cloverleaf-larry/lib/nc-engine.sh

#!/usr/bin/env bash
# nc-engine.sh — Cloverleaf engine process control. Native v3 wrapper
# around the shipped Cloverleaf binaries — modelled on v1 `bounce`,
# `bounce_processes`, `pstop`, `start`, etc.
#
# Every action goes through the journal so it's reversible. Bounces are
# journaled as paired stop+start records; the rollback executes them in
# reverse to restore prior state (best-effort — engine state can drift).
#
# Subcommands:
#   stop   <thread|process> [more...]   stop one or more processes/threads
#   start  <thread|process> [more...]   start one or more
#   bounce <thread|process> [more...]   stop then start (atomic-ish)
#   restart                              alias of bounce
#   status                               quick site status via tstat (if available)
#   resend-ob <thread> <file>            resend a file outbound (post-xlate)
#   resend-ib <thread> <file>            resend a file inbound (pre-xlate)
#   route-test <thread> <file>           run Cloverleaf route_test for a thread
#   testxlate  <xlate> <xltfile>         test an xlate against an xlt file
#   tpstest    <msgfile> <proc-args>     run a TPS test
#
# Options for stop/start/bounce:
#   --site SITE       override $HCISITE for this call
#   --confirm yes     skip Y/N prompt (still journaled)
#   --dry-run         show the binary command but do not execute
#
# Cloverleaf binaries used (auto-discovered under $HCIROOT/bin/):
#   hcienginestop  hcienginerun  hcienginerestart  hcienginestat  tstat
#   hciengineroutetest  hciengineenginesend  ...
set -o pipefail

NC_SELF="$0"
LIB_DIR="$(cd "$(dirname "$NC_SELF")" && pwd)"
JOURNAL="$LIB_DIR/journal.sh"

die() { printf 'nc-engine: %s\n' "$*" >&2; exit 1; }
warn() { printf 'nc-engine: %s\n' "$*" >&2; }

# v0.7.5: shared CR-safety primitives (coerce_int / strip_cr / read_clean).
# Source if available; otherwise inline minimal fallbacks so a partial
# install doesn't break the y/N prompts.
if [ -r "$LIB_DIR/cygwin-safe.sh" ]; then
  # shellcheck disable=SC1090,SC1091
  . "$LIB_DIR/cygwin-safe.sh"
else
  coerce_int() { local r="${1:-}" d="${2:-0}" c; c=$(printf '%s' "$r" | tr -cd '0-9'); printf '%s' "${c:-$d}"; }
  strip_cr() { local v="${1:-}"; printf '%s' "${v//$'\r'/}"; }
fi

# Source journal so journaled actions can call journal_write
[ -f "$JOURNAL" ] && . "$JOURNAL" || warn "journal.sh not available — actions will not be reversible"

resolve_binary() {
  local name="$1"
  if command -v "$name" >/dev/null 2>&1; then command -v "$name"; return; fi
  for d in "${HCIROOT:-}/bin" "${HCIROOT:-}/server/bin"; do
    [ -x "$d/$name" ] && { echo "$d/$name"; return; }
  done
  return 1
}

