#!/usr/bin/env bash
# shellcheck shell=bash
# - - - - - - - - - - - - - - - - - - - - - - - - -
##@Version           :  202609030524-git
# @@Author           :  Jason Hempstead
# @@Contact          :  jason@casjaysdev.pro
# @@License          :  LICENSE.md
# @@ReadME           :  start-runners --help
# @@Copyright        :  Copyright: (c) 2026 Jason Hempstead, Casjays Developments
# @@Created          :  Friday, Jun 05, 2026 18:14 EDT
# @@File             :  start-runners
# @@Description      :  Start act runners
# @@Changelog        :  New script
# @@TODO             :  Better documentation
# @@Other            :
# @@Resource         :
# @@Terminal App     :  no
# @@sudo/root        :  no
# @@Template         :  shell/bash
# - - - - - - - - - - - - - - - - - - - - - - - - -
# shellcheck disable=SC1001,SC1003,SC2001,SC2003,SC2016,SC2031,SC2090,SC2115,SC2120,SC2155,SC2199,SC2229,SC2317,SC2329
# - - - - - - - - - - - - - - - - - - - - - - - - -
VERSION="202609030524-git"
set -e
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
trap 'retVal=$?; emoji=$( [ -z "$NO_COLOR" ] && echo "❌ " || echo "" ); \
echo "${emoji}Fatal error occurred: Exit code $retVal at line $LINENO in command: $BASH_COMMAND"; \
kill -TERM 1' ERR
trap 'retVal=$?;if [ "$SERVICE_IS_RUNNING" != "yes" ] && [ -f "$SERVICE_PID_FILE" ]; then rm -Rf "$SERVICE_PID_FILE"; fi;exit $retVal' SIGINT SIGTERM SIGPWR
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
__help() {
	cat <<'EOF'
Usage: start-runners

Register and start one or more Forgejo act_runner daemons in parallel.

Environment:
  SERVER_ADDRESS       Forgejo server URL (defaults to http://$HOSTNAME)
  SERVER_TOKEN         Required runner registration token
  RUNNERS_LOG_DIR      Runner log directory
  RUNNERS_START        Number of runners to start (default: 1)
  RUNNER_NAME_PREFIX   Runner name prefix (default: runner)
  RUNNER_LABELS        Comma-separated runner labels
  RUNNERS_CONFIG_BASE  Base act_runner config file
  RUNNER_CACHE_HOST    External cache host (used with RUNNER_CACHE_PORT and
                       RUNNER_CACHE_SECRET when no base config is available)
  RUNNER_CACHE_PORT    External cache port
  RUNNER_CACHE_SECRET  External cache secret
EOF
}

