[EMAIL PROTECTED] (Marcin 'Qrczak' Kowalczyk) wrote,
> Sorry to reinvent the c2hs. I think it should be simple (probably just
> one small Main.hs), and it does not work to infer types of functions
> from their C types (because of typed pointers, ForeignObj, newtypes)
> so foreign import declarations are supposed to be written manually.
> The library will help with memory allocation, translating arguments
> and results; the tool will fill bits that require access to C headers.
That's fine with me - I guess, we have to explore all
interesting bits of the design space to see what works best.
However, I would suggest that we try to align the
marshalling libraries of the two tools - and ideally use as
much of the yet-to-be-defined high-level marshalling library
of the FFI itself.
I didn't react to Sven's earlier posting re the high-level
marshalling library, because I am seriously overloaded with
teaching at the moment. However, I still think that we
should find a common denominator here.
What concerns me a bit about your high-level marshalling
code that I have seen flying around earlier is that it
wasn't clear anymore whether it all fits into the standard
Haskell 98 definition of type/constructor classes. So,
maybe you can briefly comment on whether your current code
stays within Haskell 98 or not.
Cheers,
Manuel
