New issue
Advanced search Search tips
Note: Color blocks (like █ or █) mean that a user may not be available. Tooltip shows the reason.

Issue 817100 link

Starred by 1 user
Status: Assigned
Owner:
fmalita@chromium.org
Cc:
█ liyuqian@chromium.org
mstensho@chromium.org
Components:
EstimatedDays: ----
NextAction: ----
OS: Linux , Windows , Mac
Pri: 3
Type: Bug



Sign in to add a comment

Investigate border-radius-clipped-layer.html failures after Skia change

Project Member Reported by fmalita@chromium.org, Feb 27 2018 
https://chromium-review.googlesource.com/c/chromium/src/+/923767 introduces some minor anti-aliasing differences, and breaks border-radius-clipped-layer.html (ref-test).

Let's see if we can tweak the test to pass after the Skia change.

Comment 1 by fmalita@chromium.org, Feb 27 2018 

Attaching SKP captures for the test & ref.

The rounded rects are painted independently in both cases AFAICT.  Comparing a single rrect stroke:

   DrawRRect([16, 108, 208, 444], [40, 40], [40, 40], [40, 40], [40, 40])

vs.

   DrawRRect([16, 108, 208, 284], [40, 40], [40, 40], [40, 40], [40, 40])


The only thing that jumps out at me is the ref geometry is not as tall as the test (444 vs. 284).  The result looks similar because both are clipped short (236 I think).

@liyuqian do you think the geometry size could cause those diffs in anti-aliasing (even if the geometry delta is invisible due to clipping)?

The first thing I would try to fix the test is adjust the ref geometry to the same size.
border-radius-clipped-layer.skp
4.1 KB Download
border-radius-clipped-layer-expected.skp
4.0 KB Download

Comment 2 by liyuqian@google.com, Feb 27 2018 

Ah, right, I used the ratio between the number of edges and the max(width, height) of the path to decide whether the path is complex or not. I probably didn't take clip into consideration. So we probably used DAA for the shorter RRect stroke and AAA for the taller RRect stroke.
►

Sign in to add a comment