[MLton] PowerPC .deb built

[Matthew Fluet]

> 4. The unhelpful error checking /,% hit me again.
>    How often is it really the case that these operations will be wrong?

How hard is it to check?  All of the  #if defined  macros are places where 
you should be thinking (even if briefly) about whether or not it works on 
the new platform.

> PS. powerpc/linux does fail one unusual regression:
> 
> testing kitsimple
> 22c22,23
> < ~61403
> ---
> > ~61402
> > *** ERROR ***
> difference with -type-check true

That integer is the  floor  of a computed floating-point number.  Possibly 
fp rounding errors lead to the difference.