"p" tag can't have close tag?

[canadaduane]

Describe the bug

The </p> tag is not parsed correctly:

<script>
	let name = 'world';
</script>

<p>
	<pre>{name}</pre>
</p>

The result is: </p> attempted to close an element that was not open (7:0)

Information about your Svelte project:

• First-time user, just trying Svelte out
• used the REPL

Severity

• "Papercut"

Additional context

• This was a pretty surprising result for me as a first-time user--I'm converting some HTML to Svelte and didn't understand why it wouldn't parse.

[PatrickG]

The p element is not allowed to have a block element child.
See https://developer.mozilla.org/en-US/docs/Web/HTML/Element/p

That means, p is auto-closed right before the pre element.

That would work:

<script>
	let name = 'world';
</script>

<p>
<pre>{name}</pre>

[pngwn]

Yes. This is actually invalid HTML but browsers will do their best with it regardless. Svelte is much stricter and won’t parse invalid HTML. The auto closing behaviour is a little confusing, however.

[Conduitry]

I'm not sure what preferred behavior would be here. Browsers really do autoclose the <p> upon reaching a <pre>. Are you just thinking more specific compiler error messages? Like if the browser sees an erroneous closing tag for something that it had previously automatically closed, it adjust the error message?

[pngwn]

Yeah, I think so.

Browsers do autoclose paragraphs, amongst many other things they do to recover from invalid code but they also don't throw parsing errors. I don't think many of the 'lets try to make it work' features of browsers are widely known. With the above example, not only does chrome autoclose the first p but it autoopens the second.

Ideally when seeing a close tag without an open tag, after autoclosing a tag of the same type, we could give a more descriptive error. However, I don't know if the parser is setup to handle this kind of error reporting. We would need to make sure we retained that information. There are possibly a bunch of edge cases that could make this doubly confusing.

[Conduitry]

I'm removing the dev warning label because my understanding of that one is that it's for additional runtime checks that are put in the generated code only when compiling in dev mode. This issue is for an adjustment to the message that is given in a compile-time error that is unrelated to dev mode.

[Conduitry]

In 3.24.1 there's now a more helpful error message if the parser encounters a closed tag that previously had been autoclosed by opening another tag.