Invalid HTML produced when using <pre> tag in html mode

[pilotmoon]

When <pre> is used in the input in a certain way, and html mode is on, then markdown-it can produce invalid HTML output:

Demo

Input:

<h1>Heading</h1>
<pre>
First line of text

Second line of text
</pre>

Output:

<h1>Heading</h1>
<pre>
First line of text
<p>Second line of text
</pre></p>

(Note that the <p> tag is erroneously inserted around second line of text and then </pre> is inserted before closing </p> tag, violating nesting rule.)

I checked and I noted that GitHub's markdown implementation handles the same input OK .

[kliu57]

I would like to work on this.

[kliu57]

@puzrin @pilotmoon

I looked into this issue, and I found that although this behavior does not match the GitHub implementation, this is actually expected behavior according to commonmark specs.

However, if you wish for the code to not follow the commonmark specs behavior, I have a proposed fix:

Add pre to the list of valid html block names in lib/common/html_blocks.js, and the result will then match the GitHub implementation behavior. I have tested this and can provide screenshots of results.

Please let me know what behavior is desired.

[pilotmoon]

Thank you for the commonmark reference. What a strange spec! I assume there must be a reason for this decision.

In my case, I have worked around the problem by pre-treating the markdown input as follows:

  html = html.replace(/<pre/g, "\n<pre");  

By inserting a blank before the pre tag, it terminates any initial HTML block preceding it. Thus allowing the <pre block to be terminated by its own end tag instead of the blank line inside.

I shall also now add script, style and textarea to this, for completeness.

[puzrin]

As written above - works as spec says. Details about spec desitions can be seen on commonmark forum.