An <svg> element inside an html hierarchy with CSS transforms fails to redraw correctly.
Created attachment 26420 [details] Testcase
Here's what I think needs to happen: 1. SVG should avoid "absoluteTransform" computations everywhere. All bounding boxes and transforms should be relative to the SVGRoot. 2. There needs to be repaint hand-off at the SVG/HTML boundaries (i.e. SVGRoot and SVGForeignObject), so that the hosting render system can convert the repaint rects to the correct coordinates. 3. SVGForeignObject is going to need a RenderLayer (see also bug 23111, bug 23113), and RenderLayers need to be parented through SVG/HTML boundaries. I think we can avoid other RenderLayers inside SVG, but we'll need to put the accumulated transform from SVG content on the foreignObject's layer.
This was fixed by http://trac.webkit.org/changeset/42950 bug 25431. I should land your test case as a LayoutTest though.
Created attachment 30012 [details] Update test case for SVG in CSS transforms 5 files changed, 74 insertions(+), 0 deletions(-)
We currently fail the <text> section of this test, but pass all the rest.
Comment on attachment 30012 [details] Update test case for SVG in CSS transforms r=me
http://trac.webkit.org/changeset/43234