GPT on Question 14, Sep 26

In Bash, the usual pattern is to trap shell exit/signals, terminate the children, and wait for them:

#!/usr/bin/env bash

children=()

cleanup() {
    status=$?
    trap - EXIT HUP INT TERM

    if ((${#children[@]})); then
        kill -TERM "${children[@]}" 2>/dev/null || true
        wait "${children[@]}" 2>/dev/null || true
    fi

    exit "$status"
}

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

worker_one &
children+=("$!")

worker_two &
children+=("$!")

wait

This handles normal exit and catchable signals. It cannot handle SIGKILL, power loss, or the shell crashing before its traps run.

If workers create grandchildren, killing only their PIDs may not be enough. Put each worker in its own process group/session and terminate the group:

setsid worker &
pgid=$!

cleanup() {
    status=$?
    trap - EXIT HUP INT TERM
    kill -TERM -- "-$pgid" 2>/dev/null || true
    wait "$pgid" 2>/dev/null || true
    exit "$status"
}

This assumes the worker does not deliberately escape by creating another session. For production services, the most reliable approach is to use a supervisor such as systemd with KillMode=control-group; Bash alone cannot guarantee cleanup after an uncatchable parent death. On Linux, another option is having each child request PR_SET_PDEATHSIG, but that requires support in the child or a small wrapper—Bash has no built-in for it.