[Webkit-unassigned] [Bug 139582] AX: Anonymous RenderMathMLOperators are not exposed to the accessibility tree

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=139582

Joanmarie Diggs (irc: joanie) <jdiggs at igalia.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #281884|review?                     |
              Flags|                            |

--- Comment #7 from Joanmarie Diggs (irc: joanie) <jdiggs at igalia.com> ---
Comment on attachment 281884
  --> https://bugs.webkit.org/attachment.cgi?id=281884
Alternative patch based on Frédéric's

(In reply to comment #5)
> Comment on attachment 281884 [details]
> Alternative patch based on Frédéric's
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=281884&action=review
> 
> That looks good to me, especially the option to output subroles ;-) I have
> some doubts on isIgnoredElementWithinMathTree though, see inline comments.
> 
> > Source/WebCore/accessibility/AccessibilityObject.h:998
> > +    virtual bool isAnonymousMathOperator() const { return false; }
> 
> I'm not sure that this function is needed outside AccessibilityRenderObject.

It's not *yet*. I anticipate there may be a need for platforms to be able to check that for something with MathElementRole. But noted.

> 
> > Source/WebCore/accessibility/AccessibilityRenderObject.cpp:3785
> > +    return ancestorsOfType<RenderMathMLMath>(*m_renderer).first();
> 
> So IIUC you are now excluding all the non-MathML element inside a <math>
> tag, which I suspect would break accessibility in some cases, for example:

Ugh. As we discussed via chat, I made the clearly bogus assumption that we'd have MathML in MathML. So, yeah.... That will need to change.

> I'm also not sure about the performance of ancestorsOfType or of the old for
> loop.

I think that getting (and potentially creating) accessible objects just to call isMathElement() on ancestors is ideal. So perhaps doing something like your for loop, but working with the RenderObject instances rather than AccessibilityObject instances.

Anyway, based on the clearly bogus oops I made, I'm going to clear the r? flag.

Thanks for the review!

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20160623/7b926458/attachment-0001.html>