TODO.txt

WALi complaints:
  - Non-const SemElem::extend/combine; sem_elem_t should be ptr to const
    SemElem. A couple of other functions should also be const.

Non-NWA tests:

 - Except for some tests of Key stuff in the NWA tests, there are none.

   OK, now there are some things testing new WFA functionality. Still bad.


NWA tests:

 - We have some tests for the getKey and key2str functions -- other than a
   raw key_src_t. It would be kind of nice to add those.

 - We don't have tests for the functions in tables 2-4 (transition query
   functions), except insofar as they are called by other functions. (This is
   actually a majority of the non-covered code that's part of the public
   interface of the NWA portion of the library.)

 - We don't have tests of the NWA's prestar/poststar functions.

 - We don't say what happens if EPSILON labels a call or return edge, so
   I don't have any tests for it.

 - We have almost no tests for WILD.

 - We don't have any tests for print_dot, and probably not much for print or
   read_nwa (or read_nwa_proc_set).

 - WpdsToNwa, NwaToWpds*, plusWpds are not really tested at all. We call them
   with an empty NWA or empty WPDS, mostly to make sure they
   compile. (NwaToWpdsCalls() is used by languageIsEmpty() and
   getSomeAcceptedWord(), so it is actually tested reasonably decently
   as-is, except that we don't really use fancy weights so wouldn't notice if
   it's calling the wrong weightgen function.)

 - We don't test *any* of the client info callbacks in determinize or
   intersect, and we don't really test the calls to the weightgen functions
   in the NWA->WPDS conversions.

 - intersect() is lacking in coverage a little.

 - Nothing in 'details' namespaces is tested directly, nor are a number of
   functions that did not make it into the Latex documentation.