test: expand UTF-16 testing to JRuby

and make sure we're using the BOM correctly

test/xml/test_document_encoding.rb

@@ -8,6 +8,17 @@ class TestDocumentEncoding < Nokogiri::TestCase
  describe "Nokogiri::XML::Document encoding" do
  let(:shift_jis_document) { Nokogiri::XML(File.read(SHIFT_JIS_XML), SHIFT_JIS_XML) }
  let(:ascii_document) { Nokogiri::XML.parse(File.read(XML_FILE), XML_FILE) }
+ let(:utf16_document) do
+ # the document needs to be large enough to trigger a libxml2 buffer flush. the buffer size
+ # is determined by MINLEN in xmlIO.c, which is hardcoded to 4000 code points.
+ size = 8000
+ <<~XML.encode(Encoding::UTF_16)
+ <?xml version="1.0" encoding="UTF-16"?>
+ <root>
+ <bar>#{"A" * size}</bar>
+ </root>
+ XML
+ end
 
  describe "#encoding" do
  it "describes the document's encoding correctly" do
@@ -44,25 +55,37 @@ class TestDocumentEncoding < Nokogiri::TestCase
  assert_equal(Encoding::UTF_8, Nokogiri::LIBXML_COMPILED_VERSION.encoding)
  assert_equal(Encoding::UTF_8, Nokogiri::LIBXSLT_COMPILED_VERSION.encoding)
  end
+
+ it "parses and serializes UTF-16 correctly" do
+ xml = <<~XML.encode(Encoding::UTF_16)
+ <?xml version="1.0" encoding="UTF-16"?>
+ <root><bar>A</bar></root>
+ XML
+ output = Nokogiri::XML(xml).to_xml
+ output_doc = Nokogiri::XML(output)
+
+ # these are descriptive, not prescriptive. the difference is whitespace. this may change
+ # as implementations change. the intention is to verify that they're _roughly_ the right
+ # length, they're not zero or half-width or double-width.
+ expected_bytesize = Nokogiri.jruby? ? 132 : 142
+
+ assert_equal(Encoding::UTF_16, output.encoding)
+ assert_equal("UTF-16", output_doc.encoding)
+ assert_equal(expected_bytesize, output.bytesize)
+ output_doc.at_xpath("/root/bar/text()").tap do |node|
+ assert(node, "unexpected DOM structure in #{output.inspect}")
+ assert_equal("A", node.content)
+ end
  end
 
  it "serializes UTF-16 correctly across libxml2 buffer flushes" do
  # https://github.com/sparklemotion/nokogiri/issues/752
  skip_unless_libxml2
 
- # the document needs to be large enough to trigger a libxml2 buffer flush. the buffer size
- # is determined by MINLEN in xmlIO.c, which is hardcoded to 4000 code points.
- size = 8000
- input = String.new(<<~XML, encoding: "UTF-16")
- <?xml version="1.0" encoding="UTF-16"?>
- <root>
- <bar>#{"A" * size}</bar>
- </root>
- XML
- expected_length = (input.bytesize * 2) + 2 # double character width, add BOM bytes 0xFEFF
+ output = Nokogiri::XML(utf16_document).to_xml
 
- output = Nokogiri::XML(input).to_xml
- assert_equal(expected_length, output.bytesize)
+ assert_equal(Encoding::UTF_16, output.encoding)
+ assert_equal(utf16_document.bytesize, output.bytesize)
  end
  end
  end