[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Public WebGL] Initial tests available



On Thu, Jan 7, 2010 at 5:17 PM, Vladimir Vukicevic <vladimir@mozilla.com> wrote:
> On 1/7/2010 5:03 PM, Kenneth Russell wrote:
>>
>> On Thu, Jan 7, 2010 at 4:37 PM, Gregg Tavares<gman@google.com>  wrote:
>>
>>>
>>> On Thu, Jan 7, 2010 at 1:39 PM, Kenneth Russell<kbr@google.com>  wrote:
>>>
>>>>
>>>> Some initial tests have been checked in to the WebGL repository. These
>>>> are basically copies of tests from the WebKit repository used with
>>>> permission from Apple Computer and the Chromium team where
>>>> appropriate.
>>>>
>>>> You can run them interactively in a WebGL enabled browser from
>>>>
>>>>
>>>>
>>>>  https://cvs.khronos.org/svn/repos/registry/trunk/public/webgl/sdk/tests/fast/
>>>>
>>>> and check out their source code with
>>>>
>>>>    svn checkout
>>>> https://cvs.khronos.org/svn/repos/registry/trunk/public/webgl
>>>>
>>>> These aren't quite unit tests and aren't quite conformance tests which
>>>> is why they've been put in a directory called "fast".
>>>>
>>>> The -expected.txt files are currently only used by WebKit's
>>>> run-webkit-tests script, but can be used to check for regressions.
>>>>
>>>> More work is needed, for example to write a harness in JavaScript
>>>> which automatically runs each one and verifies against expected
>>>> results. Contributions are very welcome.
>>>>
>>>
>>> How about using selenium?  It's open source and it runs all the browser
>>> AFAIK. We have some existing code to do this from another project. That
>>> could would basically take a list of tests, tell the browser to run each
>>> one
>>> then wait for window.g_testResult to become defined true or false where
>>> true
>>> = that test passed.
>>>
>>> Is that an okay direction?
>>>
>>
>> If it is possible to use selenium to construct a test harness that you
>> can just point to with a web browser, that sounds fine to me. If it
>> requires command-line invocation or downloading and installation of
>> packages, then I would like to avoid it. My experience with selenium
>> on O3D, in particular debugging why O3D wouldn't load from within
>> Selenium in IE, was not good. We don't need its complexity, at least
>> not for our current tests.
>>
>> Philip Taylor's Canvas tests at
>> http://philip.html5.org/tests/canvas/suite/tests/ look nice.
>>
>
> Yep, I like Philip's tests as well, but the input format and the framework
> to generate them is.. rather ugly :-)
>
> Shouldn't be too hard to come up with something similar, though.  Perhaps a
> standardized test template, and then a driver that loads each test in a new
> iframe, one after the other?

This sounds like a good and hopefully simple solution. Is anyone
available to build a prototype?

-Ken

-----------------------------------------------------------
You are currently subscribe to public_webgl@khronos.org.
To unsubscribe, send an email to majordomo@khronos.org with
the following command in the body of your email: