https://bugs.webkit.org/show_bug.cgi?id=81866

Reviewed by Hajime Morita.

We need to expand the controlClipRect because the existing shadow
tree of a text field can be wrapped by another shadow tree. e.g.

A current situation:

<input>
  ┗ ShadowRoot
     ┗ container element
┌────────────┐<input>'s border box
│┌──────────┐│
││container box       ││
│└──────────┘│
└────────────┘

Wrapped by another ShadowRoot:

<input>
  ┗ New ShadowRoot
     ┣ <shadow> represents the existing shadow tree
     ┃  ┗ container element
     ┗ D: an additional decoration element
┌────────────┐<input>'s border box
│┌───────┬──┐│
││container box │ D  ││
│└───────┴──┘│
└────────────┘
In this case, if we clip child renderers by container box, D is not
drawn. We should clip by the content box of the <input>.

A search field has an exceptional behavior. It can have the container
box of which height is taller than the content box height. The
controlClipRect should contain both of the container box and the content
box in this case. In other cases, the container box is equivalent to the
content box. So the code always unites them.

No new tests. This won't make any behavior change for now.

* rendering/RenderTextControlSingleLine.cpp:
(WebCore::RenderTextControlSingleLine::controlClipRect):


git-svn-id: http://svn.webkit.org/repository/webkit/trunk@111660 268f45cc-cd09-0410-ab3c-d52691b4dbfc