No subject

[Henry Cejtin]

Given  ML's  decision that the language (as opposed to the library) semantics
should be completely nailed down as far as  order  of  evaluation,  shouldn't
something official be added to the MLton documentation spelling out the order
of evaluation with regards to callcc (and threads).  Is  this  necessary,  or
can  I  already  deduce it?  I was originally worried that MLton might have a
bug with respect to
    val a = (1, ???)
and what the continuation of ??? was.  The point being that it is tempting to
first   allocate  the  tuple  (before  ???),  but  then  by  re-entering  the
continuation we could mutate a tuple.  MLton does the right  thing  (although
it  costs  us  because  it  means that we have to save the result of ??? some
where else and then copy it into the tuple.