ProxmoxVE/misc/error_handler.func

#!/usr/bin/env bash
# ------------------------------------------------------------------------------
# ERROR HANDLER - ERROR & SIGNAL MANAGEMENT
# ------------------------------------------------------------------------------
# Copyright (c) 2021-2026 community-scripts ORG
# Author: MickLesk (CanbiZ)
# License: MIT | https://github.com/community-scripts/ProxmoxVE/raw/main/LICENSE
# ------------------------------------------------------------------------------
#
# Provides comprehensive error handling and signal management for all scripts.
# Includes:
#   - Exit code explanations (shell, package managers, databases, custom codes)
#   - Error handler with detailed logging
#   - Signal handlers (EXIT, INT, TERM)
#   - Initialization function for trap setup
#
# Usage:
#   source <(curl -fsSL .../error_handler.func)
#   catch_errors
#
# ------------------------------------------------------------------------------

# ==============================================================================
# SECTION 1: EXIT CODE EXPLANATIONS
# ==============================================================================

# ------------------------------------------------------------------------------
# explain_exit_code()
#
# - Canonical version is defined in api.func (sourced before this file)
# - This section only provides a fallback if api.func was not loaded
# - See api.func SECTION 1 for the authoritative exit code mappings
# ------------------------------------------------------------------------------
if ! declare -f explain_exit_code &>/dev/null; then
  explain_exit_code() {
    local code="$1"
    case "$code" in
    1) echo "General error / Operation not permitted" ;;
    2) echo "Misuse of shell builtins (e.g. syntax error)" ;;
    6) echo "curl: DNS resolution failed (could not resolve host)" ;;
    7) echo "curl: Failed to connect (network unreachable / host down)" ;;
    22) echo "curl: HTTP error returned (404, 429, 500+)" ;;
    28) echo "curl: Operation timeout (network slow or server not responding)" ;;
    35) echo "curl: SSL/TLS handshake failed (certificate error)" ;;
    100) echo "APT: Package manager error (broken packages / dependency problems)" ;;
    101) echo "APT: Configuration error (bad sources.list, malformed config)" ;;
    102) echo "APT: Lock held by another process (dpkg/apt still running)" ;;
    124) echo "Command timed out (timeout command)" ;;
    126) echo "Command invoked cannot execute (permission problem?)" ;;
    127) echo "Command not found" ;;
    128) echo "Invalid argument to exit" ;;
    130) echo "Terminated by Ctrl+C (SIGINT)" ;;
    134) echo "Process aborted (SIGABRT - possibly Node.js heap overflow)" ;;
    137) echo "Killed (SIGKILL / Out of memory?)" ;;
    139) echo "Segmentation fault (core dumped)" ;;
    141) echo "Broken pipe (SIGPIPE - output closed prematurely)" ;;
    143) echo "Terminated (SIGTERM)" ;;
    150) echo "Systemd: Service failed to start" ;;
    151) echo "Systemd: Service unit not found" ;;
    152) echo "Permission denied (EACCES)" ;;
    153) echo "Build/compile failed (make/gcc/cmake)" ;;
    154) echo "Node.js: Native addon build failed (node-gyp)" ;;
    160) echo "Python: Virtualenv / uv environment missing or broken" ;;
    161) echo "Python: Dependency resolution failed" ;;
    162) echo "Python: Installation aborted (permissions or EXTERNALLY-MANAGED)" ;;
    170) echo "PostgreSQL: Connection failed (server not running / wrong socket)" ;;
    171) echo "PostgreSQL: Authentication failed (bad user/password)" ;;
    172) echo "PostgreSQL: Database does not exist" ;;
    173) echo "PostgreSQL: Fatal error in query / syntax" ;;
    180) echo "MySQL/MariaDB: Connection failed (server not running / wrong socket)" ;;
    181) echo "MySQL/MariaDB: Authentication failed (bad user/password)" ;;
    182) echo "MySQL/MariaDB: Database does not exist" ;;
    183) echo "MySQL/MariaDB: Fatal error in query / syntax" ;;
    190) echo "MongoDB: Connection failed (server not running)" ;;
    191) echo "MongoDB: Authentication failed (bad user/password)" ;;
    192) echo "MongoDB: Database not found" ;;
    193) echo "MongoDB: Fatal query error" ;;
    200) echo "Proxmox: Failed to create lock file" ;;
    203) echo "Proxmox: Missing CTID variable" ;;
    204) echo "Proxmox: Missing PCT_OSTYPE variable" ;;
    205) echo "Proxmox: Invalid CTID (<100)" ;;
    206) echo "Proxmox: CTID already in use" ;;
    207) echo "Proxmox: Password contains unescaped special characters" ;;
    208) echo "Proxmox: Invalid configuration (DNS/MAC/Network format)" ;;
    209) echo "Proxmox: Container creation failed" ;;
    210) echo "Proxmox: Cluster not quorate" ;;
    211) echo "Proxmox: Timeout waiting for template lock" ;;
    212) echo "Proxmox: Storage type 'iscsidirect' does not support containers (VMs only)" ;;
    213) echo "Proxmox: Storage type does not support 'rootdir' content" ;;
    214) echo "Proxmox: Not enough storage space" ;;
    215) echo "Proxmox: Container created but not listed (ghost state)" ;;
    216) echo "Proxmox: RootFS entry missing in config" ;;
    217) echo "Proxmox: Storage not accessible" ;;
    218) echo "Proxmox: Template file corrupted or incomplete" ;;
    219) echo "Proxmox: CephFS does not support containers - use RBD" ;;
    220) echo "Proxmox: Unable to resolve template path" ;;
    221) echo "Proxmox: Template file not readable" ;;
    222) echo "Proxmox: Template download failed" ;;
    223) echo "Proxmox: Template not available after download" ;;
    224) echo "Proxmox: PBS storage is for backups only" ;;
    225) echo "Proxmox: No template available for OS/Version" ;;
    231) echo "Proxmox: LXC stack upgrade failed" ;;
    243) echo "Node.js: Out of memory (JavaScript heap out of memory)" ;;
    245) echo "Node.js: Invalid command-line option" ;;
    246) echo "Node.js: Internal JavaScript Parse Error" ;;
    247) echo "Node.js: Fatal internal error" ;;
    248) echo "Node.js: Invalid C++ addon / N-API failure" ;;
    249) echo "npm/pnpm/yarn: Unknown fatal error" ;;
    255) echo "DPKG: Fatal internal error" ;;
    *) echo "Unknown error" ;;
    esac
  }
fi

