#!/bin/sh
# Push this working tree to the NAS and restart it.
#
# The NAS only accepts password auth, so one ssh master connection is opened
# up front and every later step rides on it: you type the password once, not
# five times. (macOS's bundled rsync 2.6.9 cannot carry a password through
# -e at all — hence tar over ssh.)
#
#   ./deploy/push.sh [user@host] [port]
#
# The password is never stored here — ssh asks for it on your terminal. To
# stop being asked at all, run `ssh-copy-id -p 2222 ericwyuan@192.168.50.64`
# once; after that this script runs unattended.
#
# Never touches .env, .venv or the database on the far side.
set -e

HOST="${1:-ericwyuan@192.168.50.64}"
PORT="${2:-2222}"
# Needed for the sudo restart below. Prompted for rather than stored, and only
# used over the already-authenticated ssh master connection.
PASS="${NAS_PASSWORD:-}"
REPO="$(cd "$(dirname "$0")/.." && pwd)"
CTL="$(mktemp -u /tmp/garmin-deploy-XXXXXX)"

sh_() { ssh -S "$CTL" -o BatchMode=yes "$HOST" "$@"; }

cleanup() { ssh -S "$CTL" -O exit "$HOST" 2>/dev/null || true; }
trap cleanup EXIT

echo "==> connecting to $HOST:$PORT (password prompt follows, once)"
ssh -M -S "$CTL" -fN -p "$PORT" -o ControlPersist=300 "$HOST"

if [ -z "$PASS" ]; then
  # Same password as the ssh login; asked for separately because ssh consumed
  # the first one itself and sudo on the far side needs it on stdin.
  printf 'sudo password for %s (needed to restart the root-owned service): ' "$HOST" >&2
  stty -echo 2>/dev/null; read PASS; stty echo 2>/dev/null; echo >&2
fi

# The app dir has moved before; find it rather than assume it.
APP=$(sh_ 'for d in ~/apps/garmin-health-lab /volume1/web/garmin-health-lab; do
             [ -d "$d/backend" ] && { echo "$d"; break; }; done')
[ -n "$APP" ] || { echo "cannot find the app dir on $HOST" >&2; exit 1; }
echo "==> app dir: $APP"

# STATIC_DIR is ./static relative to backend/, which is where start.sh cds to.
STATIC="$APP/backend/static"
# `cmd && VAR=x` would trip `set -e` when cmd fails, so spell it out.
if sh_ "[ -f '$APP/static/index.html' ]" 2>/dev/null; then
  STATIC="$APP/static"
fi
echo "==> static dir: $STATIC"

# macOS bsdtar writes com.apple.provenance xattrs and ._ resource forks that
# the NAS's tar cannot read; it warns once per file and copies nothing useful.
TAR="tar czf - --no-xattrs"
export COPYFILE_DISABLE=1

echo "==> backend"
$TAR --exclude .venv --exclude .env --exclude __pycache__ \
    --exclude '*.db' --exclude tests --exclude .pytest_cache \
    -C "$REPO/backend" . | sh_ "tar xzf - -C '$APP/backend'"

echo "==> static (cleared first, so stale JS chunks do not pile up)"
if [ ! -f "$REPO/client/build/index.html" ]; then
  echo "client/build is missing — run 'npm run build' first" >&2
  exit 1
fi
sh_ "rm -rf '$STATIC' && mkdir -p '$STATIC'"
$TAR -C "$REPO/client/build" . | sh_ "tar xzf - -C '$STATIC'"

# The service is started at boot as root (DSM Task Scheduler -> S99garmin.sh),
# so logs/error.log and logs/access.log are root-owned. Restarting as
# ericwyuan therefore fails instantly — gunicorn cannot open its own error log
# — and this went unnoticed for a whole deploy: stop.sh could not kill a root
# process either, so the OLD master kept serving :8124, start.sh's health
# check saw a 200 and reported "started", and the new code was never loaded.
# Hence sudo, matching how the service actually runs.
SUDO="echo '$PASS' | sudo -S"
echo "==> restart (sudo: the service runs as root, started at boot)"
BEFORE=$(sh_ "pgrep -f '$APP/backend/.venv/bin/gunicorn' | tr '\n' ',' " || true)
sh_ "cd '$APP' && $SUDO sh deploy/stop.sh >/dev/null 2>&1; sleep 3; $SUDO sh deploy/start.sh" \
  || { echo "restart failed" >&2; exit 1; }

echo "==> health"
sh_ "curl -sf -m 5 -o /dev/null -w 'local api: %{http_code}\n' \
     http://127.0.0.1:8124/api/health/status" || echo "local api: unreachable"

# A 200 alone proves nothing: it is exactly what a surviving old master
# returns. The master pid must have changed for the new code to be loaded.
AFTER=$(sh_ "pgrep -f '$APP/backend/.venv/bin/gunicorn' | tr '\n' ',' " || true)
if [ -n "$BEFORE" ] && [ "$BEFORE" = "$AFTER" ]; then
  echo "the gunicorn pids did not change ($AFTER) - the old process is still" >&2
  echo "serving and your changes are NOT live. Check $APP/logs/error.log." >&2
  exit 1
fi
echo "==> gunicorn restarted: $BEFORE -> $AFTER"
echo "done. The public URL takes a few seconds longer (frp reconnecting)."