865997 - Don't prune complex subtrees for roles usually having no children

Status: RESOLVED FIXED

(Core :: Disability Access APIs, defect)

Description

[alexander :surkov (:asurkov)]

Attachment: patch

Make nsAccUtils::MustPrune smarter (it makes a part of bug Bug 249292 not necessary: http://hg.mozilla.org/mozilla-central/diff/217edac6b29f/accessible/src/base/nsAccUtils.cpp): make it to look into subtree to decide.

The reason is we were reported that we don't expose children in the following case:

<ul aria-label="Account Menu" role="menu navigation" id="userNavigation" class="navigation">
  <li role="menuitem" aria-selected="true">
    <form id="logout_form">
      <label><input type="submit" value="Log Out"></label>
    </form>
  </li>
</ul>

Comment 1

[Trevor Saunders (:tbsaunde)]

Comment on attachment 742220 [details] [diff] [review]
patch

seems ok, though adding the virtual call sucks, use ContentChildAt() maybe, or maybe I should just get around to devirtualizing GetChildAt()

btw have we  considered what it would take to just not create text leaves in these cases?

Comment 2

[alexander :surkov (:asurkov)]

(In reply to Trevor Saunders (:tbsaunde) from comment #1)
> Comment on attachment 742220 [details] [diff] [review]
> patch
> 
> seems ok, though adding the virtual call sucks, use ContentChildAt() maybe,

that should be working.

> or maybe I should just get around to devirtualizing GetChildAt()

ideas/plans? Should I care about ContentChildAt() stuff here?

> btw have we  considered what it would take to just not create text leaves in
> these cases?

I guess name computation and text ifrace are broken

Comment 3

[Trevor Saunders (:tbsaunde)]

(In reply to alexander :surkov from comment #2)
> (In reply to Trevor Saunders (:tbsaunde) from comment #1)
> > Comment on attachment 742220 [details] [diff] [review]
> > patch
> > 
> > seems ok, though adding the virtual call sucks, use ContentChildAt() maybe,
> 
> that should be working.

not sure what you mean

> > or maybe I should just get around to devirtualizing GetChildAt()
> 
> ideas/plans? Should I care about ContentChildAt() stuff here?

I think we can just special case tree stuff, but probably won't get to it particularly soon.

> > btw have we  considered what it would take to just not create text leaves in
> > these cases?
> 
> I guess name computation and text ifrace are broken

sounds right

Comment 4

[alexander :surkov (:asurkov)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/898b2c5eaa5d

Comment 5

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/898b2c5eaa5d