Jgarber RedCloth

Comments and changes to this ticket

• 

  Jason Garber March 1st, 2010 @ 08:53 AM

  • State changed from “new” to “open”

  Confirmed bug. Here's the test case

  
  

  ---
  name: three newlines after a list
  in: "* a\n* b\n* c\n\n\n"
  html: |-
    <ul>
      <li>a</li>
      <li>b</li>
      <li>c</li>
    </ul>
  

• 

  Jason Garber March 1st, 2010 @ 08:53 AM

  • Tag set to “lists”
• 

  Stephen Bannasch January 14th, 2011 @ 02:19 AM

  • Milestone order changed from “0” to “0”

  A colleague generated a pull request with a new rspec test that exposes this bug:

  https://github.com/jgarber/redcloth/pull/1

  I checked and the bug is present on master.

  I just took a look at the ragel code but it's been quite a while since I've used ragel and it's not quite as easy to parse as Ruby ;-)

  Seems like li_open in lib/redcloth/formatters/html.rb is being called incorrectly when the thrid newline is parsed.

  Here are the locations in the code where li_open is referenced:

  $ ack --all-types li_open
  
    ext/redcloth_scan/redcloth.h
    202:    ASET("type", "li_open");
  
    ext/redcloth_scan/RedclothScanService.java
    95:     ASET("type", "li_open");
  
    lib/redcloth/formatters/html.rb
    50:  def li_open(opts)
  
    lib/redcloth/formatters/latex.rb
    103:  def li_open(opts)
  
    ragel/redcloth_scan.java.rl
    94:     ASET("type", "li_open");
  
    ragel/redcloth_scan.rb.rl
    309:      ASET("type", "li_open")
  

  Here's the new spec test:

  it "should not add spurious li tags to the end of markup" do
    input         = "* one\n* two\n* three \n\n"
    failing_input = "* one\n* two\n* three \n\n\n"
    RedCloth.new(input).to_html.should_not match /<li>$/
    RedCloth.new(failing_input).to_html.should_not match /<li>$/
  end
  

  When I run the spec test in the debugger:

  $ rspec -d spec/parser_spec.rb
  

  RedCloth::Formatters::HTML.li_open(opts) is called once for each of the three list items in the first test:

  RedCloth.new(input).to_html.should_not match /<li>$/
  

  and then called again once for each list item in the second test:

  RedCloth.new(failing_input).to_html.should_not match /<li>$/
  

  and then it is called a seventh time with this argument: {:type=>"li_open", :nest=>0, :text=>""}

     50    def li_open(opts)
     51      debugger
  => 52      "#{"\t" * opts[:nest]}<li#{pba(opts)}>#{opts[:text]}"
     53    end
  

  values of argument opts:

  1: {:type=>"li_open", :nest=>1, :text=>"one"}
  2: {:type=>"li_open", :nest=>1, :text=>"two"}
  3: {:type=>"li_open", :nest=>1, :text=>"three"}
  4: {:type=>"li_open", :nest=>1, :text=>"one"}
  5: {:type=>"li_open", :nest=>1, :text=>"two"}
  6: {:type=>"li_open", :nest=>1, :text=>"three"}
  7: {:type=>"li_open", :nest=>0, :text=>""}
  

  The nesting level is 0 and the text is an empty string ...

  Is there a way to tell more about what the ragel machine has been doing and why li_open was called this last time?

• 

  Stephen Bannasch January 14th, 2011 @ 12:53 PM

  Fixed in new pull request: https://github.com/jgarber/redcloth/pull/2

• 

  Jason Garber January 23rd, 2011 @ 07:30 PM

  • State changed from “open” to “resolved”

  Accepted! Thanks!