Re: [orca-list] When pressing shortcuts outside of Orca, Orca announces the context again

[Samuel Thibault <sthibault hypra fr>]

Hello Joanmarie,

Joanmarie Diggs, on mer. 28 févr. 2018 15:00:38 +0100, wrote:
> On 02/28/2018 01:47 PM, Alex ARNAUD wrote:
> > Le 28/02/2018 à 12:31, Joanmarie Diggs a écrit :
> > > What's happening is the window is getting deactivated and then
> > > reactivated.
> >
> > It's right but it's the same window that is activated and reactivated
> > why report it to the user?
>
> I agree 100%. But imagine this scenario: You're in an accessible window.
> Something causes an inaccessible window to suddenly become activated.
> Orca doesn't present it because inaccessible windows don't emit
> accessibility events. All Orca knows is that you're no longer in the
> window you were in. You can't figure out what you're in either. But
> clearly something isn't right because Orca is no longer talking.

I have had a look inside gtk, I have submitted a patch for gtk3 to

https://gitlab.gnome.org/GNOME/gtk/merge_requests/76

perhaps you could add your comment there that we really want it to be
fixed?  Currently, orca still announces some context because there are
still frame deactivation/activation and object unfocus/focus, but
perhaps Orca could be made to ignore that case? (since in that case we
do know that we haven't switched to a non-accessible application).

Concerning gtk4, I actually realized that the current master doesn't
emit window:activate/deactivate at all any more, apparently as a side
effect of some code removal which went a bit too far.  I submitted a
patch to

https://gitlab.gnome.org/GNOME/gtk/merge_requests/77

perhaps you can also comment there that these events are really
must-have for accessibility?

Concerning avoiding the window:activate/deactivate events on keyboard
shortcut in gtk4, the gtk3 patch mentioned above won't work with
gtk4 because they removed some signaling there.  What could be
done relatively easily however, would be to at least avoid the
window:deactivate event. There would be only a window:activate spurious
event, which is hard to avoid.  Could that be enough for avoiding Orca
from being talkative? (since it wouldn't get window:deactivate, only a
spurious window:activate)

Samuel