whitespace-only alt attribute treated differently

[giacomo-petri]

In the Clarify that a whitespace-only alt attribute should be considered as the empty string PR, we clarified that alt=" " should be treated the same as alt="".

The WPT test linked in in that PR confirms that the accessible name is handled consistently, but we currently do not have a test covering the computed role. In practice, implementations are inconsistent when handling images with whitespace-only alt values.

Below are the results from a WPT test I created locally:

• Safari: Action get_computed_role completed with result none (consistent)
• Chrome: Action get_computed_role completed with result image (inconsistent)
• Firefox: Action get_computed_role completed with result image (inconsistent)

I believe this update should also have prompted implementors to verify role computation behavior.

@cookiecrook, if there are no objections, I can take care of creating a WPT issue in Accessibility Interop and submitting a PR to add my local test to html-aam/roles.html (or should I create a tentative because of inconsistencies)?

Note: @scottaohara, do you think we should also clarify in ARIA in HTML that:

If the img has an empty alt (alt="") and lacks any other img naming methods:
role=none, role=presentation

also applies when the alt value contains only whitespace characters?

[scottaohara]

could we just put in those tests into this PR web-platform-tests/wpt#52480 (it's still yet to be merged, lol!)

so it can be a test for

<img src=... alt=" ">
<img srcset=... alt=" ">
<img src=... srcset=... alt=" ">

and identify what the role should be for those cases so there is no potential divergence

[rahimabdi]

Thank you @giacomo-petri for raising the WPTs!

[giacomo-petri]

could we just put in those tests into this PR web-platform-tests/wpt#52480 (it's still yet to be merged, lol!)

so it can be a test for

<img src=... alt=" ">
<img srcset=... alt=" ">
<img src=... srcset=... alt=" ">

and identify what the role should be for those cases so there is no potential divergence

I didn't see it. Absolutely! I'll add the examples

@scottaohara @rahimabdi, can we assume that
<img src=... alt=" ">
and
<img src=... alt=" " title="test">
are already treated equally thanks to the other tests?

Or do we need to differentiate them as well?

[scottaohara]

if we don't explicitly have those written, i don't see the harm in adding them to make sure things work 'as expected'