Re: avoid running ILE programs in the DAG...

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 2010-10-06 11:35, Alan Campin wrote:

We also ran into big problems with errors in named activation groups. Every
time that an error is thrown in a named activation group the system destroys
the activation group and when you try to go back to it the program crashes
because the pointer is now pointing at a object that doesn't exist anymore.

That's odd. I think something else must be happening. I have a program
that runs in a named activation group, and writes data to a QNTC network
share. It doesn't attempt to open the output file until the first buffer
is ready to write. It's just simpler that way, since failures, while
they do happen, are rare. If an error occurs, usually because the output
file aleady exists (operator error) the program issues an escape message
to itself, and waits for the operator to investigate and respond by
canceling the program. That causes the uncommitted transaction to be
rolled back, and the activation group to be destroyed, all by default.
The operator then disposes of the pre-existing output, retrys the failed
operation, and everything is happy. The job itself remains active.
Commit doesn't run until the program completes without error, so the
program either does everything or nothing. Either is fine in terms of
system requirements. It's been working just fine for 5 years or so now.
(v5.4.0)

- --
Pete Hall
pete@xxxxxxxxxxxxxx
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)

iEYEARECAAYFAkytCfoACgkQXczQcKdXKg6qTwCfXWizVKyGZhpkh+ZxrPoQf2p6
A2oAoMHpkUXpo82sGPsLnI8XgmiiikEH
=GDYb
-----END PGP SIGNATURE-----