Handle ASCII-8BIT encoding on fragment input

If we don't know the encoding, we default to using UTF-8.  I think this
was correct behavior on 1.8, but now that Ruby is encoding aware, I
think we should raise an exception if someone passes a document encoded
as ASCII-8BIT.  However, to remain backwards compatible, we'll just
assume UTF-8 for now.

Fixes #553

lib/nokogiri/html/document_fragment.rb

@@ -6,7 +6,17 @@ class DocumentFragment < Nokogiri::XML::DocumentFragment
       def self.parse tags, encoding = nil
         doc = HTML::Document.new
 
-        encoding ||= tags.respond_to?(:encoding) ? tags.encoding.name : 'UTF-8'
+        encoding ||= if tags.respond_to?(:encoding)
+                       encoding = tags.encoding
+                       if encoding == ::Encoding::ASCII_8BIT
+                         'UTF-8'
+                       else
+                         encoding.name
+                       end
+                     else
+                       'UTF-8'
+                     end
+
         doc.encoding = encoding
 
         new(doc, tags)

test/html/test_document_fragment.rb

@@ -9,6 +9,11 @@ def setup
         @html = Nokogiri::HTML.parse(File.read(HTML_FILE), HTML_FILE)
       end
 
+      def test_ascii_8bit_encoding
+        s = String.new 'hello', encoding: Encoding::ASCII_8BIT
+        assert_equal "hello", Nokogiri::HTML::DocumentFragment.parse(s).to_html
+      end
+
       def test_inspect_encoding
         fragment = "<div>こんにちは！</div>".encode('EUC-JP')
         f = Nokogiri::HTML::DocumentFragment.parse fragment
@@ -21,7 +26,7 @@ def test_html_parse_encoding
         assert_equal 'EUC-JP', f.document.encoding
         assert_equal "こんにちは！", f.content
       end
-      
+
       def test_unlink_empty_document
         frag = Nokogiri::HTML::DocumentFragment.parse('').unlink # must_not_raise
         assert_nil frag.parent