[cells-devel] OpenMCL porting progress update

[Frank Goenninger]

On Thu, 2004-10-28 at 21:41, Kenneth Tilton wrote:
> On Oct 28, 2004, at 1:56 PM, Frank Goenninger wrote:
> 
> > Hi Kenny and all the other mostly read-onlies ;-)
> >
> > I have been digging into the GLUT-INIT issue for a couple of hours.
> > First thing to do was to update UFFI to the current stable version of
> > 1.4.28.
> >
> > This fixed a couple of things. Next surprise for me was to discover 
> > that
> > the code in the ffi-extender package was OpenMCL/MCL unaware. So this
> > got me into trouble as Kenny does a lot of things in that package that 
> > I
> > call "advanced stuff".
> 
> Oops. :) Sounds like you are past that though.

No, unfortunately I am stuck here.

Kenny, it would help if you could explain the semantical differences
between the following functions/macros:

ffx::ff-defun-callable
ffx::ff-def-call

(hmm - The question mark in my faces says: When to use which?)

Why are you not just using UFFI's def-function here?

It seems I have missed something very fundamental here...

Another PITA is the the fact that I couldn't find a way of making
callbacks to simply work as expected. I figured that defcallback is a
macro in OpenMCL to be wrapped around the body of the callback, not just
a "register" operation as used in your ffx package...

So: ???

> 
> >
> > This is where I am still just completing the OpenMCL way of doing
> > things. Along that path I had the joy of learning OpenMCL's FFI and 
> > some
> > of UFFI's internals. Becoming better with Common Lisp, guys...
> >
> > Another thing I could find was an interesting part of sample code in 
> > the
> > OpenMCL distro:
> 
> Ah, lucky us! Do they have a working OpenGL demo?
> 
> >
> > x-x-x
> > (in-package :opengl)
> >
> > ;; glut complains if it's initialized redundantly
> > (let ((glut-initialized-p nil))
> >   (defun initialize-glut ()
> >     (let ((command-line-strings (list "openmcl")))
> >       (when (not glut-initialized-p)
> >         (ccl::with-string-vector (argv command-line-strings)
> >           (rlet ((argvp (* t))    ; glutinit takes (* (:signed 32)) 
> > and (* (* (:unsigned 8)))
> > 		 (argcp :signed)) ; so why are these declared as (* t) and :signed?
> > 	    (setf (%get-long argcp) (length command-line-strings)
> > 		  (%get-ptr argvp) argv)
> > 	    (#_glutInit argcp argvp)))
> > 	(setf glut-initialized-p t))))
> >   ;; When a saved image is restarted, it needs to know that glut
> >   ;; hasn't been initialized yet.
> >   (defun uninitialize-glut ()
> >     (setf glut-initialized-p nil))
> >   )
> >
> > (pushnew #'uninitialize-glut ccl::*save-exit-functions*
> > 	 :key #'ccl::function-name)
> > x-x-x
> >
> > I am currently translating this into an UFFI-version and hope to fix
> > the problem calling GLUT-INIT.

Only a quick hack available so far. This needs a lot of polishing to be
production quality.

> >
> > That's it, unfortunately. I am still not at the speed I wish to be...
> 
> Can't wait to see what you do at full speed if you think this is slow. 
> :)
> 
> Over here I have managed to build enough under Lispworks to run 
> lesson-14, except now I have to sort out glutInit and feature out 
> glut-set-option and set up glut-wm-close etc etc., so I am right behind 
> you. Let's keep each other posted.
> 
Congrats! Yeah, but I feel we're in quite good comms.

> btw, I used the latest UFFI and this time Lispworks did not complain 
> about anything. That made me happy. :)
> 
> kt

Well, currently I am a bit frustrated. Several pieces still to be fixed
and no progress for two days... ;-)

Frank
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <https://mailman.common-lisp.net/pipermail/cells-devel/attachments/20041029/5900a4b8/attachment.sig>