[Webkit-unassigned] [Bug 53055] REGRESSION: Rendering artifacts on a rotated, pattern filled shape

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=53055





--- Comment #25 from Dirk Schulze <krit at webkit.org>  2011-08-23 01:53:22 PST ---
(In reply to comment #22)
> Niko: here's one of the cases that I was talking about; rounding down here causes us to begin to draw the pattern again (because it's not big enough) -- note the bright orange line on the right side of the image, even though the rectangles are exactly the same size.
> 
> I also question why we're drawing a significantly larger copy of the image for the rotation case; there's no scaling going on here, so it shouldn't have anything to do with pixelation when drawn on the destination. Is the target size being miscomputed?

We try to get the affected area not the screen in screen coordinates. While this works great for scaling, it looks a bit scaring on rotations and skews (like you noted above). I'm not sure if I remember correctly, but I think their were some problems to divide a transformation matrix in the two parts translation, scaling, skewing parts and the rotation. If we can divide the rotation from the other parts, we will get better results for rotated elements.

I think I can remember to have a negative example:

<pattern width="20" height="200" patternTransform="rotate(90)">
<image width="20" height="200"/>
</pattern>

IIRC we will create an image buffer with the size 200x20 and scale the image to match in this rect. The image will of course loose in quality.

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