[cffi-devel] NIL in foreign-funcall arglist

D. I. lovesan.ru at gmail.com
Sat May 16 20:07:26 UTC 2009


Hi there.

Your current implementation of "parse-args-and-types" in src/function.lisp
use destructuring clause in "loop" and tests the end(as i understood) of the
arguments list by comparing second value with nil

(defun parse-args-and-types (args)
  "Returns 4 values. Types, canonicalized types, args and return type."
  (let ((return-type :void))
    (loop for (type arg) on args by #'cddr
          if arg collect type into types
             and collect (canonicalize-foreign-type type) into ctypes
             and collect arg into fargs
          else do (setf return-type type)
          finally (return (values types ctypes fargs return-type)))))

but it also causes foreign-funcall to skip argument if it's value is NIL:
(define-foreign-type ptr-type () ()
        (:actual-type :pointer)
        (:simple-parser some-type))
(defmethod translate-to-foreign (val (type ptr-type))
  (if (null val) (null-pointer) val))

(foreign-funcall "f" :uint x some-type nil :uint y :void)
==>
(LET ((#:G13879 X))
  (LET ((#:G13880 Y))
    (CFFI-SYS:%FOREIGN-FUNCALL "f"
     (:UNSIGNED-INT #:G13879 :UNSIGNED-INT #:G13880 :VOID)
:CALLING-CONVENTION
     :CDECL :LIBRARY :DEFAULT)))

Shouldn't that function be written like this(for example)?

(defun parse-args-and-types (args)
  "Returns 4 values. Types, canonicalized types, args and return type."
  (let* ((len (length args))
         (return-type (if (oddp len) (car (last args)) :void)))
    (loop repeat (floor len 2)
          for (type arg) on args by #'cddr
          collect type into types
          collect (canonicalize-foreign-type type) into ctypes
          collect arg into fargs
          finally (return (values types ctypes fargs return-type)))))

p.s. sorry for my bad english :)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.common-lisp.net/pipermail/cffi-devel/attachments/20090517/53ed1f33/attachment.html>


More information about the cffi-devel mailing list