TR15581 - elemental assignment of structures with alloctable components

[Paul Thomas]

>If I may add my .02 Eur, the shallow copy implementation of derived
>types with ALLOCATABLE is so wrong (by default) that F2003 introduced
>MOVE_ALLOC for those situations where the user explicitly wants
>something (remotely) similar with an allocatable object.
>  
>
This is the conclusion that I had come to, as I said in my original 
message.  Since quite a lot of copying is involved, I just wanted 
confirmation that I had it right.  A longer term solution might be to 
use a smart pointer system; keep track of the number of clients for a 
patch of memory and only copy if the contents are to be modified and 
more than one client has possession.

>Which brings me back to an important (to me) issue: please, please,
>please implement MOVE_ALLOC alongside the rest of TR15581. 
>  
>
I have not forgotten your request!

>Keep up the good work
>
We'll see if it's good or not but we'll keep it up anyway!  I have a 
possibly naive expectation that if the entirity of the iso_varying_ 
string testsuite runs, the extnesion will be rather usable.

Thanks for all your remarks/comments.

Paul