153196 - Active Accessibility: support OBJID_CARET and EVENT_OBJECT_LOCATION change event for caret

Status: VERIFIED FIXED

(Core :: Disability Access APIs, defect, P2)

Description

[Aaron Leventhal]

In nsWindow.cpp, when we receive a WM_GETOBJECT message with lParam ==
OBJID_CLIENT, we need to give back an IAccessible for the caret, which reports
ROLE_CARET and the x,y,width and height of the caret and state of 0 (normal).

We also need to become a selection listener and fire an MSAA
EVENT_OBJECT_LOCATION_CHANGE whenever the caret moves.

By supporting caret in this way, we will be compatible with Microsoft Narrator,
which will be free access, although not a true screen reader.

Comment 1

[Aaron Leventhal]

Kyle, this is very similar to what you need to do with selection. How should we
work together on this? Do you have a bug open for your selection event work?

Comment 2

[Kyle Yuan]

Yes, this would definitely be same as our desire. 
We haven't filed other ATK event support except bug 152780 (for table event)and
bug 152786 (the framework).

Comment 3

[Aaron Leventhal]

Attachment: Create nsCaretAccessible which implements nsIAccessibleCaret. There is only 1 caret per window, it corresponds to the currently focused selection.

I hope this somewhat fits in with what Sun is doing with selections.

The caret is really the currently focused selection. We could expand
nsCaretAccessible to become more of an nsFocusedSelectionAccessible. Even if we
don't change the name, it could support nsIAccessibleSelectable. I don't know
if that's useful to you, but at least this shows how the caret events can be
fired.

Seeking r=

Comment 4

[Kyle Yuan]

Thanks, Aaron. It's a great job!
I'll look at this deeply next monday.

Comment 5

[Aaron Leventhal]

I don't have a clear idea how this will fit into Sun's plans for selection and
caret. The caret is really the currently focused selection -- it is a caret only
when the selection is "collapsed". Do you have any ideas?

Also, I noticed we haven't removed nsIAccessibleSelection yet. I thought we
don't need this any more.

Comment 6

[Kyle Yuan]

We have 3 events for editable text: 
1) text_selection_chage - we can catch it from nsIDOMFormListener::Select (or 
probably nsIDOMFormListener::Change, I forgot).
2) text_input - we can catch it from nsIDOMFormListener::Input.
3) text_caret_move - it's vital to ATK, we hope this fix will implement that 
for us.

It's not very clear to us that how nsIAccessibleText and 
nsIAccessibleEditableText can get benefit from nsIAccessibleCaret. I'm going to 
find the answer.

nsIAccessibleSelection has been removed yet.

Comment 7

[Kyle Yuan]

Comment on attachment 88728 [details] [diff] [review]
Create nsCaretAccessible which implements nsIAccessibleCaret. There is only 1 caret per window, it corresponds to the currently focused selection.

r=kyle

I applied the patch and tested. It works perfect! Just one thing I noticed that
we have different behavior from IE when moving caret by hold down shift key --
We send caret move event at that time, but IE does not. I don't think it's a
matter. Or, we can fix this by checking isCollapsed in
nsCaretAccessible::NotifySelectionChanged. You decide.

Comment 8

[Aaron Leventhal]

Attachment: Found some leaks. Fixed it with a few changes to the handling of WM_GETOBJECT in nsWindow.cpp.

Seeking a new r= from Kyle.
Also seeking r= from Mike Judge to the changes in nsCaret.cpp

Comment 9

[Aaron Leventhal]

Mike, could I get your r= for the changes to nsCaret.cpp?
nsCaret::GetCaretCoordinates was returning the width of the frame for the width
of the caret - I believe I have added the correct fix, by using mCaretTwipsWidth
instead.

Comment 10

[Kyle Yuan]

Comment on attachment 88909 [details] [diff] [review]
Found some leaks. Fixed it with a few changes to the handling of WM_GETOBJECT in nsWindow.cpp.

r=kyle

There are some nsXULColorPickerAccessible stuff in this patch. Is that for
another bug?

Comment 11

[Aaron Leventhal]

Yeah, I accidentally left that in. You can ignore the nsXULColorPicker stuff.

Comment 12

[Aaron Leventhal]

I would say keep the #define if it keeps the code nice and readable, and inline
won't serve your needs. However, I will defer to what Scott Collins says.

Comment 13

[mjudge]

r=mjudge on first part with caret code.

Comment 14

[Aaron Leventhal]

Frank, I would like to check to make sure this patch does not negatively affect
IME input. It fixes the width reported. GetCaretCoordinates(), and we're not
sure if IME relies on the old incorrect value for something.

Comment 15

[kinmoz]

Comment on attachment 88909 [details] [diff] [review]
Found some leaks. Fixed it with a few changes to the handling of WM_GETOBJECT in nsWindow.cpp.

sr=[e-mail]

Comment 16

[Aaron Leventhal]

checked in

Comment 17

[dsirnapalli]

-- Marking Verified.