Prepare_cmds - improvements

[Lewih]

In the documentation, the prepare_cmds are associated with the partition and it is possible to add them in the config file.

It should be possible to define something like self.prepare_cmds in the initialization method of any test like the self.prerun_cmds property.
This can be useful in some circumstances.

If somehow it is already feasible, the documentation should provide a clear example.

[vkarak]

Wouldn't the prerun_cmds in the test work?

[Lewih]

Considering the following test:

import reframe as rfm
import reframe.utility.sanity as sn


@rfm.simple_test
class HelloMultiLangTest(rfm.RegressionTest):
    lang = parameter(['c', 'cpp'])
    valid_systems = ["+default", "+login", "+test"]
    valid_prog_environs = ['-mpi']
    tags = {'calcua', 'basic', 'compilation'}
    prepare_cmds = ['echo "this is a prepare cmd"']
    executable_opts = ['> hello.out']
    sanity_patterns = sn.assert_found(r'Hello, World\!', 'hello.out')

    @run_before('compile')
    def set_sourcepath(self):
        self.sourcepath = f'hworld.{self.lang}'

the generate bash scripts are in the form

#!/bin/bash
#SBATCH --job-name="rfm_HelloMultiLangTest_ed3216b7"
#SBATCH --ntasks=1
#SBATCH --output=rfm_job.out
#SBATCH --error=rfm_job.err
#SBATCH -A my_account
#SBATCH --reservation=myreservation
module load mytoolchain
./HelloMultiLangTest_0 > hello.out

hence no prepare cmds added.

[Lewih]

Wouldn't the prerun_cmds in the test work?

This happens after the module load, I would like to execute some commands before the module load by using the prepare_cmds.

[vkarak]

Indeed, this is limitation, but we are considering a more general solution to this and similar issues by using a templating mechanism for generating the final script so that users could customize it precisely.