The objective is to detect possible errors or typos generated during the modification of the source code (bash scripts or Fortran files). The results obtained with simple configurations are known and have to be the same, whatever the version of the code we use.
The method will consist in running various test cases and comparing the results with references outputs.
For each test case:
- the number of output files generated will be checked.
- the content of the output files will be compared with reference results.
A set of scripts performing automatically the analysis and the comparisons will be developed and available under SVN. It will be accompanied with several test cases, each containing:
- an input directory with the necessary input files,
- an outputref directory with the reference result files.
The number of test cases shall be increased progressively, taking into account the new developments in the source code.
To perform basic operations and comparisons on the netCDF files, the netCDF Operator (http://nco.sourceforge.net/ NCO) will be employed.
Its installation and description can be found here.
A single data point is put in the center of a square domain.
With lot of options. Advection, variable corelation length, source, diffusion.