Fix the grammar for scanning table marker rows #55
Conversation
Scan all the way to newline. This makes some edge cases parse in accordance with upstream cmark.
| "<p>a\n-b\n-c</p>\n", | ||
| |opts| opts.ext_table = true, | ||
| ); | ||
| } |
There was a problem hiding this comment.
This would previously parse as a 1-column table with header "a" and 1 row of content, "-c", parsing "-b" as a valid table marker row.
|
This is perfect, thank you! Indeed, this now matches https://github.com/github/cmark/blob/bf28ef6e7e949157ab85a5e4a3ea9107e17497b9/extensions/ext_scanners.re#L41 correctly. The Travis CI build checks against the current version of the CM spec we're tracking, including the GFM specs; see https://travis-ci.org/kivikakk/comrak/jobs/364882300. I might need to pull a later version of upstream in to see what new failures are. |
I've pulled in the latest upstream and matched the |
2 participants
Scan all the way to newline. This makes some edge cases
parse in accordance with upstream cmark.
Looks like this might have been a typo in the grammar.
I've confirmed this doesn't regress comrak against the cm spec (it has 5 existing failures), though I don't know where the gfm spec source is.