Re: Correct memory management in -awakeAfterUsingCoder: ?
Re: Correct memory management in -awakeAfterUsingCoder: ?
- Subject: Re: Correct memory management in -awakeAfterUsingCoder: ?
- From: Greg Parker <email@hidden>
- Date: Mon, 4 May 2009 22:57:03 -0700
On May 4, 2009, at 9:45 PM, Graham Cox wrote:
I am trying to track down an over-release bug.
It looks like I'm doing something wrong in -awakeAfterUsingCoder:
which leads to an object being insufficiently retained.
I have two objects, A and B. A loads B from an archive within -
initWithCoder: while dearchiving and retains it. When B is
dearchived, it can be replaced by another object by its -
awakeAfterUsingCoder: method. If I detect that B should be replaced,
I release the current B as per the documentation and return the
replacement without doing anything regarding retain or release on
it, as it's already retained by the code that supplied it.
Subsequently it is this object that is over-released.
If I retain the object before returning it in -
awakeAfterUsingCoder:, the problem goes away, but this appears to
violate normal memory management rules, so it's not clear that this
is the correct solution. The documentation doesn't explicitly say
anything about this either.
Two answers:
1. Retain it. Conceptually, you're balancing the [self release] that
you did to the old-self. The old-self came in with one retain count
that was not your responsibility to release. That means the new-self
needs to go back out with one retain count that is not your
responsibility. So you should retain it before returning. Whatever
retain counts are held by the storage library are not sufficient by
themselves to make the accounting book balance.
Strictly speaking, old-self's retain count could have been more than
one. But if that happens you've probably already fallen afoul of
answer number two:
2. Be very careful with replacing objects in -awakeAfterUsingCoder:.
It doesn't work in the general case, so you had better be sure you're
in a specific case that does work. See this thread for some discussion:
http://www.cocoabuilder.com/archive/message/cocoa/2001/10/10/48468
(That's an old thread, but as far as I know the newer NSKeyedArchiver
didn't magically solve the problems.)
--
Greg Parker email@hidden Runtime Wrangler
_______________________________________________
Cocoa-dev mailing list (email@hidden)
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:
This email sent to email@hidden