[GAP] old attribute

[Robert Dewar]

On 10/30/2011 6:46 PM, Robert Dewar wrote:
> On 10/29/2011 4:13 PM, Jean-Pierre Rosen wrote:
>
>> But if you need a deep copy, you should define your type as controlled,
>> and the definition of Adjust will do the trick (Too lazy to check in the
>> RM, but I hope and believe that 'Old uses a regular assignment)
>
> This seems wrong to me, typically we are talking about postconditions
> in the low level routines that mess with these linked lists etc. You
> don't want casual deep copies in that context, but you do need the
> deep copy for saving the old value.
>
To see that in its most vivid form, consider that this postcondition
is being used in the routine that does an elaborate optimized deep
copy operation!