The character sequence "]]" is not written inside cdata section. One "]" is discarded.

[lepokle]

Problem

Assume you have a cdata section with the following content

  This is a ]] > test.

(This is how Atlassian Confluence "escapes" end of CDATA within a CDATA section).
If you write this sequence using XMLStreamWriter2.writeCdata you'll the the following section in the output file

This is a ] > test.

Also the initial sequence should be fine from XML point of view.

Reason

Inside ByteXmlWriter.writeCDataContents each character is checked and written to output. If it comes to a ] character it checks the next character for another ] and a finally a >. In this case the end sequence is properly handled.
In case that any other character after the second ] is detected, it will continue writing characters but without writing the first ] character to the output buffer. So it gets lost.

Solution

The other case (no > character follows the double ]) must be properly handled by writing the first detected ] to the output as well:
(line 851)

if (offset < len && cbuf[offset] == ']') {
    if ((offset+1) < len && cbuf[offset+1] == '>') {
        // Ok, need to output ']]' first, then end
        offset += 2;
        writeRaw(BYTE_RBRACKET, BYTE_RBRACKET);
        writeCDataEnd();
        // Then new start, and '>'
        writeCDataStart();
        writeRaw(BYTE_GT);
    }
     else {
        // no end found, write first bracket
        if (_outputPtr >= _outputBufferLen) {
            flushBuffer();
        }
        _outputBuffer[_outputPtr++] = (byte) ch;
    }
    continue main_loop;
}