rspec

Comments and changes to this ticket

• 

  David Chelimsky November 26th, 2007 @ 03:32 AM

  • → State changed from “new” to “open”

  Kyle - this sounds like an issue related to the fact that autotest has already loaded the environment each time it runs. Have you tried the same thing using test/unit instead of rspec? I'd be surprised if the results are any different. Not shocked, mind you :) - just surprised.

• 

  Kyle Hargraves November 26th, 2007 @ 05:09 AM

  It occurs with 'rake spec', too.

• 

  David Chelimsky November 26th, 2007 @ 05:12 AM

  Ah - missed that. Thanks.

• 

  Kyle Hargraves December 13th, 2007 @ 02:07 AM

  Reminded about this by the ML thread.

  The problem may not actually be with rspec; could Scott/anyone else confirm that this started happening without updating Rails at the same time? We're all chasing Rails edge most of the time, so I doubt anyone does that much.

  When I tried to figure this one out, I ended up printing the described type that the example groups eventually receive. It was ::FooController rather than Admin::FooController, but AFAIK ruby has no means of 'resolving' a class to some module within which it resides. It's inherently fully defined, right?

  Which I think implies that the issue is rails' const_missing or whathaveyou doesn't recognize that it's being called within a module, and searches for ::FooController. A guess, I haven't looked into Rails to find out.

• 

  Scott Taylor December 13th, 2007 @ 02:48 AM

  Just to get back to Kyle: I haven't been chasing rails trunk; I've been on rails 1.2.3 frozen, the whole time.

• 

  Kyle Hargraves December 13th, 2007 @ 02:56 AM

  So much for that idea.

  I'll try and tinker with git-bisect and figure out when this started happening, then. I've been looking for an excuse to use it more anyhow.

• 

  Kyle Hargraves December 13th, 2007 @ 03:36 AM

  This exists as far back as r1800. Reverting further beyond that is probably not worth it.

  I guess a fix would be to always go as deep into spec/controllers as possible and then load back up the directory stack, but that seems pretty lame.

  Using Admin::FooController is perfectly acceptable, no? For that matter, if people insist on putting things in modules, they can just put the fully qualified name on the line immediately before opening the module and everything's rainbows.

• 

  harm January 2nd, 2008 @ 02:00 PM

  I'm having the same problem.

  I have two controllers:

  one in the default namespace: DevicesController

  the second one is in the reseller namespace.

  i'm referencing them in my example files as follows:

  describe DevicesController do

  end

  and:

  describe Reseller::DevicesController do

  end

  However when I save the examples in the default namespace, all the examples in the reseller namespace fail. When I 'touch' the file in the reseller namespace everything is back to normal and all examples pass.

  Is this a bug? Is there a workaround available? I have lookup into the rspec source code, but could not find the source of the problem.

• 

  Thijs January 9th, 2008 @ 03:50 PM

  I'm having the same problems while using the describe Reseller::DevicesController do form to.

  It does look like this might be an issue in Rails itself. Did anybody find any clues?

• 

  Scott Taylor January 9th, 2008 @ 04:30 PM

  I'm no longer experiencing the bug with rails 2.0.2 and rspec 1.1.1, and ruby 1.8.6 (patchlevel 0). What are you guys running (I'm surprised Aslak hasn't chimed in with his boiler-plate message yet)?

• 

  Thijs January 9th, 2008 @ 04:38 PM

  I'm running ruby 1.8.6 (2007-03-13 patchlevel 0), Rails 2.0.2 gem and the newest CURRENT rspec in vendor/plugins. I still get the same errors.

• 

  Thijs January 27th, 2008 @ 04:27 PM

  Does anybody have any ideas about this issue? Or is it solved for most people?

• 

  Jason L Perry March 19th, 2008 @ 04:50 PM

  RSpec 1.1.3, Rails 2.0.2, Ruby 1.8.6 (2007-09-24 patchlevel 111) [universal-darwin9.0]

  And I still have this issue.

  I'm not sure what kyle means by:

  Using Admin::FooController is perfectly acceptable, no? For that matter, if people insist on putting things in modules, they can just put the fully qualified name on the line immediately before opening the module and everything's rainbows.

  :( Everything is not rainbows.

