Hi all,

Chris Bartram (thanks) brought to my attention a potential
problem with the current ABORTPROC implementation and I
would like your opinion on possible solutions. The new ABORTPROC
commands kills the process (and its children, if any) without
anyway to detect that the process was aborted.

Today, if you <break> and :abort a process you see CIERR 989 --
"Program aborted per user request". If the process aborts itself
you get CIERR 976 -- "Program terminated in an error state".  In
both of these situations the CIERROR variable is set (as is the
JCW variable) and can be checked in the next statement after the
RUNning of the program.

Chris made good arguments that ABORTPROC should behave similarly
when the target process is a direct child of the CI.  His point
being that the job running the process needs to know if the
process completed normally or abnormally. It could be detrimental
for a job to continue processing if the program that was run from
the job was externally aborted before completing.

Question 1: does this logic apply to any process aborted, i.e.. what
if a great grandson process is ABORTPROC'd, should JCW/CIERROR be
set, or just for the direct CI descendant process?

Question 2: do you need to distinguish between CIERR 989 (Process
aborted per user request), which I assume is not checked for in
most JCL and the ABORTPROC case.  That is, does ABORTPROC need
to define a new CIERROR value or can we use 989?  Note: a new
CIERROR value also requires a new JCW value.  The high order 2
bits of JCW are reserved for MPE and are cleared each time a process
is run by the CI and are checked each time this process terminates.
If JCW is set to %14xxxxx then CIERR 989 is reported.  If JCW is set
to %10xxxx then CIERR 976 is reported).  I will need a new JCW
value to cause a new CIERROR to be reported in the process ends
due to ABORTPROC.

Other thoughts, comments, opinions, ideas, suggestions are welcome.

Thanks,
 Jeff

p.s. ABORTPROC is a GR patch now (MPEKXR8) and will be released
before the above situation is resolved.  I decided it was
important to get the basic functionality out (and Store-to-
disk is also part of the same patch).  A new patch will be
created to resolve the above situation.