[MLton] Few problems with porting...

[Stephen Weeks]

> Yes, recent changes tickled a long-standing SSA simplifier bug, which is
> the error you are seeing.  Steve described the problem in a little more
> detail in a recent commit:
> http://www.mlton.org/commitlog

I have now discovered that in addition to tickling the SSA bug, the
code was just plain wrong when compiling with the stubs.  Matthew and
I didn't see this because we first built the early version with the
stubs and then built the later version without stubs.

In any case, I have checked in a fix.  Fortunately, this fix no longer
tickles the SSA bug, which was probably triggering due to lots of dead
code, so you can once again compile the cvs HEAD with older MLtons.

> We've been using the working versions of SML/NJ for a long time.  The
> current CVS MLton sources work with SML/NJ 110.43; the changes to work
> with SML/NJ 110.44 are trivial.

If you've done these could you go ahead and check them in.  I'd be
happy to upgrade.

> Hopefully, we will very soon transition to a point were we don't need
> SML/NJ to check the basis.

We're there I'd say.  All of the front end is in place.  All that's
left is bug reports at this point.  Which I eagerly await.