describe MyController do
  controller do
  end
end

[mhenrixon]

infere_base_class_for_anonymous_controller is true by default but even setting it so in my spec_helper does not solve the problem for me which is probably related to some other issue I noticed. I will try the other way you suggested though

[JonRowe]

Note you'll need to have correct constant name in your top level describe for it to work...

[mhenrixon]

Your suggestion don't work since there is no class ::ApplicationController and you are trying to use it.

7) ApplicationController current session when signed in returns the current session
     Failure/Error: Unable to find matching line from backtrace
     NameError:
       uninitialized constant RSpec::Rails::ControllerExampleGroup::ClassMethods::ApplicationController

The only solution for me is to create a class application_controller.rb or all my specs blow up.

[mhenrixon]

Check the test failures and it should be pretty obvious that whoever wrote that code never thought there could be a project where there is no ApplicationController in the root namespace.

[JonRowe]

Given that it's a Rails convention to have one it's a fairly safe assumption

[mhenrixon]

But won't this be a problem for engines? All namespaced projects etc as well? I find it hard to believe I will be the only one.

[JonRowe]

shrug Most projects also have a root namespace, and engines probably don't use anonymous controllers. Personally I've not see anyone use them in the "wild".

[JonRowe]

I'm not saying we won't fix it by the way.

[JonRowe]

PR's accepted, although I'd prefer an additional test (hint use hide_const) rather than what you've done to the existing tests.

[mhenrixon]

Deal! Didn't know that was what it was for

Sent from my iPhone

On 05 Feb 2014, at 01:20, Jon Rowe notifications@github.com wrote:

PR's accepted, although I'd prefer an additional test (hint use hide_const) rather than what you've done to the existing tests.

—
Reply to this email directly or view it on GitHub.

[billychan]

I think it's reasonable for rspec-rails to assume there is an ApplicationController which is generated by rails new by default.

Your tests failed because you did something customized on Rails. I think this is totally acceptable. So it would also be acceptable to do something customized on test settings instead of changing rspec-rails.

May I suggest you to add such in spec_helper if there are heavy usage of anonymous controller

config.before(:each, type: :controller) do
  stub_const('::ApplicationController')
end

If only one or two files need such you can add them directly into the files.

[JonRowe]

You need to at least do stub_const('::ApplicationController',ActionController::Base) if you take that route.

[alindeman]

Thanks! #924 addresses this, I think :)