https://bugs.webkit.org/show_bug.cgi?id=119171
<rdar://problem/14511404>

Reviewed by Simon Fraser.

Source/WebCore:

Expose a new property on AnimationController that allows newly created
animations to run even if the controller says it is suspended. See WebKit
ChangeLog for more details.

Test: transitions/created-while-suspended.html

* WebCore.exp.in: Export the new methods so WebView can use them.
* page/animation/AnimationController.cpp:
(WebCore::AnimationControllerPrivate::AnimationControllerPrivate): Initialize new flag to false.
(WebCore::AnimationControllerPrivate::startAnimationsIfNotSuspended): Check new flag is not true.
(WebCore::AnimationControllerPrivate::setAllowsNewAnimationsWhileSuspended): Expose setter.
(WebCore::AnimationController::allowsNewAnimationsWhileSuspended): "Public" getter.
(WebCore::AnimationController::setAllowsNewAnimationsWhileSuspended): "Public" setter.
* page/animation/AnimationController.h:
* page/animation/AnimationControllerPrivate.h:
(WebCore::AnimationControllerPrivate::allowsNewAnimationsWhileSuspended):
* page/animation/CompositeAnimation.cpp:
(WebCore::CompositeAnimation::CompositeAnimation): Only suspend if new flag is false. Everything else
relies on the m_suspended flag, so the real code change is this one line.

Source/WebKit/mac:

Expose a new SPI on WebView that triggers the (buggy) old behaviour
for animations, such that any newly created animations will start even
when the document is supposedly suspended. It turns out that client content was
unknowingly relying on this behaviour - e.g. suspending a view, loading a
bunch of new content, bringing the view on screen and then unsuspending. In this
situation, we were not running CSS transitions, because the page was suspended.
However, JS was still triggering them, and content was expecting a transitionEnd event.

* WebView/WebView.mm:
(-[WebView allowsNewCSSAnimationsWhileSuspended]): Calls into AnimationController.
(-[WebView setAllowsNewCSSAnimationsWhileSuspended:]): Ditto.
* WebView/WebViewPrivate.h: New methods listed above.

LayoutTests:

This is actually a test to make sure this fix didn't break anything. There is no
way to trigger the new behaviour from the test system (or from Safari).

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@153396 268f45cc-cd09-0410-ab3c-d52691b4dbfc