journal_action() {
  # Record an engine action in the journal as a synthetic "command" entry.
  # We don't snapshot files (these are runtime ops, not file edits) but we
  # write a manifest-style entry so larry-rollback.sh --list shows them.
  local action="$1" target="$2" detail="${3:-}"
  local sessdir="$LARRY_HOME/journal/${LARRY_SESSION_ID:-engine-$(date +%Y-%m-%d-%H%M%S)-$$}"
  mkdir -p "$sessdir" 2>/dev/null
  # v0.7.5: coerce_int on wc output — Cygwin wc.exe can suffix the count
  # with \r on a CRLF terminal which then crashes the surrounding $(( + 1 )).
  local _cnt; _cnt=$(coerce_int "$(find "$sessdir" -name '[0-9]*.engine' 2>/dev/null | wc -l)" 0)
  local idx; idx=$(printf '%03d' $((_cnt + 1)))
  local entry="$sessdir/${idx}_${action}_${target//\//_}.engine"
  {
    printf 'action: %s\ntarget: %s\nwhen: %s\nhost: %s\nhciroot: %s\nhcisite: %s\ndetail: %s\n' \
      "$action" "$target" "$(date -Iseconds 2>/dev/null || date)" \
      "$(hostname 2>/dev/null || echo unknown)" "${HCIROOT:-?}" "${HCISITE:-?}" "$detail"
  } > "$entry"
  # Also append to a flat engine log for quick listing
  local elog="$LARRY_HOME/journal/engine-actions.tsv"
  [ -f "$elog" ] || printf 'when\tsession\taction\ttarget\thciroot\thcisite\n' > "$elog"
  printf '%s\t%s\t%s\t%s\t%s\t%s\n' "$(date -Iseconds 2>/dev/null || date)" \
    "${LARRY_SESSION_ID:-?}" "$action" "$target" "${HCIROOT:-?}" "${HCISITE:-?}" >> "$elog"
}

run_action() {
  local action="$1" target="$2"; shift 2
  local site="${HCISITE:-}"
  local confirm=""
  local dry=0
  while [ $# -gt 0 ]; do
    case "$1" in
      --site)    shift; site="$1" ;;
      --confirm) shift; confirm="$1" ;;
      --dry-run) dry=1 ;;
    esac
    shift
  done

  local binary cmd label
  case "$action" in
    stop)    binary=$(resolve_binary hcienginestop)    || die "hcienginestop not found"; cmd="$binary -p $target";        label="STOP" ;;
    start)   binary=$(resolve_binary hcienginerun)     || die "hcienginerun not found";  cmd="$binary -p $target";        label="START" ;;
    bounce|restart)
             binary=$(resolve_binary hcienginerestart) \
             && cmd="$binary -p $target" && label="BOUNCE" \
             || {
                 # Fallback to stop + start
                 local sbin; sbin=$(resolve_binary hcienginestop) || die "hcienginestop+hcienginerestart both missing"
                 local rbin; rbin=$(resolve_binary hcienginerun)  || die "hcienginerun missing"
                 cmd="$sbin -p $target && $rbin -p $target"
                 label="BOUNCE"
             } ;;
    *) die "unknown action: $action" ;;
  esac

  printf '\n%s%s%s  thread/process=%s  site=%s\n' "${C_YELLOW:-}" "$label" "${C_RESET:-}" "$target" "${site:-?}"
  printf '  $ %s\n' "$cmd"

  if [ "$dry" = "1" ]; then
    printf '  [dry-run] not executed\n'
    return 0
  fi

  if [ "$confirm" != "yes" ]; then
    printf '  proceed? [y/N]: '
    read -r ans </dev/tty 2>/dev/null || ans=""
    # v0.7.5: strip CR so a Cygwin pty's `Y\r` matches `^[Yy]$`.
    ans="${ans//$'\r'/}"
    [[ "$ans" =~ ^[Yy]$ ]] || { echo "  DENIED by user"; return 1; }
  fi

  journal_action "$action" "$target" "$cmd"
  HCISITE="$site" eval "$cmd"
  local rc=$?
  if [ "$rc" -eq 0 ]; then echo "  ✓ ok"; else warn "  exit $rc"; fi
  return $rc
}

cmd_status() {
  local site="${HCISITE:-}"
  local binary
  binary=$(resolve_binary hcienginestat) || binary=$(resolve_binary tstat) || die "no engine-status binary on PATH (looked for hcienginestat, tstat)"
  HCISITE="$site" "$binary" "$@"
}

