Re: GNOME's testing strategy for GUIs

[Willie Walker <William Walker Sun COM>]

David Bolter wrote:
> To my mind, adding regression tests is like spending 1 hour to save 100. 

Definitely.  :-)  It also gives me piece of mind when a refactor is 
done, and it never forgets those special cases that we sometimes forget 
about when doing manual testing.  You can also build them up over time 
as you fix bugs.

> Perhaps what we need is a regression test 
> evangelist/documentation/mentor? Will, I really appreciated your phone 
> call last year answering my questions about the Orca test harness. It 
> sounds like a really great setup, although I'm sure there are other 
> great options.

I'm sure there are, too.  I'm serious when I say my goal here is not to 
push what we've done in Orca.  It's very specific to the Orca project 
and was never meant to be generalized.  We just needed *something* and 
the things available at the time didn't really meet our unique needs.

> I'd like to see documentation on 'GNOME recommended automated testing' 
> for all the kinds of projects we see in GNOME (including for the various 
> languages). I think this thread is a great way to try and get community 
> consensus and to collect information on what various projects use. I 
> suspect a lot of projects use none or very little (sadly, including GOK).
>
> IMHO this needs to change.

Agreed.  We tried to get some momentum back at GNOME Boston 2006 
(http://live.gnome.org/TestingUsingAtSpi), but we never really gained 
traction.  It may that the community wasn't ready then, but it might be 
ready now.

Will