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

Issue 681652 link

Starred by 2 users
Status: Assigned
Owner:
senorblanco@chromium.org
Cc:
pdr@chromium.org
mtklein@chromium.org
fmalita@chromium.org
█ wkorman@chromium.org
Components:
EstimatedDays: ----
NextAction: ----
OS: Linux , Mac
Pri: 2
Type: Bug-Regression



Sign in to add a comment

Line on tile border of SVG image

Project Member Reported by danakj@chromium.org, Jan 16 2017 
Chrome Version: 55.0.2883.87
OS: Linux

What steps will reproduce the problem?
(1) Open the page
(2) See line

I think this is a raster problem in skia but pdr@ can you have a look and confirm. Attached a minimal-ish repro.

I found this on the opening slide of an internal training program. :)
Screenshot from 2017-01-16 14:27:41.png
16.8 KB View Download
Screenshot from 2017-01-16 14:27:39.png
17.8 KB View Download
repro.html
1.9 KB View Download

Comment 1 by pdr@chromium.org, Jan 16 2017

Cc: fmalita@chromium.org wkorman@chromium.org
Labels: -Type-Bug OS-Mac Type-Bug-Regression
Owner: senorblanco@chromium.org
Status: Assigned (was: Untriaged)
Bisected down to: https://chromium.googlesource.com/chromium/src/+log/bcfd8a99f9f695e0028d9daabfbc249acbe59d86..75af515a0347cc0c10a3685f1eafed1d7479a1f2

This could be either of these changes:
"Remove SK_SUPPORT_SRC_BOUNDS_BLOAT_FOR_IMAGEFILTERS #ifdef."
https://chromium.googlesource.com/chromium/src/+/d9ceaad90369f2840adfb88515298c816afb5c72

"Revert of Pass along Blink's visual rect for display items."
https://chromium.googlesource.com/chromium/src/+/5e0672938d40de5fb513d5f041d664cb77f2476a

I think the first one is more likely.

Comment 2 by kavvaru@chromium.org, Jun 27 2017 

 Issue 736423  has been merged into this issue.

Comment 3 Deleted

Comment 4 by pdr@chromium.org, Jul 11 2017

Owner: pdr@chromium.org
I think my original blame of senorblanco's change could be wrong.

Comment 5 by pdr@chromium.org, Jul 19 2017 

Minimized repro
repro.html
533 bytes View Download

Comment 6 by pdr@chromium.org, Jul 23 2017 

I re-did the regression with this repro and found we regressed at this commit:
Image filters: optimize crop rect application.
https://chromium.googlesource.com/skia.git/+/6db0a7bdceb6be85721bfb0db8dea7fd27db5970

Comment 7 by pdr@chromium.org, Jul 24 2017

Cc: pdr@chromium.org
Components: -Internals>Compositing>Rasterization Internals>Skia
Owner: senorblanco@chromium.org
With the minimized repro and bisect results, I now think this is a skia regression of merge filter bounds.
►

Sign in to add a comment