🚀 go-pugleaf

 
----------------------------------------------------------------------------
>The problem arises when a TDynamicArray is cloned while an Iterator which was
>used for the TDynamicArray is stil in scope. TDynamicArray has a field named
>fIteratorPtr, which contains a linked list of Iterators currently in use for
>the TDynamicArray. When a TDynamicArray gets cloned, both (the old and the
>cloned Array's) fIteratorPtrs point to the same Iterator. If the
>CArrayIterator
>is destructed, the fIteratorPtr of the cloned TDynamicArray points to an
>invalid Memory-Address.
 
    By golly, you're right!  How they ever let one this obvious slip through
I'll never know.  TDynamicArray should have a Clone override that sets
fIteratorPtrs = NULL.
 
    What the heck are you doing that creates a clone of the array while you're
iterating it?  It seems to me that the only time that linked list of iterators
get used is when iterations are nested -- unless you are instantiating
iterators as seperate objects and not by the usual method on the stack.
 
    My other comment would be the same that I usually give to people that ask
me about subclassing, cloning, or file-storing TList and other such
TDymamicArray thingies:  "DON'T DO THAT".  Usually, these things are used by
(referenced by) other objects which are what you really want subclassed,
cloned, or stored.  Let their clone methods do the work.  Also, note that
cloning the TDynamicArray does NOT clone the objects refered to in the array.
 
    For example, I usually create my own list classes which encapsulate TList
with my own functionality.  When I want to clone something that has a
referenced TList, I always let the encapsulating object ask for a new list and
then, in the clone method of my object, iterate the cloned list to add all of
the old elements to the clone list.  That process creates a clone of the list,
but not of the contents (elements) of the list (i.e., the new list has
references to all the old elements).  If you really want a deep clone (a
seperate new list with its own new elements) then you must clone each element
and add its clone to the clone list rather than adding the old elements.
 
    I guess that explains why I (and others) never ran into the problem you
found.
 
G. Gordon Apple    D4887
Advanced Communications Engineering, Inc.
Redondo Beach, CA
 
------ Where we are not having any floods, hurricanes, blizzards, freezing,
volcanic eruptions, famines, or tidal waves -- just the usual earthquakes and
riots.  I think I'll go for a walk on the beach (two blocks away) in my shirt
sleeves this afternoon.