bug in bdyini : straight BDY only works with rim width of 1

[molines]

Context

NEMO_3.6_STABLE with straight BDY (as old OBC) and rim width > 1 produce a Seg. fault due to out of bound on dta_global2 array

Analysis

In this case bdy data are organized as a vector with all the data within the rim put sequentially from the outermost to innermost. Therefore it seems to me that in bdyini.F90 we must have

jpbdtas = MAX(jpbdtas, (nbdyend - nbdybeg + 1) * nn_rimwidth(ib_bdy)  )

instead of

jpbdtas = MAX(jpbdtas, (nbdyend - nbdybeg + 1))

Further more, the allocation of dta_global2 should be (not tested):

ALLOCATE( dta_global2(jpbdtas, 1, jpk) ) 

instead of

ALLOCATE( dta_global2(jpbdtas, nrimmax, jpk) )

Fix

The only change of jpbdtas as sugested above fix the problem of outbound.

[mdunphy]

Hi,

I ran into this same issue (out of bounds for the structured case). My conclusion was that my input netcdf file was wrong -- in the structured case we should use a two-dimensional netcdf file, i.e.,

yb = nn_rimwidth ;
xbT = nbdyend - nbdybeg + 1 ;

such that the input routine correctly loads it into the 3D array dta_global2. Meanwhile it's the unstructured case that expects data in vector form, and loads into dta_global.

[jamesharle]

After an email exchange with Jean-Marc, the conclusion is that this may have indeed been an data format issue and not a bug in the code.