[Webkit-unassigned] [Bug 42605] New: New border-radius path-based drawing code has some issues with corner-joins

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=42605

           Summary: New border-radius path-based drawing code has some
                    issues with corner-joins
           Product: WebKit
           Version: 528+ (Nightly build)
          Platform: PC
        OS/Version: Mac OS X 10.5
            Status: NEW
          Severity: Normal
          Priority: P2
         Component: CSS
        AssignedTo: webkit-unassigned at lists.webkit.org
        ReportedBy: bdakin at apple.com


The new path-based border-radius drawing code in CG has resulted in a corner-join regression. This is a good example of the regression being bad:

http://ie.microsoft.com/testdrive/Graphics/IE%20Logo/Default.html

This is because GraphicsContext::clipConvexPolygon() is currently written to be an anti-aliased clip. (That clipping sets up all the corner joins.) When we anti-alias that clip, we sometimes get these bad corner joints. In the long term, the solution to the problem is to be able to re-write the code so that it is capable of drawing multiple sides of the border at once, because this would not only prevent bad joins in the ie-logo case, but it would also fix our badness with alpha borders (see fast/borders/border-radius-wide-border-03.html). In fact, we have this ialpha-border issue with straight borders too! So that is not specific to border-radius. Right now, doing that seems like a major project, and it also seems messy in it's own way because I can't think of a way to write the code that doesn't involve branching for every possible combination of possibly-same-colored edges. Anyway, that would be good to figure out some day because it would make our corners look much better
  on borders with alpha. 

But since it is such a large project (and the alpha issue is not a regression nor is it unique to border-radius), I think it is beyond the scope of fixing this regression. I have a simple patch to fix just this issue that I will upload shortly.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.