[Kde-accessibility] A possible bug in QT At-spi bridge the tree role handling

[Frederik Gladhorn]

Hi Attila,

On 04/05/2012 02:25 PM, Hammer Attila wrote:
> Hy,
>
> I experienced an interesting issue with Skype and Qt At-spi related in
> Ubuntu 12.04 when I used 3.4 and master Orca screen reader version.
> Hopefuly I describe right place the following issue:
> If I logged in my Skype account, the contact list tree is working
> right. Orca get QT At-spi the real selected contact names.
> Unfortunately when my logged out contact friend logged in again in
> Skype, this is not happening. If I moving contacts list in up arrow
> key to search my friend contact name with I would like talking, I not
> found the list.
> I see only two "echo/soundtest service" contact name. The first this
> named contact is real the test call contact, but the top of list item
> is not, this contact my logged in friend contact name. I verifyed this
> with following test:
> 1. With up arrow key select the first "echo/soundtest service" contact
> name, press application button my keyboard, and choose rename menu
> item. The edit box show first the real "echo/sound test service"
> contact name.
> 2. Press up arrow to jump the second mistake named new logged in
> contact, and repeat the test. When I press up arrow key, Orca spokened
> me the "echo/soundtest service", this text present my braille display.
> But when I choosed the rename menu item, the edit box right presenting
> my friend real contact name.
>
> If I restarted with Orca Screen Reader, all works right again.
> In Ubuntu 12.04 the qt-at-spi 0.2.0-0ubuntu1 packaged version awailable.
> I no, Skype is a binary softvare, but possible doing any fix this
> problem related with qt-at-spi bridge?
This is a know and bad issue in qt-at-spi. The problem is that the Gnome
stack caches every accessible to the last detail.
So currently Orca and the at-spi2 libs keep track of every information
they ever get. Even marking trees as "manages descendants" which should
prevent this behavior does not seem to help.
Now the qt-at-spi bridge is not updating the tree in the way expected
since that is quite tricky and I'm still not sure how to do it properly
in an elegant way.
This leads to the Gnome side thinking that the tree is in some state it
had before. That is why restarting Orca helps.
I am convinced that sending updates for trees all the time to enable
proper caching is not a good idea and one reason why the performance of
at-spi 2 is frowned upon. That is why I refuse to even try to implement
it the same way as atk does. I don't have a good solution at the moment
though.

Frederik

>
> I reported this issue for Orca in Bugzilla with following report:
> https://bugzilla.gnome.org/show_bug.cgi?id=673556
>
> Attila
> _______________________________________________
> kde-accessibility mailing list
> kde-accessibility at kde.org
> https://mail.kde.org/mailman/listinfo/kde-accessibility