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

Issue 697969 link

Starred by 1 user

Issue metadata

Status: Assigned
Owner:
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 3
Type: Bug



Sign in to add a comment

Blur filter is rendered differently on main and impl

Project Member Reported by sunxd@chromium.org, Mar 2 2017

Issue description

Chrome Version: 56.0.2924.87 (stable)

1) Open WebKit layout test: compositing/masks/mask-with-added-filters.html (https://cs.chromium.org/chromium/src/third_party/WebKit/LayoutTests/compositing/masks/mask-with-added-filters.html?q=mask-with-added+package:%5Echromium$&l=1)

What is the expected result?

The left column and right column should look the same.

What happens instead?

The right one looks larger than the left.

If we open devtools, remove the mask properties, we can see pure gradient divs. Then try repeating adding and removing the composited class from the right quickly, we can see that it is rendered slightly different.


 
Cc: danakj@chromium.org jaydasika@chromium.org weiliangc@chromium.org enne@chromium.org
Owner: ajuma@chromium.org
Status: Assigned (was: Untriaged)
I am guessing this is because either or both of visible rect and clip rect now takes into account of filter expansion?

Also if you resize window enough to have vertical scrollbar, you can see the compositted div move when you scroll.


On the other hand, even without the the blur filter, I can still see difference in rendering on the column of stars on the right. So maybe we have more than one problem here.


Cc: sunxd@chromium.org vmp...@chromium.org
Talked w/ ajuma@, turns out there are three things at play here:
1. There was a visible rect bug that is fixed later, that explains the bouncing around when scrolling.

2. On compositor thread, image-mask is applied after blur filter (with expanded rect), while on main thread (and other browsers) image-mask is applied before blur filter.

3. The difference in the last row of stars is because of how we scale while raster. It is visible when we turn off both image-mask and filter, and the picture is a 2 by 2 image, scaled to fill the diff, and that looks different.

Comment 3 by ajuma@chromium.org, Apr 24 2017

Cc: ajuma@chromium.org
Owner: sunxd@chromium.org

Sign in to add a comment