[Accessibility] Re: [Accessibility-atspi] Re: a11y API/ABI Testing

[David Bolter]

George, thanks very much for your excellent answers and points.

Janina,

This is starting to sound like a Mozilla proposal.  We just need to 
figure out who best can do the work.

cheers,
David

George Kraft wrote:
> On Thu, 2007-11-15 at 21:46 -0500, David Bolter wrote:
>   
>> I agree Accerciser and Macaroon should be noted if we are to mention 
>> Dogtail and LDTP, but I'm not sure we want to mention any of them on the 
>> wiki page in question. It depends on our scope. I've done a little 
>> research and thought about what you said on the call and I think your 
>> approach of testing via mock application with custom objects at one end, 
>> atk/gail/atspi in the middle, and mock AT at the other is reasonable 
>> (what I mean by testing across 'layers'). Do you still agree with this 
>> approach or has it hit any snags? Do you see advantages to swapping the 
>> existing mock AT with other similar tools?
>>     
>
> I think it is a mixture of straight API testing (221/222) and API
> testing using mock ATs (9/222).  The questions are:
>
> 1) are the 221 new straight API tests acceptable?
> 2) are the 9 prototyped mock AT tests acceptable?
>
>   
>> Where are the tests run?  Is there a cron job somewhere that checks out 
>> of gnome svn, builds, runs, and reports?
>>     
>
> I don't know of any build/test systems running these tests, nor any
> posted results.
>
>   
>> If migration from the at-spi-registryd daemon approach to a dbus 
>> implementation happens, we can hopefully catch a lot of problems but 
>> sticking it between our mock apps and mock ATs which will be very handy 
>> of course.
>>     
>
> Agreed.
>
>   
>> Maybe my question is really: is there anything left to do?  Perhaps just 
>> community building and evangelism?
>>     
>
> Evaluate the above mentioned ATK tests and build a consensus of
> confidence in them.  Fix the ATK bugs found by the LSB's new ATK tests.
>
> http://linuxtesting.org/results/impl_reports?action=component_bug&name=gtk-atk
>
> Perhaps start working on an AT-SPI test suite that would definitely use
> mock AT tests...  This would require a spec review for sufficient
> details.
>
>