Tests failing in a fresh checkout

[pdubroy]

I just cloned master and ran node test, and got the following failure:

#10. def_blocks.text failed at offset 20. Near: "<blockquote><p>hello</p><p><ahref="foo">1</a>:hell".


Got:
<blockquote><p>hello</p><p><ahref="foo">1</a>:hell


Expected:
<blockquote><p>hello[1]:hello</p></blockquote><hr>

Not sure if this is a known issue or not, but thought I'd open a bug in case you weren't aware of it.

[guypursey]

+1 Ditto.

[guypursey]

Line 2 in ./test/tests/def_blocks.text

> [1]: hello

is being treated like a reflink because of line 20

[1]: foo

One can see this is expected behaviour by comparing ./test/tests/links_reference_style.text with ./test/tests/links_reference_style.html.

If def_blocks.text is to pass, the reflink inline lexer (lines 626-641 in ./lib/marked.js) needs to somehow skip reflinking when we are in a blockquote. But having only just looked into the code I can't quite see how to do that. It might be simple but I can't see how to perform the check within the function; it looks like a string is simply passed to the InlineLexer.prototype.output function to be formatted without context.

Apologies if this is all stating the obvious. Happy to keep looking but thought I would document this here in case someone more familiar with the code can get to a fix quicker.

[guypursey]

It might be that want to keep reflinking within blockquotes, of course, in which case it's simply a matter of rewriting ./test/tests/def_blocks.html. One to discuss?

[andre-artus]

Has any decision been made on this?

I see two issues here:

1. Ref links should require two sets of square brackets.
   As @guypursey pointed out the first instance of [1] is treated as a ref link to the second instance.
   This should not happen on a single pair of brackets.
2. The test is broken (def_blocks.text, and def_blocks.html).
   Assuming that this test pair are intended to contain ref links then both the input (def_blocks.text)
   and expected result (def_blocks.html) ought to be modified accordingly.

[KostyaTretyak]

It looks like you just need to delete these lines and fix the tests like this.

[joshbruce]

Closing as stale. "Fresh" was three years ago. Things have most likely changed a great deal.

See also #956

And #958

Please do contribute solutions back to the main line.

[KostyaTretyak]

Things have most likely changed a great deal.

But the fact that the current version marked (from master) does not pass this test in the same way. For three years, this simple test has been failing all the time...

[joshbruce]

Agreed. There are two failing tests in the PR referenced. If you know how to fix them (as it seems you do??), please do so and contribute back. I don't know enough about Marked to do it myself; otherwise, I would.

One person can't maintain this library. As evidenced by #956

[KostyaTretyak]

Done #974

[Feder1co5oave]

From the commonmark spec, I see that link definitions are allowed inside blockquotes too:
http://spec.commonmark.org/0.28/#example-181

So wrt commonmark, the def_blocks test looks wrong to me.