193803 - Fix ROLE_PANE events and object hierarchy

Status: RESOLVED FIXED

(Core :: Disability Access APIs, defect)

Description

[Aaron Leventhal]

When we have an iframe, browser or frame a lot of work is done to create only
one accessible object. It is more natural to have 2, one for the outer <iframe>,
<browser> or <frame> tag, and one for the inner document root.

In fact, Internet Explorer uses 2 objects - ROLE_CLIENT on the outside and
ROLE_PANE on the inside. Also, having only 1 object is causing problems when an
event occurs on the object that we're trying to remove.

Therefore, there is no reason at all to have only 1 accessible object there.

Comment 1

[Aaron Leventhal]

Attachment: Use natural outer/inner hierarchy, don't override accessible navigation

Seeking r=kyle

Comment 2

[Kyle Yuan]

Comment on attachment 114751 [details] [diff] [review]
Use natural outer/inner hierarchy, don't override accessible navigation

I prefer changes to the class name as well, because the current names seem
weird - nsHTMLIFrameRootAccessible is a child of nsHTMLIFrameAccessible. We
could use nsHTMLIFrameOutertAccessible/nsHTMLIFrameInnerAccessible instead.

Comment 3

[Kyle Yuan]

the file name should be changed to nsHTMLIFrameAccessible.cpp if you changed the
class names.

Comment 4

[Aaron Leventhal]

Attachment: Same patch, fixed some whitespace problems

Kyle, do you mind if we fix the naming in another patch? Right now I have a lot
of frame/doc related work to do, we can get to the renaming another time, if
that's okay.

Comment 5

[Kyle Yuan]

Comment on attachment 114877 [details] [diff] [review]
Same patch, fixed some whitespace problems

r=kyle, that's okay with me.

Comment 6

[Henry Jia]

Comment on attachment 114877 [details] [diff] [review]
Same patch, fixed some whitespace problems

1. There should be indent in the second line of the below.
+	 mListener->HandleEvent(nsIAccessibleEventListener::EVENT_REORDER, 
+	 parentAccessible? parentAccessible:this, nsnull);
Also, there should not be a space after the question mark in
"parentAccessible? parentAccessible:this"

2. Pay attention to the indent of
+  *aNumChildren = 1;
+	return NS_OK;

sr=Henry except the above.

Comment 7

[Aaron Leventhal]

checked in

Comment 8

[Firefox Bug Husbandry Bot]

https://bugzilla.mozilla.org/show_bug.cgi?id=1527381