[Webkit-unassigned] [Bug 92477] REGRESSION (r123767): platform/gtk/accessibility/object-with-title.html failing on GTK

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=92477


Mario Sanchez Prada <msanchez at igalia.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #154981|0                           |1
        is obsolete|                            |
 Attachment #154981|review?                     |
               Flag|                            |
 Attachment #155132|                            |review?
               Flag|                            |




--- Comment #4 from Mario Sanchez Prada <msanchez at igalia.com>  2012-07-28 08:24:26 PST ---
Created an attachment (id=155132)
 --> (https://bugs.webkit.org/attachment.cgi?id=155132&action=review)
Patch proposal

(In reply to comment #3)
> (From update of attachment 154981 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=154981&action=review
> 
> > Source/WebCore/accessibility/gtk/AccessibilityObjectAtk.cpp:77
> > +    if (!accessibilityDescription().isEmpty())
> 
> i might change this to 
> 
> if (!accessibilityDescription().isEmpty() || !title.isEmpty())
> 
> as the "accessible" name of an object is split between these two concepts, which are modeled on the Mac API
> where title is what you get visually on the screen if available, and axDescription is what you get when the
> description is not really visible (think alt tag on image)

Good point.

Actually, there was another issue with that solution (which was causing some tests to fail) and it's that calling to title() / accessibilityDescription() from this function will cause infinite recursion since accessibilityIsIgnored() will be called again at some point after calling those functions.

So, and considering that helpText() is something that seems to be used only on the Mac, I made a change directly in AccessibilityRenderObject::accessibilityIsIgnored(), so we decide not to ignore an object if helpText is not empty when on the Mac, and when (title || accessibilityDescription) is not empty in other platforms.

This seems to fix the issue indeed, while not causing any regression, at least in the GTK port.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.