* Extend method documentation
  * Mention it in actionpack/CHANGELOG
  * Update release notes

In most cases, when setting variant specific code, you're not sharing any code
within format.

Inline syntax can vastly simplify defining variants in those situations:

  respond_to do |format|
    format.js { render "trash" }
    format.html do |variant|
      variant.phone { redirect_to progress_path }
      variant.none  { render "trash" }
    end
  end

Becomes:

  respond_to do |format|
    format.js         { render "trash" }
    format.html.phone { redirect_to progress_path }
    format.html.none  { render "trash" }
  end

@responses hash needs to be initialized with mime types that we get from
Collector#collect_mimes_from_class_level. Mime::Type class as key and nil as
value. This need to happen before content negotiation. Before that, it was
looping though mime types and executing mime-type-generated method inside
collector (see
AbstractController::Collector#generate_method_for_mime). That approach resulted
in 2 unnecessary method calls for each mime type
collected by Collector#collect_mimes_from_class_level.

Now hash is initialized in place, without usage of Collector#custom method.

This reverts commit 18616114, reversing
changes made to cad9eb17.

In most cases, when setting variant specific code, you're not sharing any code
within format.

Inline syntax can vastly simplify defining variants in those sitiations:

  respond_to do |format|
    format.js { render "trash" }
    format.html do |variant|
      variant.phone { redirect_to progress_path }
      variant.none  { render "trash" }
    end
  end
`
Becomes:

  respond_to do |format|
    format.js         { render "trash" }
    format.html.phone { redirect_to progress_path }
    format.html.none  { render "trash" }
  end

@responses hash needs to be initialized with mime types that we get from
Collector#collect_mimes_from_class_level. Mime::Type class as key and nil as
value. This need to happen before content negotiation. Before that, it was
looping though mime types and executing mime-type-generated method inside
collector (see
AbstractController::Collector#generate_method_for_mime). That approach resulted
in 2 unnecessary method calls for each mime type
collected by Collector#collect_mimes_from_class_level.

Now hash is initialized in place, without usage of Collector#custom method.

Avoid one-liner conditionals when they are too big. Avoid concatenating
strings to build error messages. Improve messages a bit.

By default, variants in the templates will be picked up if a variant is set
and there's a match. The format will be:

  app/views/projects/show.html.erb
  app/views/projects/show.html+tablet.erb
  app/views/projects/show.html+phone.erb

If request.variant = :tablet is set, we'll automatically be rendering the
html+tablet template.

In the controller, we can also tailer to the variants with this syntax:

  class ProjectsController < ActionController::Base
    def show
      respond_to do |format|
        format.html do |html|
          @stars = @project.stars

          html.tablet { @notifications = @project.notifications }
          html.phone  { @chat_heads    = @project.chat_heads }
        end

        format.js
        format.atom
      end
    end
  end

The variant itself is nil by default, but can be set in before filters, like
so:

  class ApplicationController < ActionController::Base
    before_action do
      if request.user_agent =~ /iPad/
        request.variant = :tablet
      end
    end
  end

This is modeled loosely on custom mime types, but it's specifically not
intended to be used together. If you're going to make a custom mime type,
you don't need a variant. Variants are for variations on a single mime
types.

This PR fixes #13064 regression bug introduced by the #8085

Now in _process_format when the format is a Mime::NullType nothing is written in self.content_type.
In this way the method Response#assign_default_content_type_and_charset can
write the the default mime_type.

Revert "Used Yield instead of block.call" -- this causes all of atom_feed_helper_test.rb to fail with "SystemStackError: stack level too deep".

This reverts commit d3a1ce1c.

Ruby 1.8 legacy. Since 1.9 hash preserves insertion order. No need for additional array to achieve this

constant.

`view_assigns` can use the precalculated sets and remove instance
variables without allocating any extra arrays

This reverts commit c2b5a8e6, reversing
changes made to 1918b12c.

See: https://github.com/rails/rails/pull/9660#issuecomment-27627493

Fixes a typo in the description for the call class method in Responder.

After this fix application config become available when calling helper outisde of view

config/application.rb

    #...
    config.asset_host = 'http://mycdn.com'
    #...

Somewhere else

    ActionController::Base.helpers.asset_path('fallback.png')
    # => http://mycdn.com/assets/fallback.png

Fixing the typo which is formed a not required link.  Check
here http://api.rubyonrails.org/classes/ActionController/Base.html
under paramters section
keeping it under tt tag

gets reverted here ec8ef1e1

This fixes an issue where the respond_with worked directly with the given
options hash, so that if a user relied on it after calling respond_with,
the hash wouldn't be the same.

Fixes #12029

The previous regex was allowing `_` in the URI scheme, which is not
allowed by RFC 3986. This change brings the regex in line with the RFC.

In some instances, `assert_redirected_to` assertion was returning an
incorrect and misleading failure message when the assertion failed.
This was due to a disconnect in how the assertion computes the redirect
string for the failure message and how `redirect_to` computes the
string that is actually used for redirection.

I made the `_compute_redirect_to_loaction` method used by `redirect_to`
public and call that from the method `assert_redirect_to` uses to
calculate the URL.

The reveals a new test failure due to the regex used by
`_compute_redirect_to_location` allow `_` in the URL scheme.

Previously it was raising a NilException