Handle illegal self-closing according to spec

[apeschar]

Browsers will parse

<div />
<span>Hello, World!</span>
</div>

like this (proof)

<div>
<span>Hello, World!</span>
</div>

Whereas this parser will parse it like this

<div></div>
<span>Hello, World!</span>

This is because the Tokenizer emits endTag after all self-closing tags. It should really only do this for tags that may be self-closed (ie, void elements and foreign elements).

[goetas]

This makes sense but most probably is a BC break 😿

[mundschenk-at]

Is it rally a BC break if the parser gets closer to spec (or browser behavior, as an extension of the spec)?

[apeschar]

Right. The README starts with "HTML5 is a standards-compliant HTML5 parser". It isn't quite, but fixing this bug would get it a little closer.

[apeschar]

The title of this issue is a bit confusing because I did not understand that this browser behaviour is according to spec until I investigated further. See the URL I linked in my PR. Self-closing tags do not exist in HTML5, are an error, and therefore ignored by browsers, except for foreign elements.

See also: https://www.w3.org/TR/html5/syntax.html#start-tags