Re: [Chicken-users] FFI Question

[felix winkelmann]

On 5/4/06, Heath Johns <address@hidden> wrote:

> I've got a noob question: Why does this segfault?  Should I be doing
> this differently?  I'm using 2.3 stable on linux...

Hi!

No noob question at all. You've found a genuine bug: the result
type conversion didn't  do the right thing for null pointers (#f).
Here is a patch to compiler.scm:

891,893c891,900
<                                           ,@(if (memq rtype 
'(nonnull-c-string c-string))
<                                                 `((##sys#make-c-string (let 
() ,@(cddr lam))))
<                                                 (cddr lam)) )
---
>                                                ,@(case rtype
>                                                    ((nonnull-c-string) 
> `((##sys#make-c-string (let () ,@(cddr lam)))))
>                                                    ((c-string*)
>                                                     (##sys#syntax-error-hook
>                                                      "`c-string*' is not a valid 
> resulte type for callback procedures"
>                                                      name) )
>                                                    ((c-string)
>                                                     `((let ((r (let () ,@(cddr 
> lam))))
>                                                         (and r 
> (##sys#make-c-string r)) ) ) )
>                                                    (else (cddr lam)) ) )

BTW, I recommend using `define-external' (as suggested by Thomas).
In fact I undocumented `foreign-safe-wrapper' and it wil be
deprecated in future releases, since it is kludgy and not really
useful.

(that c-string* works is a mere coincedence - the special treatment of
strings as callback-results didn't take `c-string*' into account).

Thanks for reporting this!


(felix)