#!/usr/bin/env bash
# ===================================================================================
#  mediabox-bootstrap — first-boot orchestrator
#
#  Runs ONCE from mediabox-firstboot.service, but every stage is idempotent so
#  you can re-run this by hand at any time:
#      sudo /srv/mediabox-bootstrap/bootstrap.sh
#
#  Or run a single stage:
#      sudo /srv/mediabox-bootstrap/scripts/20-cifs.sh
#
#  DESIGN RULE: no stage may leave the box unbootable or unreachable. Every
#  stage that can fail, fails soft and logs. sshd is already up by the time
#  this runs; keeping it that way is the whole safety net on a headless box.
# ===================================================================================
set -uo pipefail

REPO_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
STATE_DIR="/var/lib/mediabox"
LOG="/var/log/mediabox-bootstrap.log"

mkdir -p "$STATE_DIR"
exec > >(tee -a "$LOG") 2>&1

say()  { printf '\n\033[1;36m=== %s\033[0m\n' "$*"; }
ok()   { printf '\033[1;32m  [ok]\033[0m %s\n' "$*"; }
warn() { printf '\033[1;33m  [warn]\033[0m %s\n' "$*"; }
fail() { printf '\033[1;31m  [FAIL]\033[0m %s\n' "$*"; }

[ "$(id -u)" -eq 0 ] || { fail "must run as root"; exit 1; }

FAILED=()

run_stage() {
  local script="$1" name
  name="$(basename "$script")"
  say "$name"
  if [ ! -x "$script" ]; then chmod +x "$script" 2>/dev/null || true; fi
  if "$script"; then
    ok "$name complete"
  else
    fail "$name exited $? — continuing (see $LOG)"
    FAILED+=("$name")
  fi
}

say "mediabox bootstrap starting $(date -Is)"
echo "host: $(hostname)  kernel: $(uname -r)  ip: $(hostname -I)"

run_stage "$REPO_DIR/scripts/00-preflight.sh"
run_stage "$REPO_DIR/scripts/10-secrets.sh"
run_stage "$REPO_DIR/scripts/20-cifs.sh"
run_stage "$REPO_DIR/scripts/30-nvidia.sh"
run_stage "$REPO_DIR/scripts/40-shell-mcp.sh"

say "bootstrap summary"
if [ ${#FAILED[@]} -eq 0 ]; then
  ok "all stages completed"
  touch "$STATE_DIR/firstboot.done"
else
  warn "stages needing attention: ${FAILED[*]}"
  warn "NOT marking first-boot done — fix and re-run $0"
fi

# ---- announce the result on ntfy (best-effort, never blocks) ---------------------
# The box tells Mike's phone directly — publish path verified from 10.0.1.20
# (HTTP 200) on 2026-08-01. A notification failure must never fail the bootstrap.
NTFY_URL="https://ntfy.thewichersfamily.com/arrsstack-alerts"
MOUNTS_OK="$("$REPO_DIR/scripts/20-cifs.sh" --verify >/dev/null 2>&1 && echo yes || echo no)"
if [ ${#FAILED[@]} -eq 0 ]; then
  NTFY_TITLE="mediabox: first boot COMPLETE"
  NTFY_MSG="All bootstrap stages OK on $(hostname) ($(hostname -I | awk '{print $1}')). NAS mounts: ${MOUNTS_OK}. Next: message Claude 'box is up' to start the D: copy."
else
  NTFY_TITLE="mediabox: first boot NEEDS ATTENTION"
  NTFY_MSG="Stages failed: ${FAILED[*]}. NAS mounts: ${MOUNTS_OK}. SSH in (thethreemagi@10.0.1.20) or tell Claude which stage failed. Log: /var/log/mediabox-bootstrap.log"
fi
curl -s -m 10 -H "Title: ${NTFY_TITLE}" -d "${NTFY_MSG}" "$NTFY_URL" >/dev/null 2>&1 \
  || warn "ntfy notification failed (non-fatal)"

cat <<'NEXT'

--------------------------------------------------------------------------------
  REMAINING STEPS — deliberately NOT automated (2026-07-31 flow)
--------------------------------------------------------------------------------
  1. NAS credentials: the CIDATA stick copy of user-data now writes
     /etc/cifs/korval.cred on first boot, so the mounts normally come up on
     their own. If they did not, write the cred file by hand and verify:
         printf 'username=<nas-user>\npassword=<nas-pass>\ndomain=WORKGROUP\n' \
           > /etc/cifs/korval.cred
         chmod 600 /etc/cifs/korval.cred
         systemctl daemon-reload
         /srv/mediabox-bootstrap/scripts/20-cifs.sh --verify
     The six library shares must REFUSE writes, media_pc must accept them.

  2. Verify Quick Sync before anything Plex-related:
         vainfo --display drm --device /dev/dri/renderD128

  3. Copy EVERYTHING off D: to the NAS media_pc share, stage via the SSD,
     format D: to ext4, then restore and remap:
         /srv/mediabox-bootstrap/migration/README.md   (sections M2, M3, M4)
     Do NOT claim the server. The migration preserves its identity.
     The 337 GB Media/ cache lives on /srv/data (the former D:), NOT the SSD.

  4. Repoint NPM proxy host 42 from mediabox-mcp:8103 to 10.0.1.20:8103,
     then delete the mediabox-mcp container and its key on arrsstack.

  5. Soak 1-2 weeks. Keep the NAS media_pc backup — pruning it later is a
     deliberate decision, not a cleanup step.
--------------------------------------------------------------------------------
NEXT

exit 0