IGCM_comp_Finalize

[acosce]

With Xios in server mode we don't need anymore rebuild for output netcdf files. So, as I understand, we use this test in IGCM_comp_Finalize :

# Rebuild has been done online or it was not needed

(...)
if ( [ ! X${thereisapatch} = X${NULL_STR} ] && [ X${AsynchronousRebuild} = Xtrue ] && [ -f ${file_in} ] ) ; then

IGCM_sys_Mv ${file_in} ${RUN_DIR}/REBUILD_${PeriodDateBegin}
(...)

else

(...)
if [ X${Pack} = Xtrue ] ; then

IGCM_sys_PutBuffer_Out ${file_in} ${file_out}

else

IGCM_sys_Put_Out ${file_in} ${file_out}

fi

fi

but "[ ! X${thereisapatch} = X${NULL_STR} ]" is true if we don't ask any post-treatment for a file (ex : (fileout.nc, ${R_OUT_SRF_O_M}/${PREFIX}_1M_fileout.nc, NONE)) and in this case we will move fileout.nc in REBUILD directory, even if we don't need this REBUILD.

Maybe is there a possibility to cancel all rebuild for file allready rebuild by xios.

Anne

[sdipsl]

Yes we move it in REBUILD in case we need to apply a patch. And we will apply this patch even in that case because we don't want to live with imperfect files:
(fileout.nc, ${R_OUT_SRF_O_M}/${PREFIX}_1M_fileout.nc, NONE)

The drawback is minimal. The rebuild job will be submitted and will just move files around and patch them if necessary. We still need rebuild.job for the meshmask.nc, and we still need the pacth for coupler files (that we barely never use).

May be renaming rebuild.job will be enough to close the ticket.

Let it open for now until further discussion.

[sdipsl]

Wont rename rebuild.job because the impact is too high on the existing material (documentation, ...)