201180 - mozilla's AtkUtil implementation doesn't support atk_add_key_event_listener

Status: RESOLVED FIXED

(Core :: Disability Access APIs, defect)

Description

[Bill Haneman]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.1) Gecko/20021003
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.1) Gecko/20021003

If I add a "toolkit key event listener" via a call to
atk_add_key_event_listener(), I should be getting callbacks from the ATK 
implementing toolkit (in this case, Mozilla) for every keypress received
by the toolkit/application.

In Mozilla I don't get any event callbacks.  The "atk-bridge" registers 
such a listener so that events received by the toolkit can be 
intercepted and (potentially) consumed.  See bugzilla.gnome.org bug
100850 for more information.

Reproducible: Always

Steps to Reproduce:
1.obtain a running gnome-2.2 stack with accessibility "turned on"
2.run the 'key-listener-test' client or run gnopernicus with mozilla 
3.focus mozilla, note that gnopernicus or another client of at-spi
which is listening for key events at the 'toolkit level' won't get 
key notifications.

Actual Results:  
expected key events were not received by the AT client via at-spi 
and the atk-bridge.

Expected Results:  
Mozilla should implement the atk_key_event_listener API in its
implementation of AtkUtil, and send key notifications via the
callback mechanism there.  If the callback consumes the event,
Mozilla should remove that event from the key event queue/stream
which would go onwards to the Mozilla component.

Comment 1

[Aaron Leventhal]

-> Sun

Comment 2

[Kyle Yuan]

=> Bolian

Comment 3

[Bolian Yin]

I'll come to this after the atk accessibility reorg bugs (see 202388)

Comment 4

[Bill Haneman]

202388 doesn't seem to be the right number for the "accessibility reorg" bug
what is the right number?

Comment 5

[Bolian Yin]

Sorry. The right bug for accessibility reorg is: bug 199060 and bug 202085 (for atk)

Comment 6

[Bolian Yin]

Attachment: patch

Comment 7

[Kyle Yuan]

Comment on attachment 123112 [details] [diff] [review]
patch

Bolian, I can't see how your patch listens to mozilla's keypress event and
passes it to AT apps. Can you give me more details?

Comment 8

[Bolian Yin]

this patch implement two atk "virtual functions", like this:
+    atk_class->add_key_event_listener = mai_util_add_key_event_listener;
+    atk_class->remove_key_event_listener = mai_util_remove_key_event_listener;

these two functions are used for AT apps to add/remove key listeners

The main point in mai_util_add_key_event_listener is:
+        key_snooper_id = gtk_key_snooper_install(mai_key_snooper, NULL);
that line add mai_key_snooper as a listener for gdk key press event. So when
there is key press event, "mai_key_snooper" get called in gtk main loop before
that event pass to Mozilla. "mai_key_snooper" will in turn translate and pass
that event to AT app listeners.

Comment 9

[Kyle Yuan]

Comment on attachment 123112 [details] [diff] [review]
patch

Okay, that's fine :) r=kyle

Comment 10

[Bolian Yin]

checked in trunk.