Bug 252964 – AX: Make m_deferredFocusedNodeChange a single item rather than a list to make AXObjectCache::handleFocusedUIElementChanged easier to reason about

• Component: Accessibility
• Status: RESOLVED FIXED

[Tyler Wilcock]

Make m_deferredFocusedNodeChange a single item rather than a list to make AXObjectCache::handleFocusedUIElementChanged easier to reason about

[Radar WebKit Bug Importer]

<rdar://problem/105950128>

[Tyler Wilcock]

Created attachment 465186 [details]
Patch

[chris fleizach]

Comment on attachment 465186 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=465186&action=review

> Source/WebCore/accessibility/AXObjectCache.cpp:1380
> +            m_deferredFocusedNodeChange->second = newNode;

Do we need to handle case where 

Old = a
New = b

Then 2nd
Old = b
New = a

Do we avoid any work here?

[Tyler Wilcock]

(In reply to chris fleizach from comment #3)
> Comment on attachment 465186 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=465186&action=review
> 
> > Source/WebCore/accessibility/AXObjectCache.cpp:1380
> > +            m_deferredFocusedNodeChange->second = newNode;
> 
> Do we need to handle case where 
> 
> Old = a
> New = b
> 
> Then 2nd
> Old = b
> New = a
> 
> Do we avoid any work here?
Ah, good point. I would interpret that sequence as no focus change happening (because we never were actually able to commit `b` as the new focused element before `a` regained focus). I'll update the patch.

[Tyler Wilcock]

Created attachment 465188 [details]
Patch

[Tyler Wilcock]

Created attachment 465189 [details]
Patch

[Tyler Wilcock]

Created attachment 465193 [details]
Patch

[Andres Gonzalez]

(In reply to Tyler Wilcock from comment #7)
> Created attachment 465193 [details]
> Patch

--- a/Source/WebCore/accessibility/AXObjectCache.cpp
+++ b/Source/WebCore/accessibility/AXObjectCache.cpp

         // Recompute isIgnored after a focus change in case that altered visibility.
-        recomputeIsIgnored(deferredFocusedChangeContext.first);
-        recomputeIsIgnored(deferredFocusedChangeContext.second);
+        recomputeIsIgnored(m_deferredFocusedNodeChange->first.get());
+        recomputeIsIgnored(m_deferredFocusedNodeChange->second.get());

in the case of a -> b -> c, in rapid submission, you end up with (a, c), so you never do above for b. Would that have any side effect?

--- a/LayoutTests/accessibility/ios-simulator/focus-change-notifications.html
+++ b/LayoutTests/accessibility/ios-simulator/focus-change-notifications.html

+var rootElement = 0;

+var rootElement = null;

While you're at this test, can we modernize it?

[Tyler Wilcock]

Created attachment 465222 [details]
Patch

[Andres Gonzalez]

(In reply to Tyler Wilcock from comment #9)
> Created attachment 465222 [details]
> Patch

--- a/LayoutTests/accessibility/ios-simulator/focus-change-notifications.html
+++ b/LayoutTests/accessibility/ios-simulator/focus-change-notifications.html

+successfullyParsed = true;

I don't think this variable is needed at all.

[Tyler Wilcock]

Created attachment 465224 [details]
Patch

[Tyler Wilcock]

Created attachment 465229 [details]
Patch

[Tyler Wilcock]

Created attachment 465231 [details]
Patch

[Tyler Wilcock]

Created attachment 465232 [details]
Patch

[EWS]

Committed 261037@main (c87b5bf70b93): <https://commits.webkit.org/261037@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 465232 [details].