RE: Run test setup button on Test 1 for radio button touching index changes content

Hi James,

Thankx for the response.  My big question is, can I more forward with 
testing, logging my results as is?

 I was trying to get this testing done before end of year to help this 
group meet its goals.


I'm happy to keep testing or wait until a revision.  I have time this week 
to test and possibly even early next week as well. So just let me know 😊

Thankx,
  Joe Humbert

-----Original Message-----
From: James Scholes <james@pac.bz>
Sent: Tuesday, December 17, 2024 3:14 PM
To: Joe Humbert A11y (Android) <joe@a11yeval.com>; public-aria-at@w3.org
Cc: isa.delcastillo5@gmail.com
Subject: Re: Run test setup button on Test 1 for radio button touching index
changes content

Hi all,

Thanks, Joe, for filing this feedback.

To provide some context: The test pages for both radio groups (roving
tabindex and active descendant) have a "Pizza Crust" heading which is used
as the group label.  Specifically, the heading is the first child element
inside the group, and the ID of the heading is used as the value of
aria-labelledby on the parent div with role="group".

To avoid screen reader users hearing duplicate elements/information, and to
test a screen reader's pure ability to convey the group name as the group
name rather than as a heading, the relevant setup scripts have always set
the heading to hidden via display:none.  This avoids a workflow like the
following example:

1. A tester lands on a link before the radio group.
2. They move forwards on the page, let's say by two elements, and encounter
the heading followed by the first radio.
3. They mark the group name assertion as passing, only because they heard
some text as a heading rather than actually as the group name.

To phrase it slightly differently: If a screen reader fails to convey the
group name when the user doesn't explicitly encounter it as text in their
virtual cursor, that should be considered a failure.

The complication here is that we're hiding the target element of an
aria-labelledby, and as such, we're actually testing the wrong thing: Does a
screen reader honour aria-labelledby when the referenced element is hidden?

It might be better to find a way to place the user's focus outside of the
group, but after the heading, and then not hide the heading, but that's
currently complicated by the heading behind positioned inside the group.

Hopefully this helps to set the scene.  Happily to discuss further in the
new year, or by text.

Regards,

James Scholes (he/him)
Director of Digital Accessibility, Prime Access Consulting, Inc.
https://www.pac.bz/

On 16/12/2024 at 19:48, Joe Humbert A11y (Android) wrote:
> Hi All,
>
> I filled a GitHub issue https://github.com/w3c/aria-at/issues/1170
> <https://github.com/w3c/aria-at/issues/1170>
>
> Pressing the "Run Test Setup" button for test 1 causes the <h3> to have
> display none.  This makes accurately running the test extremely difficult.
>
> I have stopped running any more tests because I do not want to have to
> rerun all tests a second time.
>
> There were also issues with VO bot out (extra spaces and no output when
> there was output) so I have filled two more GitHub issues (1171 and 1172)
>
> Please let me know when this is fixed so I can continue testing.
>
> Happy to help test other patterns while this gets fixed.
>
> Thanks,
> Joe Humbert