On Jan 29, 2013, at 00:51 , Bob Cromwell <[email protected]> wrote:
> if below code valid ? All sample codes I could find invoke "self = [super
> init]" first and then check the input parameter inside "if (self)" block.
>
> @implementaion Bla
>
> - (id)initWithFoo:(NSString *)foo
> {
> if (foo == nil) {
> return nil;
> }
> self = [super init];
> if (self) {
> _foo = foo;
> .....
> }
> return self;
> }
>
> I had thought there would be memory leak. "Bla * b = [Bla alloc ]
> initWithFoo:nil];" . However under ARC using Instruments Leaks, there are no
> leak found.
No, it's not valid. You can see why if you rewrite this using manual
retain/release conventions:
if (foo == nil) {
[self release];
return nil;
}
self = [super init];
You must (and ARC does) release 'self' before returning nil, to avoid a memory
leak. The problem is that releasing 'self' will invoke 'dealloc', and you have
no idea whether it's safe to invoke 'dealloc' if you haven't invoked done
'[super init]' yet.
Even if your class doesn't have a 'dealloc', you don't know whether a
superclass or a subclass might have one, now or in the future.
In practice, you *may* get away with doing this, but it's a dangerous idiom to
use. Much better to check the parameter a bit later:
if (self) {
if (foo == nil)
return nil;
_foo = foo;
.....
}
Note that this isn't guaranteed safe either. The class, or a subclass, may have
a dealloc that isn't safe if the initialization wasn't completed**. However,
most dealloc's are written to avoid such problems.
** Here's an example that will crash with an early return:
if (self) {
if (foo == nil)
return nil;
_foo = foo;
_bar = malloc (100);
.....
}
…..
- (void) dealloc {
free (_bar);
}
_______________________________________________
Cocoa-dev mailing list ([email protected])
Please do not post admin requests or moderator comments to the list.
Contact the moderators at cocoa-dev-admins(at)lists.apple.com
Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com
This email sent to [email protected]