Got hit by this bug today. Credits should go to Josep M. Bach firstname.lastname@example.org, see #881.
Fix bug when Rails.application is defined but is nil. See #881
I remember having this issue while working on rails-api with @spastorino, the solution iirc was to actually define an application (ie inherit from Rails::Application somewhere in our test_helper).
@marcandre do you have a small example that shows the issue? I don't remember exactly how it happened at that time. Thanks!
it will probably happen in any gem that defines a railtie and uses integration testing.
It's possible I wasn't doing the right thing, but I needed an ActiveRecord based model in one test, so I created one (https://github.com/marcandre/table_helper/blob/master/test/unit/model_test.rb ) and that had impact on another test (https://github.com/marcandre/table_helper/blob/master/test/helpers/table_helper_test.rb ) because now the TestCase could not be instantiated anymore.
Maybe there is something else that could be fixed, but I feel the patch should be accepted. It's very possible that defining an application works, but it shoudn't be mandatory if it can be avoided.
Agreed, I just wanted to remember what was the use case, thanks for clarifying, I think mine at that time was a bit different, but I've got the same error. Anyway, the patch looks fine, thank you.
Don't run tests in Rails 3.0/3.1 at the moment
Reason: rails/rails#6429 seems to not have been backported for 3.0/3.1.