AX: align AccessibilityRole::Switch with ::Checkbox

[annevk]

5ce55c6

AX: align AccessibilityRole::Switch with ::Checkbox
https://bugs.webkit.org/show_bug.cgi?id=263008
rdar://116805287

Reviewed by Tyler Wilcock.

A switch should essentially be a checkbox except that it doesn't
support the intermediate/mixed state.

As such, this makes the following changes:

- We make switch support the required attribute.
- We make interaction regions support switches.

* LayoutTests/accessibility/aria-required-expected.txt:
* LayoutTests/accessibility/aria-required.html:
* LayoutTests/inspector/dom/getAccessibilityPropertiesForNode-expected.txt:
* LayoutTests/inspector/dom/getAccessibilityPropertiesForNode.html:
* LayoutTests/interaction-region/aria-roles-expected.txt:
* LayoutTests/interaction-region/aria-roles.html:
* LayoutTests/platform/gtk/inspector/dom/getAccessibilityPropertiesForNode-expected.txt:
* Source/WebCore/accessibility/AccessibilityNodeObject.cpp:
(WebCore::AccessibilityNodeObject::supportsRequiredAttribute const):
* Source/WebCore/page/InteractionRegion.cpp:
(WebCore::shouldAllowAccessibilityRoleAsPointerCursorReplacement):

Canonical link: https://commits.webkit.org/269311@main

179d374

Misc	iOS, tvOS & watchOS	macOS	Linux	Windows
✅ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	✅ 🛠 wincairo
✅ 🧪 bindings	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	🧪 wpe-wk2
✅ 🧪 webkitperl	✅ 🧪 ios-wk2	✅ 🧪 api-mac	✅ 🛠 gtk
	✅ 🧪 ios-wk2-wpt	✅ 🧪 mac-wk1	✅ 🧪 gtk-wk2
	✅ 🧪 api-ios	✅ 🧪 mac-wk2	✅ 🧪 api-gtk
	✅ 🛠 tv	🧪 mac-AS-debug-wk2
	✅ 🛠 tv-sim	✅ 🧪 mac-wk2-stress
✅ 🛠 🧪 merge	✅ 🛠 watch
	✅ 🛠 watch-sim

[cookiecrook]

Will you also confirm computedrole with a new subtest in WPT?
https://github.com/web-platform-tests/wpt/blob/master/html-aam/roles.html

[cookiecrook]

Oh the tests PR is already in. Commented there.

[twilco]

The purpose checkboxOrRadioRect() is to unify the rect of the control with the rect of its associated label. But right now this only works with native <label for="id-of-control">Foo label</label>, or containing labels:

<label>
   Foo bar label for switch
   <div role="switch"></div>
</label>

But I'm wondering if either of these methods of pairing labels are "allowed" or recommended patterns for a role="switch"? Certainly for <input type=checkbox switch>, but role="switch" I'm less confident in.

@cookiecrook what do you think?

[annevk]

Hmm, but the logic here also does this for <label><span role=checkbox></span></label>, right? Perhaps it shouldn't be using isCheckboxOrRadio() to begin with?

[twilco]

Yes you're right, it would behave that way for role=checkbox, but arguably shouldn't. Let's remove this part and we'll address it separately later.

[annevk]

#19048

[cookiecrook]

Yes label for/id should only work on native inputs, not ARIA controls.

[annevk]

@cookiecrook this PR is mainly to align role=switch with role=checkbox. <input type=checkbox switch> is a bigger change that will take several PRs tracked as dependencies of https://bugs.webkit.org/show_bug.cgi?id=259380. And yeah, I'll make sure to add a computed role test there.

[annevk]

Update here:

1. It seems that checkboxOrRadioRect() should only apply to genuine <input> checkboxes and radio buttons. This is probably best tackled separately as it no longer is about aligning role=switch with role=checkbox.
2. I'd still appreciate some guidance with respect to testing the "required" attribute (aria-required presumably?) and shouldAllowAccessibilityRoleAsPointerCursorReplacement(). I didn't find any observable differences of these changes myself, which makes that a bit tricky.

[twilco]

2. I'd still appreciate some guidance with respect to testing the "required" attribute (aria-required presumably?) and shouldAllowAccessibilityRoleAsPointerCursorReplacement(). I didn't find any observable differences of these changes myself, which makes that a bit tricky.

It would be great to add a testcase to LayoutTests/accessibility/aria-required.html. It's interesting that there are no VoiceOver behavior differences. supportsRequiredAttribute() is used by VoiceOver to determine whether it should bother asking for the required state of the element in future requests. But that doesn't seem to be the case here. Regardless, I'd still make your change as-is.

One place there will be a behavior difference is in inspector. Please add a role="switch" aria-required="true" testcase to inspector/dom/getAccessibilityPropertiesForNode.html. Without your patch, inspector will not report the required status.

[annevk]

@twilco the changes I made to accessibility/aria-required.html pass with or without this change. inspector/dom/getAccessibilityPropertiesForNode.html is impacted by the supportsRequiredAttribute() change.

And interaction regions is a visionOS-specific fix it seems. @etiennesegonzac helpfully provided the test expectations there (that EWS will ignore, but hopefully that's good enough for everyone here to get r+).

[webkit-early-warning-system]

EWS run on current version of this PR (hash 179d374)

Misc	iOS, tvOS & watchOS	macOS	Linux	Windows
✅ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	✅ 🛠 wincairo
✅ 🧪 bindings	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	🧪 wpe-wk2
✅ 🧪 webkitperl	✅ 🧪 ios-wk2	✅ 🧪 api-mac	✅ 🛠 gtk
	✅ 🧪 ios-wk2-wpt	✅ 🧪 mac-wk1	✅ 🧪 gtk-wk2
	✅ 🧪 api-ios	✅ 🧪 mac-wk2	✅ 🧪 api-gtk
	✅ 🛠 tv	🧪 mac-AS-debug-wk2
	✅ 🛠 tv-sim	✅ 🧪 mac-wk2-stress
✅ 🛠 🧪 merge	✅ 🛠 watch
	✅ 🛠 watch-sim

[webkit-commit-queue]

Committed 269311@main (5ce55c6): https://commits.webkit.org/269311@main

Reviewed commits have been landed. Closing PR #18938 and removing active labels.