# ==============================================================================
# SECTION 2: ERROR HANDLERS
# ==============================================================================

# ------------------------------------------------------------------------------
# error_handler()
#
# - Main error handler triggered by ERR trap
# - Arguments: exit_code, command, line_number
# - Behavior:
#   * Returns silently if exit_code is 0 (success)
#   * Sources explain_exit_code() for detailed error description
#   * Displays error message with:
#     - Line number where error occurred
#     - Exit code with explanation
#     - Command that failed
#   * Shows last 20 lines of SILENT_LOGFILE if available
#   * Copies log to container /root for later inspection
#   * Exits with original exit code
# ------------------------------------------------------------------------------
error_handler() {
  local exit_code=${1:-$?}
  local command=${2:-${BASH_COMMAND:-unknown}}
  local line_number=${BASH_LINENO[0]:-unknown}

  command="${command//\$STD/}"

  if [[ "$exit_code" -eq 0 ]]; then
    return 0
  fi

  local explanation
  explanation="$(explain_exit_code "$exit_code")"

  printf "\e[?25h"

  # Use msg_error if available, fallback to echo
  if declare -f msg_error >/dev/null 2>&1; then
    msg_error "in line ${line_number}: exit code ${exit_code} (${explanation}): while executing command ${command}"
  else
    echo -e "\n${RD}[ERROR]${CL} in line ${RD}${line_number}${CL}: exit code ${RD}${exit_code}${CL} (${explanation}): while executing command ${YWB}${command}${CL}\n"
  fi

  if [[ -n "${DEBUG_LOGFILE:-}" ]]; then
    {
      echo "------ ERROR ------"
      echo "Timestamp : $(date '+%Y-%m-%d %H:%M:%S')"
      echo "Exit Code : $exit_code ($explanation)"
      echo "Line      : $line_number"
      echo "Command   : $command"
      echo "-------------------"
    } >>"$DEBUG_LOGFILE"
  fi

  # Get active log file (BUILD_LOG or INSTALL_LOG)
  local active_log=""
  if declare -f get_active_logfile >/dev/null 2>&1; then
    active_log="$(get_active_logfile)"
  elif [[ -n "${SILENT_LOGFILE:-}" ]]; then
    active_log="$SILENT_LOGFILE"
  fi

  if [[ -n "$active_log" && -s "$active_log" ]]; then
    echo "--- Last 20 lines of silent log ---"
    tail -n 20 "$active_log"
    echo "-----------------------------------"

    # Detect context: Container (INSTALL_LOG set + /root exists) vs Host (BUILD_LOG)
    if [[ -n "${INSTALL_LOG:-}" && -d /root ]]; then
      # CONTAINER CONTEXT: Copy log and create flag file for host
      local container_log="/root/.install-${SESSION_ID:-error}.log"
      cp "$active_log" "$container_log" 2>/dev/null || true

      # Create error flag file with exit code for host detection
      echo "$exit_code" >"/root/.install-${SESSION_ID:-error}.failed" 2>/dev/null || true
      # Log path is shown by host as combined log - no need to show container path
    else
      # HOST CONTEXT: Show local log path and offer container cleanup
      if declare -f msg_custom >/dev/null 2>&1; then
        msg_custom "📋" "${YW}" "Full log: ${active_log}"
      else
        echo -e "${YW}Full log:${CL} ${BL}${active_log}${CL}"
      fi

