[MLton] re: commit r4541

[Matthew Fluet]

>> When it comes to silencing warnings, I think it's better to leave a
>> correct warning alone.
>
> I disagree.  I think that having warnings around in the default
> compile makes it more likely that we will miss new warnings in the
> future.  I'm open to fixes either way that remove the warnings in this
> particular case.  But in the absence of that, I'd prefer warnings on
> Darwin to warnings on Linux.

I think we can push the const qualifier through the runtime.  Especially, 
in light of the fact that the use of the 4-byte aligned 'pointer' type for 
code addresses and program counter isn't correct on x86.