[m-dev.] export declarations

[Andrew Bromage]

G'day.

Peter Schachte wrote:

> There is no good reason I can think of for the Mercury language (as
> distinguished from the current implementation) to imply that C's calling
> conventions are the only ones.

Uhm... I may be missing something here, but isn't a pragma by definition
not part of the language, as distinguished from an implementation "extra"?

(Now that I think about it, if this is true, perhaps our implementation
should warn about unknown pragmas rather than signal an error as we
currently do.)

Cheers,
Andrew Bromage