[Mono-osx] MonoMac retain/release

[Geoff Norton]

On 2010-04-20, at 1:45 PM, Duane Wandless wrote:

> That sounds like the retain/release is then being automatically handled.  Once the object is referenced inside C# the retain occurs.  Then when the C# object is GCed the release should occur.  Is that correct?  Which if it is then all I should have to do is set the C# variable to null or call Dispose and I have the same basic functionality I listed.

This is correct (the Disposer is what cleans up as well so if you want to be deterministic with the native object graph you call Dispose).

> 
> It would probably also be worthwhile making retainCount public.  As this is a useful debugging technique for memory leaks.

Considering how much we monkey with this (multiple managed references are in 1 slot, additionally with nsautorelease pool) I'm not sure I want this in the public api.  You can get at it yourself trivially:

Messaging.uint_objc_msgSend (objc.Handle, Selector.GetHandle ("retainCount"));

> 
> I have yet to use MonoMac so I'm still guessing a bit on how the interaction with native obj-c code will work.  I just want to ensure I'm not loosing obj-c functionality.  Having control of the retain cycle is an important 'feature' of obj-c.

Its not a 'feature' its a requirement, and we aim to remove that requirement by letting people use the GC.

-g

> 
> Duane
> 
> On Tue, Apr 20, 2010 at 1:37 PM, Geoff Norton <gnorton at novell.com> wrote:
> Whenever we bubble a proxy object, it takes a reference to the underlying native object, as such in the circumstance you described the object could not "go away"
> 
> -g
> 
> On 2010-04-20, at 1:24 PM, Duane Wandless wrote:
> 
> > I noticed that retain/release/autorelease our internal methods.  I'm not fan of doing memory management as that takes away time from solving business needs.  Memory management is required with obj-c.  And I believe these should be public.
> >
> > There are times when an object is passed as an argument into my C# code that must be retained while the C# code manipulates the object (typically a UI object)... then release that object when done.  The most common usage is the UI is a delegate for the async C# method.  At times the UI object will go away while the C# code is running its async task.  The retain of course prevents sending a selector to a deallocated object.  And there are times when the UI delegate selector is not invoked so I must be able to release the UI object from the C# code.
> >
> > I think one of the fundamental design patterns of MonoTouch and now MonoMac is that all the code will be written in C#.  I do think this assumption is limiting.  Giving the C# code full access to the obj-c objects allows for truly the best of both worlds.  The advantages of the C# language can be leveraged.. but without limiting the functionality of the various NS classes within C#.
> >
> > Again my usage is a native obj-c app for the UI with the business logic in C#.  There are advantages to fully developing within C# MonoMac . I just want to push the design so that is not required.
> >
> > Duane
> > _______________________________________________
> > Mono-osx mailing list
> > Mono-osx at lists.ximian.com
> > http://lists.ximian.com/mailman/listinfo/mono-osx
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.ximian.com/pipermail/mono-osx/attachments/20100420/4d94ac0c/attachment-0001.html