Improve error for unexpected closing xml tag

[tobil4sk]

#12015 (comment)

On a side note, the error seems odd too, I don't know if a plain </xml> should complain about being unterminated...

[kLabz]

Will have to check with @ncannasse about that, iirc he wanted inline xml to have very few limitations on terms of syntax

[tobil4sk]

Note: the reason the lexer continues past </ is because empty tags are allowed: 68fdc5a. Otherwise it would give invalid character.

[Simn]

Will have to check with @ncannasse about that, iirc he wanted inline xml to have very few limitations on terms of syntax

The point here is that this isn't really an xml-literal in the first place. It's not like you can close it with <//xml> so there's nothing you can do with this syntax, unless I'm missing something.

[tobil4sk]

It's not like you can close it with <//xml> so there's nothing you can do with this syntax, unless I'm missing something.

You can close it with </> though, since it's parsed as an empty tag 😅

[Simn]

I thought an empty tag was <>, why is </ parsed like that?

[tobil4sk]

For example with: </</>

The tag name is the empty string between the first < and /, the / is the middle content and the </> is the closing tag.

It's just a side effect of how empty tags are handled for not_xml, it doesn't really make sense.

[Simn]

Something is definitely wrong here, there's even a comment in lexer.ml that says (* don't allow self-closing fragments *).

[tobil4sk]

Something is definitely wrong here, there's even a comment in lexer.ml that says (* don't allow self-closing fragments *).

In general this is consistent with how xml literals behave.

<tag...</tag> is allowed, and </</> is just an edge case of that with an empty tag name. It's the same as < </> or <.</> just with a slash in the middle, which happens to produce </ at the beginning.

This patch just disallows an empty tag name if it is immediately followed by /, in order to catch the case of a rogue closing tag, but the cost is breaking </...</>, so it's a question of whether that is an acceptable constraint or not.

[Simn]

Right, I had forgotten how silly all this is...