See attached video. This is on macOS, 2018-09-27 with gfx.webrender.all set to true. Load https://belen-albeza.github.io/scroll-demos/parallax.html, change the browser theme to "A Web Browser Renaissance", and scroll down. The misplaced box-shadow (that's bug 1494453) ends up showing up on top of the UI, which is bad. Maybe bug 1494453 will fix this by fixing the box-shadow positioning, but really we should have stronger safeguards against this sort of thing.

Hi Dzmitry, This one is concerning. When do you think you'd likely get to this (after you clear your current queue)?

I think this is slightly more urgent than the plane-splitting issues, I'll switch to this bug once the backface-visibility is sorted (no pun!).

I can't repro this on "64.0a1 (2018-10-22) (64-bit)" / Linux. Can you see it, kats? > really we should have stronger safeguards against this sort of thing I agree that our current rendering constraints for iframes are not strong enough. We basically rely on the: 1. local clip begin computed correctly (including the combined local clip from the clip-scroll tree) 2. it's provided to the shader correctly and used there appropriately It comes down to the fact iframes don't exist as a concept for the renderer. It sees the whole document as a bunch of render tasks and batches. Document separation is strong though.

I can't reproduce it at the moment. If you want I can try harder by backing out some stuff that has affected the behaviour on that page.

kats, it would certainly be useful to understand what as going on, but it's less of a priority given a chance that we don't have this issue any more

bug 1494453 comment 5 might have fixed comment 0. But bug 1496019 is still reproducible.

Un-assigning myself and lowering the priority, since it's not actionable right now.