ORCA2 + land_suppression failed to run

[mathiot]

Context

I run the sette ORCA2_ICE_PISCES test and land suppression activated (jpnij=27, jpnj=4, jpni=7) compiled with debug options. It failed after the first time step because of NaN in the top right processor (the top left is removed). It try ln_nnogather=.true., it doesn't change anything. Without land suppression, it run fine.

Analysis

It appears that in mpp_nfd_generic.h90 some part of the temporary array ztabr,ztabl (nnogather) and ztab are not initialised in this particular case (land value 'overlapping' the missing processor I think).

Fix

Initialise these arrays to 0._wp.

[mathiot]

In 9945:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

[smasson]

I agree,

however, initializing arrays to 0 is often a bad solution that is potentially hiding a real bug!

In this specific case:

• when there is no land-processor suppression, all values of ztab* arrays should be defined. If it is not the case, this means there is a bug somewhere...
• when some processors of the north fold are suppressed, values of ztab* arrays corresponding to these suppressed domain won't be defined and we need a default definition to 0.

I propose to add an if in from of these initializations to make sure we are in a "suppressed land-processors" case. For example

IF ( jpni*jpnj /= jpnij )

I don't know a simple way to have a better test: testing if "suppressed land-processors" belongs to the north-pole folding...

[mathiot]

In 9971:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found