The element under mouse should be in hover state even if its not inActiveChain.

The element under mouse should be in hover state even if its not inActiveChain.

BUG=122746

[pals, 2014-06-05 07:35:58]

PTAL.

[Rick Byers, 2014-06-12 18:24:06]

Sorry for the delay.  I had to find some time to get up-to-speed on this code
and behavior in other browsers.  There's a fair amount of complexity here - we
need to be careful that we're not making anything worse for web developers. 
Adding a couple other folks who may have context on the behavior here.

https://codereview.chromium.org/313183004/diff/1/Source/core/dom/Document.cpp
File Source/core/dom/Document.cpp (right):

https://codereview.chromium.org/313183004/diff/1/Source/core/dom/Document.cpp...
Source/core/dom/Document.cpp:5481: if (curr->node() && !curr->isText() &&
(!mustBeInActiveChain || curr->node()->inActiveChain() || curr == oldHoverObj))
This doesn't look like the right fix to me.  How does your patch change the
behavior of this (more general) test case: http://jsbin.com/rider/2/?  If we're
going to fix this bug then I'd expect Chrome's behavior to match Firefox -
namely that hover is cleared on all nodes that are no longer under the mouse and
applied to all nodes that are now under the mouse.  You seem to be special
casing the innermost node here.

Also the behavior we have today seems to be intentional given this comment
above: 
  // If the mouse is down and if this is a mouse move event, we want to restrict
changes in
  // :hover/:active to only apply to elements that are in the :active chain that
we froze
  // at the time the mouse went down.

Changing this seems reasonable to me (I see no reason to freeze the hover chain,
but freezing the active chain makes sense).  But you should update the comment
and overall structure of this code to do that, and make sure you're handling all
the different sub-cases consistently.  Perhaps you just need to remove the
'mustBeInActiveChain' logic?

[pals, 2014-08-07 05:32:22]

I tried to re-factor the logic of this Document::updateHoverActiveState() as
suggested but could not cover all the use-cases. I think someone who understand
this part code should take up this issue.