SYNOPSIS: BUG IN ONE OF THE EXAMPLE DRIVERS
	  SEMI-BUG WITH RANDOM NUMBER INITIALIZATION

                                               Boulder, 14/02/96
Greetings PIKAIA Users;

   Some problem were encountered by users attempting to run some
   of the examples codes for the problems discussed in chapter 
   5 of the User's Guide; investigating these has revealed a
   problem present in *all* the driver codes in the examples
   subdirectory, as well as in the driver included in the
   pikaia.f file. The problem is not with the PIKAIA subroutine
   itself. The following detail what is to be done

-----------------------------------------------------------------
                  ***** BUG REPORT #2 *****

   SYMPTOM:    results independent of the value of the random seed
   SUBROUTINE: all drivers (main in pikaia.f, xpk1a.f, xpk1b.f xpk2.f 
	       xpk3.f)
   PROBLEM:    superfluous line, leftover from an earlier random
	       number generator used by us for code development

   CORRECTIVE ACTION: remove the line

      seed=-abs(seed)

   in all drivers listed above. Examination of subroutine rninit
   will reveal that this line will force the generator to always
   be initialized with iseed=123456, independently of the seed
   value originally specified by the user

               ***** END OF BUG REPORT *****
------------------------------------------------------------------

   There is also an additional bug to be fixed in the driver code
   (xpk1b.f) for the nonlinear least-squares example of section 5.4,
   for those users who have grabbed copies of the drivers and fitness
   functions in the examples subdirectory:

-----------------------------------------------------------------
                  ***** BUG REPORT #3 *****

   SYMPTOM:    lack of acceptable convergence in running the
	       nonlinear least square example problem
   SUBROUTINE: finit (called by xpk1b.f)
   PROBLEM:    the number of fourier mode (m) to be used for the
               fit is left undefined !

   CORRECTIVE ACTION: add the following line

      m=5

   somewhere in subroutine finit; with m left undefined, most
   compilers will set m=0, causing the fit to proceed exactly as
   in the linear least squares problem of section 5.2, thus the
   lack of acceptable convergence.

               ***** END OF BUG REPORT *****
------------------------------------------------------------------

   I also did some silly things while renaming demo files for
   inclusion in example subdirectory, relating to the various
   include statements; the include statements pertaining to the
   fitness functions (fit*.f) have a superfluous .txt suffix,
   which must be removed. Also driver xpk1b.f should include fitness
   function fit1b.f instead of fit2.f, and driver xpk2.f should
   include fitness function fit2.f instead of fit4.f (which does
   not even exist!). Some of you may also have received a copy of the
   installation code where the seed is set automatically, without  any
   prompt being issued to the user for a seed value, contrary to what
   is described insection 2.2 of the User's guide; the required lines
   are commented out, which represents yet another leftover
   of a last minute test prior to official release. Our thanks to
   Richard Boivin (CERCA/Montreal, CAN) and Luciano Mantegazza
   (INFN/Pavia, IT) for calling various pieces of this mess to our
   attention. You may also notice that the original driver xpkaia
   and fitness function twod must be stripped from the original
   pikaia.f file if the examples drivers are to be used as is,
   i.e., with the include "pikaia.f" statement.

   Last but not least, the fitness function TWOD provided in
   the pikaia.f main file is NOT that discussed in Section 5.1
   of the User's Guide, but rather a very similar function which
   also has a maximum at (x,y)=(0.5,0.5), where f(x,y)=1, surrounded
   by rings of secondary extrema. It also makes a fine installation
   check problem...  but Mea culpa on that one at any rate. The
   difference with the function plotted on Fig 5.1 of the Guide
   has to do with the overall envelope of the function, which is
   biquadratic instead of exponential.

   All of the above problems have been corrected as of Wed 14 Feb 1996,
   mid-afternoon Colorado time.

   A few rather minor mistakes have been found in the User's
   guide by Steve Tomczyk (HAO/NCAR). These have now been
   corrected, and a file
   
   userguide_errata.txt

   has been put in the main pikaia directory, describing the
   changes.

   That's it for today. Actually, hopefully for this month...

							   Paul