Twice Bitten...

Question

Can someone explain to me why following the recommendations causes problems?
It may be a 3.0 vs 2.2.1 issue. I'm throwing that out just as food for thought.
But, I think I've been bitten by this on both 2.2.1 and 3.0.
For memory managment, anything you alloc/init, you release.
Anything you retain using:@property (retain) object* obj;
you are supposed to (in 3.0, under viewDidUnload, in 2.x under dealloc) provide aself.obj = nil;
as well as in dealloc add [obj release];
Well, I do that, and I run into EXCBADACCESS errors.
IF I comment out, either the self.obj = nil, or the release statements, all is cool.
I only have an iPod Touch to test on device, and it is still on OS 2.2.1, since it's "borrowed" and the owner and I have not discussed the $10/$5 upgrade fee to bring it current.
I've run into this EXCBADACCESS on the simulator before and the solution was to comment out the self.obj = nil lines, and that solved the problem. When I started device testing this week, however, the problem reared it's ugly head again, and so I commented out the code in all of my viewDidUnload blocks. For 2.2.1 testing, viewDidUnload does not exist, so those are only compiled under
#ifdef _IPHONE_30
#endif
code blocks to allow testing on older devices.
So,... can someone explain why this happens?
Since release is supposed to work if the value is nil, and setting the pointer to nil does an implicit release (on the pointer, before setting it to nil), this should only be a problem if the dealloc is called before viewDidUnload.
Oh wait... doesn't viewDidUnload get called after, oh right.
This doesn't explain why the following in dealloc also seems to crash....
self.obj = nil;
[obj release];
I'm sorry, but the apple documentation seems to botch the explanation completely and seems to have it backwards, at least from how it works operationally.
I'm happy to code it whatever way works. I'd just like to understand why!
Thanks.

Answer 1

reststop wrote:
For memory managment, anything you alloc/init, you release.
Anything you retain using:@property (retain) object* obj;
you are supposed to (in 3.0, under viewDidUnload, in 2.x under dealloc) provide aself.obj = nil;
as well as in dealloc add [obj release];
Don't do both, your else you will get...
Well, I do that, and I run into EXCBADACCESS errors.
... exactly.
IF I comment out, either the self.obj = nil, or the release statements, all is cool.
According to the "dealloc" section in the Objective C 2.0 language guide you are supposed to use "release", but "If you are using the modern runtime and synthesizing the instance variable, however, you cannot access the instance variable directly, so you must invoke the accessor method". Of course, that is kind of backwards logic. What it really means is you should use "self.obj = nil".
I've run into this EXCBADACCESS on the simulator before and the solution was to comment out the self.obj = nil lines, and that solved the problem.
There isn't any specific "EXECBADACCESS" error. That just means your program has one or more bugs. There could be any number of changes you need to make. You should review that code because "self.obj = nil" is something that you need.
I'm sorry, but the apple documentation seems to botch the explanation completely and seems to have it backwards, at least from how it works operationally.
It does look backwards, but when deallocating memory, doing it twice isn't twice as good