Contents of <pre> tags wrapped in <div> tags are not left alone if they contain empty lines

[aral]

Contents of HTML preformatted text blocks (<pre>…</pre>) are, correctly, left alone:

e.g.,

<pre>a

b
</pre>

is rendered correctly as, with the contents untouched:

<pre>a

b
</pre>

However, if you wrap that in a <div>:

<div>
  <pre>a

  b
  </pre>
</div>

the commonmark parser starts parsing the contents of the <pre> tag following the empty line, which corrupts the resulting HTML:

<div>
  <pre>a
<p>b
</pre></p>
</div>

Notice the erroneously added <p> tag prior to b that closes following the </pre> tag.

I encountered this in Markdown-it and was able to narrow it down to this test case which is also reproducible in the commonmark parser at https://spec.commonmark.org/dingus/.

Related:

• markdown-it/markdown-it#238

[jgm]

Do you have a suggestion to make about how the spec might be altered to deal with this case (without introducing too much complexity)?

[aral]

Do you have a suggestion to make about how the spec might be altered to deal with this case (without introducing too much complexity)?

I’m sorry, I’m not familiar with the details of the spec and the project so I’m not sure how helpful I can be at the moment.

That said, at first glance, it looks like this was a known/planned limitation:
https://github.com/commonmark/commonmark-spec/blob/master/spec.txt#L2425

My initial naïve gut feeling is that the rule for HTML blocks of type 6 should be modified so that they do not end at the first blank like if the parser is inside a container element that preserves formatting (this should also affect <script>, <style>, and <textarea> in addition to <pre>). In this special case, newlines should be preserved.

[aral]

I’ve worked around this now in Kitten by replacing empty lines in this scenario with a token that is replaced post render.