Issue metadata
Sign in to add a comment
|
Regression:'show page contents' text is seen misplaced on scrolling the page of www.quirksmode.org/html5/tests/video.html |
||||||||||||||||||||
Issue descriptionChrome Version:63.0.3239.20/10032.17.0 dev-channel Candy,Daisy,Reks OS:Chrome OS What steps will reproduce the problem? (1)Sign into user -> Launch Chrome and navigate to https://www.quirksmode.org/html5/tests/video.html (2)Click on 'show page contents' text present on top-right corner ->Scroll the page and Observe the misplacement of that text(kindly refer video) Actual:'show page contents'text is seen misplaced when page is scrolled Expected: Instead, the text doesn't get misplaced on scrolling the page. This is a Regression issue as same is working fine in 61.0.3163.120/9765.81.0 dev channel Daisy Note: 1. Issue is not seen in Windows, Linux OS 2. Issue is also seen in latest M-64
,
Oct 27 2017
,
Oct 27 2017
Reproduces with --enable-prefer-compositing-to-lcd-text on Linux.
,
Oct 27 2017
Scratch comment 3. It repros with devtools mobile emulation on. Bisecting now.
,
Oct 27 2017
Bisect: https://chromium.googlesource.com/chromium/src/+log/68374d6fc868984795f7ecb5457f7e4009d16de8..f470dd6dfd2e88e40e344e5bd914c90c999ddbe4 Suspecting a655402.
,
Nov 3 2017
Not blocking initial stable on this but willing to merge for stable refresh if safe CL is available
,
Nov 9 2017
I don't think my CL could have introduced this regression. I think there is either a site bug related to this particular setup of composited scrolling, or a bug in the input system in which it doesn't pass enough scroll events or with the right parameters. Does the input system guarantee eventual consistency of composited scrolling? Meaning: even though the scroll events come async to the actual scroll, they add up to the total scroll offset when quiescent. Untriaging and sending to the scroll component.
,
Nov 9 2017
,
Nov 13 2017
Seems that onScroll isn't invoked sometimes. flackr@ do you know of any bugs that cause a commit to happen on the cc side and the main thread isn't told about it?
,
Nov 16 2017
flackr@, ping.
,
Nov 17 2017
flackr@ for triage.
,
Nov 17 2017
I'm aware of the scroll offset synced back to main being off by one (issue 681951) but that shouldn't cause this as far as I can tell. I will investigate.
,
Feb 14 2018
,
Sep 17
This still reproduces on 69.0.3497.92 (Official Build) (64-bit) Linux, using devtools mobile emulation (Responsive, 1200x800, using touch-based scrolling.)
,
Yesterday
(47 hours ago)
Looking at this in dev tools, this appears to be missing an invalidation / layer property update. Observing the #header element the top value seems to be updating correctly on every scroll, but the layer doesn't always visually update its position. Hovering over the element in the devtools elements panel triggers a repaint which fixes the position. Chris, can you look into the invalidation issue. |
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by rkalavakuntla@chromium.org
, Oct 26 2017