Yes, there is a way to do this.  However, the only way I know uses
Priv Mode.  There is a storage area on the stack of all processes
(including the CI) where you can load a PLABEL of a procedure that
will be called when a process terminates.  It is used by Image to
close open databases if a process terminates.  Since the CI does not
open Image databases (at least not normally) you could load it with a
procedure that sends messages where ever you want.  I have used this
technique for years with my own in house job scheduler.   Keep in mind
that the procedure will be running as part of the CI in PRIV MODE and
as the CI is terminating.  If you violate any rules of MPE your system
WILL crash (note: not MIGHT crash).  This function may be available
through the Procedure Exits function or similar system interface
function calls.

If you are looking for a MPE supported command, you mya be out of
luck.  I like the idea of James Clark of having a "Monitor Job" use
the WAIT until the job logs off and then send a message to your
process.  This requires two jobs running at the same time: one which
streams and monitors the "real" job; and the "real" job.  But it
should be able to be implemented without any of the fancy Priv Mode.

Rgds
Mel Rees
[log in to unmask]
On Tue, 21 Dec 1999 08:21:50 -0500, Noam Koren
<[log in to unmask]> wrote:

>Hello all.
>
>I would like to know if there is a way to be informed when a Job starts or
>ends NOT to the console but to another application (i.e. a way to ask the OS
>to invoke something when the job starts or ends).
>
>I currently use a command that I insert into the Job itselt at the beggining
>and the end of the job file.
>But - I have a problem when something in the job goes wrong and the Job is
>Flushed - the last command is not invoked so that I dont get a notice.
>
>Example:
>!job jjj,us.ac
>!jobmsg.bin.acc                      <-- send message through tcp/ip
>!
>! other commands
>
>oops - ERROR - Job flushed
>!
>! other commands
>!
>!jobmsg.bin.acc                      <-- send message through tcp/ip actualy
>!EOJ                                     not invoked
>
>
>Is there a solution for this?
>TIA for any idea.
>
>Noam
>