GPT on Question 14, Sep 21

There is no fully reliable “after the parent dies” mechanism in portable Bash. Orphans are immediately reparented, and an EXIT trap cannot run if the script receives SIGKILL or crashes.

The usual pattern is:

• 1.Run children in a dedicated process group/session.
• 2.Trap normal exit and catchable signals.
• 3.Kill the entire child process group, not just the immediate child.

#!/usr/bin/env bash
set -Eeuo pipefail

setsid my_worker --arg foo &
worker_pgid=$!

cleanup() {
    trap - EXIT INT TERM HUP

    # Negative PID means “signal this process group”.
    kill -TERM -- "-$worker_pgid" 2>/dev/null || true

    # Optional grace period, then force termination.
    sleep 2
    kill -KILL -- "-$worker_pgid" 2>/dev/null || true

    wait "$worker_pgid" 2>/dev/null || true
}

trap cleanup EXIT
trap 'exit 130' INT
trap 'exit 143' TERM
trap 'exit 129' HUP

wait "$worker_pgid"

Avoid relying on pkill -P "$$": it only finds current direct children and is vulnerable to children forking or being reparented.

If cleanup must also happen when the parent is killed with SIGKILL, Bash alone cannot guarantee it. Use one of:

• •systemd service/scope with KillMode=control-group
• •a container or cgroup supervisor
• •Linux prctl(PR_SET_PDEATHSIG, SIGTERM) in the child, usually via a small wrapper
• •a dedicated watchdog/supervisor

For production services, a supervisor/cgroup is generally the idiomatic and most reliable solution.