The tokens T_COMMENT_START, T_TEXT and T_COMMENT_END have been merged into a single token: T_COMMENT. This simplifies both the lexer and the parser.