Re: [Pyobjc-dev] KV Observers caused Abort Trap (or: when to remove them)

[Clive J. <cl...@bl...>]

Can I post TGZs to the list (I'll probably have to include a nib for 
the example) or do I have to post it to a web site and then post the 
link to the list?
C
On 6 Apr 2005, at 09:43, Bob Ippolito wrote:

>
> On Apr 6, 2005, at 12:25 AM, Clive Jevons wrote:
>
>> In one of the apps I'm writing I use KVO in order to track when a 
>> property of a data object is changed so that I can mark it dirty. 
>> That part works really nicely - in the init() method of the object, I 
>> attach an observer (itself) for each property which is stored in the 
>> DB. Everything worked fine until I re-loaded the data, or deleted an 
>> object. Then the whole thing would crash out without giving any hint 
>> or sign as to why. As the KVO was the only thing I'd changed, I knew 
>> it had to be that. After a quick think, I figured it was because I 
>> hadn't detached them before releasing all references to the object in 
>> python, and that there were probably invalid pointer or references 
>> hanging around in the Objective-C part responsible for KVO. Once I 
>> added calls to removeObserver_forKeyPath_ in all the appropriate 
>> places everything was fine again.
>>
>> I *did* try to do the removing of the observers in the dealloc (which 
>> told me I wasn't to use that method) and the __del__ ... but the 
>> latter wasn't called before the app bombed out with an 'Abort Trap'.
>>
>> The solution I currently have works, but I find the code quite untidy 
>> - on the whole, the reason I love Python is that it enables me to 
>> express a solution to a problem without loads of technical grunt work 
>> - like memory management - so having to intersperse KVO cleanup code 
>> with my business code is really going against the grain :( ... can 
>> anyone suggest a cleaner solution to this?
>
> If you post a simple runnable example that breaks, I'll look into it.
>
> -bob
>
>