[Webkit-unassigned] [Bug 84936] [chromium] Compute highlight for touch targets

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=84936





--- Comment #4 from W. James MacLean <wjmaclean at chromium.org>  2012-04-26 08:58:23 PST ---
(In reply to comment #3)
> I couldn't find chromium account for the others, could you help me CC them?
> 
> This is an early WIP. I did open some pages and hover over some links to make sure it doesn't crash, but the actual generated paths are not verified. I believe there are still some bugs need to be fixed before requesting for review.
> 
> Also there are still some design concerns:
> 
> 1. When do we generate the highlight, and how do we attach them to the GraphicsLayer?
> The best timing is probably do it when painting the texture, but is it safe to keep a WebCore::Node* reference to the touch target? The node can be removed by javascript in the meantime.

This is being worked out in https://bugs.webkit.org/show_bug.cgi?id=84487

> 2. How do we remove the highlight from the GraphicsLayer?

Ditto.

> 3. We probably want to refactor RenderInline::absoluteQuads.
> I added another virtual function RenderObject::containerQuads that works similar to absoluteQuads except that it return the quads in a container RenderBoxModelObject's cords instead of the absolute cords. For RenderInline it is generating too much duplicate code. Better make RenderObject::absoluteQuads non-virtual and just call into containerQuads with NULL container.
> 
> 4. Need to make the function more efficient.
> It probably doesn't need to generate quads all descendant nodes... For example if there is a <div style="overflow:hidden"> then all its descendants can be skipped (except for fixed/absolute positioned descendant). I'm still thinking a clean way to do this.
> 
> 5. absoluteQuads is probably not the perfect way to determine hit test region. There is an interesting example. See the first and second touch target in the attached example. The space between "Hello" and "link" is still occupied by the line box of the relative-positioned <div>, but it is only clickable when applied with a background color. Can't figure out how to do this right. :(
> 
> By the way there is an interesting bug in WebKit rendering. See the 4th touch target in the example. If you click on it, the blue box can escape from its parent's clipping.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.