Re: Describing an application state in ATK, without Gtk

[Alejandro Piñeiro <apinheiro igalia com>]

On 25/06/15 19:16, Luca Saiu wrote:
> Hello Alejandro, and the other people on the list.  You've been kind the
> last time, so I'm taking the liberty of writing you again.
>
> I did make progress, but I'm still not quite there.
>
> Before binding everything to Compiz I'm playing with a trivial SDL
> application consisting in a single window which reacts to keyboard and
> mouse events.  I want to make that accessible, without using any
> toolkit.
>
> I've modified your "dummyatk" code from the at-spi2-atk testsuite (I had
> to get the git version, for I was bitten by
> https://bugzilla.gnome.org/show_bug.cgi?id=751138 ); I implemented the
> Text role in MyAtkComponent or at least most of its methods (by the way,
> I would like to know which are necessary; do I need all the
> non-deprecated ones?),

No. By definition, if they are deprecated, you don't need to implement
them. Take a look to the documentation of the deprecated methods, as
they should explain what you should implement.

>  and initialized the bridge correctly as you
> suggested.  Now accerciser shows my widget tree, and so do the programs
> in at-spi2-examples, thanks to your suggestions.  Good.
>
> The problem is that my application window is still not "seen" as
> accessible. 

On atk/at-spi it is assumed that you would provide a root object for the
accessibility object hierarchy. On ATK you do that by providing a
implementation of AtkUtil. Yes, the name is misleading, seems to suggest
an optional object, but is in fact needed. The tests at at-spi2-atk
provide an example:
https://git.gnome.org/browse/at-spi2-atk/tree/tests/test-application.c#n81

If you want more examples, this is how gtk do that:
https://git.gnome.org/browse/gtk+/tree/gtk/a11y/gtkaccessibilityutil.c#n100

BR

-- 
Alejandro Piñeiro (apinheiro igalia com)