The release note builder parses pull request descriptions, looking for a "# Summary" header. It will recognize it at any depth, and will then look forward from there for the next line that starts with that many or fewer hash symbols (to try and anticipate the next header of the same or higher level).

This works fine, until you have a code block with a comment in it, e.g.

## Summary

blah blah blah
blah blah blah

```ruby
# This comment will be interpreted as markdown header by the parser!
def do_something
end
```

blah blah blah

The parser will treat the comment in that code block as a markdown header, note that it is higher level than the "## Summary" header, and terminate the summary at that point. We need to teach the parser to ignore code blocks.