280027 - SVG gradients incorrect if stops not at 0.0 and 1.0

Status: RESOLVED FIXED

(Core :: SVG, defect)

Description

[chris]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8b) Gecko/20050127 Firefox/1.0+
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8b) Gecko/20050127 Firefox/1.0+

Using the SVG GDI+ build of Firefox 1.0+, gradients display correctly if there
is at minimum a stop with offset 0.0 and a stop with offset 1.0 (plus zero or
more intermediate stops). However, if the offsets are not exactly 0.0 and 1.0, a
default red to green gradient displays instead. Although most content does in
fact include stops at the extremes of the offset range, this is not required by
the spec
http://www.w3.org/TR/SVG11/pservers.html#GradientStops
and some content does use gradient stops some way from the ends, to give a
little padding. The defauly value for the spreadMethod attribute on
linearGradient and radialGradient is pad, so the color at the stop with the
lowest offset is used for the portion 0 to stop, and the color of the stop at
the highest offset is used for the portion stop to 1.0

Reproducible: Always

Steps to Reproduce:
1. Load attached file blue-red-yellow.svg into an SVg-enabled build of Firefox
2. 
3.

Actual Results:  
Three gradients are displayed. The upper and lower gradients are red to green

Expected Results:  
Three gradients displayed, all blue to red to yellow, with small differences due
to the precise stop offsets.

Note that the lowest gradient of the three has one stop at 0.0, so the bug is
produced when one of the stops is at the extreme of the offset range and the
other is not.
The bug happens with the percentage values of offset, too.

Comment 1

[chris]

Attachment: test case demonstrating the bug

Shows three gradients, the middle one is correct and the top and bottom ones
are incorrect.

Comment 2

[tor]

Looks to be a "feature" of GDI+ - testcase works fine with cairo.

Comment 3

[tor]

Attachment: ensure extremes, range change, ordering per spec

Comment 4

[tor]

Checked in.