User-Agent: Mozilla/5.0 (X11; Linux i686; rv:2.0) Gecko/20100101 Firefox/4.0 Build Identifier: Mozilla/5.0 (X11; Linux i686; rv:2.0) Gecko/20100101 Firefox/4.0 For feSpotLight the light source position and the limitingConeAngle are interpreted wrong and there is an unspecified clipping as well (maybe for a similar reason than for bug 647687), respecitive alternatively the coordinate system for the filter could be wrong. In the file the animated effect of the limitingConeAngle is indicated with white circles, these circles indicate the cone angles related to the opacity values within the file. The position of the light source and the pointing direction are chosen in such a way, that the horizontal middle axis of the document and filtered object is a symmetry axis for the filter effect, this means, the top part is a mirrored correspondence of the part below and vice versa. Reproducible: Always Steps to Reproduce: 1. View the file given at the URI above 2. Compare with description within the file and the specification 3. Compare with the behaviour of the adobe plugin or Opera (Opera has only a slightly wrong limitingConeAngle) Actual Results: Apart from the wrong clipping effect respectively filter area, the effect is not symmetrical to the central horizontal axis and the limitingConeAngle does not fit to the indicating white circles. Expected Results: The display has to be symmetrical als explained above, without a clipping effect and the limitingConeAngle results in circular borders of the magenta areas, outside the limitingConeAngle the filter area is expected to be black. Borders are indicated with the white circles respectively for the used opacity values of the filtered object. Compare with the adobe plugin or approximately with the behaviour of Opera for details.

Sounds like some, perhaps all of this is the same as bug 619992

I can see similar position problems for tests with feSpecularLighting. The clipping bug dominates most of my filter tests with light sources, therefore I did not look for more details in them yet...

This is fixed by the patches in bug 619992

Fixed by bug 619992 and bug 647687