A new OPTION to force pointer declaration

IBM will use this request as input to planning but no commitment is made or implied. This request will be updated in the future if IBM implements it. IBM will use votes and comments from others in the community to help prioritize this request.

This feature would most likely be implemented as a new value for the DCLOPT keyword, such as DCLOPT(*NOIMPLICIT). It would apply to other implicit declarations (the parameters for the COMMIT and DEVID file definition keywords).

We're thinking that if we implement this item, it would probably be a new option for the DCLOPT Control keyword rather than a new value for the OPTION parameter and OPTION keyword.

Mark, if there were some new option to prevent the compiler from implicitly defining a BASED pointer, I think it would probably not even be allowed to define a local variable based on a global pointer.

The compiler can't start using a global pointer if the local pointer is not defined.

But even aside from the issue I mentioned, there's also the issue Niels mentioned, where the programmer mistypes the pointer for the BASED keyword.

I see the issue, but how would an option correct this particular problem? by causing a compile time error? At least the programmer would know that the global pointer wouldn't be automatically attached as the basing pointer. I'm torn between yea and nay on this one. In general I don't like globals, but they do have a limited place. Maybe a better way to handle this is if you would just use the global if it were defined as a pointer. That is really what most people would expect since the basing pointer need not create a new pointer. Or maybe implicit definitions should always use a like defined global if one exists. They use a like defined local if one exists and even error out if the like named local is defined differently. We are getting into some complicated name resolution routines here.

In case anyone is not aware of the issue here, RPG implicitly defines a pointer if it is used with the BASED keyword. This can cause unexpected behaviour if you use the BASED keyword in a procedure expecting the variable to be based on a global pointer. Instead, the compiler defines a local pointer.

dcl-s g_ptr pointer;
dcl-s storage char(10);
g_ptr = %addr(storage);
p1();
return;

dcl-proc p1;
dcl-s based_fld char(10) based(g_ptr); // A local g_ptr will be implicitly defined by the compiler
dsply based_fld; // error, pointer not set
end-proc;

@NielsLiisberg not that type of declarations, only to prevent run time errors when you declare something like:
Dcl-s CodeList like(code) dim(10) based(p_CodeList);
Dcl-s x_CodeList pointer inz(%addr(CodeFields));
Note x_ and p_

Can you elaborate on this: Do you think about strongly types pointers like in C / C++ . for example "pointer to int"