Rietveld Code Review Tool
Help | Bug tracker | Discussion group | Source code | Sign in
(2666)

Issue 5989070: Reinstate early out from antialiased path rendering if we anticipate overflow (Closed)

Can't Edit
Can't Publish+Mail
Start Review
Created:
12 years ago by TomH
Modified:
8 years, 3 months ago
Reviewers:
reed1
Base URL:
http://skia.googlecode.com/svn/trunk/
Visibility:
Public.

Description

If the intersection of the bounds of the path & the bounds of the clip have any coordinates >= 8192, we'll overflow & render incorrectly if we go through the antialiased code, so fall back to SkScan::FillPath(). Works around http://code.google.com/p/chromium/issues/detail?id=121405

Patch Set 1 #

Patch Set 2 : Fix layout tests, respond to verbal review #

Patch Set 3 : Now with early exit! #

Unified diffs Side-by-side diffs Delta from patch set Stats (+44 lines, -3 lines) Patch
M src/core/SkScan_AntiPath.cpp View 1 2 2 chunks +44 lines, -3 lines 0 comments Download

Messages

Total messages: 8
TomH
Looks fine in test case (just turns antialiasing off once we pass x=8192); currently running ...
12 years ago (2012-04-09 20:13:25 UTC) #1
TomH
This breaks 4 layout tests: fast/box-shadow/* Also, todos from verbal review: - write helper function ...
12 years ago (2012-04-10 14:09:28 UTC) #2
TomH
Should be good to go now. If we have an inverse path fill, we may ...
12 years ago (2012-04-11 21:03:03 UTC) #3
reed1
doesn't pathClippedOut mean nothing will draw?
12 years ago (2012-04-12 12:54:02 UTC) #4
TomH
I sure think so - if you think so too, I'll try throwing away the ...
12 years ago (2012-04-12 13:21:42 UTC) #5
TomH
Early exit passes both gms and LayoutTests.
12 years ago (2012-04-12 14:14:53 UTC) #6
reed1
sweet -- lgtm
12 years ago (2012-04-12 14:17:05 UTC) #7
TomH
12 years ago (2012-04-12 14:24:23 UTC) #8
Committed as r3658.
Sign in to reply to this message.

Powered by Google App Engine
RSS Feeds Recent Issues | This issue
This is Rietveld f62528b