Testing gnu parallel behaviour

random_duration_script.py

#!/usr/bin/python3
import sys
from time import sleep
import random

for i in range(random.randint(3, 8)):
    sleep(random.random() / 10)
    print(f"{sys.argv[1]}, step {i}")
    if i == 2 and ("arg1" in sys.argv[1] or "arg2" in sys.argv[1]):
        raise ValueError(f"{sys.argv[1]} raised an exception")

Removing the commands wrapping the call to random_duration_script.py confirms that the error code is otherwise preserved even when processes complete successfully following a failing process:

$ parallel './random_duration_script.py {}' ::: arg1 arg2 arg3 arg4 arg5 2&> /dev/null; echo $?
2

(note that I pipe stdout and stderr to /dev/null here as it is unimportant in this exploration of error code handling of 'parallel')

We can retain our error code AND still echo our 'endgroup' with a minor change:

$ parallel 'echo ::group::{}; ./random_duration_script.py {} && echo ::endgroup:: || (echo ::endgroup:: && exit 1) ' ::: arg1 arg2 arg3 arg4 arg5 2&> /dev/null; echo $?
2