      # Offer to remove container if it exists (build errors after container creation)
      if [[ -n "${CTID:-}" ]] && command -v pct &>/dev/null && pct status "$CTID" &>/dev/null; then
        # Report failure to API before container cleanup
        if declare -f post_update_to_api &>/dev/null; then
          post_update_to_api "failed" "$exit_code"
        fi

        echo ""
        echo -en "${YW}Remove broken container ${CTID}? (Y/n) [auto-remove in 60s]: ${CL}"

        if read -t 60 -r response; then
          if [[ -z "$response" || "$response" =~ ^[Yy]$ ]]; then
            echo -e "\n${YW}Removing container ${CTID}${CL}"
            pct stop "$CTID" &>/dev/null || true
            pct destroy "$CTID" &>/dev/null || true
            echo -e "${GN}✔${CL} Container ${CTID} removed"
          elif [[ "$response" =~ ^[Nn]$ ]]; then
            echo -e "\n${YW}Container ${CTID} kept for debugging${CL}"
          fi
        else
          # Timeout - auto-remove
          echo -e "\n${YW}No response - auto-removing container${CL}"
          pct stop "$CTID" &>/dev/null || true
          pct destroy "$CTID" &>/dev/null || true
          echo -e "${GN}✔${CL} Container ${CTID} removed"
        fi
      fi
    fi
  fi

  exit "$exit_code"
}

# ==============================================================================
# SECTION 3: SIGNAL HANDLERS
# ==============================================================================

# ------------------------------------------------------------------------------
# on_exit()
#
# - EXIT trap handler
# - Cleans up lock files if lockfile variable is set
# - Exits with captured exit code
# - Always runs on script termination (success or failure)
# ------------------------------------------------------------------------------
on_exit() {
  local exit_code=$?
  # Report orphaned "installing" records to telemetry API
  # Catches ALL exit paths: errors (non-zero), signals, AND clean exits where
  # post_to_api was called ("installing" sent) but post_update_to_api was never called
  if [[ "${POST_TO_API_DONE:-}" == "true" && "${POST_UPDATE_DONE:-}" != "true" ]]; then
    if declare -f post_update_to_api >/dev/null 2>&1; then
      if [[ $exit_code -ne 0 ]]; then
        post_update_to_api "failed" "$exit_code"
      else
        post_update_to_api "failed" "1"
      fi
    fi
  fi
  [[ -n "${lockfile:-}" && -e "$lockfile" ]] && rm -f "$lockfile"
  exit "$exit_code"
}

# ------------------------------------------------------------------------------
# on_interrupt()
#
# - SIGINT (Ctrl+C) trap handler
# - Displays "Interrupted by user" message
# - Exits with code 130 (128 + SIGINT=2)
# ------------------------------------------------------------------------------
on_interrupt() {
  # Report interruption to telemetry API (prevents stuck "installing" records)
  if declare -f post_update_to_api >/dev/null 2>&1; then
    post_update_to_api "failed" "130"
  fi
  if declare -f msg_error >/dev/null 2>&1; then
    msg_error "Interrupted by user (SIGINT)"
  else
    echo -e "\n${RD}Interrupted by user (SIGINT)${CL}"
  fi
  exit 130
}

# ------------------------------------------------------------------------------
# on_terminate()
#
# - SIGTERM trap handler
# - Displays "Terminated by signal" message
# - Exits with code 143 (128 + SIGTERM=15)
# - Triggered by external process termination
# ------------------------------------------------------------------------------
on_terminate() {
  # Report termination to telemetry API (prevents stuck "installing" records)
  if declare -f post_update_to_api >/dev/null 2>&1; then
    post_update_to_api "failed" "143"
  fi
  if declare -f msg_error >/dev/null 2>&1; then
    msg_error "Terminated by signal (SIGTERM)"
  else
    echo -e "\n${RD}Terminated by signal (SIGTERM)${CL}"
  fi
  exit 143
}

# ==============================================================================
# SECTION 4: INITIALIZATION
# ==============================================================================

# ------------------------------------------------------------------------------
# catch_errors()
#
# - Initializes error handling and signal traps
# - Enables strict error handling:
#   * set -Ee: Exit on error, inherit ERR trap in functions
#   * set -o pipefail: Pipeline fails if any command fails
#   * set -u: (optional) Exit on undefined variable (if STRICT_UNSET=1)
# - Sets up traps:
#   * ERR → error_handler
#   * EXIT → on_exit
#   * INT → on_interrupt
#   * TERM → on_terminate
# - Call this function early in every script
# ------------------------------------------------------------------------------
catch_errors() {
  set -Ee -o pipefail
  if [ "${STRICT_UNSET:-0}" = "1" ]; then
    set -u
  fi

  trap 'error_handler' ERR
  trap on_exit EXIT
  trap on_interrupt INT
  trap on_terminate TERM
}