Wheel Event Not Fired For `body,html { height:100% }`

[Brent Fulgham]

Created attachment 259898 [details]
Test Case showing the improper behavior

From Chrome Developers:
This *was* one of the most starred bug for Chromium (800 stars). With the help of the community we got this FIXED in Chrome v45.

https://code.google.com/p/chromium/issues/detail?id=500144

We NEED WebKit to fix it as well!

Wheel event should work like other mouse events (click) and fire in the popular
 `body,html { height:100% }` case.

Steps to Reproduce:
Test page: http://output.jsbin.com/kapogi 
(also attached as a file)

/* 
 * The problem is caused by an element 
 * bigger than viewport, and body,html
 * having 100% height.
 */
html, body {
  height: 100%;
}

#page {
  position: absolute;
  width: 50%;
  height: 1000%;
}

Have an `onwheel` or `onmousewheel` handler on the document.

Expected Results:
Wheel events like all other mouse events should fire on the sides (the test page shows how click fires correctly).

Actual Results:
After passing the html,body { height:100% } limit the wheel events are no longer fired (on the sides).

[Brent Fulgham]

<rdar://problem/22376037>

[Brent Fulgham]

Created attachment 259907 [details]
Patch

[Beth Dakin]

I think we need to understand this fix more. I don't understand why it works, and it seems like an indirect and maybe incorrect way to get the desired behavior.

[Sam Weinig]

Comment on attachment 259907 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=259907&action=review

> Source/WebCore/page/EventHandler.cpp:2717
> +    HitTestRequest request(HitTestRequest::ReadOnly | HitTestRequest::Active| HitTestRequest::DisallowShadowContent);

Your changelog says that this will "Fall back to document", but it is unclear why.  Can you explain in a bit more detail what is going on here? Also, you are missing a space after Active.

[Simon Fraser (smfr)]

Comment on attachment 259907 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=259907&action=review

> Source/WebCore/ChangeLog:12
> +        This change was adapted from Blink r198210:
> +        http://src.chromium.org/viewvc/blink?view=rev&rev=198210  

I would like an explanation for why setting Active gives the desired behavior. The Blink code has a FIXME here.

[Simon Fraser (smfr)]

rdar://problem/22756010

[Simon Fraser (smfr)]

Created attachment 277673 [details]
Patch

[Brent Fulgham]

Comment on attachment 277673 [details]
Patch

r=me. \0/

[WebKit Commit Bot]

Comment on attachment 277673 [details]
Patch

Clearing flags on attachment: 277673

Committed r200247: <http://trac.webkit.org/changeset/200247>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.