[lsb-discuss] ATK conformance testing

Stew Benedict stewb at aysenterprises.com
Wed May 2 04:27:05 PDT 2007

On Tue, 1 May 2007, George Kraft IV wrote:

> Stew,
> Currently, the LSB Desktop specifies "libATK" and does a little testing
> in "lib-gtkvst"; however, it appears that those tests don't do
> much.  :-)
> http://refspecs.freestandards.org/LSB_3.1.0/LSB-Desktop-generic/LSB-Desktop-generic/libatk-1.0.html
> http://bzr.linux-foundation.org/lsb/3.1/desktop-test?cmd=inventory;rev=pqm%40freestandards.org-20070405180010-5eikl96uu0cfaggx;path=lsb-gtkvts/src/tests/functions/Atk/
> The prototype ATK Conformance Tests that I have proposed exercise the
> entire a11y ecosystem by having mock ATs (assistive technologies) query
> test applications which are using libATK.
> http://lists.freestandards.org/pipermail/accessibility-atspi/2007-March/000390.html
> http://developer.gnome.org/projects/gap/tech-docs/SPIBlockDiagram.png
> http://live.gnome.org/GAP/AtkGuide/Gtk?action=AttachFile&do=get&target=ATKObjectArchitecture.jpg
> Yes, this testing technique drags in the at-spi-registryd daemon and
> libspi; however, those are necessary to make accessibility work in the
> real world.  Think of it like needing TCP/IP and an X-server to test
> libX11.
> I would like to discuss this testing technique with the LSB desktop team
> during the conference call this Wednesday.
> PS: This is only for the "LSB Desktop" that covers Gtk applications.
> LSB core for servers would not require desktop accessibility.

OK, thanks, the diagrams clarify things a bit. 

I'm still struggling with the concept of exercising a set of interfaces
that aren't part of the LSB spec. If a test fails, where does the 
distribution look to fix the problem?  A different libspi? A different 
GNOME stack?

Putting it in the context of many of the other tests, it's fairly clear 
what LSB specified interface is being tested, and what assertion fails 
when the test fails, which usually leads one to either an issue with the 
test itself of a change in the library behavior. In the atktests, it's not 
as clear to me where I would look in the case of a failure.

True, we do have a somewhat similar issue with the X tests, where one 
needs a "special" Xvfb build to be able to get good test results. We hope 
to address this by providing a lsb-Xvfb that the tests will use in the 

Stew Benedict

More information about the lsb-discuss mailing list