`sourcepos` is incorrect for `<script>` tags

[GabeIsman]

I've encountered an odd issue. I'm using the Ruby wrapper (Commonmarker) and inspecting the AST shows some strange behavior around the sourcepos of <script> tags. AFAICT this is only related to script tags, and only ones that appear all on one line at that.

Commonmarker.parse("<script></script>")
=> #<Commonmarker::Node(document):
  source_position={:start_line=>1, :start_column=>1, :end_line=>1, :end_column=>17}
  children=[#<Commonmarker::Node(html_block):
       source_position={:start_line=>1,
        :start_column=>1,
        :end_line=>0,
        :end_column=>0}>]>

note the end_line and end_column are both zero. In general I've observed that end_line is start_line - 1, and end_column is always zero.

Let me know if I should open an issue on Commonmarker instead, but I don't see how this could be caused by the wrapper. I just lack the rust expertise to test it directly in rust, sorry!

[kivikakk]

Thanks for the report! This is definitely a Comrak thing. No guarantees on when I'll be able to look into this, but it shoooooould be fairly simple. 🤞

[digitalmoksha]

@kivikakk it looks like this is occurring for the tags <script>, <pre>, <textarea>, and <style>. They get recognized in html_block_start,

comrak/src/scanners.re

Line 152 in 42f4594

[<] ('script'|'pre'|'textarea'|'style') (spacechar | [>]) { return Some(1); }

and returns a 1, and then

comrak/src/parser/mod.rs

Lines 1928 to 1938 in feaf5cf

	AddTextResult::HtmlBlock(block_type) => {
	self.add_line(container, line);
	let matches_end_condition = match block_type {
	1 => scanners::html_block_end_1(&line[self.first_nonspace..]),
	2 => scanners::html_block_end_2(&line[self.first_nonspace..]),
	3 => scanners::html_block_end_3(&line[self.first_nonspace..]),
	4 => scanners::html_block_end_4(&line[self.first_nonspace..]),
	5 => scanners::html_block_end_5(&line[self.first_nonspace..]),
	_ => false,
	};

handles it.

I'm not sure why these would get treated differently than any other HTML block tag, such as a <p>?

[kivikakk]

See https://spec.commonmark.org/0.31.2/#html-blocks.

[digitalmoksha]

Of course, I should've looked at the spec 🤦