[racket] FFI: Giving a double pointer to a function

Stepping through the program when debugging, it (the Racket runtime)
now crashes when running (cpArbiterGetShapes arb) with the
modifications you suggested.
(also I just pushed the changes to the GitHub repo for the bindings:
https://github.com/Freezerburn/Rhipmunk-Physics if you want to take a
look at them directly)

On Sun, Dec 2, 2012 at 8:23 PM, Ryan Culpepper <ryan at cs.utah.edu> wrote:
> Based on the documentation I found and the CP_ARBITER_GET_SHAPES macro, it
> looks like the double-pointer arguments are just used for multiple outputs.
> So you should use the following type for the function:
>
> (_fun _cpArbiter-pointer
>       (out1 : (_ptr o _cpShape-pointer))
>       (out2 : (_ptr o _cpShape-pointer))
>       -> _void
>       -> (values out1 out2))
>
> The FFI takes care of allocating the necessary temporary space that the
> function writes into. The Racket-side function takes just one argument (for
> the arbiter), and it returns two _cpShape-pointer values. So you would use
> it like this:
>
>   (let-values ([(water poly) (cpArbiterGetShapes arb)])
>     ___
>     (let ([count (cpPolyShapeGetNumVerts poly)])
>       ___))
>
> Does that work?
>
> Ryan
>
>
>
> On 12/02/2012 07:59 PM, Vince Kuyatt wrote:
>>
>> So I recently started to poke at my humble little bindings to the
>> Chipmunk library again. (someone sent me an email asking for help,
>> which reminded me that I hadn't worked on it in a while) The problem
>> is that I'm still stumbling over the same block that made me stop in
>> the first place a number of months back. One of the function in the
>> Chipmunk library (cpArbiterGetShapes) takes two double pointers to a
>> cpShape. (cpShape**) So far, I have been unable to figure out how to
>> make this work without a future function that accesses the cpShape
>> crashing the entire Racket runtime, including the IDE window I have
>> open.
>>
>> So the function definition for cpArbiterGetShapes is:
>> static inline void cpArbiterGetShapes(const cpArbiter *arb, cpShape
>> **a, cpShape **b)
>>
>> And my current binding in Racket is:
>> (_fun _cpArbiter-pointer
>>                (_ptr io _cpShape-pointer)
>>                (_ptr io _cpShape-pointer)
>>                -> _void)
>>
>> And the (pertinent) code where I use it is:
>> (let*
>>        ([water (cast (malloc _cpShape-pointer) _pointer _cpShape-pointer)]
>>         [poly (cast (malloc _cpShape-pointer) _pointer _cpShape-pointer)]
>>         [water-ptr (cast (malloc _cpShape-pointer) _pointer
>> _cpShape-pointer)]
>>         [poly-ptr (cast (malloc _cpShape-pointer) _pointer
>> _cpShape-pointer)])
>>      (memcpy water-ptr water (ctype-sizeof _cpShape-pointer))
>>      (memcpy poly-ptr poly (ctype-sizeof _cpShape-pointer))
>>      (cpArbiterGetShapes arb water-ptr poly-ptr)
>>      (let*
>>          (...
>>           [count (cpPolyShapeGetNumVerts poly)] ;Note that it crashes
>> on this line
>>
>> Which is following the buoyancy demo included in Chipmunk by default
>> as close as possible. Also note that cpPolyShapeGetNumVerts is
>> supposed to take a _cpShape-pointer. I'm really not sure how to do
>> what I need in Racket given this situation, and all I've really seen
>> in the Racket documentation is just single pointers. I've also tried
>> just allocating water and poly with not water-ptr and poly-ptr and
>> just using those, but that crashes Racket just the same. I would
>> really, really, really appreciate any help with this. It's bugging me
>> that I can't get this solved and I really want this to work.
>>
>> Thank you so much in advance for any assistance provided. (also I
>> apologize if this ends up badly formatted. I'm not sure of formatting
>> rules for a listserv when sending through email)
>> ____________________
>>    Racket Users list:
>>    http://lists.racket-lang.org/users
>>
>