Correction Author: David A. Pearlman
Correction Date: 11/13/91

Programs: gibbs

Severity: slight

Problem:  If the user runs a free energy simulation where 

               1) NRUN > 1, and
               2) the simulation terminates because the entire lambda range
                  0<->1 has been spanned on some run < NRUN

     then the "MICSTAT" free energy summary file will either not be printed
     (if fewer than NDMPMC windows have been calculated), or else will be
     incomplete (if more than NDMPMC windows have been calculated; NDMPMC
     is by default set to 100).

     The bug will occur only if both of the above conditions are met.

     This bug does *not* affect energies, trajectories or any other feature
     of the program. In any case, the requisite energy information will 
     still be printed in the standard output file, and can be extracted from 
     there.
     
Affects: No "MICSTAT" file will be written, or the "MICSTAT" file will be
     incomplete (missing some entries at the end).

Cause: A logical expression used to determine if the buffers which store
     the information for the MICSTAT file should be dumped to the file and
     cleared was incomplete.

Fix: In routine RUNMD (in file amber4/src/gibbs/giba.f), make the
     following change:

        old line:
           IF (LASTRN .OR. TIMDNE.OR.ERROR)

        new line:
           IF (LASTRN .OR. TIMDNE.OR.ERROR .OR. DONE)

Temporary workarounds: You can ensure that the simulation will
     complete on the NRUNth run by setting NRUN and NSTLIM appropriately. 
     Or you can set NRUN=1 (and set NSTLIM to be appropriately large or -1). 
     In this case, you ensure that you will never to encounter the bug.

Routines affected:

     GIBBS :  Routine RUNMD in file .../amber4/src/gibbs/giba.f