event.target shouldn't be retargeted as the event bubbles into a slot
https://bugs.webkit.org/show_bug.cgi?id=157369

Reviewed by Antti Koivisto.

Source/WebCore:

When an event bubbles up from an assigned node to its assigned slot, we shouldn't be adjusting
event.target to point to the slot. Since a shadow tree should have access to nodes outside
the shadow tree, event.target is accessible inside the shadow tree of such a slot.

New behavior matches the behavior of Google Chrome Canary as well as the shadow DOM specification:
http://w3c.github.io/webcomponents/spec/shadow/#dfn-retargeting-algorithm

Test: fast/shadow-dom/event-inside-slotted-node.html

* dom/Event.cpp:
(WebCore::Event::deepPath):
* dom/EventContext.h:
(WebCore::EventContext::isUnreachableNode): Use Node::isUnclosedNode instead of isReachable.
(WebCore::EventContext::isReachable): Deleted.
* dom/EventPath.cpp:
(WebCore::EventPath::EventPath): Don't set the target to the slot when entering a slot. Also moved
the code to adjust the target as we exit a shadow tree to the end of the outer loop for clarity.
(WebCore::isUnclosedNodeOf): Deleted. Renamed to Node::isUnclosedNode.
(WebCore::EventPath::setRelatedTarget):
(WebCore::EventPath::computePathUnclosedToTarget): Renamed from computePathDisclosedToTarget.
(WebCore::moveOutOfAllShadowRoots): Extracted from RelatedNodeRetargeter::RelatedNodeRetargeter.
(WebCore::RelatedNodeRetargeter::RelatedNodeRetargeter): Fixed a bug that we were exiting early
without setting m_hasDifferentTreeRoot true when target and relatedNode are disconnected from
a document.
(WebCore::RelatedNodeRetargeter::currentNode):
(WebCore::RelatedNodeRetargeter::checkConsistency): Updated to match the spec with one exception.
We don't use null as the adjusted related target when the (original) related target and the target
are in two distinct disconnected trees since such a behavior is not Web compatible. This spec bug is
tracked by https://github.com/w3c/webcomponents/issues/494
* dom/EventPath.h:
(WebCore::EventPath::eventTargetRespectingTargetRules): Returns Node* instead of EventTarget* since
we need a Node in RelatedNodeRetargeter::checkConsistency.
* dom/Node.cpp:
(WebCore::Node::isUnclosedNode): Moved from RelatedNodeRetargeter.cpp
* dom/Node.h:

LayoutTests:

Updated test cases to expect the target to be not adjusted to a slot element when the event path
enters one as this didn't match the spec or the behavior of Google Chrome Canary. Both WebKit and
Chrome passes the test with this change.

* fast/shadow-dom/event-inside-slotted-node.html:


git-svn-id: http://svn.webkit.org/repository/webkit/trunk@200464 268f45cc-cd09-0410-ab3c-d52691b4dbfc
9 files changed