Re: [Chicken-users] SWIG and c-callback with arguments

[Bastian Müller]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Am 20.08.2007 um 01:38 schrieb felix winkelmann:
> On 8/20/07, Bastian Müller <address@hidden> wrote:
> > >
> > The code doesn't crash. The callback get's called. That's not the
> > problem.
>
> Oh, well. Then everything is fine.
>
> > What's missing, is the support for argument passing.
>
> C_save should work ok. What exactly does happen?

In ode.i:
static void
EGG_near_callback (void *closure_root, dGeomID o1, dGeomID o2) {
  C_save(o1);
  C_save(o2);
  C_callback(CHICKEN_gc_root_ref(closure_root), 2);
}

scheme callback:
(define (near-collision geom1 geom2)
  (write "near collison") (newline)
  (let ((body1 (ode-geom-get-body geom1))
        (body2 (ode-geom-get-body geom2))
        ...

>>>
"near collison"
Error: Type error in argument #1: expected struct dxGeom *|dGeomID:  
"bad argument type"

I think EGG_near_callback gets called in the ODE internals with real
dGeomIDs and near-callback get's real dGeomIDs. The problem is,
that all swig-interface functions don't handle the real data but
swig-pointers to the actual data. That means, ode-geom-get-body
expects an swig-pointer to an "struct dxGeom *|dGeomID" and not
the actual dGeomID. I think the error-message is in this case somehow
misleading.

Is that correct? If so, any idea how to fix that?

Thank you very much!

- --
Bastian Müller - http://turbolent.com/


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFGyWuEUFTPKTPeTPURAvf2AJ0a2IWpxs1H9MV166u+syFGyX63bgCeOaOI
J15kSKex4qWURkofgSyudmk=
=oSWJ
-----END PGP SIGNATURE-----