********>Bugfix 16:
Author: Dave Case
Date: 07/29/2004

Programs: sander

Description: One of the broadcasts from the master nodes to the slaves
             (in routine setup) had a bad length.  This potentially could
             cause spurious results (first discovered on MacOSX systems).
             If your system appears to be working, and passes the parallel
             tests, then you should not need to worry, although applying the
             patch is still a good idea.

Fix:  apply the following patch to amber8/src/sander/parallel.f

------------------------------------------------------------------------------
*** parallel.f	2004/02/10 02:10:12	7.112
--- parallel.f	2004/07/29 21:25:51
***************
*** 83,89 ****
     
     !  extra.h:
     
!    call mpi_bcast(ilbopt,6,MPI_INTEGER,0,commsander,ierr)
     call mpi_bcast(lbwght,1,MPI_DOUBLE_PRECISION,0,commsander,ierr)
     
     !  box.h:
--- 83,89 ----
     
     !  extra.h:
     
!    call mpi_bcast(ilbopt,5,MPI_INTEGER,0,commsander,ierr)
     call mpi_bcast(lbwght,1,MPI_DOUBLE_PRECISION,0,commsander,ierr)
     
     !  box.h:
------------------------------------------------------------------------------

Temporary workarounds: none.