May I also cautiously suggest that you might be “implementing a language parser, the hard way.”   The CPAN library contains both very-powerful lexers/parsers that are written in Perl, and interfaces to well-known legacy systems.   Perhaps by using one of these tools you can dramatically change (for the better ...) your approach to this problem.   Several LALR-grammars for the C++ language, for example, are already available off-the-shelf.