Forum: Rails-core (closed, excessive spam) require 'application' loads it twice

I've discovered that application.rb is being loaded twice in a couple
situations.  It is not when simply running script/server, but it is
when running script/console and specs with rspec.

The first place it is always loaded is in the #to_prepare callback in
dispatcher.rb, using #require_dependency.  The second time it's being
loaded is simply using "require 'application'" in console_with_helpers:
19, and in rspec_on_rails/lib/spec/rails.rb:3

I don't think it's expected behavior that requiring 'application'
loads it a second time.

Looking at the code, it appears that #require_dependency (which calls
#require_or_load) has two different mechanisms: :load or :require.
I'm curious if anyone knows the reasoning for this, and why :load is
the default. But even if it did use #require, #require_or_load expands
the path of the filename, so simply requiring application would still
re-load it.

I'm also wondering if #require_dependency is necessary in the
dispatcher.  Would simply using #require do the trick?

I'm willing to pursue a patch to fix this, I just wanted to see if
anyone had any insights.

Thanks,
Brandon

> I don't think it's expected behavior that requiring 'application' loads it a
> second time.

No, double requiring isn't expected.

> Looking at the code, it appears that #require_dependency (which calls
> #require_or_load) has two different mechanisms: :load or :require.  I'm
> curious if anyone knows the reasoning for this, and why :load is the
> default. But even if it did use #require, #require_or_load expands the path
> of the filename, so simply requiring application would still re-load it.
>
> I'm also wondering if #require_dependency is necessary in the dispatcher.
>  Would simply using #require do the trick?

require_dependency has two different load mechanisms to support the
reloading we do in development mode.  But some of that code (including
require_dependency) is so old it predates active support, and I'm sure
there's some cruft in there.  Perhaps it can be simplified.

> I'm willing to pursue a patch to fix this, I just wanted to see if anyone
> had any insights.

The correct fix is to use require_dependency 'application' as it works
with reloading.  The double require is probably happening because
require_dependency normalises the path.

What are you doing in the console which triggers this?   It's probably
a bug in rspec too.

> Thanks,
> Brandon
>



--
Cheers

Koz

I'm experiencing the same exact problem (application.rb is loaded
twice when using script/console and running rspec). Using the latest
stable Rails (2.1.0) and RSpec (1.1.4). I'm looking into it but have
nothing helpful to contribute at the moment. This is on a project that
we're bringing up to speed after it's been dormant for a few months.

I also have been puzzled by this. It can be reproduced in one minute:
Create a fresh rails app named 'fred':  rails fred
Edit apps/controller/application.rb  and insert a constant: X = 2
Type: script/console

> I'm experiencing the same exact problem (application.rb is loaded
> twice when using script/console and running rspec). Using the latest
> stable Rails (2.1.0) and RSpec (1.1.4). I'm looking into it but have
> nothing helpful to contribute at the moment. This is on a project that
> we're bringing up to speed after it's been dormant for a few months.


http://github.com/rails/rails/commit/292501c7e0199...

That was easy :).

Does this fix it for everyone here?

--
Cheers

Koz

It fixes it in script/console  but not when I run rake spec.

That sounds like an rspec problem.  Is there a require 'application'
somewhere in their source?



On Fri, Jul 11, 2008 at 10:36 PM, randy <randy.j.parker@gmail.com> 
wrote:
>
> It fixes it in script/console  but not when I run rake spec.
> >
>



--
Cheers

Koz

Yes, the same line does appear in rspec-rails/lib/spec/rails.rb
And when I make Koz's fix, it does fix it for "rake spec"

Thanks!!

On Fri, Jul 11, 2008 at 10:48 PM, randy <randy.j.parker@gmail.com> 
wrote:
>
> Yes, the same line does appear in rspec-rails/lib/spec/rails.rb
> And when I make Koz's fix, it does fix it for "rake spec"

awesome.  Remember to submit it upstream :)


--
Cheers

Koz

On Jul 11, 2008, at 4:50 PM, Michael Koziarski wrote:
> On Fri, Jul 11, 2008 at 10:48 PM, randy <randy.j.parker@gmail.com>  
> wrote:
>>
>> Yes, the same line does appear in rspec-rails/lib/spec/rails.rb
>> And when I make Koz's fix, it does fix it for "rake spec"
>
> awesome.  Remember to submit it upstream :)

http://rspec.lighthouseapp.com/projects/5645/ticke...

Thanks Koz.

Brandon