Re: Segfault in -[NSObject dealloc]

[Markus Hitter]

Am 18.05.2006 um 07:18 schrieb Richard Frith-Macdonald:

> On 18 May 2006, at 05:31, Michael Gardner wrote:
>
> > I'm allocating objects with class_createInstance(), storing them  
> > on a stack (well, just a plain NSArray), and popping them off at a  
> > later time. When I pop an object from the stack, it gets released  
> > (as I expect), but I get a segfault from -[NSObject dealloc] that  
> > I can't track down.
> >
> > I'm pretty sure I'm not double-releasing the object since I've  
> > sprinkled printf statements in my dealloc methods, and I've tried  
> > looking through NSObject.m to find the culprit, but can't see  
> > where the segfault is coming from (the inline functions hinder  
> > debugging somewhat). What kind of things can cause segfaults in - 
> > [NSObject dealloc]? Could this be a GNUstep bug?
>
> You can't use class_createInstance() to create instances of GNUstep  
> classes (unless you build NSObject.m in the base library specially  
> to permit it).
> The way to allocate OpenStep/GNUstep objects is using the  
> NSAllocateObject() function.

Can I add "+[myClass alloc] would be even better"?

I'm really wondering what's the big difference between GNUsteppers  
and Cocoa developers. According to the mailing lists, GNUsteppers are  
regularly using low level functions (objc_...()), +load and  
thisalike. On Cocoa lists, I've yet to read a single line about this  
stuff; they obviously use NSObject and above and are happy with this.

Is GNUstep missing some important piece of the Obj-C cake or are  
GNUsteppers simply more comfortable with the C language?


Markus

- - - - - - - - - - - - - - - - - - -
Dipl. Ing. Markus Hitter
http://www.jump-ing.de/