From owner-bug-octave at bevo dot che dot wisc dot edu  Thu Oct 17 02:54:46 1996
Subject: Bug in Octave's LSODE-function
From: Janne Karonen <janne dot karonen at kemira dot com>
To: bug-octave at che dot utexas dot edu
Date: Thu, 17 Oct 1996 08:48:51 -0700

BUG REPORT

Author:	Janne Karonen
E-mail:	janne dot karonen at kemira dot com
Date:	17.10.1996

NOTE:	This bug report has nothing to do with Kemira Oy, author is
	just speaking for himself.


BACKGROUND

This bug report deals with Octave's (version 1.1.1) built-in function
LSODE and it's use for simulation of set of ODEs.

System description

Computer:	486 DX/33- clone
OS:		Linux, kernel 1.2.13 (Slackware 3.0)
Octave version:	1.1.1 (binary distribution)


PROBLEM DESCRIPTION

Introduction

Problem with the LSODE function occurred when it was used to simulate
laboratory scale heat exchanger. Heat-exchanger model consists of
set of temperature derivatives. Model was verified with real input
data which was passed to the model as global variables because LSODE
doesn't allow more than two arguments for m-file model.

Because heat exchanger model is quite large package of files a similar
case was conducted to demonstrate the bug.


Problem

To recreate the bogus behaviour issue the following commands in Octave's
command window. Before that make sure you have m-file bug.m in the
working directory.

	lsode_options("relative tolerance",1e-3)
	lsode_options("absolute tolerance",1e-3)

	data=lsode("bug",[0],(t=linspace(0,20,21)'));

This will lead to the following error messages.

	error: matrix index = 23 exceeds maximum dimension = 21
	error: evaluating index expression near line 9, column 3
	error: evaluating assignment expression near line 9, column 2
	error: called from 'bug' in file '/home/jkaronen/octave/bug.m'
	error: lsode: evaluation of user-supplied function failed

Simulation of the 'bug.m' can be performed if the maximum step size
is set to 1 with the following command.

	lsode_options("maximum step size",1)

When 'bug.m' file is modified commenting out the line which returns
the simulation time 't' on every round it can be seen that the last
value of 't' is 22,289. This is for some unknown reason over the end
point of the simulation.


M-FILE

---------------------------------------------------------------------
function f=bug(x,t)

a=[5 5 5 6 6 6 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2]';

# Comment this out to see how Octave chooses simulation time steps.
# t

index=(t-(t-floor(t)))+1;

b=a(index);

f=1/(1+x^2)-b*x^2;

endfunction
---------------------------------------------------------------------