Re: Minutes, Paris F2F day 4 (September 8, 2010)

Hi everyone,


>   <trackbot> [16]http://www.w3.org/Graphics/SVG/WG/track/actions/2812
[...]
> http://dev.w3.org/SVG/profiles/1.1F2/ua-tests/filter-feconvolvematrix-bias-ISSUE-2812.svg
[...]
>   ED: Opera and Firefox same on bias
>   ... In batik it drew nothing
>   ... color-interpolation-filters should affect the result, so
>   inkscape 0.48 is wrong
>
>   CL: Often filter implementations seem to not get the colour space
>   right for calculations (at least, at first) until they see what a
>   difference it makes
>
>   batik svn says 'rendering document failed' on that test

FWIW, I've crawled a bit and Batik doesn't support bias other than 0.
An exception is thrown in the console ("Only bias equal to zero is
supported in ConvolveMatrix."). It's listed in the implementation that
"Does not support bias other than zero - pending 16bit pathway" [1] so
it's a known limitation.

I'm CC'ing Thomas since he was the author and probably might be
curious to known about this. :-)


[...]
>   CL: I extended the test to add an explicit bias="0.0"
>   ... checking in now
>
>   [19]http://dev.w3.org/SVG/profiles/1.1F2/ua-tests/filter-feconvolvem
>   atrix-bias-ISSUE-2812.svg

This will "help" Batik in the case of
"convolve-with-cif-linearrgb-nobias" filter. The other three filters
present in the test case (with non-zero bias) will still cause Batik
to abort rendering.


Cheers,
 Helder


[1] http://svn.eu.apache.org/viewvc/xmlgraphics/batik/trunk/sources/org/apache/batik/ext/awt/image/renderable/ConvolveMatrixRable8Bit.java?annotate=478363#l49

Received on Sunday, 12 September 2010 08:01:38 UTC