DocumentFragment.parse of certain attribute values mangles document on JRuby

[jmcnevin]

Nokogiri 1.5.5
JRuby 1.6.7.2

There seems to be an issue on JRuby with attribute values beginning with something that looks like a protocol (namely, characters, and then a colon). This causes the document to be parsed incorrectly.

I created a small script to test this:

require 'nokogiri'

test_string = <<-EOF
<p>This is a sample document that has been created as an example of a link to a file that is not an .html document.</p>
<p>
  <img src="embedded:image1.png" alt="image1.png" />
</p>
EOF

doc = Nokogiri::HTML::DocumentFragment.parse(test_string)

puts doc.to_s

CRuby:

<p>This is a sample document that has been created as an example of a link to a file that is not an .html document.</p>
<p>
  <img src="embedded:image1.png" alt="image1.png"></p>

JRuby:

<p>This is a sample document that has been created as an example of a link to a file that is not an .html document.</p>
<p>
  <img image1.png="">
</p>

[yokolet]

Hello!

I got below on Nokogiri master and JRuby 1.7.0.preview2:

<p>This is a sample document that has been created as an example of a link to a file that is not an .html document.</p>
<p>
  <img alt="image1.png" src="embedded:image1.png">
</p>

So, it looks the bug has been fixed by another bug fix.

[jvshahid]

Fixed.