12 months runoff hard coded...

[smasson]

Context

Reported by Rachid: in sbc_rnf_init, the use of 12 record runoff files is hard coded...
Indeed we have a nice hard coded declaration...

      REAL(wp), DIMENSION(12)                 :: zrec             ! times records

No need to say that there is no call to any warning/stop or any comment to warn people about this hard coded definition...

This zrec is required but, in fact, not needed in iom_gettime:

         CALL iom_gettime( inum, zrec, kntime=nbrec)

nbrec is next used to allocate a 3D array zrnfcl(jpi,jpj,nbrec) without any test on the error status exit... I let you imagine what happen when nbrec = 365...

Cherry on the cake: this 3D array is not needed as the iom_get is done in a temporal loop:

         DO jm = 1, nbrec
            CALL iom_get( inum, jpdom_data, TRIM( sn_rnf%clvar ), zrnfcl(:,:,jm), jm )
         END DO

Last point, iom_gettime mix-up unlimited dimension size with a time axis variable... this routine requires an input array (that is in fact never used) which size is not know when calling iom_gettime...

Analysis

Clean up all this mess...

Recommendation

Dont't let this king of crappy code to be commited...

[smasson]

In 10522:

trunk: replace iom_gettime by iom_getszuld, see #2212

[smasson]

[10522] : first step: get rid of iom_gettime to avoid the use of an array which size is unknown...

[smasson]

In 10523:

trunk: avoid the allocation of potentialy large array in sbcrnf, see #2212