Should definitely be possible. I suspect we're going to have a long tail of issues with regards to when we can still draw with LCD-text. On issue 500962 for example, the layer created for the sticky-bar is not opaque because its layer bounds extend further than the backround for some reason.
I'm hoping to tackle some of the easier wins in this long tail, starting with opaque background scrollers of course.
Great, assigning to you then. Yep, I think we should just use UMA to help prioritize which of the not-too-hard scenarios to focus on. Note that it appears that position:fixed is MUCH more common a reason for main thread scrolling than overflow:scroll (~38% vs. 5% on Windows).
We may need more strict conditions on when we can promote but I have a patch which promotes opaque fixed position elements on low dpi here: https://codereview.chromium.org/2318243002
Will CompositeOpaqueFixedPosition be enabled? If not, can we remove the flag and associated code? I recently found this when implementing main thread scrolling reasons for BlinkGenPropertyTrees.
Comment 1 by rbyers@chromium.org
, Jul 7 2016