Failures on grid coming from treatment of runArgs.jobConfig not caught in CI

Original file: https://gitlab.cern.ch/atlas-physics/pmg/mcjoboptions/-/blob/9938e84d065b3b479dd57e063889896daa7ff9e7/521xxx/521163/MadGraphControl_TRSM_HHH.py#L21

Original MR: !2251 (merged)

Pipeline passed: https://gitlab.cern.ch/atlas-physics/pmg/mcjoboptions/-/pipelines/5008939

https://gitlab.cern.ch/atlas-physics/pmg/mcjoboptions/-/blob/9938e84d065b3b479dd57e063889896daa7ff9e7/521xxx/521163/MadGraphControl_TRSM_HHH.py#L21

This then failed on the grid: see ATLMCPROD-10348 example log: https://bigpanda.cern.ch//media/filebrowser/a8f96442-63a3-4574-aee8-53704fce19da/mc15_13TeV/tarball_PandaJob_5732336628_SiGNET/log.generate

Fix in: !2288 (merged)

New file in: https://gitlab.cern.ch/atlas-physics/pmg/mcjoboptions/-/blob/master/521xxx/521163/MadGraphControl_TRSM_HHH.py

Offending line seems to be:

f_list = os.listdir(runArgs.jobConfig[0])

where runArgs.jobConfig[0] on the grid seems to evaluate to 521163 while on the CI it would evaluate to ../521163. Not clear from which level Gen_tf.py runs on the grid.

@mborodin could you point me to the code that executes Gen_tf.py on the grid?