[MLton] Intel Mac version of MLton

[Matthew Fluet]

> Can anyone explain why John Reppy saw the problem with library
> functions (like tanh) on Intel Mac and Eric McCorkle didn't?

A simple explaination would be that John ran the regression suite, while 
Eric just performed a self-compile.  I don't believe that mlton, mllex, 
mlyacc, or mlnlffigen (the four executables produced by a top-level 
'make') need to reference the trancendental math functions.