Depth sorting glitch with CSS transforms?
Reported by
trusktr@gmail.com,
Dec 7 2016
|
|||||||
Issue descriptionUserAgent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_12_1) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/54.0.2840.71 Safari/537.36 Steps to reproduce the problem: 1. Visit https://jsfiddle.net/trusktr/r83tz09a/ 2. Observe the glitches near the wheels while the car rotates. Sometimes the wheels are completely in front of the car's chassis. What is the expected behavior? The camo surface of the car should always intersect the wheels. What went wrong? It doesn't always intersect: sometimes the wheel elements are drawn on top. Did this work before? Yes I think I recall that it worked before, but not sure which version. Does this work in other browsers? No Firefox has glitches too, but they behave differently. Chrome version: 54.0.2840.71 Channel: n/a OS Version: OS X 10.12.1 Flash Version: Shockwave Flash 23.0 r0 Here are two HTML-only snapshots. This first one shows an angle of rotation where the wheel is rendered as expected: https://jsfiddle.net/trusktr/r83tz09a/5 This second one shows one angle where the wheel is drawn on top rather than intersecting with the chassis: https://jsfiddle.net/trusktr/r83tz09a/6 The difference in angle (from 50 to 49) is small so that the difference in renderings is easy to distinguish. The screenshots show the correct and wrong behavior.
,
Dec 9 2016
Able to reproduce the issue on the latest canary(57.0.2946.0) on Windows-10, Mac OS 10.11.6 and Linux Ubuntu 14.04. This is regression issue broken in M-43. Last good build: 43.0.2330.0 First bad build: 43.0.2331.0 Changelog: ========== https://chromium.googlesource.com/chromium/src/+log/a4d00e2cf24c50e8d31b80ae77303b63689010e0..3d8eb1d841ec1ea36906d351bc65af4d7d27718a Suspecting: https://codereview.chromium.org/595593002. awoloszyn@: Could you please take a look at this and confirm if the change is related. Note: Got we don't have enough build error when tried bisecting for hasbisect-per-revision for win(32 bit).
,
Dec 9 2016
Seeing awoloszyn@ last visited> 30 days ago. CC'ing vmpstr@ for help in finding an appropriate owner for this.
,
Dec 10 2016
Peter, sending to you since awoloszyn@ doesn't work on chrome anymore.
,
Dec 13 2016
This looks like damage area computation issues, since I can't seem to reproduce it after replacing the script with direct decoration of divs with transforms. Mind you, I also don't see it on new builds with the example fiddles. I have seen this sort of error in other 3d scene recreations but I haven't managed to track it down. I can own this until I can prove it, but I'm going to finish the CSS-only example(s) and park it for a while. I'm not sure why this is P1 when it has been on 12 Milestones without noise.
,
Dec 13 2016
Here's a simplified pure CSS sample of roughly the same thing. It would be edifying to know if there is somewhere it suffers the same bug, that would mean CSS model/transform code rather than timing interactions with the script interpreting the model build.
,
Jan 3 2017
I can't get this to happen any more on stable even. Happy to re-approach of we can get repro steps, or clarification from the wild.
,
Mar 24 2017
|
|||||||
►
Sign in to add a comment |
|||||||
Comment 1 by nainar@chromium.org
, Dec 8 2016Status: Untriaged (was: Unconfirmed)