docker: add ENABLE_NQPTP to opt out of starting NQPTP

[haavar]

What this is

Adds an ENABLE_NQPTP environment variable to the Docker image's launcher (docker/run.sh) so a container can be told not to start its own NQPTP.

if [ -z "$SERVICE_TYPE" ] && { [ -z ${ENABLE_NQPTP+x} ] || [ $ENABLE_NQPTP -eq 1 ]; }; then
  echo "Starting NQPTP ($(date))"
  (/usr/local/bin/nqptp > /dev/null 2>&1) &
fi

Why

The launcher starts NQPTP for every AirPlay 2 container. When several AirPlay 2 instances run on one host they should share a single NQPTP — it needs exclusive use of ports 319/320 and is multi-client (it keeps a separate clock per shared-memory interface name). Today, running one instance per room means every container also starts its own NQPTP, and they contend for 319/320. The only way to avoid it was a custom entrypoint.

With this change, you run one dedicated NQPTP (e.g. a sidecar with entrypoint: ["/usr/local/bin/nqptp"]) and set ENABLE_NQPTP=0 on each instance container:

  kitchen:
    image: mikebrady/shairport-sync:latest
    network_mode: host
    ipc: host
    environment: [ENABLE_NQPTP=0]
    command: ["-a", "Kitchen", "--port=7000", "--", "-d", "room_kitchen"]
    depends_on: [nqptp]

Notes

• Mirrors the existing ENABLE_AVAHI convention in the same script.
• Default is unchanged — with ENABLE_NQPTP unset (or 1), NQPTP is started exactly as before, so single-instance users and existing setups are unaffected.
• Classic-only launches (--service-type=classic / airplay1) already skip NQPTP; that is unchanged.

Testing

Built container runs of the image with the modified launcher:

• ENABLE_NQPTP=0 → the launcher does not print "Starting NQPTP" and no nqptp process is started;
• default (unset) → "Starting NQPTP" is printed and NQPTP starts, as before.

sh -n docker/run.sh is clean.

This supports the multi-instance-per-host setup discussed in #2266.

[mikebrady]

Thanks for this!