parser crashes on wordgraph example

[jeisner]

wget http://cs.jhu.edu/~jason/tmp/edit.dyna
wget http://cs.jhu.edu/~jason/tmp/swapadj5.dyna
wget http://cs.jhu.edu/~jason/tmp/swap5.dyna
wget http://cs.jhu.edu/~jason/tmp/anagram5.dyna
wget http://cs.jhu.edu/~jason/tmp/subst5.dyna
cat edit.dyna *5.dyna > wordgraph.dyna

I seem to be able to load these files individually with dyna -i, but when I try to load their concatenation I get this crash:

jasoneis@a14:~$ dyna -i wordgraph.dyna
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
/home/nwf/src/dyna2/src/Dyna/Backend/Python/interpreter.py in <module>()
    688 
    689 if __name__ == '__main__':
--> 690     main()

/home/nwf/src/dyna2/src/Dyna/Backend/Python/interpreter.py in main()
    668 
    669         try:
--> 670             interp.do(plan)
    671         except DynaInitializerException as e:
    672             print e

/home/nwf/src/dyna2/src/Dyna/Backend/Python/interpreter.py in do(self, filename, initialize)
    512 
    513         # TODO: this should be a transaction.
--> 514         for k, v in env.agg_decl.items():
    515             self.new_fn(k, v)
    516 

AttributeError: 'module' object has no attribute 'agg_decl'
FATAL ERROR (AttributeError): 'module' object has no attribute 'agg_decl'
Crash log available /home/jasoneis/.dyna/crash.log
Saving crash as jasoneis:2013-07-03:18:01:13:19675.log ...

[jeisner]

Meanwhile, @timvieira reports:

$ cat *5.dyna > graph-data.dyna
$ dist/build/dyna/dyna graph-data.dyna 
Compiler panic!
This is almost assuredly not your fault!  Please contact a TA.
 Uncaught Haskell exception: stack overflow

and comments:

These files seem to crash the parser... haskell throws a stackoverflow exception then the parser exit with status 0.
@nwf can you fix the return code (it shouldn't be 0) and look into why the parser doesn't like these files?

[nwf]

The stack overflow remains mysterious despite some time investigating; I'll pick it up again tomorrow. The compiler now at least exits with a failure when it happens.