Fenced code block absorbing a blank line doesn't propagate the blank to the enclosing list's tight/loose state (other absorbing leaves do)

[numist]

Hello! Back again with a sibling of #608.

When a fenced code block (``` or ~~~) is open and absorbs a blank line as content, and the fence is then terminated by container collapse (rather than by an explicit closing fence), cmark leaves the enclosing list marked tight="true".

Input:

- - ```

  b

Expected:

<ul>
<li>
<ul>
<li>
<pre><code>
</code></pre>
</li>
</ul>
<p>b</p>
</li>
</ul>

Actual:

<ul>
<li>
<ul>
<li>
<pre><code>
</code></pre>
</li>
</ul>
b</li>
</ul>

AST:

<list type="bullet" tight="true">
  <item>
    <list type="bullet" tight="true">
      <item>
        <code_block xml:space="preserve">
</code_block>
      </item>
    </list>
    <paragraph>
      <text xml:space="preserve">b</text>
    </paragraph>
  </item>
</list>

Among the seven multi-line block constructs that absorb blanks (indented code, fenced code, HTML types 1, 2, 3, 4, 5), only fenced code is affected. The other six already update the list's looseness signal correctly when they absorb a blank.

[jgm]

Isn't this because the blank line is part of the code block?
Thus,

- - ```

  b

is the same as

- - ```

    ```
  b

[numist]

Fair enough, but is that reasoning also applicable to any unclosed type-1/2/3/4/5 opener (which cmark currently renders loose)? Or does the dedent implicitly close the code fence?

[jgm]

Or does the dedent implicitly close the code fence?

Yes. Once you have the dedent, the list item container is closed, and the code block is implicitly closed at that point.

[numist]

Makes sense, thanks!