case "${1:-}" in
	-h|--help)
		__help
		exit 0
		;;
esac

# Function to __log messages with timestamp
__log() { echo "[$(date '+%Y-%m-%d %H:%M:%S')] $1" >>"$RUNNERS_LOG_DIR/runners" 2>&1; }
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# Function to cleanup child processes on exit
__cleanup() {
	__log "Shutting down runners..."
	kill $(jobs -p) 2>/dev/null || true
	wait
	__log "All runners stopped"
}
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# Set up signal handling
trap __cleanup SIGTERM SIGINT
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# Validate required environment variables
if [ -n "$SERVER_ADDRESS" ]; then
	if [[ "$SERVER_ADDRESS" != *"://"* ]]; then
		SERVER_ADDRESS="http://$SERVER_ADDRESS"
	fi
else
	SERVER_ADDRESS=http://$HOSTNAME
fi
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
if [ -z "$SERVER_TOKEN" ]; then
	__log "ERROR: SERVER_TOKEN environment variable is required"
	exit 1
fi
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
if [ -z "$RUNNER_LABELS" ]; then
	_default_runner_labels=(
		"linux:host"
		"node14:docker://node:14"
		"node16:docker://node:16"
		"node18:docker://node:18"
		"node20:docker://node:20"
		"node22:docker://node:22"
		"node:docker://node:latest"
		"perl:docker://perl:latest"
		"ruby:docker://ruby:latest"
		"python:docker://python:latest"
		"python3:docker://python:latest"
		"php7:docker://casjaysdevdocker/php:7"
		"php8:docker://casjaysdevdocker/php:8"
		"php:docker://casjaysdevdocker/php:latest"
		"alpine:docker://casjaysdev/alpine:latest"
		"debian:docker://casjaysdev/debian:latest"
		"ubuntu:docker://casjaysdev/ubuntu:latest"
		"rhel:docker://casjaysdev/almalinux:latest"
		"redhat:docker://casjaysdev/almalinux:latest"
		"almalinux:docker://casjaysdev/almalinux:latest"
		"act_runner:docker://catthehacker/ubuntu:full-latest"
		"ubuntu-latest:docker://catthehacker/ubuntu:full-latest"
	)
	RUNNER_LABELS="$(
		IFS=,
		echo "${_default_runner_labels[*]}"
	)"
	unset _default_runner_labels
fi
# Determine number of runners to start
RUNNERS_START=${RUNNERS_START:-1}
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# Validate RUNNERS_START is a positive integer
if ! [[ "$RUNNERS_START" =~ ^[0-9]+$ ]] || [ "$RUNNERS_START" -lt 1 ]; then
	__log "WARNING: Invalid RUNNERS_START value '$RUNNERS_START', defaulting to 1"
	RUNNERS_START=1
fi
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
__log "Starting $RUNNERS_START act_runner instance(s)"
__log "Server Address: $SERVER_ADDRESS"
__log "Runner Name Prefix: ${RUNNER_NAME_PREFIX:-runner}"
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# Full base config rendered by zz-act_runner.sh. Fall back to a shared cache-only
# config when this script is invoked without a rendered Forgejo runner config.
RUNNERS_DEFAULT_HOME="${RUNNER_DEFAULT_HOME:-/config/act_runner/forgejo}"
# Keep the established registration path by default; callers may override it
# independently when relocating runner state.
RUNNERS_REGISTRATION_ROOT="${RUNNER_REGISTRATION_ROOT:-/config/act_runner/reg}"
RUNNERS_CACHE_CONFIG=""
if [ -n "$RUNNERS_CONFIG_BASE" ] && [ -f "$RUNNERS_CONFIG_BASE" ]; then
	__log "Base config: $RUNNERS_CONFIG_BASE"
elif [ -n "$RUNNER_CACHE_HOST" ] && [ -n "$RUNNER_CACHE_PORT" ] && [ -n "$RUNNER_CACHE_SECRET" ]; then
	RUNNERS_CONFIG_BASE=""
	RUNNERS_CACHE_CONFIG="/config/act_runner/runners-cache.yaml"
	cat <<EOF >"$RUNNERS_CACHE_CONFIG"
runner:
  file: REPLACE_RUNNER_STATE/.runner
cache:
  enabled: true
  external_server: 'http://${RUNNER_CACHE_HOST}:${RUNNER_CACHE_PORT}/'
  secret: '${RUNNER_CACHE_SECRET}'
EOF
else
	RUNNERS_CONFIG_BASE=""
fi
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# Create the config used by one runner, isolating its registration state.
__runner_config() {
	local runner_dir=$1
	local runner_config="$runner_dir/config.yaml"
	if [ -n "$RUNNERS_CONFIG_BASE" ]; then
		sed "s|^\([[:space:]]*\)file:.*|\1file: $runner_dir/.runner|" "$RUNNERS_CONFIG_BASE" >"$runner_config" || return 1
		echo "$runner_config"
	elif [ -n "$RUNNERS_CACHE_CONFIG" ]; then
		sed "s|REPLACE_RUNNER_STATE|$runner_dir|" "$RUNNERS_CACHE_CONFIG" >"$runner_config" || return 1
		echo "$runner_config"
	fi
}
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# Register a single runner synchronously (no daemon start)
__register_runner() {
	local runner_id=$1
	local runner_name="${RUNNER_NAME_PREFIX:-runner}-${runner_id}"
	local runner_dir="$RUNNERS_REGISTRATION_ROOT/${runner_name}"
	local config_args=()
	local runner_config=""

	mkdir -p "$runner_dir"
	[ -d "$runner_dir" ] && cd "$runner_dir" || return 1
	runner_config="$(__runner_config "$runner_dir")" || return 1
	[ -n "$runner_config" ] && config_args=(--config "$runner_config")

	if [ ! -f "$runner_dir/.runner" ]; then
		__log "Registering runner: $runner_name (ID: $runner_id)"
		act_runner register "${config_args[@]}" --instance "$SERVER_ADDRESS" --token "$SERVER_TOKEN" --name "$runner_name" --labels "$RUNNER_LABELS" --no-interactive
		if [ $? -ne 0 ]; then
			__log "ERROR: Failed to register runner $runner_name"
			return 1
		fi
	fi
	__log "Runner $runner_name registered"
}
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# Start daemon for a single runner (called in background after all are registered)
__start_runner_daemon() {
	local runner_id=$1
	local runner_name="${RUNNER_NAME_PREFIX:-runner}-${runner_id}"
	local runner_dir="$RUNNERS_REGISTRATION_ROOT/${runner_name}"
	local config_args=()
	local runner_config=""

	[ -d "$runner_dir" ] && cd "$runner_dir" || return 1
	runner_config="$(__runner_config "$runner_dir")" || return 1
	[ -n "$runner_config" ] && config_args=(--config "$runner_config")
	__log "Starting daemon for runner: $runner_name"
	exec act_runner daemon "${config_args[@]}"
}
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# Phase 1: register all runners sequentially so IDs are assigned in order
for i in $(seq 1 $RUNNERS_START); do
	__register_runner "$i" || { __log "Aborting: registration failed for runner-$i"; exit 1; }
done
unset i
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# Phase 2: start all daemons in parallel
runner_pids=()
for i in $(seq 1 $RUNNERS_START); do
	(__start_runner_daemon "$i") &
	runner_pids+=("$!")
done
unset i
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
RUNNERS_ID="${runner_pids[*]}"
__log "All $RUNNERS_START runners launched"
__log "Process IDs: $RUNNERS_ID"
printf 'All %s runners launched: %s\n' "$RUNNERS_START" "$RUNNERS_ID"
# Wait for all background processes and preserve a failure status. A daemon that
# exits is reported as failed instead of being treated as successfully started.
runner_failures=0
for runner_pid in "${runner_pids[@]}"; do
	if ! wait "$runner_pid"; then
		runner_failures=$((runner_failures + 1))
		__log "ERROR: Runner process $runner_pid exited unsuccessfully"
	fi
done
if [ "$runner_failures" -gt 0 ]; then
	__log "$runner_failures runner process(es) failed"
	printf '%s\n' "$runner_failures runner process(es) failed" >&2
	exit 1
fi