Bad parsing of code block inside quotes followed by a code block

[unix-world]

Test

Blockquotes are very handy in email to emulate reply text.
This line is part of the same quote.

H6

one
two
three

ok

let a = 2;
const b = 3;
var a = 4;
class A {
  constructor() {
  }
}

ok

``` go
func fact(n int) int {
	if n <= 1 {
		return n
	}
	return n * fact(n-1)
}
```

Quote break.

This is a very long line that will still be quoted properly when it wraps. Oh boy let's keep writing to make sure this is long enough to actually wrap for everyone. Oh, you can put Markdown into a blockquote.

This is 2nd level

and this is 3rd

Preformatted blocks are useful for ASCII art:

             ,-.
    ,     ,-.   ,-.
   / \   (   )-(   )
   \ |  ,.>-(   )-<
    \|,' (   )-(   )
     Y ___`-'   `-'
     |/__/   `-'
     |
     |
     |    -hrr-
  ___|_____________

END

[unix-world]

Please see the attached file
wrong-markdown.md

[L-four]

I think my issue is related. A code block is causing a indefinite loop.

This minimal reproduction of the issue.

https://goplay.tools/snippet/JQkkXK22Y2S

package main

import "github.com/gomarkdown/markdown"

func main() {
	html := markdown.ToHTML([]byte("```\n: "), nil, nil)
	print(html)
}

[kjk]

@L-four I fixed infinite loop at 87d094f

[kjk]

@unix-world can you check using latest master?

If it still doesn't work, please:

• clarify what "breaking the parser" means?
• minimize the test case i.e. remove as much as possible from wrong-markdown.md
• provide a test case, e.g. based on https://arslexis.io/goplayground/#txO7hJ-ibeU

[unix-world]

Still does not work.
I am using these options:

extensions := mkparser.CommonExtensions | mkparser.HardLineBreak | mkparser.Attributes | mkparser.SuperSubscript
htmlFlags := mkhtml.SkipHTML | mkhtml.LazyLoadImages

This is how gomarkdown is parsing, wrong

This is how it should be, parsed by goldmark

[unix-world]

I will attach you the test Markdown I have, I cannot use the online playground because the markdown contains both " and ` and is too complicate to escape them, so the content is within a file ...

PS: if you look in the previous message screenshots it has to do with a fenced code inside a blockquote.
The goldmark is playing nice there. The gomarkdown seems to have an issue, the fenced code still have the prefix > on each line.

markdown-test.md

[kjk]

It's way too big.

If you can't reduce repro to less than 5 lines, I'm not going to wade through hundreds of lines trying to guess which part you think is parsed wrong.

I need:

• markdown, share it as https://babelmark.github.io/ link (after you press convert, the url contains the code)
• html that we render
• html that you expect to render

[unix-world]

I just isolated the case, this part of the code is rendered wrong, as you can see in the screenshots I atatched you in the previous message.
See the attachment md here, I provided you 2 test cases.
One works, the other have the described issue.

Looks like if I put a fenced code block inside a quoted block and after, somewhere below another fenced block will have this issue.

[unix-world]

See the attachments...
markdown-wrong.md
markdown-ok.md

[unix-world]

I got a fix for this issue, in the file parser/block.go
If I commented out that code it works ...
But now is another issue, a new line is not ending the block.
I don't know what the commented code does, why you compensate that with -1 :-) but at least it fixed the bug I reported here. Please see below:

// parse a blockquote fragment
func (p *Parser) quote(data []byte) int {
	var raw bytes.Buffer
	beg, end := 0, 0
	for beg < len(data) {
		end = beg
		// Step over whole lines, collecting them. While doing that, check for
		// fenced code and if one's found, incorporate it altogether,
		// irregardless of any contents inside it
		for end < len(data) && data[end] != '\n' {
			//-- fix: this portion of the code is breaking the fenced code inside blockquote with another fenced code after the entire bockquote
		//	if p.extensions&FencedCode != 0 {
		//		if i := p.fencedCodeBlock(data[end:], false); i > 0 {
		//			// -1 to compensate for the extra end++ after the loop:
		//			end += i - 1
		//			break
		//		}
		//	}
			//-- #end fix: https://github.com/gomarkdown/markdown/issues/325#issuecomment-2516332756
			end++
		}

[kjk]

Can you confirm that this is the minimized repro:

> ~~~ javascript
> let a = 2;
> ~~~

	code block indent

~~~ go
// another code block
~~~

go playground
babelmark repro

It generates a single code block inside a blockquote.

It should generate 3 code blocks, one inside blockquote.

Other cases:

> ~~~ javascript
> let a = 2;
> ~~~

~~~ go
// another code block
~~~

The simplest case is really:

> ~~~
> // comment
> ~~~

~~~

The last ~~~ is interpreted as end of block code instead of of > ~~~.

[kjk]

Just removing fence code logic inside Parser.quote() breaks other test cases.

[kjk]

The problem is that fencedCodeBlock() called from quote() doesn't see > ~~~ as end of code block. Would probably have to pass insideQuote flag to fencedCodeBlock() and to isFencedLine().

Another case:

> ~~~ go
// comment
~~~

On babelmark those generate blockqote with codeblock, paragraph for "// comment" and another codeblock.

We have very uncommon interpretation of this as a single code block inside blockquote. But this logic breaks when there's a code block after blockquote.

We probably could fix this case without breaking those other test cases.