[Openmcl-devel] Finding the CDB interface files

[Cyrus Harmon]

I've done a bit of this with the vimage framework with reasonable luck.  
I'll try and package this up in the next couple days, but


the general idea is:

======
(require "PARSE-FFI")
(ccl::parse-standard-ffi-files :vimage)
======


and then (after appropriate package def, etc...)

======
(in-package :vimage)

(defun add-logical-pathname-translation (host log phy)
   (let ((q (assoc log (logical-pathname-translations host) :test  
#'equalp)))
     (if q
         (rplacd q (list phy))
         (setf (logical-pathname-translations host)
               (append (list (list log phy))
                       (logical-pathname-translations host)
                       )))))

(defun local-directory (&rest path)
   (print (merge-pathnames
    (make-pathname :directory (apply 'list :relative :up :up path))
    (make-pathname :directory (pathname-directory *load-truename*)))))

(print *load-truename*)
(add-logical-pathname-translation "ccl"
                                   #4P"ccl:darwin-headers;vimage;**;*.*"
                                   (merge-pathnames
                                    "**/*.*"
                                    (local-directory "darwin-headers"  
"vimage")))

(add-logical-pathname-translation "ccl"
                                   #4P"ccl:darwin-headers;vimage;*.*"
                                   (merge-pathnames
                                    "*.*"
                                    (local-directory "darwin-headers"  
"vimage")))
======

And then to use:


======
(open-shared-library  
"/System/Library/Frameworks/Accelerate.framework/Frameworks/ 
vImage.framework/vImage")
(ccl:use-interface-dir :vimage)
======


I'll try and clean up, document etc... one of these days.

Cyrus




On Jan 23, 2005, at 3:51 PM, Gary Byers wrote:

>
>
> On Fri, 21 Jan 2005, David Steuber wrote:
>
>> I keep my openmcl in /Users/david/usr/src/ccl/.  Openmcl has no
>> difficulty finding the interface files when started from SLIME or
>> running the openmcl script in Terminal.  However, when I save an image
>> with save-application and use it in an application bundle, it can't
>> find the interface files (I use slime-connect to talk to the running
>> application).  What form do I need to have evaluated to fix that  
>> issue?
>
> The Darwin interfaces are presumed to be in the directory denoted by
> the logical pathname "ccl:darwin-headers;", so the question is sort
> of "how do logical pathname translations for the logical host CCL
> get set up?", or, if you prefer, "what physical directory does the
> CCL logical host use as the base for all relative logical pathnames?"
>
> The function CCL::CCL-DIRECTORY is called each time the lisp starts
> up to try to provide an answer to this question.  (There are many
> potential answers, and it'd be nice to make this a little more
> flexible. perhaps by defining a method on the application class.)
> CCL::CCL-DIRECTORY tries to use (in order):
>
>   1) the value of the "CCL_DEFAULT_DIRECTORY" environment variable,
>      which is ordinarily set by the "openmcl" shell script;
>   2) a Resources/ccl directory, relative to the pathname of the
>      executable.  (e.g., if the executable file resides in
>      "MyApp.app/Contents/MacOS/dppccl", it will look for
>      "MyApp.app/Contents/Resources/ccl/", which IIRC can be a
>      symbolic link to the "real" CCL directory);
>   3) the current directory
>>
>> A nearly related question:  Building the FFI files does not work for  
>> me
>> because I do not have GCC 2.95 on my system.  Is there anyway for the
>> FFIGEN stuff to work with GCC 3.3?
>
> Helmut Eller got FFIGEN working with GCC 3.x several months ago.  I
> haven't had time to make binaries available (or even to help other
> people who offered to do so.)
>
>>
>> Going even further astray, has anyone got any C header file parsing
>> code in Lisp that can generate Lisp code for using the other FFI
>> functions like EXTERNAL-CALL?  And can I manually declare foreign  
>> types
>> that will work with RLET/PREF, etc?
>
> Defining foreign types is, AFAIK, documented.
>
> There should be a documented way of doing what #_ does (without
> relying on the interface database.)  IIRC, the undocumented way got
> broken quite a while ago.
>
>>
>> This is in aid of being able to use agl.h along with AGL.framework
>> without killing my fingers typing huge amounts of FFI code.
>> QuickTime.framework is in the future as well if I get that far.
>>
>
> You really should be able to do this more easily.
>
> As an experiment, I just tried taking 10 minutes away from my other
> job on a Sunday afternoon.  There are archives containing AGL and
> QT interfaces on the FTP server; so far, the earth hasn't started
> spinning off its axis.
> _______________________________________________
> Openmcl-devel mailing list
> Openmcl-devel at clozure.com
> http://clozure.com/mailman/listinfo/openmcl-devel