Re: %ADDR and possible pointer problems

[Vern Hamberg <vhamberg@xxxxxxxxxxx>]

Hello

First thought is, did the problem occur with the older version of the 
program when you ran it in 6.1? There is a conversion to a new internal 
structure for programs and service programs.

I would definitely declare the parameter with LIKEDS. I might even go so 
far as to have the DS defined with the TEMPLATE keyword, then use LIKEDS 
even for the DS you pass.

You probably know that if you leave off the VALUE keyword, that you ARE 
passing the address of the DS. Before we had LIKEDS, we had to use the 
pointer technique or use LIKE and copy the parameter into a real DS. The 
latter is also safer, I think, than the pointer technique.

Maybe the main issue with using pointers this way is, you lose data-type 
checking, which could really help you here.

Hope this helps
Vern

On 3/6/2012 5:57 AM, PAPWORTH Paul wrote:
> Since we installed Version of the RPG compiler with the attribute o/s current we have been encountering problems with a technique that previously worked ok.
>
>
>
> Module créé en édition . . . . . . . . . . . . :   V6R1M0
>
> Module créé pour édition . . . . . . . . . . . :   V6R1M0
>
>
>
> With the third parameter we pass the address to a data structure :-
>
>
>
>                    wrkCRtn = cpReadI( XmoYAccVal : XmoCValeur
>
>                                                  :  %addr(vlxopc)  )
>
>
>
>
>
>
>
> In the called function we do the following
>
>
>
>   PcpRead           B                   Export
>
>
>
>   d                 Pi                  Like( Tint )
>
>   d ii_YAcc                        1    Value
>
>   d ii_CVal                       12    Value
>
>   d PtrVlXOpc                           Like( TPointer ) Value
>
>   d                                     Options( *NoPass )
>
>
>
>
>
>
>
>
>
>   d vlxopc        e ds                  extname(dvlxopc)
>
>   d                                     Based(PtrVlxOpc)
>
>
>
>
>
> Has anybody had problems with this technique ?
>
>
>
> Yesterday I got a messages that an objet had been partially or entirely destroyed and could not be used followed by a recursion exception !!
>
>
>
> I recompile the program and it now works ! I suppose the allocation of memory may be a bit different and as such the (I suspect) pointer error no longer occurs.
>
>
>
> I am thinking of identifying and where possible replacing this technique ( %addr ) by passing the ds itself.
>
>
>
>
>
> I did read a comment my Hans Boldt that this technique should be avoided. I should perhaps leave the use of this technique for use with API access only.
>
>
>
> What do you all think