Implement hit testing bits of inert subtrees

• Component: DOM
• Status: RESOLVED FIXED

[Tim Nguyen (:ntim)]

Depending on whether we want to affect JS APIs or not, we can either:
* Change hit testing code to treat inert nodes as pointer-events: none, like in: https://github.com/nt1m/WebKit/commit/991684e0f3b810e566329e0c031342de613933b0
-> This is close to Firefox's implementation
-> Firefox argues the spec should be this way: https://github.com/whatwg/html/issues/5650

* Treat inert nodes more like isDisabledFormControl() (e.g. having similar checks when triggering events)
-> More similar to Chrome's implementation
-> Closer to latest inert attribute PR by the Chromium team: https://github.com/whatwg/html/pull/4288


In the context of <dialog>, I think both approaches are acceptable wrt. the spec and similarly easy. Hit testing also doesn't matter much generally since everything is behind a backdrop in a modal dialog context. This is more relevant for the inert attribute implementation.

[Radar WebKit Bug Importer]

<rdar://problem/82186260>

[Tim Nguyen (:ntim)]

Created attachment 436919 [details]
Patch

[Tim Nguyen (:ntim)]

Created attachment 438657 [details]
Patch

[Antti Koivisto]

Comment on attachment 438657 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=438657&action=review

> Source/WebCore/rendering/style/StyleRareInheritedData.h:169
> +    bool isInert;

You should include this into the bitfield with the rest of the booleans.

[Tim Nguyen (:ntim)]

Created attachment 438661 [details]
Patch

[Tim Nguyen (:ntim)]

Created attachment 438978 [details]
Patch

[Tim Nguyen (:ntim)]

Created attachment 439020 [details]
Patch

[Tim Nguyen (:ntim)]

Comment on attachment 439020 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=439020&action=review

> Source/WebCore/style/StyleSharingResolver.cpp:283
> +            return false;

Note to self, this could use an inert attribute feature flag check (though not sure if that slows down the SharingResolver or not)

[Tim Nguyen (:ntim)]

Comment on attachment 439020 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=439020&action=review

> Source/WebCore/style/StyleSharingResolver.cpp:283
> +            return false;

Note to self, this could use an inert attribute feature flag check (though not sure if that slows down the SharingResolver or not)

[Antti Koivisto]

Comment on attachment 439020 [details]
Patch

Please split this up to pieces that can be reviewed separately (along the lines of "Add effectiveInert bit to RenderStyle")

[Tim Nguyen (:ntim)]

Created attachment 439039 [details]
Patch

[Tim Nguyen (:ntim)]

Created attachment 439042 [details]
Patch

[Antti Koivisto]

Comment on attachment 439042 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=439042&action=review

> Source/WebCore/style/StyleAdjuster.cpp:526
> +    bool isInertSubtreeRoot = (m_document.activeModalDialog() && m_element == m_document.documentElement())
> +        || (m_document.settings().inertAttributeEnabled() && is<HTMLElement>(m_element) && m_element->hasAttribute(HTMLNames::inertAttr));

This would read better as a lambda.

> Source/WebCore/style/StyleAdjuster.cpp:532
> +    if (isInertSubtreeRoot)
> +        style.setEffectiveInert(true);
> +
> +    // Make sure the active dialog is interactable when the whole document is blocked by the modal dialog
> +    if (m_element == m_document.activeModalDialog())
> +        style.setEffectiveInert(false);

What should happen when you have an active <dialog inert>?

[Tim Nguyen (:ntim)]

Created attachment 439262 [details]
Patch

[Tim Nguyen (:ntim)]

Committed r283079 (242137@main): <https://commits.webkit.org/242137@main>

[Tim Nguyen (:ntim)]

*** Bug 229410 has been marked as a duplicate of this bug. ***