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

On 27/10/2009, at 1:22 AM, Scott Klement wrote:

There's a difference specifically in the way 1-byte return values are
passed. When ILE was implemented, C and CL passed them one way,

Yes, I'd forgotten that. I now recall (see the archives) saying on some earlier occasion that CL made the wrong decision on that behaviour. I think it's more that C widens return values and CL chose to assume the same behaviour from all called procedures rather than the more natural alternative.

and RPG
and Cobol chose to pass them differently. (I think C and CL pass them
as an integer, and RPG and Cobol pass them as character strings... and
there's some sort of a difference in the mechanism somewhere deep under
the covers.)

Unfortunately, I don't know exactly what that difference is.

however, I believe *CNOWIDEN tells RPG to use the same convention that C
(and CL) use for 1 character return values.

Yes, I guess that's the trick. *CNOWIDEN changes RPG behaviour to accommodate C although doesn't that actually seem counter-intuitive? I mean if you specify nothing then RPG doesn't widen but the return value is messed up. If you specify *CWIDEN then RPG widens and the return value is OK. If you specify *CNOWIDEN then RPG doesn't widen but the return value is still OK. Sounds like a side-effect to me--or special case.

I guess my real complaint is with C not providing a way to control widening of return values. Ah, well no amount of complaining will correct that situation but at least I can see what RPG is attempting by use of *CNOWIDEN. I still have to work around the behaviour but that's OK.

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 / \
--------------------------------------------------------------------