A few of our tests attempt to test controllers with production-style rescuing. Before the current rspec/Rails 2.2.2 combo, it was fairly easy to set up for these tests: it "should throw a 404 for strange actions" do controller.use_rails_error_handling! get :strange_action_name response.should be_missing response.body.should include("You've hit our custom 404 page.") end As of trunk rspec/Rails 2.2.2, this doesn't work. It turns out that ActionController::TestCase has invented its own hook for testing. Without this hook set, you end up in 'local development' mode. So to get into production mode, you need to do this: it "should throw a 404 for strange actions" do controller.use_rails_error_handling! # rspec no longer throws the error without hitting Rails logic rescue_action_in_public! # ActionController::TestCase no longer considers this request a local request get :strange_action_name response.should be_missing response.body.should include("You've hit our custom 404 page.") end Of course, if you have 'rescue_from' handlers, these will happen no matter what. I guess my question is, now that the 'rescue_action_in_public!' hook exists in Rails, does it still make sense for rspec-rails to have its own method? Might it make sense to merge the two functionalities in some way? =Nathan
on 2009-02-14 00:55
on 2009-02-18 00:12
On Fri, Feb 13, 2009 at 5:20 PM, Nathan Wilmes <email@example.com> wrote: > > this request a local request > in Rails, does it still make sense for rspec-rails to have its own method? > Might it make sense > to merge the two functionalities in some way? Nathan, Thanks for bringing this all up. I've made some changes and additions based on this and a few other conversations. Here's how it works now (latest in git and latest github gem, and the next formal release): 1. controller.use_rails_error_handling! now works as it did in 1.1.12 2. controller.use_rails_error_handling! is now deprecated, because rescue_action_in_public! does exactly the same thing without rspec-rails having to monkey patch rails 3. added bypass_rescue, which you can use to bypass any rescued errors (defined in rescue_from) This allows you to do any of the following: 1. specify un_rescued errors with "should raise_error" (default) 2. specify un_rescued errors with status codes (using rescue_action_in_public!) 3. specify rescued errors with bypass_rescue with "should raise_error" 4. specify rescued errors with bypass_rescue with status codes (using rescue_action_in_public!) Both rescue_action_in_public! and bypass_rescue can be declared per-example, or in before(:each). Questions?