Generating Function Prototypes

[Simon Marlow]

> Simon Marlow wrote:
> > [...] But I don't see a way around it!  [...]
> 
> What's wrong with  Alastair's suggestion to post-process the
> generated assembler code? We mangle it anyway with -f-via-C,
> and the additional mangling required looks quite stable and
> easy to implement.

I forgot about that.  Hmm.

I can see one drawback: if mangling is required in order to compile FFI
code, then it means we won't be able to do unregisterised compilation.
Or at least unregisterised compilation would need some extra machinery,
like header files or auto-generated FFI wrappers.  That's a bit of a
downer for portability.  But if this problem can be solved, then I guess
I'm not against the idea, if someone wanted to implement it...

Cheers,
	Simon