  /cry

• 

  Jason L Perry March 19th, 2008 @ 05:01 PM

  ok so grok this:

  I did nothing but change:

  require File.dirname(__FILE__) + '/../../spec_helper'
  
  describe Admin::SchoolsController do
  

  to:

  require File.dirname(__FILE__) + '/../../spec_helper'
  Admin::SchoolsController
  describe Admin::SchoolsController do
  

  And 'rake spec' passes all examples again. I removed the line, and they still continued to pass. shrug I'm go to stop thinking about this now.

• 

  Robby Russell May 16th, 2008 @ 07:25 PM

  I'm seeing the same behaviour here as well. Added

  Admin::UsersController

  to my controller spec and `rake spec` works again.

• 

  Thijs May 16th, 2008 @ 07:52 PM

  I've noticed that this problem mainly pops up when you have a model with the same name that a namespace has.

  Do you have a model called Admin?

• 

  Robby Russell May 16th, 2008 @ 07:56 PM

  I've noticed that this problem mainly pops up when you have a model with the same name that a namespace has.

  Do you have a model called Admin?

  Indeed! We do have an Admin model.

• 

  Kyle Hargraves May 16th, 2008 @ 08:12 PM

  I have never had an Admin model on a project, and it's always the Admin namespace that bites me. =\

• 

  Scott Taylor May 16th, 2008 @ 08:25 PM

  In reply to Jason's comment above,

  I thought that the following would work:

  describe Admin::FooBarController { ... }

  but the following wouldn't:

  module Admin

  describe FooBarController { ... }

  end

  Am I mistaken? If I'm correct, than I would expect this is about load order regarding Rails using Module.autoload.

• 

  Phil Orwig July 1st, 2008 @ 05:41 PM

  • → Tag changed from “” to “bug namespaces rails”

  Is there anything I can do to help this along? I'm running into the same issue with namespaced controllers, similar to the usual Admin issue listed here.

  This is with rspec 1.1.4 and Rails 2.1.

• 

  Thijs July 1st, 2008 @ 05:43 PM

  Do you by any chance have a model called admin?

• 

  Phil Orwig July 1st, 2008 @ 06:17 PM

  No -- in fact, everything here is namespaced under Analyst, and I don't have an Analyst model.

• 

  David Chelimsky July 1st, 2008 @ 06:19 PM

  • → Assigned user changed from “” to “David Chelimsky”
• 

  Tony Perrie August 6th, 2008 @ 07:35 PM

  I'm seeing the same issue here. We have a model called Admin, however, removing that model seems not to resolve the issue. As a workaround, I just added a require for the particular controller in the RSpec test itself. I only see this issue occurring on Ubuntu Linux and not OS X. Is this perhaps something to do with the order that the filesystem returns the list of controllers (HPFS vs ext3)?

• 

  Robby Russell August 11th, 2008 @ 07:55 PM

  Tony,

  One way around this is to name your model Administrator. There was some conflicts in Rails as well with the namespacing. So, moving to calling the model Administrator (instead of the abbreviated Admin) did wonders.

  Robby

• 

  Tony Perrie August 12th, 2008 @ 05:03 AM

  Robby,

  I've actually tried removing the model already, and that seemed not to have an effect. I can try renaming it, but since removing the Admin model seemed not to matter, I suspect a rename would be irrelevant as well.

  Also, this issue only occurs on Ubuntu Linux. I tested the actual Rails application itself on Ubuntu, and it seems to not have the same problem as RSpec. Does rake boot the controllers in the same order that Rails and script/runner?

• 

  Scott Taylor August 12th, 2008 @ 05:12 AM

  AFAIK, rails uses const_missing to load (or require) anything which isn't already loaded (or required).

  Tony - what version of ruby, rubygems, and rspec are you using on ubuntu? Was the version of ruby installed via source or through apt-get?

• 

  Tony Perrie August 12th, 2008 @ 05:17 AM

  Scott,

  Ubuntu Feisty Fawn is running a fully updated Ruby 1.8.5. Our Rails version is 2.1. We saw the same issue on an another Linux box running Ruby 1.8.6 though. So, it seems not to be a 1.8.5 or 1.8.6 issue. The strange thing is, we don't see the problem at all on Ruby 1.8.6 under OS X 10.5.