[MLton] Intel Mac version of MLton

[Eric McCorkle]

On Oct 16, 2006, at 2:55 PM, Stephen Weeks wrote:

>
> Ah OK.  I had thought Eric had run the regressions, but perhaps he
> hadn't.  Eric?

No, I only did a self compile.  The port was the first thing I've  
done with the project, so I didn't know about the regressions.

Looking at John's post, that agrees with what I've seen.  GCC outputs  
something like, say, __dyld_stub_printf (not to be taken for a strict  
adherence to the convention) in place of _printf, which ultimately  
does the liking (this is where I caught it choking on the misaligned  
stack pointer on a movapq instruction).  As far as I can tell, it  
overwrites the stubs with actual _printf calls.

I didn't think about MLton's own output and this convention.  I'll  
look into exactly what Mach-O expects, though I might not be the best  
to actually implement this, as I lack knowledge of the compiler  
internals.

A point I should mention: misaligned stack pointers will show up as  
illegal instruction traps (as opposed to something like a bus error,  
which would make more sense).  Don't be fooled.

-- 
Eric McCorkle,
Brown University
CS Graduate Student