Hi, See topic :) It seems like the big gradient renders over everyhing else? Cheers, Rob.
Created attachment 4987 [details] Reduced testcase This testcase has script and style elements removed. Also just one svg shape(rect) and a form that should display over the svg rect.
I bet my disabling of RenderLayer creation for svg elements is causing the z-ordering system to be confused...
This will crash recent development builds due to an ASSERT.
Removing: virtual bool requiresLayer() { return false; } should fix this. The problem then becomes making sure that they layer code doesnt' cause double-opacity in certain instances, since the SVG render objects handle their own opacity instead of relying on RenderLayer.
This seems to be a CSS selector problem at this point.
Actually, I was wrong. This is still a z-order issue.
It seems that webkit renders it fine now, except for slice.
It's still not clipping 100% correctly. The top and bottom extend too far.
Created attachment 14836 [details] even better test case This seems to be a CDF interaction problem. Somehow the SVG thinks it is taller than it should think it is.
Created attachment 15030 [details] Fix z-index problems This should fix the last issue. It still needs a testcase, but that should be easy. What is more annoying is that it changes most DRT results. So we may need to wait with it and make RenderSVGRoot output RenderSVGRoot instead of RenderSVGContainer in the DRT in one fell swoop. Cheers, Rob.
Comment on attachment 15030 [details] Fix z-index problems Perfect. The final patch will need to include test updates as well as this new test. However the code change itself looks great. Rob will land this later this week.
Comment on attachment 15030 [details] Fix z-index problems Rob said there is an issue with SVG + z-indexing that he needs to discuss with Hyatt before this can be landed. Marking as r- to get it out of the commit queue for now.
Fixed in r24292.