Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!bloom-beacon!think!ames!amdahl!dlb!dana!rap
From: rap@dana.UUCP (Rob Peck)
Newsgroups: comp.sys.amiga
Subject: Bug in audiotools
Message-ID: <221@dana.UUCP>
Date: Mon, 14-Sep-87 18:40:12 EDT
Article-I.D.: dana.221
Posted: Mon Sep 14 18:40:12 1987
Date-Received: Wed, 16-Sep-87 06:42:02 EDT
Organization: Dana Computer, Inc., Sunnyvale, CA
Lines: 26
Keywords: audio tools bug

I found a bug in the latest release of my audiotools... it has to do with
the calculation of the time to play a waveform.  Basically, any waveform
that is to play more than 1.18 seconds (1180 thousandths of a second)
will get an arithmetic overflow.  I was trying to retain the largest
number of digits of precision possible, so specified the duration in
1000ths of a second.  Part of the duration calculation has the
following calculation:  ((3579545 * duration)/1000) and this is
what overflows a 32-bit value.   To fix the problem, use the following
calculation instead:

	long value;	/* resultant value */

	long frac_part;	/* fractional part of the duration value in 1000ths */
	long int_part;  /* integer part of the duration value, in seconds */

	int_part = duration/1000;	/* truncates to an integer */

	frac_part = duration - (int_part * 1000);

	value = 3579545 * int_part +
		(3579545 * frac_part)/1000;

This gives the same value, but does not result in an arithmetic overflow,
while retaining the number of significant digits that you started with.

Rob Peck			...ihnp4!hplabs!dana!rap