On Jan 15, 2014, at 01:29 , Greg Parker <gpar...@apple.com> wrote:
> It sounds more like *B's* retain count had become zero
I’m sure it hadn’t. C’s dealloc was the only place that the CFRetain was
balanced. I believe I tried commenting out the CFRetain, to see if the crash
went away, and duly observed B's memory being leaked.
> So if you examine the weak variable's storage via forbidden paths
FWIW, my comment about the debugger was just an aside. This is the actual code:
> - (void) dealloc {
> if (_undoProperty)
> CFRelease ((__bridge CFTypeRef)
> _undoProperty.objectReference->mutableObject);
> }
>
> - (id) initWithValue: (id) value forProperty: (BLSTransactionProperty*)
> property {
> self = [super init];
>
> _undoValue = value;
> _undoProperty = property;
> CFRetain ((__bridge CFTypeRef)
> _undoProperty.objectReference->mutableObject);
>
> return self;
> }
and the “objectReference” property is synthesized from:
> @property (nonatomic,weak,readonly) BLSTransactionObjectReference*
> objectReference;
There *may* be a forbidden path in the dealloc, but it seemed to me that 'if
(_undoProperty)’ was true, I should have been able to retrieve the pointer to
‘mutableObject’, but this wasn’t so. _undoProperty tested true but the object
was unmessage-able.
Maybe this is a compiler defect in that it’s using the raw ivar value for the
‘if’ but using ‘objc_loadWeak’ for the message-send.
_______________________________________________
Cocoa-dev mailing list (Cocoa-dev@lists.apple.com)
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 arch...@mail-archive.com
