fix: guard Plex finish kill when no PID is found

[OmarB97]

Summary

Guard the initial kill -15 $pids call in root/etc/services.d/plex/finish so it only runs when a Plex Media Server PID was found.

Why

If Plex has already exited by the time the finish script runs, pids is empty and the shell executes kill -15 without a PID. That prints the kill: usage... message during container shutdown. The later SIGKILL block already guards against an empty PID list, so this makes the TERM path match the existing stuck-process path.

Validation

• bash -n root/etc/services.d/plex/finish
• Reproduced in a downstream Unraid container using plexinc/pms-docker:latest, where the shutdown logs showed kill: usage...; applying the guard removed that message while Plex restarted healthy.

[cilindrox]

Thanks @OmarB97 - some comments regarding improving the globbing/splitting of the unquoted vars.

[cilindrox]

Can we quote the $pids var for safety here and on L18 below?

Suggested change

	kill -15 $pids
	kill -15 "$pids"

[OmarB97]

Thanks @OmarB97 - some comments regarding improving the globbing/splitting of the unquoted vars.

Thanks, pushed c506e49 to address this.

I quoted the PID handling, but used a small kill_pids helper instead of calling kill -15 "$pids" directly because the command substitution can return multiple newline-separated PIDs. Quoting the whole value in a direct kill call would turn multiple PIDs into one argument; the helper reads each PID line and calls kill "$signal" "$pid", so each PID is quoted safely while preserving multi-process behavior.

I applied the same path for both the SIGTERM and SIGKILL calls and checked it with bash -n plus a stubbed multi-PID helper test.