Re: gec: Is it possible to specify a signature for a external C call?

>>>>> "Eric" == Eric Bezault <ericb-D6Qt/9opevxWk0Htik3J/[email protected]> writes:

    >> You think this is a hard piece to add to gec?

    Eric> It is probably not hard to add. It's just that I'm not sure
    Eric> this is the right thing to do. If you don't want to write a
    Eric> header file, then I think that it would be better to use
    Eric> ECMA's "C inline" external routines.

Which isn't an option if you want to link to Pascal or Fortran libraries.

    Eric> The fact that it works with SmartEiffel and ISE Eiffel is just
    Eric> luck in my opinion.

It works, because they emit a prototype. That had to be coded, so it's
not luck.

    Eric> Does it work as well when you add assertions to these
    Eric> external routines?

Yes. The created prototypes are just for the called external function.


    Eric> Are these prototypes only added when the external routines do
    Eric> not specify any header files, or do they appear also when
    Eric> there is a header file specified?

Here SE's 1.2r7 code:

  /* Extra external prototype for line 34 of /home/berend/src/eposix/src/capi/spec/se/capi_time.e:*/
  T5 posix_difftime(T2 a1,T2 a2);

If I put in a (dummy) header:

		external "C use <time.h>"

I get this:

/*l34c12/capi_time.e*/#ifdef __cplusplus
}
#endif
#include <time.h>
#ifdef __cplusplus
extern "C" {
#endif
/*l34c12/capi_time.e*//*POSIX_TIME*/


For ISE it generates the prototype in a .c file:

  extern EIF_REAL_64 posix_difftime(EIF_INTEGER_32, EIF_INTEGER_32);


If I add a header with:

		external "C use <time.h>"

No such prototype is created.

So their behaviour is intentional and it allows everyone to link to
libraries without having to know how to write header files (or having to
distribute them if you distribute your C code).

-- 
Cheers,

Berend de Boer


[Non-text portions of this message have been removed]