Gemini on Question 14, Sep 17

The idiomatic way to handle this in Bash is to use the trap builtin command to catch the EXIT pseudo-signal.

When you set a trap on EXIT, Bash will execute your cleanup code regardless of whether the script finishes naturally, errors out, or is terminated by a standard signal (like SIGINT from Ctrl+C, or SIGTERM).

Here are the three most common and idiomatic ways to do it, depending on your exact needs:

1. The Job Control Method (Best for general use)

If your script spawns background tasks using &, you can use jobs -p to get a list of all currently running background process IDs spawned by the shell, and pass them to kill.

#!/bin/bash

# Suppress error messages in case the jobs already finished naturally
trap 'kill $(jobs -p) 2>/dev/null' EXIT

echo "Starting background jobs..."
sleep 100 &
sleep 200 &

echo "Doing work in the foreground..."
sleep 5
echo "Done. Exiting will now kill the background sleeps."

2. The Array Method (Best for targeting specific processes)

If you only want to kill certain background processes, or if you are doing complex things where jobs might not be reliable, track the Process IDs (PIDs) manually in an array.

#!/bin/bash

pids=()

# Set the trap to kill only the PIDs in the array
trap 'kill "${pids[@]}" 2>/dev/null' EXIT

echo "Starting specific jobs..."
sleep 100 &
pids+=($!) # $! is the PID of the last background command

sleep 200 &
pids+=($!)

echo "Doing work..."
sleep 5

3. The Process Group Method (Best for deep process trees)

If your script spawns a child, and that child spawns grandchildren, Methods 1 and 2 will leave the grandchildren orphaned. To kill the entire process tree, you send a signal to the whole Process Group by using kill 0.

#!/bin/bash

# kill 0 sends a SIGTERM to all processes in the current process group
trap 'trap - EXIT; kill 0' EXIT

echo "Starting deep process tree..."
# Even if this script spawns other scripts, kill 0 gets them all
bash -c 'sleep 100' & 

sleep 5

Note: trap - EXIT unregisters the trap before triggering kill 0 to prevent infinite recursion/weird behavior, since kill 0 also sends the signal to the parent script itself.

⚠️ The One Crucial Caveat: kill -9

No script can catch SIGKILL (kill -9).
If someone runs kill -9 <your_script_pid>, the Linux kernel instantly terminates your Bash script. The trap will not fire, and your child processes will be orphaned.

If your application absolutely cannot tolerate orphans even in a kill -9 scenario, you must use Linux-specific system calls (like prctl(PR_SET_PDEATHSIG, SIGTERM)) which cannot be done natively in pure Bash. You would need to wrap your child commands in a utility like prlimit or a tiny Python/C wrapper. For 99% of Bash scripts, however, the trap ... EXIT method is the accepted standard.