Parsing programming languages is an essential
component of the front-end of most tools for
program analysis, manipulation, comprehension, 
visualization, maintenance and reverse-engineering.
One problem with the construction of a front-end
for a programming language is the wide-spread 
ad-hoc treatment of grammarware development,
even for the parsing phase, which is typically
more structured than subsequent phases of
language development.

In this paper we describe our approach to augment 
the GNU gcc C++ parser to generate a parse tree 
for the input program in XML format. We have chosen 
C++ because of the well-documented difficulty 
in developing a parser for the language. To validate 
the generated parse tree, we traverse the tree and 
regenerate the source code, comparing the original 
source with the regenerated source.  We apply our 
approach to three different test suites including 
the gcc test suite that accompanies the compiler 
and two other test suites including the Mozilla 
browser. We evaluate the adequacy of these test suites 
by measuring their coverage of the non-terminals and
the production rules imbedded in the grammar of the
C++ parser. Finally, we exploit mutation testing to 
evaluate the structural correctness of the generated 
parse tree.