ORCA2_LIM3_PISCES crashes after 179 timesteps in dev_merge_2017

[cbricaud]

Context

ORCA2_LIM3_PISCES crashes after 179 timesteps in dev_merge_2017 ( SETTE tests, but 1500 TS instead of 150 for LONG run)

ORCA2_LIM3_PISCES is also not reproducible ( diff in run.stat after 71 timesteps )

Analysis

dev_METO_MERCATOR_2017 : same problem
dev_METO_2017 : no problem
dev_MERCATOR_2017 : same problem

dynspg_ts uses rn_scal_load.
rn_scal_load value is defined and read in tideini, even if ln_tide=F
As rn_scal_load is not initialized, its value is read in namelist ( rn_scal_load = 0.094 in namelist_ref )

Recommendation

I propose to force rn_scal_load = 0. when ln_tide=F

  IF( ln_tide ) THEN
      [...]
  ELSE
      rn_scal_load = 0._wp 
      ! rn_scal_load is used in dynspg_ts, even if ln_tide=F ;
      ! if (rn_scal_load is =/ 0. .AND. ln_tide=F in namelist), it will modify results.
      IF(lwp) WRITE(numout,*)
      IF(lwp) WRITE(numout,*) 'tide_init : tidal components not used (ln_tide = F)'
      IF(lwp) WRITE(numout,*) '~~~~~~~~~ '
      IF(lwp) WRITE(numout,*) '         Fraction of SSH used in scal. approx.   rn_scal_load = ', rn_scal_load
      RETURN
   ENDIF

With this fix, ORCA2LIM3PISCES "LONG SETTE run" runs 1500 timesteps.
But it is still not reproducible ( diff in run.stat after 71 timesteps )

[nicolasmartin]

Switch to 'trunk' version, 4.0 is not officially released.
I don't even know how you set it, the entry should not be available from the drop-down menu.

[jchanut]

In 9337:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

[jchanut]

In 9338:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found