Re: Request for new testcases for HTML5 elements

Kris Krueger wrote:
> Phillip are you still interested in contributing your Canvas tests?
> In March this was discussed and you were going to update some of your canvas tests.

I'm still interested, but I'm afraid I'm not good at making time to work 
on it. But I've made a few minor changes now to 
http://philip.html5.org/tests/canvas/suite/tests/ :

* Removed _getPixel's workarounds for browser bugs (it assumes that 
getImageData is implemented correctly or is not implemented at all)

* Fixed some 2d.pattern.* tests that used 'NULL' instead of 'null'

* Fixed 2d.path.arc.nonfinite using wrong number of arguments

* Fixed 2d.imageData.object.round to expect the correct rounding

* Updated http://philip.html5.org/tests/canvas/suite/tests/results.html 
with results from IE9 (thanks to Simon Pieters for running the tests). 
(All these results come from before the above fixes, so they're slightly 
out of sync with the test cases and shouldn't be taken entirely seriously.)

If these tests are submitted to the W3C repository, what files should be 
included? (I don't remember if there's already been any conclusions or 
decisions on this). There are various parts like:

* The source data containing the original version of all the tests: 
http://philip.html5.org/tests/canvas/suite/tests2d.yaml etc

* The Python scripts that generate the HTML files: 
http://philip.html5.org/tests/canvas/suite/gentest.py etc

* The generated output (almost 3000 files, 5MB), with three different 
versions of each test case (of varying levels of verbosity) and a PNG 
for each, and the generated test harness pages

* The generated annotated spec: 
http://philip.html5.org/tests/canvas/suite/tests/spec.html . (This is 
based on the WHATWG version of the spec; I don't know how its license 
interacts with the required test licensing.)

I'd be happy to put all of the files into W3C Hg and use that as the 
primary source for any future updates I make. But I don't know if other 
people are happy with a load of barely-comprehensible Python and YAML in 
there, or if it would be preferable to simply have one HTML file per 
test case and leave out all the other complexity.

-- 
Philip Taylor
pjt47@cam.ac.uk

Received on Thursday, 24 June 2010 14:55:50 UTC