Ppk's pos fixed tests
Rick Byers
Hey Dave,
I just stumbled across this page where ppk describes our old fixed-pos zoom behavior as "correct": http://www.quirksmode.org/blog/archives/2013/12/position_fixed_1.html. I think what he's saying is that consistently positioning / sizing relative to the visual viewport is at least rational and consistent with some other browsers.
His test page (http://www.quirksmode.org/css/css2/positionfixed.html) doesn't appear to do anything in Chrome 40+ (presumably because he's expecting fixedpos elements to relayout on zoom), but I haven't dug into it yet. Can you take a look to make sure there are no surprises here?
Perhaps we should reach out to ppk to see what he thinks of the new behavior?
Rick
Rick Byers
Discussion with ppk: https://twitter.com/ppk/status/556818108560982017
David Bokan
Took a look. This is an instance of the problem in crbug.com/437034. The page is very wide so we size the outer viewport to match it's width and set the height to match the screen's aspect ratio. This means the outer viewport spans all the content. The problem comes up on pages that are so wide that we can't zoom out far enough to fit all the entire width of the content.
This should already be fixed on ToT. I haven't seen it cause an issue on any real pages so I'm not sure if we want to merge it back to 41. It's too late for 40 at this point. It can be surprising when you see it on pages like this though.
Jacob is right that the confusion is that ppk's layout viewport is the ICB where as we consider it scrollable. I'm curious to hear what his thoughts are here.