This is the list of changes since Dec 28, 1997.

June 1, 2008: I've found a bug in the realification routines (file
   irex.cc). It only shows up if you work with terms of order 34 or
   higher (for this reason I didn't realize of the problem before).
   The reason is that the binomial coefficients are stored in standard
   integers (32 bits long), and this is not enough for order 34 (and
   higher). Now it is corrected.

Sep 10, 2007: I've introduced a few very minor changes in some files,
   to supress some warnings issued by my compiler (gcc 4.1.2).
   Apparently, none of them had any other side effect apart from the
   warnings when compiling.

Aug 1, 2002: Rytis Paskauskas <rytis@cns.physics.gatech.edu> reports a
   bug in the realification routines (see also the note of Jan 17,
   1998 below). I think the bug is corrected now.

Jan 21, 1998: I've introduced a small change in the routine that
   realifies the hamiltonian reduced to the centre manifold (it is
   contained in the file rcm6s.cc). Now the routine calls function
   kill_cm to know which monomials have been killed by the normal form
   process. In the previous version, the killing criterion was also
   coded in the realifying routine rcm6s. This modification allows to
   change the way of computing the normal form (there are several ways
   of doing that) by simply changing the function kill_cm.

Jan 17, 1998: I've discovered a bug in the routines remo6s and
   remo6p. It affects to the realification of power series in the
   following situation: imagine we have computed a normal form (or a
   partial normal form) where some of the variables are hyperbolic.
   Then, the hyperbolic variables have not been complexified for the
   normal form process and, hence, they don't need to be realified.
   This is taken into account by routines remo6s and remo6p but, due
   to a programming mistake, this was not done in the proper way. The
   result was that the hyperbolic variables were altered (so that the
   result was incorrect). This bug didn't show up in the tests
   explained in the paper doc/mac.ps because there, we only deal with
   hyperbolic directions when computing the reduction to the centre
   manifold. In this case, after computing the realifications for the
   changes of variables, we set the hyperbolic directions to zero (to
   only work with the centre manifold), that cancels the wrong
   coefficients in the expansions.
