Fix-up of: "Add support for IA2_ROLE_LANDMARK #10110"

[LeonarddeR]

Link to issue number:

Fixes issues introduced by #10110

Summary of the issue:

The following issues arose after merging #10110

1. In braille, the landmark type was no longer shown.
2. In braille, when at the start of the line, the braille cursor was positioned at the landmark role text and not at the actual start of the content. This was caused by landmarks being treated as markers, not as containers
3. The landmark role was announces for regions

Description of how this pull request fixes the issue:

1. The code that calculates the roleText for a landmark is now abstracted in the aria module. It also covers suppressing of the landmark role for regions
2. Landmarks now have presentation type container. In the braille module, an exception is made not to show the end of landmarks, as in speech. There is currently some debate on whether this is appropriate, see Entering and exiting landmarks #10420

Testing performed:

Tested the following example:

data:text/html, <p>paragraph</p><div role="region">Region without label</div><div role="region" aria-label="this is a label">Region with a label</div><div role="main">This is a main landmark</div><article>This is an article</article><p>This is a paragraph, again</p>

I've only been able to do it in Firefox yet.

Known issues with pull request:

We now announce end of article, we don't announce end of landmark. DO we agree this is OK?

Change log entry:

None

[LeonarddeR]

Some additional thoughts coming to mind here.

Regions are very similar to groupings in that we only want them reported if they have a name. The aria module has the following comment:

# Strictly speaking, region isn't a landmark, but it is very similar.

With the grouping support about to be added, I wonder whether we should consider decoupling region reporting from landmarks. It has the following benefits:

1. It follows the ARIA spec more closely. A region just isn't a landmark
2. We don't have to make an exception for regions in our landmark processing code. A landmark is a landmark, so we can always announce the landmark role.

@michaelDCurran If you agree with this approach, I think I prefer merging #10425 into this.

[michaelDCurran]

If it makes the code easier and more understandable for regions to be separate from landmarks, that sounds good to me. However, note that for labeled regions at least, move to next/previous landmark currently finds them.

[LeonarddeR]

Superseded by #10462