Re: Quicktime/CoreImage memory leak

FROM : Daniel Hazelbaker
DATE : Sat Aug 12 22:46:23 2006

On Aug 11, 2006, at 4:25 AM, Fabien Roy wrote:

> Maybe "leaks" could help.


Thanks for the thought, but MallocDebug does what leaks does, plus a 
little better.  Specifically leaks does not include memory leaks by 
Carbon API's (whats up with that?  A leak is a leak isn't it?). 
Interestingly, leaks only shows a few minor things (as in a few K 
instead of many MBs).  MallocDebug shows many MB, which leads me to 
think all the lost memory is being allocated by NewHandle() and there 
is something I am doing/not doing that is causing it not to be freed. 
Grr.

leak man page states:

> WEAKNESSES
>      Memory allocated via Carbon's NewHandle() function and then 
> leaked will
>      not be noted by leaks. Thus, running leaks on a Carbon 
> application will
>      show only a subset of all possible leaks.  The leaks reported 
> will always
>      be true leaks.
>
>      MallocDebug will correctly find leaked blocks that were 
> allocated via
>      NewHandle, and permits easier browsing of leaked blocks. 
> However, Mal-
>      locDebug does not detect leaks in circularly-linked structures 
> or iden-
>      tify groups of leaked, connected nodes; leaks's pointer 
> analysis can cor-
>      rectly identify such leaks.



Daniel

>
> ---
> Fabien Roy
> 3426 N. 32nd Street, #64
> Phoenix, AZ 85018, USA
> <<email_removed>>
> Tel:    602 476-1914
> Cell:  602 312-2948
>
>
>