From help-octave-request at che dot utexas dot edu  Sun Jul 10 12:26:14 1994
Subject: 1 does not equal 1
From: Ted Harding <Ted dot Harding at nessie dot mcc dot ac dot uk>
To: andrew at watsci dot uwaterloo dot ca
Cc: help-octave at che dot utexas dot edu
Date: Sun, 10 Jul 1994 18:25:34 +0100 (BST)

" To: help-octave at che dot utexas dot edu
" Subject: float->double conversion problem?
" 
" Perhaps someone can help me here: it seems that one is not equal to one...
" 
" Octave, version 1.0.  Copyright (C) 1992, 1993, 1994 John W. Eaton.
" This is free software with ABSOLUTELY NO WARRANTY.
" For details, type `warranty'.
" 
" octave:1> format long e
" octave:2> 1
" ans =  1.00000047683363e+00
" octave:3> 1/1
" ans =  1.00000047683363e+00
" octave:4> quit
" 
" So, what exactly does this mean? All attempts at setting a variable to exactly
" 1.0 have failed... is this a bug in the printing subroutines, or is a 
" conversion between a float and a double going on somewhere that twiddles with
" the low order bits on the floating point number?
" 
" -Andrew. (andrew at watsci dot uwaterloo dot ca)
----------------------------------------------------------------------------
Compare the above with the following (copied from the output):
=====================================
Octave, version 1.0.  Copyright (C) 1992, 1993, 1994 John W. Eaton.
This is free software with ABSOLUTELY NO WARRANTY.
For details, type `warranty'.

octave:1> format long e
octave:2> 1
ans =  1.00000000000000e+00
octave:3> 1/1
ans =  1.00000000000000e+00
octave:4> 1.0
ans =  1.00000000000000e+00
octave:5> quit
=====================================
This was produced by a pre-compiled octave binary (as found in
ftp.che.utexas.edu:/pub/octave/BINARIES). If yours was compiled by you,
there may be a compiler problem. What you observe is also produced by
a well-known trap in C with printf and scanf. If 'x' is a double
then scanf("%f",&x) stores the input as a _float_ in the top half of 'x'
and does not touch the bottom half, which may contain garbage. If you then
do printf("%f",x) you get the garbage as well. Printf is supposed to
automatically cast from float to double before formatting for printing,
so the trouble should not occur in printf. The cure for the C trap is
to make sure that you use scanf("%Lf",&x). Something analogous may be
happening in your 'octave'.
Ted Harding                                        Ted dot Harding at nessie dot mcc dot ac dot uk