Rendering of 'feSpecularLighting' in 0.47 differs from 0.46

[Hadmut Danisch]

Hi,

after upgrading from Ubuntu 9.04 with inkscape 0.46-5ubuntu4 to Ubuntu 9.10 with inkscape 0.47~pre4-0ubuntu1, an important function in inkscape is broken.

With the older version the attached svg produced the expected result a threedimensional looking box with shadown. With the older version, the same svg produces a box that looks darker and flat. I'll attach the svg and both outputs.

[su_v (suv-lp)]

reproduced with Inkscape 0.46-2 and Inkscape 0.46+devel r22575 on OS X 10.5.8:
the same object with the filter 'glanz3D' is rendered darker in 0.47(pre) than in 0.46.

filter in question:
<filter id="glanz3D" filterUnits="objectBoundingBox" x="-50%" y="-50%" width="200%" height="200%">
  <feGaussianBlur in="SourceAlpha" stdDeviation="4" result="specBlur"/>
  <feSpecularLighting in="specBlur" result="specOut" lighting-color="#bbbbbb" surfaceScale="1" specularConstant="0.5" specularExponent="20">
    <fePointLight x="-5000" y="-10000" z="20000" />

[su_v (suv-lp)]

possibly related to the discussion about "color-interpolation and color-interpolation-filters" in inkscape-devel?
<http://thread.gmane.org/gmane.comp.graphics.inkscape.devel/30566/focus=30571>

[Hadmut Danisch (hadmut)]

Any estimation when the bug will be fixed?

[ScislaC (scislac)]

Hadmut,

We really don't ever have estimates of when certain bugs will be fixed unless someone is actively working on them, and no one is working on this one as far as I know.

[Hadmut Danisch (hadmut)]

:-(

[jazzynico (jazzynico)]

Also reproduced on Windows XP, Inkscape 0.48 (and working correctly with 0.46).

[su_v (suv-lp)]

The issue seems to be with the effect of the 'specularConstant' setting for 'feSpecularLighting':
The old filter effect created in 0.46 used '0.5', with 0.47 and later versions the same effect is achieved with '0.9'.

Since other SVG renderers (Squiggle (Batik 1.7), Opera 11.52, Chromium) agree with Inkscape's current rendering, this was possibly an incorrect initial implementation in Inkscape 0.46 (needs verification).

[ScislaC (scislac)]

I tried in Inkscape trunk and then in Firefox 25 and a Chrome 33 beta, they render as Inkscape 0.46 used to render it. I'd say the original implementation would be correct.

[ScislaC (scislac)]

Okay... new comparison with image. Reference info...

Chrome 38 beta
Firefox 32 beta
Ink-old is Hadmut's reference image from 0.46
Ink-new is Hadmut's reference image from 0.47
r13527 is from trunk
r13527a is trunk but with the filter modified as suggested by ~suv

My lesson? Don't eyeball it instead of doing a side-by-side, my bad from the previous comment.

It looks to me like current trunk is even darker than the supposedly "bad" image. After adjusting the filter, it looks closer to how those other mentioned renderers are doing it. So yes, while the newer behavior is more correct, it still not quite right.