Created attachment 8670803 [details] accessible-event listener: showing.py Steps to reproduce: 1. Run the attached accessible-event listener in a terminal 2. Perform a task (see below) triggering an accessible alert/doorhanger(?) Expected results: Only one object:state-changed:showing (true) event emitted. Actual results: Sometimes (see below) there is a duplicate event emitted. Sample tasks: 1. Begin typing in the Location bar causing an alert asking if you want to improve your search experience with suggestions. (Result: 1 event. Good.) 2. View https://people.mozilla.org/~mnoorenberghe/w3c_notifications.htm and press the 'Request permission' button. (Result: 2 events. Bad.) 3. Submit a form with an input element with type="email" in which the value is not an email address, causing an alert asking you to please type an email address. (Result: 1 event. Good.) 4. Submit a form with an input element with type="password", causing an alert asking if you want to remember the password. (Result: 2 events. Bad.)
Created attachment 8670805 [details] test case for examples 3 and 4 from OR: duplicate-events.html Type "foo" in the first input field and press the submit button to get one event as expected. Type some string in the second input field and press the submit button to get two events when only one is desired.
is my understanding correct that this one is a recent regression?
(In reply to alexander :surkov from comment #2) > is my understanding correct that this one is a recent regression? Perhaps ask Jamie? As a reminder: When I was in Toronto last week, we discussed the fact that Windows has an "alert" event which ATK/AT-SPI2 lacks. I suggested that object:state-changed:showing would be the closest alternative I could think of, and you agreed to emit it at the same time you emitted the Windows "alert" event. Thanks!! :) Before (i.e. with stable/shipping Firefox), I got no alert/showing events for the items in the sample tasks. This lack of showing events makes it hard for Orca to properly present them to users. Now I sometimes get one event (perfect) and sometimes two (the bug being reported here). Jamie, on the other hand, presumably has been getting alert events for some time now. Thus my suggestion above that you ask him if duplicate events is a recent regression.
You need to log in before you can comment on or make changes to this bug.