The main styles were restricted to panel body, but guides and others
were containers, not panels, so the styles did not get applied.

Now guides, videos, comments are all panels.

Some previous <pre>-related style tweaks resulted in <pre> line height
being no more applied to the CodeMirror editor.  Now explicitly stored
the line-height style.

CodeMirror just defines the rule:

    .CodeMirror pre { ... }

This selector has quite low specificity, so it's easy to accidentally
override it.  And this indeed happened when the selector for normal pre
elements was changed from:

    .class-overview pre { ... }

to:

    .class-overview .x-panel-body pre { ... }

The result was that a <pre> inside CodeMirror editor received unwanted
rounded corners etc, rendering the editor pretty-much unusable.

Resolved this by only styling pre.prettyprint and pre.notpretty.
(The latter class is now added to all not-to-be-prettified pre-s, which
distinguishes them from pre-s used inside CodeMirror component.)

Use Ext.clone when passing #data config to store to avoid the
destructive behavior.

In addition to stack trace the name of the file causing the problem is
printed.

It took hours to track this down... and I'm out of ideas of why we need
to require this class over there.

First I'm constraining the styles of the main class-overview container
inside the body-area - otherwise the styles will also have unpredictable
effects on toolbar (especially the styles for tables as in 4.1 the
search field inside toolbar is wrapped in two tables).

Secondly the search field styles are applied differently as the DOM that
makes up the field has changed considerably.  At the moment this results
in the field looking a bit different in 4.1.  Needs more work...

Also added width to the trigger-field which otherwise would change its
width when trigger is shown/hidden.

A really strange bug indeed. Hope it gets fixed in 4.1 final.

Previously the when menu buttons were created the menus were also
created right away.  Ext JS 4.1 didn't like that and placed the menus
somewhere at the top and refused to hide them initially.  Didn't really
managed to find out why, but doing it this way is probably better
anyway.

Created LoginHelper class for generating the HTML and binding the
submit listener in one single spot.

Best to `rm -rf node_modules`, then re-run `npm install`.

Instead of EXT_DIR --> EXT_BUILD.
Additionally added TOUCH_BUILD.

Also now using the constants themselves all over the code, not mapping
them into class variables.  This means the build will only fail if a
particular variable is needed.  So you don't need to define ANIMATOR_DIR
if you're not planning to build animator docs.

I'm not adding a new warning type to allow turning this off.
Instead everybody should just upgrade. No exceptions.

In 4.1 the View class also handles data attribute and things
will crash if we have our own special-purpose data in there.

At creation of Tabs component the overflow button creation was
triggered twice - first by afterrender, then by resize.  Removed
the first as having it twice caused a crash in Ext JS 4.1.

Previously it happened to be required by search controller.
But one shouldn't rely on such a coincidence - each view should
declare all the classes it depends on.

The Settings store happened to be instantiated by Classes controller.
This was just a coincidence that Docs.Settings class relied upon.
Now the store is created inside the Settings class itself when it's
initialized.

Removing the "#" from the beginning of history tokens was a mistake.
Also the case of no previous history token was unhandled.

Now comments can be marked as read from normal comments list of a
class/member.

The "read" status of comments is shown in all comment lists when
user is moderator.

Changed the styling of "read" badge into more prominent.

The code relied on the order of events being fired.  It expected that
when clicking tab close button the "click" handler for close button
will be invoked before the "click" handler for the whole tab.  This
worked in the majority of browsers but not in IE8 where the events
fire in the opposite order.

Now checking the event target inside tab click handler to see if
close button was clicked instead of relying on the close button handler
to set a flag.

This also fixes the issue of tabs sometimes not activating when clicked
in IE8.  Was caused by justClosed flag being set but not unset.

The return value of Array.indexOf was bogusly compared to false,
but #indexOf returns -1 not false on failure.