Makes it easier to find the close-all-tabs feature.

Handling the internals of tabs inside the view and leaving the
handling of higher-level events to the controller.

It makes sense for the tabs component to adjust itself when it's resized,
rather than having a controller tell it.

Monitor the clicks on menu items inside the view.

This way the whole interaction with tab-menu is entirely encapsulated inside view.

Instead access the menu from this.overflowButton.menu.

Replace it with classname.

Use class name instead.

In supporting Google Closure Compiler syntax.

For the record types to even reach type parser the DocParser needed
to be modified to parse the {...} type blocks by taking account the
need to balance the braces.

It completely puzzles me why Google Closure Compiler requires
varargs type name in function argument context to be wrapped inside
square braces, while in normal context no square brackets are needed.

But here we go...

And better structured tests.

Also now only allow optionality suffix "=" be used inside function type
arguments.

Needed for function type parameter types.

In supporting Google Closure compiler syntax.

- string
- number
- boolean
- null
- undefined

When "string" is used as type and String class exists, a link to
String class is created.  Otherwise just text "string" is shown.

In supporting Google Closure Compiler syntax.

To support Google Closure Compiler syntax for optional parameters.

    @param {Type=} name

This is detected at DocParser not in TypeParser like the rest of the
type definition syntax - all other way of defining optional parameters
are also implemented in DocParser, so it keeps things simpler to have it
all in there at one spot.

In supporting Google Closure Compiler syntax.

In supporting Google Closure Compiler syntax.

The syntax used by Google Closure Compiler.

Supported by Google Closure Compiler.

It was deprecated in 3.2.1 in December 2011.

Should be OK to remove now.

We can now have input files in any encoding as long as --encoding
option specifies it correctly.  When encoding differs from UTF-8,
we convert the text to UTF-8 internally and always output in UTF-8
also.

This option only works in Ruby 1.9.  Under 1.8 it has no effect.

In Ruby 1.9 reading files through builtin IO class attempted to
auto-detect the encoding which sometimes failed.

Created JsDuck::IO to be used instead of it, which forces all files
to be treated as UTF-8.  It also behaves well in Ruby 1.8 - just
doing nothing with the input and causing no errors.

Recent changes in server now allow any links to be posted.

For now it's just dead code. Better remove before it becomes a problem.

Instead of supplying it in every request config object.

It deals with extracting data from form, and the form is described in
view.  So it really is the specifics that should belong to view.

So the view now does the extracting of data and then calls Auth.login().

Using the same function now in both Auth and Comments controllers.

Rename several methods:

- loggedIn -> setLoggedIn
- loggedOut -> setLoggedOut
  For setting the state of login form.

- getSession -> retreveSession
  "get" is a bad start for method that doesn't return a value.

- showLogin -> showLoginForm
  Previously it was hard to distinguish it from showLoggedIn.

- loggedIn -> isLoggedIn
  Now the method for checking logged in status is named the same
  way in both Auth and AuthHelpers.  This also differenciates it
  from the event "loggedIn".

Currently used nowhere, not supported by backend either.

All Ajax and JsonP request for comments data are now performed
through a custom #request method that appends server and session ID data
to the given URL.

I didn't notice before that there was a #showError method already, so I
created #notify method to notify of subscription changes.  Merging these
two methods now into #notify to cover both uses.

The method now takes a third argument to allow for anchoring on
different sides.

One downside is, that I threw out the smooth fade-in and fade-out of
subscription notification.  It was a bit of a too much trouble for such
a small feature, so I guess the overall simplicity in code is a win.