cmd_resend() {
  local kind="$1" thread="$2" file="$3"; shift 3
  [ -n "$thread" ] && [ -f "$file" ] || die "usage: resend-{ib,ob} <thread> <file>"
  local cmd
  case "$kind" in
    ob) cmd="$thread resend_ob $file" ;;
    ib) cmd="$thread resend_ib $file" ;;
    *)  die "bad resend kind: $kind" ;;
  esac
  printf '\nRESEND-%s  thread=%s  file=%s\n  $ %s\n  proceed? [y/N]: ' "${kind^^}" "$thread" "$file" "$cmd"
  read -r ans </dev/tty 2>/dev/null || ans=""
  # v0.7.5: strip CR so a Cygwin pty's `Y\r` matches `^[Yy]$`.
  ans="${ans//$'\r'/}"
  [[ "$ans" =~ ^[Yy]$ ]] || { echo "  DENIED"; return 1; }
  journal_action "resend-$kind" "$thread" "file=$file"
  eval "$cmd"
}

cmd_route_test() {
  local thread="$1" file="$2"
  [ -n "$thread" ] && [ -f "$file" ] || die "usage: route-test <thread> <file>"
  local cmd="$thread route_test $file"
  printf '\nROUTE-TEST  thread=%s  input=%s\n  $ %s\n  proceed? [y/N]: ' "$thread" "$file" "$cmd"
  read -r ans </dev/tty 2>/dev/null || ans=""
  # v0.7.5: strip CR so a Cygwin pty's `Y\r` matches `^[Yy]$`.
  ans="${ans//$'\r'/}"
  [[ "$ans" =~ ^[Yy]$ ]] || { echo "  DENIED"; return 1; }
  journal_action "route-test" "$thread" "file=$file"
  eval "$cmd"
}

cmd_testxlate() {
  local xlate="$1" xltfile="$2"
  [ -n "$xlate" ] && [ -f "$xltfile" ] || die "usage: testxlate <xlate> <xltfile>"
  local cmd="testxlate $xlate $xltfile"
  printf '\nTESTXLATE xlate=%s file=%s\n  $ %s\n  proceed? [y/N]: ' "$xlate" "$xltfile" "$cmd"
  read -r ans </dev/tty 2>/dev/null || ans=""
  # v0.7.5: strip CR so a Cygwin pty's `Y\r` matches `^[Yy]$`.
  ans="${ans//$'\r'/}"
  [[ "$ans" =~ ^[Yy]$ ]] || { echo "  DENIED"; return 1; }
  journal_action "testxlate" "$xlate" "file=$xltfile"
  eval "$cmd"
}

cmd_tpstest() {
  local msgfile="$1"; shift
  [ -f "$msgfile" ] || die "usage: tpstest <msgfile> <proc-args...>"
  local procs; procs="$*"
  local cmd="tpstest $msgfile $procs"
  printf '\nTPSTEST msgfile=%s procs=%s\n  $ %s\n  proceed? [y/N]: ' "$msgfile" "$procs" "$cmd"
  read -r ans </dev/tty 2>/dev/null || ans=""
  # v0.7.5: strip CR so a Cygwin pty's `Y\r` matches `^[Yy]$`.
  ans="${ans//$'\r'/}"
  [[ "$ans" =~ ^[Yy]$ ]] || { echo "  DENIED"; return 1; }
  journal_action "tpstest" "$msgfile" "procs=$procs"
  eval "$cmd"
}

SUB="${1:-help}"
case "$SUB" in
  stop|start|bounce|restart)
    shift
    [ $# -ge 1 ] || die "usage: $SUB <target> [more...] [--site SITE] [--confirm yes] [--dry-run]"
    # Separate targets from flags
    targets=(); flags=()
    while [ $# -gt 0 ]; do
      case "$1" in --*) flags+=("$1" "${2:-}"); shift 2 ;; *) targets+=("$1"); shift ;; esac
    done
    for t in "${targets[@]}"; do run_action "$SUB" "$t" "${flags[@]}"; done
    ;;
  status)      shift; cmd_status "$@" ;;
  resend-ob)   shift; cmd_resend ob "$@" ;;
  resend-ib)   shift; cmd_resend ib "$@" ;;
  route-test)  shift; cmd_route_test "$@" ;;
  testxlate)   shift; cmd_testxlate "$@" ;;
  tpstest)     shift; cmd_tpstest "$@" ;;
  help|-h|--help) sed -n '2,30p' "$NC_SELF" ;;
  *)           die "unknown subcommand: $SUB" ;;
esac