Bug 211512 – [Wheel event region] Add basic support for generating accurate wheel event listener region

• Status: RESOLVED
• Resolution: FIXED
• Component: Layout and Rendering

[Antti Koivisto]

To EventRegion

[Antti Koivisto]

Created attachment 398611 [details]
patch

[Antti Koivisto]

Created attachment 398614 [details]
patch

[Simon Fraser (smfr)]

Comment on attachment 398614 [details]
patch

View in context: https://bugs.webkit.org/attachment.cgi?id=398614&action=review

> Source/WebCore/rendering/style/StyleRareInheritedData.h:158
> +    unsigned eventListenerRegionTypes : 2; // OptionSet<EventListenerRegionType>

It's going to be really easy to forgot to change this when we add more EventListenerRegionType bits. I think maybe just eat the 8 bits right now, or add a compile-time assert that things will fit.

> Source/WebCore/style/StyleAdjuster.cpp:219
> +    auto findListeners = [&](auto eventName, auto type, auto nonPassiveType) {
> +        auto* eventListenerVector = element.eventTargetData()->eventListenerMap.find(eventName);
> +        if (!eventListenerVector)
> +            return;
> +
> +        types.add(type);
> +
> +        bool passiveOnly = true;
> +        for (auto& listener : *eventListenerVector) {
> +            if (!listener->isPassive())
> +                passiveOnly = false;
> +        }
> +
> +        if (!passiveOnly)
> +            types.add(nonPassiveType);
> +    };

Maybe this should be a function on Element

[Daniel Bates]

Comment on attachment 398614 [details]
patch

View in context: https://bugs.webkit.org/attachment.cgi?id=398614&action=review

>> Source/WebCore/rendering/style/StyleRareInheritedData.h:158
>> +    unsigned eventListenerRegionTypes : 2; // OptionSet<EventListenerRegionType>
> 
> It's going to be really easy to forgot to change this when we add more EventListenerRegionType bits. I think maybe just eat the 8 bits right now, or add a compile-time assert that things will fit.

Patch is ok as is. No change needed.

Just for me: Could this storage also be used to track the editable region? Usually only elements with userModify != ReadOnly are in that region, but there's one except for HTMLTextFormControl elements that is currently handled by plumbing an overrideUserModifyIsEditable book through EventRegionContext::unite(). This storage could be used to eliminate that bool. I didn't want to add storage to every element's style before, but now that this storage exists it could be used. Just a thought....

[Antti Koivisto]

Created attachment 398625 [details]
patch

[Antti Koivisto]

> Maybe this should be a function on Element

Nah, element is the same for all callers.

[Antti Koivisto]

Created attachment 398629 [details]
patch

[Antti Koivisto]

> Just for me: Could this storage also be used to track the editable region?

Probably.

[EWS]

Committed r261239: <https://trac.webkit.org/changeset/261239>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 398629 [details].

[Radar WebKit Bug Importer]

<rdar://problem/62937481>

[Ryan Haddad]

As indicated by mac-debug-wk1 EWS, this change caused fast/events/wheel-event-outside-body.html to assert 

SHOULD NEVER BE REACHED
./rendering/RenderLayerBacking.cpp(2948) : void WebCore::RenderLayerBacking::paintIntoLayer(const WebCore::GraphicsLayer *, WebCore::GraphicsContext &, const WebCore::IntRect &, OptionSet<WebCore::PaintBehavior>, WebCore::EventRegionContext *)

https://bugs.webkit.org/show_bug.cgi?id=211536

[Ryan Haddad]

Reverted r261239 for reason:

Caused fast/events/wheel-event-outside-body.html to assert on macOS WK1

Committed r261253: <https://trac.webkit.org/changeset/261253>

[Antti Koivisto]

Created attachment 398717 [details]
patch

[EWS]

Committed r261279: <https://trac.webkit.org/changeset/261279>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 398717 [details].