HTML mode throws on mismatched tags

[dabbler0]

Parse5 is able to parse the following document:

<b>
  <i>
</b>

But when the HTML mode tries to blockify it it thows Improper parser: block cannot nest immediately inside another block.

[sakagg]

Seems to be a problem with parse5 location info.
I'll try to fix this in parse5 itself.

[sakagg]

On further analysis I've found the cause of the problem
The parser isn't parsing it correctly.
I've opened an issue there
inikulin/parse5#57

[davidbau]

From that issue, it looks like this parse5 behavior is intentional, and I agree, that's what is specified in the HTML5 specification and how browsers actually behave. When tags are misnested, the parser can synthesize extra elements where no explicit tag was given.

This is an interesting case to try to render. We could either
(a) omit showing as blocks any elements that are missing their begin-tag or
(b) try rendering elements-without-a-begin-tag as blocks anyway.

I think (a) might make more sense. If we are in such a situation and there is no begin-tag but there is an end-tag for an element, we might synthesize an "error block" to surround just the end-tag.

[davidbau]

Another minimal test document that we should untangle:

abc</b>def

When we have an end-tag without a begin-tag we seem to lose track of things. I think we should probably highlight it as an error-tag block.

[sakagg]

I didn't realize that's how HTML5 spec has it! If its part of the spec, it should be part of this program
Both of them seem to be interesting problems.

[sakagg]

We can close this now.