Re: dynamic ffi and C struct

[Nala Ginrut]

And if you want to dereference the returned struct pointer, use parse-c-struct
say, there's a returned C struct like:
aa { (int)1 ,(int)2 }

you may convert it to list like this:
(parse-c-struct aa (list int int))
==> (1 2)

Then the rest is list operation, has no part in C.


On Sun, May 20, 2012 at 8:45 AM, Mark H Weaver <address@hidden> wrote:
> cong gu <address@hidden> writes:
>> Is it possible to use the current dynamic ffi to call a C function
>> whose parameter is a C struct (not a pointer to a C struct) ?
>
> Although it is not documented in the manual, I see from the source code
> that a C struct parameter can be specified using a nested list in the
> foreign type passed to 'pointer->procedure'.
>
>> For example, what should I do for the following function?
>>
>> typedef struct {
>>   int dat[2];
>> } foo_t;
>>
>> int func ( foo_t arg );
>
> For this example, there's an additional complication of the inline array
> within the struct, which is not supported by libffi, upon which Guile's
> dynamic FFI is based.  I'm not an expert on this subject, but I strongly
> suspect that on most (if not all) platforms, the struct above has the
> same layout as the following one:
>
>  typedef struct {
>    int dat_0;
>    int dat_1;
>  } foo_t;
>
> Assuming this is the case, here's one way to wrap 'func':
>
>  (use-modules (system foreign))
>
>  (define foo-library-obj (dynamic-link "libfoo"))
>  (define foo-struct-type (list int int))
>
>  (define (make-foo-struct dat-0 dat-1)
>    (make-c-struct foo-struct-type (list dat-0 dat-1)))
>
>  (define func
>    (pointer->procedure int
>                        (dynamic-func "func" foo-library-obj)
>                        (list foo-struct-type)))
>
> and here's an example of how to use it:
>
>  (func (make-foo-struct 4 5))
>
> It is also possible to wrap C functions that return structs, by passing
> a list to the first parameter of 'pointer->procedure', and then using
> 'parse-c-struct' to extract the fields of the returned struct.
>
> Hope this helps!
>
>    Mark
>