IBM can't reproduce this. I'm hoping someone else can.
That didn't come out right. I don't really want to you have this
problem. That would be mean. I just hope this august body can help.
Simplest case:
CDS.BP M
01 LOOP
02 EXECUTE '* anything' CAPTURING X
03 REPEAT
>RUN CDS.BP M
Memory fault(coredump)
3046$ls -l core
-rw------- 1 cds33896 tech 279499648 Mar 17 17:11 core
This is on HP Itanium, HPUX11.23, UV10.2.6
Anticipating some of your questions:
* If I get rid of CAPTURING X it works fine.
* PERFORM works fine.
* It doesn't matter what the command executed is.
I executed a comment ('* anything')above, because what is more benign?
* ps -l shows memory increasing (SZ column) as the process executes.
(Monitoring it from a 2nd session.)
* EXECUTE "SH -c'[any unix command]'" CAPTURING X
gets slower & slower each iteration.
( Because ever increasing memory is duplicated for child process? )
* Other flavours of the CAPTURING syntax are the same:
EXECUTE cmd CAPTURING X
EXECUTE cmd,OUT>X
EXECUTE cmd,/OUT.>X
They compile identically.
* Behaviour is the same in IDEAL- & PICK-flavoured accounts.
* It doesn't matter if I null out X.
X = ''
EXECUTE cmd CAPTURING X
* It doesn't matter if I assign X to an unassigned variable.
X = Y ;* (where Y has deliberately never been assigned.)
EXECUTE cmd CAPTURING X
* My simple case fails at about
1/4 million iterations to fail on my smaller dev system,
1 million on the 2 larger production system.
* The memory leak applies to the session, not the program.
If I limit the number of iterations in the program,
but repeatedly execute it from TCL, eventually it fails.
This fails just the same:
VOC M.LOOP
O1 PA
02 LOOP
03 EXECUTE "RUN CDS.BP M
04 SH -c"ps -lp $$PID"
05 REPEAT
where M is limited to 100,000:
CDS.BP M
01 FOR I = 1 TO 100000
02 EXECUTE "* comment" CAPTURING X
03 REPEAT
* My old HP PA/risc, HPUX 11.11i, UV 10.0.16 system works fine.
Thank-you in advance,
Charles Stevenson
Getronics
Costa Mesa, CA
714-641-5943 desk
714-396-2771 cell
-------
u2-users mailing list
u2-users@listserver.u2ug.org
To unsubscribe please visit http://listserver.u2ug.org/
