Olaf Matthes

Hi, the pattrsdk PDF file says about object_new():

"This routine returns a new instance of the object class if successful, 

event = object_new(CLASS_NOBOX, gensym("event"), 1, a);

    error("creation of object 'event' failed");

object_register(gensym("_myetest"), name = symbol_unique(), event);

object_attach(gensym("_myetest"), name, x);

As you can see, I'm calling object_free() immediately after registering 

the object, so without messing around with it.

What's happening is that the _new function of the object gets called and 

object_new() does NOT return NULL. However, the object_free() call says:

Why's that? Yes, my object does have a free routine and looks quite okay 

"This routine returns a new instance of the object class if successful, 
or NULL if unsuccessful."

void *event;
t_symbol *name;
t_atom a[3];
atom_setlong(a, 0);
event = object_new(CLASS_NOBOX, gensym("event"), 1, a);
if (!event)
{
	error("creation of object 'event' failed");
	return;
}
object_register(gensym("_myetest"), name = symbol_unique(), event);
object_attach(gensym("_myetest"), name, x);
object_free(event);

As you can see, I'm calling object_free() immediately after registering 
the object, so without messing around with it.
What's happening is that the _new function of the object gets called and 
object_new() does NOT return NULL. However, the object_free() call says:

Why's that? Yes, my object does have a free routine and looks quite okay 
to me. This is on Windows XP.

pattrsdk-object_new-trouble

On Dec 4, 2006, at 11:30 AM, Olaf Matthes wrote:

> Why's that? Yes, my object does have a free routine and looks quite 

This is difficult to know without seeing your event object struct 

p.s. for future compatibility and interoperablity with other external 

developers, I might also use class names, or a class namespace which 

are less generic. You can use anything you like for your own use, it 

> "freeobject: 53f3738: bad object"
>
> Why's that? Yes, my object does have a free routine and looks quite  
> okay to me. This is on Windows XP.

This is difficult to know without seeing your event object struct  
definition and constructor.

p.s. for future compatibility and interoperablity with other external  
developers, I might also use class names, or a class namespace which  
are less generic. You can use anything you like for your own use, it  
doesn't have to be CLASS_NOBOX.


> On Dec 4, 2006, at 11:30 AM, Olaf Matthes wrote:

>> Why's that? Yes, my object does have a free routine and looks quite 

> This is difficult to know without seeing your event object struct 

Okay, here is the stripped down version that still dosn't work:

/**************************************************************

static void event_notify(event_t *x, t_symbol *s, t_symbol *msg, void 

    if (msg == _sym_free)    // this message is sent when the object is freeing

        object_detach(gensym("_at_event"), s, x);

        object_unregister(sender);

        post("somebody said good bye");

    /* we use object_alloc here, rather than newobject */

    if (x = (event_t *)object_alloc(event_class))

    c = class_new("event", (method)event_new, (method)event_free,

        (short)sizeof(event_t), 0L, A_CANT, 0); // A_CANT so that we can have 

    /* initialize the common symbols, since we want to use them */

    /* register the byte offset of obex with the class */

    class_obexoffset_set(c, calcoffset(event_t, obex));

    /* add a notify method, so we get notifications from child objects */

    class_addmethod(c, (method)event_notify, "notify", A_CANT, 0);

    /* we do not need this class to instantiate inside of the Max UI; ergo 

> p.s. for future compatibility and interoperablity with other external 

> developers, I might also use class names, or a class namespace which 

> are less generic. You can use anything you like for your own use, it 

Ah, so this does not really tell whether it has a box around or not? I 

was already thinking whther my use of class name 'event' collides with 

anything, since it's such an obvious name. So do I understand it 

correctly, that within the same class namespace all classes are 'chared' 

>
> On Dec 4, 2006, at 11:30 AM, Olaf Matthes wrote:
>
>> "freeobject: 53f3738: bad object"
>>
>> Why's that? Yes, my object does have a free routine and looks quite  
>> okay to me. This is on Windows XP.
>
>
> This is difficult to know without seeing your event object struct  
> definition and constructor.

-----------------------------------
#ifndef _AT_EVENT_H_
#define _AT_EVENT_H_

#include "ext.h"
#include "ext_obex.h"
#include "commonsyms.h"

typedef struct _event
{
	t_object *obj;
	void *obex;
} event_t;

#endif
-----------------------------------

-----------------------------------
#include "event.h"

/**************************************************************
*
*  event class
*
*/
static void event_notify(event_t *x, t_symbol *s, t_symbol *msg, void 
*sender, void *data)
{
	if (msg == _sym_free)	// this message is sent when the object is freeing
	{
		object_detach(gensym("_at_event"), s, x);
		object_unregister(sender);
		post("somebody said good bye");
	}
}

void *event_new(long ac, t_atom *at)
{
	event_t *x;
	t_symbol *name;

	/* we use object_alloc here, rather than newobject */
	if (x = (event_t *)object_alloc(event_class))
	{
		post("event_new");
	}
	return (x);
}

static void event_free(event_t *x)
{
	object_obex_free(x);
}

	c = class_new("event", (method)event_new, (method)event_free,
		(short)sizeof(event_t), 0L, A_CANT, 0); // A_CANT so that we can have 
arbitrary args

	/* initialize the common symbols, since we want to use them */
	common_symbols_init();

	/* register the byte offset of obex with the class */
	class_obexoffset_set(c, calcoffset(event_t, obex));

	/* add a notify method, so we get notifications from child objects */
	class_addmethod(c, (method)event_notify, "notify", A_CANT, 0);

	/* we do not need this class to instantiate inside of the Max UI; ergo 
CLASS_NOBOX */
	class_register(CLASS_NOBOX, c);
	event_class = c;
}
-----------------------------------

>
> -Joshua
>
> p.s. for future compatibility and interoperablity with other external  
> developers, I might also use class names, or a class namespace which  
> are less generic. You can use anything you like for your own use, it  
> doesn't have to be CLASS_NOBOX.

Ah, so this does not really tell whether it has a box around or not? I 
was already thinking whther my use of class name 'event' collides with 
anything, since it's such an obvious name. So do I understand it 
correctly, that within the same class namespace all classes are 'chared' 
among objects?

This should be a t_object, not a pointer to a t_object.

Am 05.12.2006 um 00:15 schrieb Olaf Matthes:

> This should be a t_object, not a pointer to a t_object.

Arrghh, I knew it was a SILLY bug! Thanks alot!!!

pattrsdk: object_new() trouble