Bug 112214 – [Shadow Dom]: Non Bubbling events in ShadowDOM dispatch in an incorrect order

• Status: RESOLVED
• Resolution: FIXED
• Component: UI Events

[Daniel Freedman]

In the demo shown, a Custom Event is created in the shadow dom that is bubbles: false. The event is dispatched to first child in the shadow, which sets the stop propagation flag. The host event listener fires afterwards.
This seems to be incorrect according to step 8 of http://dom.spec.whatwg.org/#concept-event-dispatch.

If the CustomEvent has bubbles: true, the event listener on host does not fire.

[Daniel Freedman]

After more review, it turns out I was mistaken about stopPropagation. Instead, for non-bubbling events, the listeners for an event at AT_TARGET phase are fired in capture order (top down, DOM order).

Given a DOM construction A->{SR}->B, where A and B are DOM nodes with listeners for "focus" and {SR} is a ShadowRoot, if a focus event is dispatched to B, the listeners will fire in order A, B. This seems incorrect.

[Hayato Ito]

Confirmed.

The implementation does not seem to follow the change of spec.
https://dvcs.w3.org/hg/webcomponents/rev/f46cfd9b0832

Let me fix that.

[Hayato Ito]

Created attachment 193228 [details]
Fixed the order.

[WebKit Review Bot]

Comment on attachment 193228 [details]
Fixed the order.

Clearing flags on attachment: 193228

Committed r145873: <http://trac.webkit.org/changeset/145873>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Daniel Freedman]

Sorry about the delay in looking at this, but I think the patch is backwards.

As a reference, the URL I provided in the bug report (http://jsfiddle.net/dfreedm/EMgtb/) shows that for DOM construction host->{SR}->inner, with this patch applied, host will *always* see an event fired on inner *before* inner has a chance to see it.

The expected result is that host will see all events, bubbling or not, on inner *after* inner sees it.

[Hayato Ito]

Thank you for raising a question.

(In reply to comment #6)
> Sorry about the delay in looking at this, but I think the patch is backwards.
> 
> As a reference, the URL I provided in the bug report (http://jsfiddle.net/dfreedm/EMgtb/) shows that for DOM construction host->{SR}->inner, with this patch applied, host will *always* see an event fired on inner *before* inner has a chance to see it.
> 

That's an expected behavior in the spec. I think we should discuss this on Shadow DOM spec.

> The expected result is that host will see all events, bubbling or not, on inner *after* inner sees it.

Note that, before this patch, the parent node of host can see events on inner at its capturing phase before inner sees it.
So there is still a chance that nodes on the outer tree can cancel an event on inner anyway with/without this patch.

[Daniel Freedman]

(In reply to comment #7)
> Thank you for raising a question.
> 
> (In reply to comment #6)
> > Sorry about the delay in looking at this, but I think the patch is backwards.
> > 
> > As a reference, the URL I provided in the bug report (http://jsfiddle.net/dfreedm/EMgtb/) shows that for DOM construction host->{SR}->inner, with this patch applied, host will *always* see an event fired on inner *before* inner has a chance to see it.
> > 
> 
> That's an expected behavior in the spec. I think we should discuss this on Shadow DOM spec.

Will do.

> 
> > The expected result is that host will see all events, bubbling or not, on inner *after* inner sees it.
> 
> Note that, before this patch, the parent node of host can see events on inner at its capturing phase before inner sees it.
> So there is still a chance that nodes on the outer tree can cancel an event on inner anyway with/without this patch.

Agreed. I was imprecise in my last comment. I meant that I would expect that event listeners for the bubbling phase should fire in the inner, host order, no matter if event.bubbles is true or false.

[Daniel Freedman]

(In reply to comment #8)
> (In reply to comment #7)
> > Thank you for raising a question.
> > 
> > (In reply to comment #6)
> > > Sorry about the delay in looking at this, but I think the patch is backwards.
> > > 
> > > As a reference, the URL I provided in the bug report (http://jsfiddle.net/dfreedm/EMgtb/) shows that for DOM construction host->{SR}->inner, with this patch applied, host will *always* see an event fired on inner *before* inner has a chance to see it.
> > > 
> > 
> > That's an expected behavior in the spec. I think we should discuss this on Shadow DOM spec.
> 
> Will do.
> 
> > 
> > > The expected result is that host will see all events, bubbling or not, on inner *after* inner sees it.
> > 
> > Note that, before this patch, the parent node of host can see events on inner at its capturing phase before inner sees it.
> > So there is still a chance that nodes on the outer tree can cancel an event on inner anyway with/without this patch.
> 
> Agreed. I was imprecise in my last comment. I meant that I would expect that event listeners for the bubbling phase should fire in the inner, host order, no matter if event.bubbles is true or false.

Filed https://www.w3.org/Bugs/Public/show_bug.cgi?id=21404 on Shadow DOM

[Hayato Ito]

Let me close this bug.

The discussion is on-going https://www.w3.org/Bugs/Public/show_bug.cgi?id=21404 .

Looks like the desired behavior on the spec side is different than either before/after this patch.
We should file a new bug in addition to updating the spec.

[Hayato Ito]

Filed as https://bugs.webkit.org/show_bug.cgi?id=113676.

(In reply to comment #10)
> Let me close this bug.
> 
> The discussion is on-going https://www.w3.org/Bugs/Public/show_bug.cgi?id=21404 .
> 
> Looks like the desired behavior on the spec side is different than either before/after this patch.
> We should file a new bug in addition to updating the spec.