[MLton] patch for CM cc variable support?

[Stephen Weeks]

> May I request this small change to the way that CM works in mlton? 
> Right now, there are no defined symbols whatsoever, . which (as far
> as I can tell) makes it impossible to use the conditional
> compilation system to make a CM file that works with both mlton and
> the newest sml/nj.

According to http://mlton.org/CompilationManager, the following should
work.  I just checked, and indeed it does.

  In SML/NJ, the preprocessor defines the symbol SMLNJ_VERSION. In
  MLton, no symbols are defined. So, to conditionally include foo.sml
  when compiling under SML/NJ, one can use the following pattern.

  # if (defined(SMLNJ_VERSION))
  foo.sml
  # endif

  To conditionally include foo.sml when compiling under MLton, one can
  negate the test.

  # if (! defined(SMLNJ_VERSION))
  foo.sml
  # endif

> It might also make sense to modify line 408 of lexer.sml to return
> SOME 0 for undefined symbols, as SML/NJ apparently does. 

I guess you mean line 404?  I have no strong objection to such a
change, if that's what people in the CM world would expect.  But I
worry that such a change would cause more confusion than reporting an
error, as conditionals would be silently taken (or not).  Since MLton
doesn't define any variables, I don't really see the use for such a
thing, or any useful overlap with SML/NJ.  Also, with the above idiom,
it should be possible to wrap stuff that uses SML/NJ CM variables so
that it only runs under SML/NJ, right?  Anyone else have any thoughts?