[webkit-reviews] review requested: [Bug 48110] getBoundingClientRect: Do not truncate the coordinates to integers : [Attachment 72622] Fix and layout test

[bugzilla-daemon at webkit.org]

Cosmin Truta <ctruta at chromium.org> has asked  for review:
Bug 48110: getBoundingClientRect: Do not truncate the coordinates to integers
https://bugs.webkit.org/show_bug.cgi?id=48110

Attachment 72622: Fix and layout test
https://bugs.webkit.org/attachment.cgi?id=72622&action=review

------- Additional Comments from Cosmin Truta <ctruta at chromium.org>
In response to Niko's comment in bug 48640:

> Hmm, as I wrote in my last comment, the usage of enclosingIntRect is the
actual problem. We shouldn't do that at all, but give back a FloatRect,
according to the CSS OM API?

Done.

I am still puzzled by the discrepancy between zoomCount=2 and zoomCount=3. With
zoomCount=1,2 the results are well-behaved, within a very small distance from
the expected value (with delta=1.0e-5 -- very decent!). On the other hand, with
zoomCount=3,4,5... the errors grow wilder, suddenly, and even delta=0.5 is not
good enough for some of the coordinates.

I wonder if some precision is being lost accidentally, somewhere inside WebKit.