svg/filters/shadow-on-rect-with-filter.svg is currently failing the pixel tests because we started using a transparency layer for shadows and clipping appropriately. Turns out the same bug already existed for opacity which does the same thing; in other words, if you replace the shadow style in that test case with "opacity: 0.5," we clip too much in the same way.
Yes. http://trac.webkit.org/browser/trunk/WebCore/rendering/SVGRenderSupport.cpp#93 and http://trac.webkit.org/browser/trunk/WebCore/rendering/SVGRenderSupport.cpp#L102 SVGRenderSupport uses the boundingbox. We might use the union of filter rect and objectboundingbox for the clipping, if filters are present.
But atm the filter rect is calculated at http://trac.webkit.org/browser/trunk/WebCore/rendering/SVGRenderSupport.cpp#L135.
*** Bug 32702 has been marked as a duplicate of this bug. ***
It needs a redesign, but I already have a patch for it. Just needs a bit more clean-up and testing.
Created attachment 45154 [details]
fix for the clipping bug
This fixes the failing pixel tests. Sadly object->repaintRectInLocalCoordinates() seems to change some local values of Renderers, so many filter related LayoutTests have different results.
Also we still have to calculate the filterRect a second time on drawing. But this is not avoidable, since we call the filters by id and one filter can be used by different object's.
*** Bug 32765 has been marked as a duplicate of this bug. ***
There are other affected tests (run pixel tests with --tolerance 0, and you'll see the errors):
(In reply to comment #6)
> There are other affected tests (run pixel tests with --tolerance 0, and you'll
> see the errors):
^^^^^^^^^^^^^^^^^ Beth already reported that one, oops.
Closed the bug 32584 that I just filed, it's a duplicate of this bug.
All have the same problem -- we _do_ need a pixel tests bot :(
Beth, can you do anything from the Apple side about this?
The clipping bug is fixed with r52647.
I'm not sure, if the patch also fixes:
If not, we have to open another bug.