Write tests for SVG-AAM role mapping

[cookiecrook]

We need SVG element computedrole tests for:
https://w3c.github.io/svg-aam/#mapping_role_table

Related spec issues that I plan on filing [have now filed] that do not block (or only block a portion of) the initial testing planned for Interop 2023.

• Define computedrole for the SVG roles table w3c/svg-aam#24
• Does "no role may be applied" mean "UAs MUST NOT apply a role even if an author supplies one and gets the element rendered"? w3c/svg-aam#29
• Should concrete shape/object roles like circle have a more specific role description? w3c/svg-aam#30
• Need tests for Core-AAM #roleMappingComputedRole example 3 (<video> -> 'html-video' ) #70
• Incorporate host-language unique computedrole string for reserved native elements with no direct ARIA equivalent w3c/html-aam#511

[cookiecrook]

Trying to list the currently testable rules here for a v1 set of WPT tests from:

• https://w3c.github.io/svg-aam/#exclude_elements
• https://w3c.github.io/svg-aam/#include_elements

This is my shorthand version, so it may be imprecise, but good enough as a starting point to write the tests.

1. the following should be treated as generics by default: shape elements (circle, ellipse, line, path, polygon, polyline, rect), use, g, image, mesh, text formatting elements (textPath, tspan), and foreignObject
2. All elements (including those above) should be included IF the element is both rendered (many element types won't ever be rendered), and one of the following is true:
   • it has a non-empty direct child[sic] title element or desc element
   • it has a non-empty aria-label or aria-roledescription attr.
   • valid aria-labelledby or aria-describedby
   • valid integer tabindex
   • non-generic role (not generic/none/presentation/etc)
   • other criteria from Core-AAM Including Elements in the Accessibility Tree
   • question/todo: find where text elements (not just formatters) fit into these rules, or file a spec issue
3. Unless… it should not be included, b/c one of the following exclusion rules applies (and so it's not exposed)
   • It's generic/none/presentation/etc (expose the descendant contents but not it)
   • aria-hidden="true" (other hidden types are accounted for with the "rendered" criteria above)
   • ancestor element is Children Presentational: True (such as a button)
4. Unless… it's focusable or becomes focused, then it is exposed.

[cookiecrook]

Some outstanding issues that may need to be cross-referenced in the WPT comments.

• Define computedrole for the SVG roles table w3c/svg-aam#24
• Browsers off spec? SVG maps to "group" but the spec suggests it should map to "graphics-document" w3c/svg-aam#22
• Should the mappings of a nested 'svg' be different from that of a root 'svg'? w3c/svg-aam#18
• SVG-AAM should address which elements should map to ARIA generic role w3c/svg-aam#21
• expectations for image tag with aria-label  w3c/svg-aam#25
• Does "no role may be applied" mean "UAs MUST NOT apply a role even if an author supplies one and gets the element rendered"? w3c/svg-aam#29
• Should concrete shape/object roles like circle have a more specific role description? w3c/svg-aam#30
• SVG-AAM Section 8.1 Name and Description needs rewrite w3c/svg-aam#31
• SVG with ARIA roles w3c/svg-aam#11

[cookiecrook]

Pretty confident this last 2% of the interop investigation score is taking up more than 2% of the effort.

[cookiecrook]

Also testable, the title element and xlink:title attr fallbacks in the SVG-AAM name comp variations (noting the caveats in w3c/svg-aam#31)

[cookiecrook]

Draft WPT PR #43196 is up. A lot is blocked. Some areas could be filled out a little bit more with additional exhaustive test cases, but I think the current version is a good size to value ratio. Any larger, and I'm going to have trouble getting reviewers to commit their time.

All the blocked stuff is tracked in #80. I will look at this Issue/PR again tomorrow to make sure I didn't miss anything obvious, then ask for reviews.