Re: Compiler bug with *YEAR ? - For Barbara Morris ?

On Mon, Jan 3, 2022 at 4:42 AM Martin Neugebauer via RPG400-L
<rpg400-l@xxxxxxxxxxxxxxxxxx> wrote:

Happy new year to all !

Same to you!

Is this a compiler bug?

No. It is arguably unintuitive behavior, though.

**freedcl-s c1 char(4);dcl-s c2 char(4);dcl-s c3 char(4);dcl-s YNum packed(4);
YNum = *year;
c1 = %editc(*year - 1 : 'X'); // Error?
c2 = %editc(YNum - 1 : 'X'); // Error?
YNum = YNum - 1;c3 = %editc(YNum : 'X'); // expected result
*inlr = *on; //-------------------------------------------------------------------
Results:EVAL c1 C1 = '0202' EVAL c2 C2 = '0202' EVAL c3 C3 = '2021' //-------------------------------------------------------------------

The expressions being assigned to c1 and c2 are 5 characters long,
equal to '02021'. This is due to (in my opinion) arcane precision
rules for intermediate results within expressions.

So, the 5 characters thing is one area of probably unintuitive behavior.

The implicit truncation feature of EVAL (like using MOVEL on character
fields) is documented, so is also not a bug and perhaps not even
"unexpected". But personally, given how much safety was prioritized in
the %BIFs and freeform aspects of RPG IV, I would have personally
expected a compile-time error (if the too-long character value is
known at compile-time) or a run-time error (if the length cannot be
known at compile-time, but turns out to be too long at run-time).

To me, the simplest way to adjust the code to do what you want is to
use EVALR instead of the naked equals sign (which is shorthand for
EVAL).

John Y.