Created attachment 106989 [details] Simple testcase When using a linear gradient as the background image for an element, stop positions that are specified in pixels are inconsistent. For a vertically-oriented gradient, the positions change as the height of the element increases. At very large heights, the positions as rendered can be extremely different than the specified values, but even relatively smaller dimensions cause some shifting. In the Confirmed on WebKit r94906 Mac, Safari 5.1 Mac, and countless Chromes and Chromiums on several platforms.
Created attachment 106990 [details] Simple testcase, 2nd attempt Attaching a URL didn't seem to have the effect I expected.
Confirmed on Webkit r96763 and Chrome 16.0.901.0 canary.
We investigated this, and WebCore is passing correct values down to the platform graphics layers. There appear to be bugs in the linear gradient rasterisers in both CoreGraphics (seen on Safari on Mac 10.6 and 10.7) as well as Skia (for Chrome). Closing as wontfix, as this is not a bug in WebCore.
Cross-referenced on the underlying Skia bug: http://crbug.com/117140