ed,

I would actually recommend against your approach. Not because it isn't a good one, but because Ada is a HUGE language, and (as you have found) tricky to parse. Also, there are alot of Ada tools out there already, and some that may just fit your needs. Here is what I found on Google:

• http://www.grammatech.com/products/
• http://www.littletree.com/products.html
• Rational Software Ada Analyzer
• java-doc-like-tool for Ada
• http://www.scitools.com/uada.html
• http://dmoz.org/Computers/Programming/Languages/Ada/Tools/

Again, I would really recomend using a pre-existing tool, your likely to get much more accurate results/information. And given that you are likely working for either a Defense contractor or Aerospace firm of some kind (who the hell else uses Ada), accuracy of information, I imagine, is of paramount importance.

Again, let me re-iterate this is nothing wrong with you approach, I just don't recomend it for Ada. IMHO Ada is just to large and complex a language for this kind of approach.

NOTE:
The grammer you found is also not 100% standard Backus-Naur form either. This page describes the differences. So along with the conversion from a slightly off Backus-Naur you would need to massage it for either yacc, Parse::YAPP or Parse::RecDescent. I see alot of opportunity for loss of imformation.

-stvn