Repository navigation
Definition + setext underline/thematic break, creates loose empty paragraph #262
Description
I think this is a bug in commonmark.js:
[a]: b
---
[a]: b
===
On the dingus, it yields:
<p></p>
<hr />
<p>===</p>Rendered here (cmark-gfm + more GH stuff, likely same as just cmark):
<p>---</p>
<p>===</p>---
===
- I don’t think a random empty
<p></p>should show up, that’s a clear bug. - I probably think, but this is more vague, that
<p>---</p>makes more sense here, as definitions/paragraphs/setext heading underlines interplay weirdly (Mention that link reference definitions are constructed from paragraphs commonmark-spec#605).
Activity
jgm commentedon Nov 30, 2022
One observation: if you change the setext heading line regex to something that will never be matched, the empty p tag disappears. So this is due to some interference with setext heading parsing.
wooorm commentedon Nov 30, 2022
Similar to what you note, I think the area around definitions/setext headings/lists is very hard to think about. One example: wooorm/markdown-rs#22 (comment). I have no idea what should happen in these cases. Which comes first? Which construct has precedence, and what happens if it doesn’t match
jgm commentedon Nov 30, 2022
I remember there were some discussions about this -- you could check the tracker (including closed issues) fro commonmark-spec.
Add regression cases from commonmark#262 for definitions before setex…
Add regression cases from #262 for definitions before setext underlines
jgm commentedon Oct 13, 2026
Fixed by #314.