Bug 165279 – Implement inert attribute (affects keyboard, events, and accessibility)

• Component: Accessibility
• Status: RESOLVED FIXED

[James Craig]

WICG discussion happening here:

https://discourse.wicg.io/t/inert-attribute/1838

[Radar WebKit Bug Importer]

<rdar://problem/29467435>

[James Craig]

There are obvious benefits for accessibility. 

Are there concerns from WebKit implementors about the functionality change or implementability of for events handling and keyboard/focus behavior?

[James Craig]

Also discussion re: naming conflict with template: https://github.com/WICG/inert/issues/39

[James Craig]

*** Bug 159748 has been marked as a duplicate of this bug. ***

[James Craig]

Relating to <dialog> bug 84635

[Chris Rebert]

See also: https://github.com/whatwg/html/pull/1474

[James Craig]

Google this shipping behind a flag now.

Enable:
chrome://flags#enable-experimental-web-platform-features

Explainer:
https://github.com/WICG/inert/blob/master/README.md

Chrome "Intent to Implement" from February:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/N--HhuYFJQI

[Carolyn MacLeod]

FF bug was reopened in May/17: https://bugzilla.mozilla.org/show_bug.cgi?id=921504

Edge bug opened in Sept/17 (already has 187 votes): https://wpdev.uservoice.com/forums/257854-microsoft-edge-developer/suggestions/31606522-support-the-inert-attribute

[Michael Puckett]

Please consider adding to the feature status list here:

https://webkit.org/status

[Frédéric Wang (:fredw)]

Created attachment 396647 [details]
testcase

[Frédéric Wang (:fredw)]

Created attachment 396649 [details]
WIP Patch

Just a quick proof-of-concept patch. It seems to at least work for links. Probably there are several cases to check and things to verify, but that gives a starting point.

[Frédéric Wang (:fredw)]

Comment on attachment 396649 [details]
WIP Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=396649&action=review

> Source/WebCore/dom/Element.h:473
> +    virtual bool isDisabledFormControl() const { return isInert(); }

Obviously, this should also be done in derived class... it seems they don't call their parent.

[Carolyn MacLeod]

Nudge. :)

Chrome/Edge/Opera and Firefox all have this ready and waiting behind a flag: https://caniuse.com/mdn-api_htmlelement_inert

The HTML spec has an approved pull request waiting: https://github.com/whatwg/html/pull/4288

So it seems everybody is waiting on webkit/Safari.

[Carolyn MacLeod]

Hmmm, I guess people are also waiting on https://github.com/whatwg/html/issues/5650 and https://github.com/w3c/html-aam/issues/295. If you can help move those conversations forward from a webkit point of view, please do.

[Tim Nguyen (:ntim)]

I'm going to start implementing inert subtrees in bug 110952 in the next weeks, though only the minimum needed for <dialog> element. I suspect it won't be hard to get the attribute working after that's properly done.

[Tim Nguyen (:ntim)]

Created attachment 435977 [details]
Patch

[Tim Nguyen (:ntim)]

Created attachment 435994 [details]
Patch

[Sam Weinig]

Comment on attachment 435994 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=435994&action=review

> Source/WebCore/html/HTMLElement.idl:60
> +    // Experimental: https://github.com/WICG/inert/blob/master/README.md
> +    [CEReactions, Reflect, EnabledBySetting=InertAttribute] attribute boolean inert;

I would put this above the Non-standard attributes (and below the attachInternals commented out attribute).

[Simon Fraser (smfr)]

Comment on attachment 435994 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=435994&action=review

> Source/WebCore/dom/Node.cpp:2640
> +        const Element* element = is<Element>(this) ? downcast<Element>(this) : parentElement();
> +        while (element) {
> +            if (is<HTMLElement>(element) && element->hasAttribute(HTMLNames::inertAttr))
> +                return true;
> +            element = element->parentElement();

Doing an ancestor tree walk every time this is called doesn't seem great for perf.

