Alistair wrote [snip]

...

ps I still think we're better off removing header files completely and having the Haskell type completely determine the calling convention employed. But since I'm not getting any takers on that, I'll settle for pinning down the spec as tightly as possible.

[snip] Perhaps I'm missing something here, but in that case how on earth is the poor user to find out if the Haskell type really corresponds to the C type, before the program coredumps?

I see there is a problem here in that no such check can be carried out when the compilation is not done via C. Would it be possible for the compiler in such a case to spit out a C file which typechecks if and only if the Haskell and C prototypes match, and ask gcc to typecheck it? You could even use this approach in all cases, then C code compiled from Haskell would never need to include user-supplied header files.

Yes, I believe something along these lines could be done. C compilers spit out information about the calling convention in a variety of cunning encodings: .stab debugging entries, warning messages, error messages, etc. It's just a question of figuring out which one. -- Alastair