Re: DTAARA Lock Issue?

On 4/17/2017 11:33 AM, Roger Harman wrote:

Two jobs retrieved the same number.

NOTE: Not my code - I just have to work with it. Why it has the add/z-add combo is unknown.

** Data structure for next system number
** & next temporary reference.
D NextNum DS 14
D NSYSNA 1 7
D NSYSNN 1 7 0
D TMPRNA 8 14
D TMPRNN 8 14 0

C SrGetSys Begsr

C *DTAARA DEFINE SYSNBR NextNum

C *LOCK IN NextNum
C NSYSNA IFEQ *BLANKS
C Z-ADD 1 NSYSNN
C ENDIF

C NSYSNN ADD 1 UPDNBR
C Z-ADD UPDNBR NSYSNN
C OUT NextNum

C Endsr

Thanks for posting the code; I found it helpful. As a preface, I've
used *LOCK IN in many dozens of programs; over the 3 decades or so
probably a few billion update cycles. I've never, not once, seen an
issue with the OS. It's always been the program. That said, there is a
gotcha with this sample.

The data area is storing /two/ 'unique' numbers, right? You've been
focussed on two jobs that are trying to get a unique NSYSNN, but what if
there's a third job that's trying to get a unique TMPRNN? It could be
putting back an /old/ copy of NSYSNN as it increments TMPRNN.

Another possibility is a *PSSR/INFSR that somehow gets control between
the IN and the OUT. Seems unlikely, but we're looking for something
non-obvious, so I bring it up. The point is that job 1 might get number
555 and not increment it properly.