Thanks for bringing this to our attention.  Please point to the relevant HIG docs to tell us what to do correctly.  Thanks!

(In reply to comment #1)
> Thanks for bringing this to our attention.  Please point to the relevant HIG
> docs to tell us what to do correctly.  Thanks!

Eeks!  I didn't mean to sound snarky on this.  I really want to know the right thing to do and I really do appreciate this being brought to our attention.

Adding Calum, our HCI expert, to the CC.

I'm taking a look at them now.  One quick observation: the Quit dialog probably ought not to be in  a Glade file, you should just call GtkMessageDialog at runtime to get all the right spacing etc.  However, I'll update the Glade file anyway to show you what a HIG-compliant Quit alert should look like :)

Created attachment 84058 [details]
Updated glade file for setup window

Updated setup dialog.  I've tried to tidy up the spacing and resizing behaviour, and ditch as many GtkAlignments as I could find that weren't really needed.  I didn't really have time to tackle the Magnifier tab quite as much as I'd have liked (it's a bit of a monster!), so I might have missed one or two things there.

You'll probably need to check that everything still works, as Glade has a habit of not preserving widget names as well as it should when you start hacking things around.

This dialog probably still needs some more general work really, e.g. you should never really need to use the word "enable" in a checkbox label, but every checkbox on the Key Echo tab does just that :)  And the resize behaviour on the Magnifier tab is a bit odd.

Created attachment 84059 [details]
Updated glade file for quit alert

As noted above, this shouldn't really need to be in a glade file at all, being acheivable with a single gtk call, but it should look something like this.  (Feel free to amend the secondary text as necessary, I didn't know exactly what the consequences of quitting Orca would be...)

Created attachment 84060 [details]
Updated glade file for find dialog

Just a very minor spacing tidy-up here.

Calum,

I just had a quick look at your new Orca quit Glade file.
I noticed you modified it from a top-level window to a
popup. You also suggest using GtkMessageDialog instead.

I don't think either of these approaches will work because
Orca needs to speak/braille it's own quit dialog, and
adjusting it to be a popup dialog or to use GtkMessageDialog
seems to prevent that.

This seems to need special casing because it's a screen reader.

Would in be okay to leave it as a top-level window in a Glade
file but just adjust to use the other changes you made?

Thanks.

PS: Haven't looked at the other two Glade files yet).

> I just had a quick look at your new Orca quit Glade file.
> I noticed you modified it from a top-level window to a
> popup.

Oops, didn't actually mean to do that, I remember toying with it but I thought I'd set it back.

> I don't think either of these approaches will work because
> Orca needs to speak/braille it's own quit dialog, and
> adjusting it to be a popup dialog or to use GtkMessageDialog
> seems to prevent that.

Hmm, worth filing a bug against GtkMessageDialog perhaps?  (I presume *all* GtkMessageDialogs aren't similarly invisible to Orca?  If so, we've got a lot of broken message dialogs on our desktop that need fixing...)

Anyway, fine by me to do whatever works for now, and rattle any appropriate cages further down the stack if need be.

Created attachment 85523 [details]
New version of the setup glade file reinstating numerous missing labelled-by and label-for relationships.

I've committed the three new Glade files. Please try them out and comment back.
Putting the bug into "[pending]" state before closing out as FIXED.

The only thing I'm noticing is that in the Quit dialog, Orca tends to speak the Cancel button twice when the dialog first appears:

1. Launch Orca
2. Press Insert Q

Interesting. I'm not seeing (hearing) that at all. Could others
please try it and let me know what they get. Joanie, could you
create a full debug and attach to the bug report please? I gotta
guess it's getting a couple of focus: event or something...
Thanks.

Created attachment 85538 [details]
debug.out requested by Rich

Here you go!  And, yeah, it looks like I'm getting one object:state-changed:focused event and one focus: event.  Seems like everyone is telling me to focus these days.... ;-)

Created attachment 85544 [details]
Another full debug.out for quitting Orca

Here's my equivalent. I too get the two kinds of "focus" events:

object:state-changed:focused  -- which speaks the "cancel button"
focus:                        -- which doesn't.

Hmmm. I wonder what the difference is. I'll investigate some more tomorrow.

I've had a look at the two debug.out files.

With mine, I get an "object:state-changed:focused" event for the
Cancel button on the Orca Quit dialog. This is almost immediately
followed by a "focus:" event for the same. orca_state.locusOfFocus
is already set to the new event.source, so the orca.setLocusOfFocus()
routine just returns without speaking/brailling anything new.

Joanie, with yours, there are several "object:state-changed:focused"
events for gnome-terminal in between the the "object:state-changed:focused"
and the "focus:" events for the Cancel button. When orca.setLocusOfFocus()
is called for the "focus:" event, it would appear that the focus had changed
in the meantime, and so Orca spoke/brailled the Cancel button again.

To confirm this, can you try quitting Orca on a quiescent system and see
what happens? 

Thanks.

Created attachment 85573 [details]
debug.out when nothing was running other than Orca

Hi Rich.  I quit all applications, launched Orca from the Applications menu, and then quit it.  I still get two "Cancel button"s.

Joanie added some debug statements I sent her to orca.setLocusOfFocus(),
and got the following:

--

Seems my Cancel button has multiple personality disorder.... (After running your patch, I decided to add name into the mix):

orca.setLocusOfFocus entered.
orca.setLocusOfFocus: obj:  <orca.atspi.Accessible instance at 0xb6041b8c>
orca.setLocusOfFocus: obj role:  push button
orca.setLocusOfFocus: obj name:  Cancel
orca.setLocusOfFocus: orca_state.locusOfFocus:  <orca.atspi.Accessible instance at 0xb60bb80c>
orca.setLocusOfFocus: orca_state.locusOfFocus role:  push button
orca.setLocusOfFocus: orca_state.locusOfFocus name:  Cancel
default: onFocus: calling orca.setLocusOfFocus for newFocus: <orca.atspi.Accessible instance at 0xb60bb98c> 

--

I'm totally baffled why there should be two different instances of the
same Cancel button.

Still looking.

Perhaps I'm having a senior moment, but didn't the shortcuts for the combo boxes in Orca Preferences -> Speech used to work?

I don't recall the combobox mnemonics ever working.  :-(

Created attachment 85740 [details] [review]
patch to add combo box and slider mnemonics

I assume the attached is cool from an HIG point of view, but since what I know about HIG could happily line dance with 10,000 of its closest friends on the head of a pin....

Patch committed. Thanks Joanie.