SVG movement causes artifacts when background is set with background-blend-mode and background color
Reported by
salva...@gmail.com,
Apr 12 2018
|
||||||||||
Issue descriptionUserAgent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_13_4) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/65.0.3325.181 Safari/537.36 Example URL: https://codepen.io/anon/pen/MVRVpe Steps to reproduce the problem: 1. Open the Codepen linked, OR copy and paste this Gist: https://gist.github.com/zedee/e5af3ba433c5c15c74833e17f154a5d6 into an html file and open it. 2. Click anywhere on the screen. 3. Observe if the moving box drops artifacts (leaving a trace of lines) when moving. What is the expected behavior? The box should move without leaving any trace nor artifacts What went wrong? This bug doesn't happen consistently between Macs. It happens, though, with all Macs with an embedded Intel Graphics (Intel Graphics HD4000, Intel Iris Plus Graphics 640), and some Macs with dedicated graphics card. It did not happen with Macs with a Radeon dedicated graphics card. Does it occur on multiple sites: Yes Is it a problem with a plugin? No Did this work before? No Does this work in other browsers? Yes Chrome version: 65.0.3325.181 Channel: stable OS Version: OS X 10.13.4 Flash Version: Running Chrome/Chromium with the flag --disable-gpu seems to solve the issue, so it has to be something up with the GPU.
,
Apr 12 2018
Caused by GPU rasterization - disabling it in chrome://flags "fixes" the bug. Bisect info: 405364 (good) - 405381 (bad) https://chromium.googlesource.com/chromium/src/+log/7c6bf957..96bea8c0?pretty=fuller Suspecting r405379 = 5ac42f32e7b4142cd578082424a85a26794cf0ab = https://crrev.com/2110083004 by ericrk@chromium.org "Partial raster for GPU" Landed in 54.0.2796.0 Since it wasn't enabled by default in Chrome 54, command line switch "--enable-gpu-rasterization" was used.
,
Apr 12 2018
,
Apr 12 2018
,
Apr 13 2018
Able to reproduce the issue on reported chrome version 65.0.3325.181 and on the latest version 67.0.3395.0 using Mac 10.12.6. As the issue is seen from M60(60.0.3112.0) considering it as Non-Regression and marking it as Untriaged. Note: Issue is not seen on Ubuntu 14.04 and Mac 10.12.6. Thanks!
,
Apr 13 2018
ericrk can you take a look? It looks like this might be related to https://crrev.com/2110083004.
,
Apr 16 2018
It looks like the border pixels when partial rastering are not being generated correctly when a background-color blend mode is in use - seems like the edges of the partial raster region end up having *only* the background-color, not the image that is supposed to be blended with that color. Will try to capture an SKP.
,
Apr 17 2018
It appears that, in the scenario we're hitting here, Skia is drawing outside of the clip set on the canvas by 1 pixel, and filling these pixels incorrectly. At the end of raster, but before popping the raster clip from the canvas, I asked Skia to draw a huge solid-color rect. Although this rect is much larger than the clip, it does not cover the border pixels we're seeing, indicating that Skia has somehow filled outside of the clip. hcm@ can you help triage?
,
May 1 2018
I think I need to get this to Mike for a look..thanks for the analysis
,
May 1 2018
inval bounds?
,
Aug 2
,
Jan 16
(6 days ago)
|
||||||||||
►
Sign in to add a comment |
||||||||||
Comment 1 by salva...@gmail.com
, Apr 12 2018