Wrap the EncodingReader in a java wrapper to handle the case when encoding detection succeed and the stream is missing the first 1K of data (in which case we cannot use the underlying stream directly, and instead rely on the EncodingReader::read() to do the right thing). the other case is when the encoding detection doesn't succeed and EncodingReader doesn't store the first chunk of data. In this case the wrapper will store the first chunk of data and return it the next time NokogitiEncodingReaderWrapper::read() is called.
…ename the tests to test_<parser_name>_entity_reference_<options>.
…G_JAVA was set.
…s due to duplicate declaration of notion and entities.
…nd is skipped.
…havior under C nokogiri vs. Java Nokogir.
…ssed from Nokogiri::XML::Node.
…t have a reference node correctly decorated.
… has cut one 1.9 error off, too.
… its text method.
See related commit 5c5e1b5
… that we can test JRuby on Travis-CI. See https://groups.google.com/group/nokogiri-core/browse_thread/thread/2229680cdb8d93bb for background.
…ction available, skips the test.
Merging text nodes when adding a new node has the following two special cases: 1. Try to merge with the previous node if this is one of the following: add_child, add_next_sibling or replace 2. Try to merge with the next node if this is one of the following: add_previous_sibling or replace