wait for Element Upgrade in connectedCallback: FireFox and Chromium differences - callback

Question

Update March 2021:
FireFox bug fixed, now behaves the same as Chromium and Safari.
That means waiting for the JS EventLoop to be empty (with setTimeout or requestAnimationFrame) in the connectedCallback is now a cross-browser method
connectedCallback(){
setTimeout(()=>{
// can access lightDOM here
}); // ,0 not required
}
What the heck is the Event Loop? - Philip Roberts
https://www.youtube.com/watch?v=8aGhZQkoFbQ
Update Oct. 28 2020:
Now reported by Mozilla engineer Anne van Kesteren as a bug in FireFox:
FireFox invokes the connectedCallback too late:
https://bugzilla.mozilla.org/show_bug.cgi?id=1673811
First post May. 2020:
Bitten again by this Chrome Element upgrade issue, after spending a week in FireFox.
Forgot to wrap code in a setTimeout before delivering to Chromium browsers.
FireFox prints: ABCD
Chromium prints: ADCD
Question: Why the difference?
<script>
customElements.define('my-element', class extends HTMLElement {
connectedCallback() {
console.log(this.innerHTML);// "A" in FireFox, "" in other Browsers
if (this.innerHTML == "A")
this.innerHTML = this.innerHTML + "B";
else
setTimeout(() => this.innerHTML = this.innerHTML + "D");
}
})
</script>
<my-element>A</my-element><my-element>C</my-element>
Related answers over the past years:
How to have a 'connectedCallback' for when all child custom elements have been connected
How to wait for Custom Element reference to be "upgraded"?
Update #1
Apple/Safari: prints: ADCD (same as Chromium)
note: Chromium Blink engine is a fork of Apples (WebKit)WebCore code!!
Update #2
With Supersharps reference we found the related threads:
(2016) connectedCallback timing when the document parser creates custom elements
https://github.com/w3c/webcomponents/issues/551
(2019) Need a callback for when children changed or parser finished parsing children
https://github.com/w3c/webcomponents/issues/809
order of callbacks in FireFox versus Chromium:
source: https://jsfiddle.net/WebComponents/9p5qyk1z/

Answer 1

I think the Chrome/Safari behaviour is less intuitive for the beginners, but with some more complex scenarios (for example with child custom elements) then it is much more consistant.
See the different examples below. They act strangely in Firefox...
Another use case that I don't have the courage to code: when a document is parsed, maybe you don't have the end of the document yet. Therefore, when a custom element is created, you cannot be sure you get all its children until you get the closing tag (that could never arrive).
According to Ryosuke Niwa for WebKit:
The problem then is that the element won't get connectedCallback until
all children are parsed. For example, if the entire document was a
single custom element, that custom element would never receive
connectedCallback until the entire document is fetched & parsed even
though the element is really in the document. That would be bad.
So it's better no to wait and connect the custom element as soon as it is created, that means with no child.
<script>
customElements.define( 'c-e', class extends HTMLElement {} )
customElements.define('my-element', class extends HTMLElement {
connectedCallback() {
console.log(this.innerHTML, this.childNodes.length)
let span = document.createElement( 'span' )
if (this.innerHTML.indexOf( 'A' ) >= 0 )
span.textContent = 'B'
else
span.textContent = 'D'
setTimeout( () => this.appendChild( span ) )
}
})
</script>
<my-element>A</my-element><my-element>C</my-element>
<br>
<my-element><c-e></c-e>A</my-element><my-element>A<c-e></c-e></my-element>
<br>
<my-element><c-e2></c-e2>A</my-element><my-element>A<c-e2></c-e2></my-element>
As far as I understand, there was a consensus on it that led to adjust the spec that (Chrome/Safari) way:
Fixes w3c/webcomponents#551 by ensuring that
insertions into the DOM trigger connectedCallback immediately, instead
of putting the callback reaction on the the backup element queue and
letting it get triggered at the next microtask checkpoint. This means
connectedCallback will generally be invoked when the element has zero
children, as expected, instead of a random number depending on when the
next custom element is seen.
We can conclude that Firefox also follow the spec... yes, but we should not rely on the content in connectedCallback for the reasons discussed above.