Re: RPG to C and C's stupid concept of widening data types

[Simon Coulter <shc@xxxxxxxxxxxxxxxxx>]

On 27/10/2009, at 2:42 AM, Barbara Morris wrote:

> Scott explained how *CNOWIDEN controls more than just widening.

Not described in the RPG reference material as far as I can see. Just  
a vague imprecation to use *CWIDEN or *CNOWIDEN when communicating  
with C but little about its actual effect.

> It also
> controls whether the 1-byte return value is treated as an aggregate
> (character string) or as a scalar (integer).

My understanding is that aggregates are data structures and arrays  
(including multi-byte fixed-length character types) and scalar is  
everything else (1-byte char, short, int, long long, pointers of all  
types). The distinction being whether the variable can hold multiple  
values at the same time. A numeric variable of any type can hold only  
one value therefore scalar. A pointer can hold only one address  
therefore scalar. A 1-byte char can hold only a single character  
therefore scalar. Structures and arrays can hold multiple different  
values and types. It seems the idea that a scalar is also numeric is  
yet another C peculiarity--much like its weird view that char is a  
number.

> You are correct that when a C function returns char and the RPG
> prototype is defined without any special EXTPROC value, then the RPG
> prototype would have to define the return value as a 1-byte unsigned
> integer, and then the caller would have to convert the returned  
> value to
> a 1-byte char.
>
> By the way, you can write a C function to return a 1-byte aggregate by
> defining the return value as a struct with a 1-byte subfield.

Thanks, I might use this method because it will allow me to keep all  
the changes in the same language and source module. It might confuse  
the C weenies but they won't have to look at it or use it and it will  
be transparent to RPG and COBOL consumers.

> Also by the way, starting in V6R1, this difference regarding single- 
> byte
> return values disappears.  Aggregate return values up to 16 bytes get
> handled by the system in the same way as scalar return values.

True but doesn't that require recompiling of both caller and callee to  
get the new behaviour? I recall some previous discussion on this for  
CL (adds archive search to list of things to do today).

Regards,
Simon Coulter.
--------------------------------------------------------------------
   FlyByNight Software         OS/400, i5/OS Technical Specialists

   http://www.flybynight.com.au/
   Phone: +61 2 6657 8251   Mobile: +61 0411 091 400        /"\
   Fax:   +61 2 6657 8251                                   \ /
                                                             X
                 ASCII Ribbon campaign against HTML E-Mail  / \
--------------------------------------------------------------------