Re: [cssom-view] getBoundingClientRect() when the first rectangle is empty

The spec has rather peculiar wording around this, as if it is making special affordances to make sure the first rectangle is always in the return value.

Do we know why the current version is written this way?

> On May 5, 2016, at 6:06 AM, Koji Ishii <kojiishi@gmail.com> wrote:
> 
> I'd like to propose a change to getBoundingClientRect algorithm[1] from:
> 
> 3. Otherwise, return a static DOMRect object describing the smallest rectangle that includes the first rectangle in list and all of the remaining rectangles of which the height or width is not zero.
> 
> to:
> 
> 3. If all rectangles in the list have zero in its width or height, return the first rectangle in the list.
> 4. Otherwise, return a static DOMRect object describing the smallest rectangle that includes all of the rectangles of which the height or width is not zero.
> 
> Background
> ----------
> 
> Blink recently changed getBoundingClientRect() implementation to follow the spec. We received feedback saying that it produces undesirable result when the first rectangle is empty and the second is not.
> 
> I talked to roc and he confirmed that Gecko actually does what proposed above.
> 
> As far as I tested, Trident unions all rectangles including width or height is zero. WebKit excludes empty rects, but returns (0,0,0,0) when all rectangles are empty.
> 
> Changing the spec this way produces reasonable results when only the first rectangle is empty, and makes Gecko and Blink (with a coming fix) conform to the spec.
> 
> [1] https://drafts.csswg.org/cssom-view/#dom-element-getboundingclientrect
> 
> /koji
>