[Cffi-devel] Fw: Fw: What is the proper way to do GC in CFFI

[Joeish W]

If you get some free time I would love to see a concrete example, but I do appreciate you time so far and don't want to put you out in any way.. I'm building a big library and the better I can make each individual component the better

On Sunday, April 13, 2014 8:26 AM, Willem Rein Oudshoorn <woudshoo at xs4all.nl> wrote:
 
Joeish W <joeish80829 at yahoo.com> writes:
>
>> I already figured out a great finalizer, I could just use help on one thing.
>> The below finalizer is for the %mat defcfun below that.
>> How do I update the finalizer below to be for the mat-data defcfun at the 
>> bottom of the page. So where do I put the rows cols params, there seems 
>> to be no place for them in the defstruct wwhere %mat is called. Any
>> help is appreciated 
>
>Hm, I feel this is not the right way to approach this. 
>However, there is a bug in your finalize logic.
>
>1. If you call make-cvmatrix with enable-finalizer true you attach a
>   finalize on the new instance of cvmatrix.  
>2. However you return the SAP.  
>3. As a consequence the cvmatrix instance can be immediately garbage
>   collected, even if the returned SAP still has references to it.
>
>        
>
>> (defstruct (cvmatrix (:constructor %make-cvmatrix)) 
>>   (sap (%mat) :type sb-sys:system-area-pointer :read-only t)) 
>>  
>>  (defun make-cvmatrix (&optional enable-finalizer) 
>>   (let* ((matrix (%make-cvmatrix)) 
>>           (sap (cvmatrix-sap matrix))) 
>> (when enable-finalizer 
>>     (tg:finalize matrix (lambda () (del-mat sap)))) 
>>     sap)) 
>
>I do not think that this is what you want.
>
>Two remarks for future improvement:
>
>1. You use sb-sys:system-area-pointer as type, and this will tie you to
>   SBCL.   There is no reason for that.
>
>2. You should read up on `translate-to-foreign' and
>   `translate-from-foreign' in the cffi documentation.  It will really
>   help to make the code simpler, cleaner and more robust.
>
>If I have more time I might give some more detailed suggestions.  But I
>hope this helps a bit.
>
>Wim Oudshoorn.
>
>
>_______________________________________________
>Cffi-devel mailing list
>Cffi-devel at common-lisp.net
>http://common-lisp.net/cgi-bin/mailman/listinfo/cffi-devel
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.common-lisp.net/pipermail/cffi-devel/attachments/20140413/d587e50d/attachment.html>