[Webkit-unassigned] [Bug 108386] [Chromium] WebWidget::selectionBounds should return the bounds in document space

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=108386





--- Comment #5 from Chris Hopman <cjhopman at chromium.org>  2013-01-31 10:51:42 PST ---
(In reply to comment #3)
> Does this fix a specific problem?

So here's the issue: when page scale is applied in the compositor, there are essentially 2 different window spaces, the scaled one and the unscaled one. WebKit deals with the unscaled and then when points or rects are passed between WebKit and the renderer, if they are passed in window space, they need to be scaled or unscaled.

It is unclear (at least to me) whose responsibility this is (i.e. at what boundary should it be done). Currently, it looks like we do this in WebKit (look for applyPageScaleFactorInCompositor in WebFrameImpl.cpp or WebViewImpl.cpp). I don't like this because it becomes confusing when a public API is called within WebKit. For example, WebWidget::selectionBounds() returns two rects. If WebKit applies the pagescale, then when selectionBounds is used in WebKit (for example WebViewImpl::computeSCaleAndScrollForFocusedNode) you have to know that the rects are not actually in WebKit's viewport space, they are in this other scaled space. And so now, we have this layer that needs to know all about an extra space.

My thought was that document space could be this common language that could be used across the boundary into and out of WebKit, so that neither side needed to know about the other's viewport space (and thus never had to distinguish between points in unscaled vs scaled viewport space).

I guess I can just switch to what we've already done in other places and make selectionBounds instead scale these rects when necessary (and its users in WebKit can be updated to know that the rects are scaled).

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