Debug mode on HEAD v4 failed

[mathiot]

Context

Before looking for an issue appearing in my eORCA025 config poping up after my upgrade to NEMO 4.0.4, I decided to run the HEAT of v4 in debug mode on the UGA cluster, and I failed to run ORCA2_ICE_PISCES.

Compilation options used are: -i4 -r8 -O0 -fpe0 -g -traceback -fp-model precise -fno-alias -CB -ftrapuv -check uninit -debug -init=arrays,snan,huge

Analysis

• Some partially initialised variables (level jpk not intialised) lead to an error in lbcnfd.f90 or in p4zsed.F90

error (182): floating invalid - possible uninitialized real/complex variable.

This affect: zwz in dynvor.F90, ze3divh in sbcwave.F90 and nitrpot in p4zsed.F90

• numrtr is not initialised to -1, which lead to an out of bound when try to call iom_varid

• divu_i not initialised if ln_icedyn is off in icewri

Recommendation

• for unitialized variables dynvor.F90, sbcwave.F90 and p4zsed.F90, we can simply intitialised the jpk value to 0._wp.
• numrtr should be set to -1 in trc.F90
• divu_i should only be printout if ln_icedyn is set to true

Extra information:

• not tested in the trunk (need to be checked)
• other sette configurations not tested

ICE/icewri.F90

@@ -252 +252 @@
       CALL iom_rstput( 0, 0, kid, 'snwpre', sprecip      )   ! Snow precipitation
       CALL iom_rstput( 0, 0, kid, 'sisali', sm_i         )   ! Ice salinity
       CALL iom_rstput( 0, 0, kid, 'sivolu', vt_i         )   ! Ice volume
-      CALL iom_rstput( 0, 0, kid, 'sidive', divu_i*1.0e8 )   ! Ice divergence
       CALL iom_rstput( 0, 0, kid, 'si_amp', at_ip        )   ! Melt pond fraction
       CALL iom_rstput( 0, 0, kid, 'si_vmp', vt_ip        )   ! Melt pond volume
       CALL iom_rstput( 0, 0, kid, 'sithicat', h_i        )   ! Ice thickness
@@ -259 +258 @@
       CALL iom_rstput( 0, 0, kid, 'siconcat', a_i        )   ! Ice concentration
       CALL iom_rstput( 0, 0, kid, 'sisalcat', s_i        )   ! Ice salinity
       CALL iom_rstput( 0, 0, kid, 'snthicat', h_s        )   ! Snw thickness
+      IF ( ln_icedyn )  CALL iom_rstput( 0, 0, kid, 'sidive', divu_i*1.0e8 )   ! Ice divergence
 
     END SUBROUTINE ice_wri_state
 

OCE/DYN/dynvor.F90

@@ -221 +221 @@
          IF(lwp) WRITE(numout,*) '~~~~~~~~~~~'
       ENDIF
       !
+      zwz(:,:,jpk) = 0._wp
       !
       SELECT CASE( kvor )                 !==  volume weighted vorticity considered  ==!
       CASE ( np_RVO )                           !* relative vorticity
@@ -573 +574 @@
          IF(lwp) WRITE(numout,*) '~~~~~~~~~~~'
       ENDIF
       !
+      zwz(:,:,jpk) = 0._wp
       !                                                ! ===============
       DO jk = 1, jpkm1                                 ! Horizontal slab
          !                                             ! ===============

OCE/SBC/sbcwave.F90

@@ -219 +219 @@
       !
       !                       !==  vertical Stokes Drift 3D velocity  ==!
       !
+      ze3divh(:,:,jpk) = 0._wp
       DO jk = 1, jpkm1               ! Horizontal e3*divergence
          DO jj = 2, jpj
             DO ji = fs_2, jpi

TOP/PISCES/P4Z/p4zsed.F90

@@ -491 +491 @@
       !
       IF( p4z_sed_alloc /= 0 )   CALL ctl_stop( 'STOP', 'p4z_sed_alloc: failed to allocate arrays' )
       !
+      ! initialisation
+      nitrpot(:,:,jpk) = 0._wp
+      !
    END FUNCTION p4z_sed_alloc
 
    !!======================================================================

TOP/trc.F90

@@ -24 +24 @@
    INTEGER, PUBLIC ::   numtrc_cfg = -1   !: reference passive tracer namelist_top_cfg
    INTEGER, PUBLIC ::   numonr     = -1   !: reference passive tracer namelist output output.namelist.top
    INTEGER, PUBLIC ::   numstr            !: tracer statistics
-   INTEGER, PUBLIC ::   numrtr            !: trc restart (read )
+   INTEGER, PUBLIC ::   numrtr     = -1   !: trc restart (read )
    INTEGER, PUBLIC ::   numrtw            !: trc restart ( write )
 
    !! passive tracers fields (before,now,after)

[mathiot]

Part of these issue has already been reported in the trunk in 13526, 13527, 13546, 13558. These commit has to be carefully check to see if they apply into NEMO4.

[mathiot]

Fix has been done in the trunk but not implemented in the trunk. It correspond to commit 13526, 13527, 13546, 13558 (from personal discussion with Seb).

[clem]

In 14588:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found