Bug 218753 – Support AbortSignal in addEventListenerOptions to unsubscribe from events

• Status: RESOLVED
• Resolution: FIXED
• Component: DOM

[Benjamin Gruenbaum]

See https://github.com/whatwg/dom/issues/911 and the spec PR https://github.com/whatwg/dom/pull/919 

I am working on adding support for AbortController in various web APIs. One of those is adding support for passing `AbortSignal`s to `addEventListener` and have an easy way to remove a lot of listeners.

As per the WHATWG process - I am filing a WebKit bug.

Please feel free to ask any questions or make suggestions regarding this.

I will follow up with WPTs in that PR.

[Radar WebKit Bug Importer]

<rdar://problem/71258012>

[Chris Dumez]

Created attachment 418185 [details]
Patch

[Chris Dumez]

Created attachment 418187 [details]
Patch

[Chris Dumez]

Created attachment 418199 [details]
Patch

[EWS Watchlist]

Thanks for the patch. If this patch contains new public API please make sure it follows the guidelines for new WebKit2 GTK+ API. See https://trac.webkit.org/wiki/WebKitGTK/AddingNewWebKit2API

[Chris Dumez]

Created attachment 418202 [details]
Patch

[Darin Adler]

Comment on attachment 418202 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=418202&action=review

> Source/WebCore/dom/AddEventListenerOptions.h:41
> +    AddEventListenerOptions(bool capture = false, Optional<bool> passive = WTF::nullopt, bool once = false, RefPtr<AbortSignal>&& signal = nullptr)
> +        : EventListenerOptions(capture)
> +        , passive(passive)
> +        , once(once)
> +        , signal(WTFMove(signal))
> +    {
> +    }

It’s nice to have structs with default values but without constructors when we can. I guess there is code that depends on this?

> Source/WebCore/dom/EventListener.h:37
> +class EventListener : public RefCounted<EventListener>, public CanMakeWeakPtr<EventListener> {

I’m assuming there are not cases where an EventListener was:

1) not already doing the weak pointer thing
2) a small object allocated many, many times

But I didn’t audit all the uses the way I often would when asking myself a question like that.

> Source/WebCore/dom/EventListenerOptions.h:33
> +    EventListenerOptions(bool capture = false)
> +        : capture(capture)
> +    { }

It’s nice to have structs with default values but without constructors when we can. I guess there is code that depends on this?

[Chris Dumez]

Comment on attachment 418202 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=418202&action=review

>> Source/WebCore/dom/AddEventListenerOptions.h:41
>> +    }
> 
> It’s nice to have structs with default values but without constructors when we can. I guess there is code that depends on this?

There are call sites that call addEventListener with a simple boolean like so:
addEventListener(type, listener, true)

Using struct initialization here would be a bit awkward here I think, especially considering it involves struct inheritance.

[Chris Dumez]

Created attachment 418299 [details]
Patch

[EWS]

Committed r271806: <https://trac.webkit.org/changeset/271806>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 418299 [details].