Serialize the children of void elements as the empty string

[Zirro]

Proposed solution to #4220, which makes the algorithm consistent with the current implementations in Chrome and Safari. The intention is for the innerHTML of void elements to return the empty string even in cases where child nodes are present. I changed the steps to explicitly return a value in two locations as it made most sense for this change.

There are tests for this in https://wpt.fyi/results/html/syntax/serializing-html-fragments/serializing.html which previously disagreed with the specification. This change would align the specification with the tests.

---

/parsing.html ( diff )

[annevk]

How about we put VOIDLIST in a variable, so we only have to enumerate it once in this algorithm? (FWIW, I checked, only this algorithm makes use of it within HTML and I suspect that means everywhere.) Looks good otherwise, appreciate the algorithm prose cleanup too.

[annevk]

Do you also want to file a bug against Firefox?

[Zirro]

@annevk Variable added. Here's the Firefox bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1514499

[domenic]

I pushed some editorial tweaks. @Zirro and/or @annevk, can you double-check my work? This seems good to merge to me now, but we should get a second set of eyes.

[Zirro]

LGTM, with one suggested change.

[domenic]

I wrote this commit message:

Fixes #4220, by making the algorithm consistent with the current
implementations in Blink and WebKit, as well as the existing web
platform tests. Now, innerHTML of void elements returns the empty string
even in cases where child nodes are present.

Tests (preexisting): https://wpt.fyi/results/html/syntax/serializing-html-fragments/serializing.html

But realized that you don't seem to be in the acknowledgments, @Zirro. Would you like to add yourself?

[Zirro]

Thank you for asking, but I'm actually listed there under my real name after a previous contribution.

[domenic]

Ah thanks! I think my Ctrl+F was malfunctioning...

[annevk]

Beautiful!