Re: Tests required for recent changes to SVG 1.1 Second Edition

On Fri, 29 May 2009 08:01:55 +0200, Cameron McCormack <cam@mcc.id.au> wrote:

> Erik Dahlström:
>> The 'type' attribute doesn't decide the encoding in HTML5. What it's
>> used for is deciding if the script is supported or not. HTTP headers,
>> the BOM of the external script resource and the host document encoding
>> may all also have some influence on the chosen char-encoding for
>> external script resources.
>
> OK.  Is SVG’s behaviour consistent with this?

In Opera yes, it's handled the same way. The 'script' element in HTML has @charset, which SVG doesn't have, and it will affect the encoding. It would be easy to add to SVG as well.

> …
>> I'll be doing tests for the filters module anyway sooner or later, so
>> I can take the ones below here.
>
> OK thanks.
>
> Cameron McCormack:
>> > * Test kernelUnitLength on feSpecularLighting (since it is
>> >   now defined). However there aren’t any tests at all for
>> >   kernelUnitLength on either of the lighting filter primitives (or on
>> >   feConvolveMatrix)!
>>
>> Any ideas for good pass criteria for such tests would be welcome.
>
> Filters make my head hurt.  I’ll try and work through an example and see
> what you can reasonable test with a different of kernelUnitLength.  (You
> should be able to require particular pixel values as long as you don’t
> scale any of the bitmaps involved, so you’re not relying on particular
> resampling algorithms, right?)

Pixelvalues, well...maybe, that's the "must exactly match the reference image" criteria? Since we have no way (in the spec) of extracting the pixel data it still might be hard to verify. However, writing this now I wonder if it might be possible to generate the resulting image as a raster, and to use another filter for adding this referenceimage to the test filter output in order to get some image that can be more easily verified.

-- 
Erik Dahlstrom, Core Technology Developer, Opera Software
Co-Chair, W3C SVG Working Group
Personal blog: http://my.opera.com/macdev_ed

Received on Friday, 29 May 2009 11:09:18 UTC