Last week I posted a note about my difficulty in getting octave
to print out a nice, clean 1.000000000000000000e+00.
The problem seems to be in the printing routines:
octave:1> format long e
octave:2> x = 1
x = 1.00000047683363e+00
octave:3> x .^ 100
ans = 1.00000047683363e+00
seems to be about the norm.
I am using the precompiled octave available from ftp.che.utexas.edu
for irix 4.0.5.
As long as calculations are done accurately internally, this isn't really
a major bug. What concerns me is if you are relying on saving and loading
floating point data often... the printf and scanf routines could be
doing some strange things then...