defunctirozaion?

[Allyn Dimock]

Stephen Weeks writes:
 > 
 > Hi Allyn.  If the problem is just with missing stuff from the basis library,
 > your best bet to stick with 1999-3-19 version and to insert appropriate stubs in
 > the basis library sources that are distributed with MLton.  If the problem is
 > with a bug in the defunctorizer itself, I can probably find and fix it.  I have
 > lots of old snapshots, and the 1999-7-12 version probably still had a working
 > defunctorizer, but I don't think it would be much different from 1999-3-19.
 > After that, I don't think I have a snapshot with a working defunctorizer.  I
 > haven't kept the defunctorizer up to date, and I doubt a working version will
 > make into a release of MLton again.
 > 
 > If you want, send me the mlton -v output of the example(s) that fail and I'll
 > take a look and advise you how to patch the basis library.

As I recall, the parser for 1999-3-19 was a bit delicate in its
handling of some constructs.  I seem to recall commenting out a lot of
signatures by hand in some (I forget which) benchmark before MLton
1999-3-19 would accept the program.

I wanted to put MLton defunctorization in series with the CIL core ML
compiler to automatically process whole programs containing module
level constructs.  However, I can always pass the input through MLton
1999-3-19 and have the user play with any signature problems that may
arise.

-- Allyn