[Webkit-unassigned] [Bug 39004] Crash on SVG feLigthing effects

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=39004





--- Comment #10 from Zoltan Herczeg <zherczeg at webkit.org>  2010-05-19 09:53:23 PST ---
> You know the current position in the (4, 4) input, so if your matrix takes a point outside of this rect, just set the value to zero. I don't see the problem.

That is the problem :) It should have a 1.0 instead of 0.0, which badly affects the normal vector. There is no algorithm to guess the non-existing pixels. The input algorithm should only use the incoming pixels, and nothing more. The non-incoming pixels should be set to 0,0,0,0 RGBA value. I think.

> I don't understand the problem here either. You calculate the value of a certain point. And the matrix should calculate the value relative to this point. So it shouldn't matter how much pixel you use around the (4, 4) effect. The result shouldn't change.

It changes, unfortunately. The standard has different for algorithm for edge pixels. I.e: (0,0) point is the top-left corner, the normal vector is calculated from (0,0) (0,1) (1,0), (1,1) pixels. However, if (0,0) is an interior pixel, the 8 pixel around it should be used to determine the normal vector.

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