Parameter question (long)

["Wim" <midrange-l@xxxxxxxxxxxxx>]

Hi All,

This problem is about passing parameters through command return values from one 
program to another and how they should be defined and handled in both the 
program that uses the command and the CPP that processes the return value.


scenario: CL program using a RTV*command that has an ILE RPG as CPP. 

The idea is a "keyword=value" parser for an IFS file. I have created a command 
with a RTNVAL(*YES) PARM. This returns a value into a variable. The command 
gets the filename, the keyword to locate in the file, my return variable and 
the length of the return variable. Four in all. 

The CPP all defines the fields to their maximum lengths. The maximum length for 
a keyword value is 2048 bytes. 

The programs I wish to use the command in, defines the variables at their 
expected length. The length parameter is passed to avoid writing in unknown 
areas. For example, the <TYPE> keyword/value combination in the file returns a 
maximum keyword length of 10 positions: "TYPE=PROGRAM, TYPE=FILE, etc". So I 
declare the &TYPE var to LEN(10). 
I now create a program that uses the command, like so: 

RTVQ FILE('/qopensys/etc/file.inf') KEY('TYPE') KEYVAL(&TYPE) KEYVALLEN(10) 

I see that my CPP behaves as expected by filling the value of the return value 
up to the number of positions provided in KEYVALLEN. Then the program Returns. 
The CPP then does some closing things (there is a undebugable section (where 
you press f10 several time but nothing seems to happen)) and I get a "Scalar 
operand contains a value that is not valid" error (MCH5003). If I extend the 
value of &TYPE to 2048 and prefill it with some random value, all runs fine and 
the random value is replaced by blanks, even after the position of KEYVALLEN 
(as I expected).

I suppose the CPP tries to copy the value of the parameter back into the 
returnvalue and crosses some boundary. I hoped that the pointer to &TYPE was 
simply passed to the CPP and that the CPP was in charge of that pointer but 
that is obviously not the case. Is there some inbetween buffer in the command 
space where the parameter is passed by value? I tried to trick this command 
space by defining 1 position for the KEYVAL but the OS seems persistent in 
copying back all values from the parameter length as defined by the CPP. 

The latter indicates that the RPG ILE program wants to return all parameter 
info. Which in turn indicates that some proxy field is used to pass the 
parameter information rather then the field i define in my PLIST. In the RPGILE 
manuals there is a comment stating that "at closing, the factor 2 of the *ENTRY 
parm is moved into the result field of the PARM" however, there is no factor 2. 

Could it be that a command using an ILE Pgm as CPP is passing command RTNVAL 
variables by reference indirect?

When I move a CLP (OPM) in between the ILE and the command doing nothing other 
then pass the parameters to the intended CPP, all runs fine. 

O Midrange-l, enlighten me!

Wim 

www.remain.nl