#646 Bugs with Packed Decimal Scaled (P) Data

[Marty Heyman]

Attached: data_packed1.at and testsuite.log.

data_packed1.at contains 9 tests at present. It is three groups of three tests each. It tests for correct values stored in scaled COMP-6, COMP-3, and PACKED-DECIMAL variables with pictures of 999P, 999PP, and 999PPP. Eight of the nine tests fail. While I intend to extend this series of tests, I felt the initial findings were worth registering with the project. This was triggered by @rdubner's reports some time ago on the "devel" list.

I feel this feature is sufficiently obscure that this can be deferred to 4.0.

[Edward Hart]

I had begun writing a patch for this - however, it seems I forgot to save a patch and then reverted all my changes. Such is life when you use Subversion.

I do at least remember the one big, subtle issue I came across: there is currently no way for libcob to distinguish between packed-decimals with a PIC of VPPP999 and VPPPP99 (for example). Consider their cob_fields:

• scale: both 6.
• size: both 2.
• digits: both 6 digits.
• pic: both absent (since they're not numeric-edited items).
• flags: identical (except possibly for NO_SIGN_NIBBLE, which doesn't help).

I suggest adding to cob_attr a field for the number of P's in the PIC.

[Ron Norman]

I have done some work on this for the 4.0 version. It likely has never worked before.
It should now work for COMP-3, COMP-6 as well as COMP-5.