Re: Update A File After A Couple Other Chains -- RPG400-L

I agree with Buck. I will do a CHAIN(n) for on-line maintenance so thatthe lock doesn't hang some other production job, but always check to seeif the record has changed after during the straight CHAIN (Before andAfter images). If it has changed, say, "Sorry, you took too long. Tryagain."

For batch jobs, though, I just do the straight CHAIN because I know therecord will be released in, say, 2 microseconds; my timeout variable islonger than that.

But Mike said he had a "read". As long as he's checking for eof or(strange thought) not accessing the same table again for some reason, heshouldn't get the "tried update before record retrieved" error. So, onthe surface, it would appear that one of those things is (or is not)being done.



        * Jerry C. Adams
*IBM System i5/iSeries Programmer/Analyst
B&W Wholesale Distributors, Inc.* *
voice
        615.995.7024
fax
        615.995.1201
email
        jerry@xxxxxxxxxxxxxxx <mailto:jerry@xxxxxxxxxxxxxxx>



Buck wrote:

Make the chain a chain with an (n) no-lock extender. Then, chain againat the time of updating. If you have several lines of intervening code,as you have here, someone someday will mess it up and it will be a bearto find the problem.
Ooohh danger! There's nothing wrong with this advice as long as you'reaware that the contents of the record may have changed in the interim(by some other job.)
chain()
 (credit score = 800)
...
time passes
...
a different job changes the record -- remember, it wasn't locked! Thecredit score in the file is now 100.
...
if credit score > 700...
  (this asserts TRUE, despite the actual value being 100)

chain
(at this moment, the credit score value field in the RPG program ischanged to 100. That might be a problem, since a bunch of calculationswere based on the assertion that the credit score was high.)
Often, when doing a CHAIN(N), then a long running set of operations(like EXFMT) you need to re-evaluate the contents of the record whendoing the CHAIN for update. Either that, or have some sort ofcomparison to be sure the record is really the same both times you CHAIN.
  --buck