[MI400] Still more fun with nullmaps

[jamesl@xxxxxxxxxxx]

This is a multi-part message in MIME format...
--
To: mi400@midrange.com
From: jamesl@hb.quik.com
X-Advert: http://emumail.com
Reply-To: jamesl@hb.quik.com
Date: Mon, 29 Apr 2002 13:11:21 EDT
X-Mailer: EMUmail
Subject: Still more fun with nullmaps

Fellow MI programmers:

The nullmap-handling code I developed with your help
last November has been
working fine until now.

We have recently received
reports from two different end-users, who have been
having problems with
MCH0601 errors occurring in the code that accesses the
nullmap regions of
the file buffers, one while adding a record to a specific
file (one that is
known to contain no null-capable fields), and the other
during a read
operation for which I have no details. We have been unable to
duplicate the
problem at all (our machines vary from V2R3 to V4R4) Diagnostic
code has
shown that the numbers for the size and buffer position of the nullmap
were
calculated correctly, and were not corrupted; the space offsets listed in

the MCH0601 errors (x'00000917' for one, and x'00000920' for the other)
appear
to make no sense in terms of the record lengths.

It gets
worse.

I sent out a beta-test PTF that bypassed the nullmap-access code for
files
known to have no null-capable fields, which appeared to solve the
problem with
the crash-on-add, but it later turned out to cause data
corruption in other
files.

Any idea what could be going wrong?

(or to
coin a phrase,

HHHHEEELLLLLLLLLLLLLLLLLLLPPPPPPPPPPPPPPPP
MMMMMMEEEEEEEEEEEEEEEEEEEEEEE!!!!!!!

--
J.Lampert