Revisit moving the "orphaned role" tests to .tentative while ARIA issue discussed

[cookiecrook]

FYI, we've received some additional implementation concerns about the "orphaned roles" prose. Seems like one engineer may be considered working around this in the engine test context primarily, which IMO defeats the purpose of testing… I'll be opening an ARIA spec issue and possibly moving some of these tests to .tentative files.

Originally posted by @cookiecrook in #109 (comment)

Core-AAM Spec prose:

When an element has a role but is not contained in the required context (for example, an orphaned listitem without the required accessible parent of role list), User Agents MUST ignore the role token, and return the computedrole as if the ignored role token had not been included.

<div role="listitem"> <!-- Author error: orphaned listitem. computedrole returns "generic" -->

<div role="list"> <!-- computedrole returns "list" -->
  <div role="listitem"> <!-- computedrole returns "listitem" in the required context. -->

Cite: https://w3c.github.io/core-aam/#roleMappingComputedRole

[nmlapre]

one or more implementations may be working around this in the test context primarily

Can you expand on this? The only implementation I see passing these tests on the wpt.fyi dashboard is Firefox, but we definitely didn't implement it in Gecko just in a test context.

[cookiecrook]

Clarified and edited the description above from "may be working around this" to "considered working around this." It did not land.

[cookiecrook]

And FWIW, since you did the implementation work already, if you think they are good requirements and tests, it's a completely valid response to say, "I disagree with the issue. Please keep the requirements and tests." I'm not trying to create additional work for people, but hoping we can all arrive at the same, best behavior.

[cookiecrook]

ARIA issue is now here, by the way:

• Implementation concerns over Orphaned Role requirements in Core-AAM, HTML-AAM… and possibly elsewhere. w3c/aria#2166

[jcsteh]

since you did the implementation work already, if you think they are good requirements and tests

As noted in web-platform-tests/wpt#45715 (comment), I don't agree with these requirements, but we implemented according to the spec regardless, assuming it had been specified this way for good reason. FWIW, @nmlapre did some profiling on some reasonably large pages and didn't see any observable performance regressions, but of course, that doesn't mean we won't see any in future. If the spec is changed, that's probably for the best, though it's unfortunate that we didn't come to this consensus sooner to avoid some wasted work.

[cookiecrook]

Yes, that is unfortunate. Each implementation has changed in the last year, and the specs have been changed in some scenarios too. The ARIA group's new merging process puts a lot more onus on the spec editors in order to avoid unnecessary implementation work in the future. In particular, the only-recently testable via WPT is intended to help solve this problem. As we get closer to having more test methods for ensuring interoperability, I'm hopeful we'll see a lot less wasted work and speculative features of prior decades.

[zcorpan]

I suppose the correct action here is to make the tests tentative until the spec is fixed.