The debugger-output cut-and-pasted below has me completely at a loss
(sorry about the wrapping problems).

DV is computed at line 2164.  Line 2166 tests DV for strictly positive;\
one should be able to get to line 2168 only for *positive* values of DV.
But dbx's "print dv, dh" at that point shows a *negative* DV.

What the ??!?!

The debugger is sunstudio12.1/bin/dbx, with "ident" reporting version
"$FreeBSD: src/usr.bin/yacc/skeleton.c,v 1.28 2000/01/17 02:04:06 bde 
Exp $".  The compiler is sunstudio12.1/bin/sunf95 "Sun Fortran 95 8.4 
Linux_i386 2009/06/03".

The code is compiled with "-g -C -xarch=native -xtarget=native -m64 
-xdebugformat=dwarf -xs -M. -xfilebyteorder=big8:%all -openmp=noopt 
-stackvar -c", in a run with OMP_NUM_THREADS=1.  Both DV and DH are
PRIVATE in the enclosing parallel loop, by the way, as you can see
from the "larger context" section at the bottom...

"uname -a" gives "Linux hilbert 2.6.27.5-server-2mnb #1 SMP Thu Nov 20 
16:22:43 EST 2008 x86_64 Intel(R) Core(TM)2 Duo CPU     E8400  @ 3.00GHz 
GNU/Linux".


-- Carlie Coats

------------  cut here for debugger output  ---------------------------

t@139859468928896 (l@6097) stopped in _$d1J2151.mod_route.routechan_ at 
line 2168 in file "mod_route.f90"
  2168                               DH         = 1.0E3 * DV / WAREA( K )
(dbx) print dv, dh
dv = -9067.942
dh = 0.0
(dbx) list 2164,2170
  2164                           DV = CVOLV( K ) - BFVOL( K )
  2165
  2166                           IF ( DV .GT. 0.0 ) THEN
  2167
  2168                               DH         = 1.0E3 * DV / WAREA( K )
  2169                               CVOLV( K ) = BFVOL( K )
  2170                               HEADV( K ) = BFDEP( K )

------------  Larger context  ----------------------------------------

         IF ( CHNFDBK ) THEN

!$OMP       PARALLEL DO DEFAULT( NONE ),                               &
!$OMP                    SHARED( NFLOWS, RCHMTHD, N0, N1, CVOLV, BFVOL,&
!$OMP                            WAREA, HEADV, BFDEP, STRWET, NCOLS2,  &
!$OMP                            SFCHEADRT, NSTOVL ),                  &
!$OMP                   PRIVATE( F, K, L, M, C, R, DH, DV ),           &
!$OMP                  SCHEDULE( DYNAMIC )

             DO F = 1, NFLOWS

                 IF ( RCHMTHD( F ) .EQ. DIF_WAVE ) THEN

                     DO K = N0( F ), N1( F )

                         DV = CVOLV( K ) - BFVOL( K )

                         IF ( DV .GT. 0.0 ) THEN		!  overflow

                             DH         = 1.0E3 * DV / WAREA( K )
                             CVOLV( K ) = BFVOL( K )
                             HEADV( K ) = BFDEP( K )

                             DO L = STRWET( K-1 ) + 1 , STRWET( K )
                                 M = STRWET( NSTOVL+L ) - 1
                                 C = 1 + MOD( M , NCOLS2 )
                                 R = 1 +      M / NCOLS2
                                 SFCHEADRT( C,R ) = SFCHEADRT( C,R ) + DH
                             END DO

                         END IF      !  if overflow

                     END DO          !  end loop on cells K in this reach

                 END IF              !  if flow is diffusive-wave

             END DO                  !  end loop on reaches F

         END IF          !  if chnfdbk

On Nov 6, 11:56 am, Carlie Coats <car...@jyarborough.com> wrote:
> The debugger-output cut-and-pasted below has me completely at a loss
> (sorry about the wrapping problems).

I can think of several reasons dbx might produce the
results you indicate, but I do not have enough
information to reproduce the problem.  Please file
a bug report and include a test case that
demonstrates the problem.  If you wish, you can use
http://bugs.sun.com to file the bug report.

Bob Corbett