[Tim Nguyen (:ntim)]

(In reply to Simon Fraser (smfr) from comment #19)
> Comment on attachment 435994 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=435994&action=review
> 
> > Source/WebCore/dom/Node.cpp:2640
> > +        const Element* element = is<Element>(this) ? downcast<Element>(this) : parentElement();
> > +        while (element) {
> > +            if (is<HTMLElement>(element) && element->hasAttribute(HTMLNames::inertAttr))
> > +                return true;
> > +            element = element->parentElement();
> 
> Doing an ancestor tree walk every time this is called doesn't seem great for
> perf.

Simon suggested adding a bit on Document that is set anytime one of the elements get the inert attribute, and only doing the ancestor walk in that case.

[Tim Nguyen (:ntim)]

Created attachment 436191 [details]
Patch

[Tim Nguyen (:ntim)]

Created attachment 436192 [details]
Patch

[Chris Dumez]

Comment on attachment 436192 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=436192&action=review

> Source/WebCore/dom/Document.h:1635
> +    bool hasInertAttributeBeenUsed() { return m_hasInertAttributeBeenUsed; }

should be const.

> Source/WebCore/dom/Node.cpp:2636
> +        const Element* element = is<Element>(this) ? downcast<Element>(this) : parentElement();

Seems like you should be using lineageOfType:

for (auto& element : lineageOfType<HTMLElement>(*this)) {
    if (element->hasAttribute(HTMLNames::inertAttr))
        return true;
}

> Source/WebCore/html/HTMLElement.cpp:486
> +        document().setHasInertAttributeBeenUsed();

How does that work when an element gets moved from one document to another?

[Tim Nguyen (:ntim)]

Created attachment 436200 [details]
Patch

[Chris Dumez]

Comment on attachment 436200 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=436200&action=review

r=me

> Source/WebCore/dom/Node.cpp:2636
> +        auto* element = is<Element>(this) ? downcast<Element>(this) : parentElement();

is<Element>(*this) to avoid an unnecessary null-check.

[Tim Nguyen (:ntim)]

Created attachment 436203 [details]
Patch

[EWS]

Found 30 new test failures: accessibility/accessibility-crash-focused-element-change.html, accessibility/accessibility-crash-setattribute.html, accessibility/accessibility-node-reparent.html, accessibility/mac/abbr-acronym-tags.html, accessibility/mac/accessibility-make-first-responder.html, accessibility/mac/accesskey.html, accessibility/mac/attributed-string/attributed-string-for-range-with-options.html, accessibility/mac/attributed-string/attributed-string-for-range.html, accessibility/mac/search-text/search-text.html, accessibility/mac/select-text/select-text-1.html ...

[Ryosuke Niwa]

Comment on attachment 436203 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=436203&action=review

> Source/WebCore/dom/Node.cpp:2636
> +        auto* element = is<Element>(*this) ? downcast<Element>(this) : parentElement();

Please use RefPtr.

[Tim Nguyen (:ntim)]

Created attachment 436228 [details]
Patch

[Ryosuke Niwa]

Comment on attachment 436228 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=436228&action=review

> Source/WebCore/dom/Node.cpp:2637
> +        auto* element = is<Element>(*this) ? downcast<Element>(this) : parentElement();
> +        if (element) {

Please use RefPtr. Also declare the variable inside if like so:
if (RefPtr element = ~)

[Emilio Cobos Álvarez (:emilio)]

Comment on attachment 436228 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=436228&action=review

> Source/WebCore/dom/Node.cpp:2639
> +                if (is<HTMLElement>(ancestor) && ancestor.hasAttribute(HTMLNames::inertAttr))

Shouldn't this use the flat tree?

[Tim Nguyen (:ntim)]

Created attachment 436265 [details]
Patch

[EWS]

Committed r281490 (240864@main): <https://commits.webkit.org/240864@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 436265 [details].