@jonit-dev/night-watch-cli 1.8.4-beta.4 → 1.8.4-beta.6

package/dist/scripts/night-watch-cron.sh

@@ -102,24 +102,12 @@ latest_failure_detail() {
 }
 
 # ── Global Job Queue Gate ────────────────────────────────────────────────────
-# Acquire global gate before per-project lock to serialize jobs across projects.
-# When gate is busy, enqueue the job and exit cleanly.
+# Atomically claim a DB slot or enqueue for later dispatch — no flock needed.
 if [ "${NW_QUEUE_ENABLED:-0}" = "1" ]; then
   if [ "${NW_QUEUE_DISPATCHED:-0}" = "1" ]; then
     arm_global_queue_cleanup
-  elif acquire_global_gate; then
-    if queue_can_start_now; then
-      arm_global_queue_cleanup
-    else
-      release_global_gate
-      enqueue_job "${SCRIPT_TYPE}" "${PROJECT_DIR}"
-      emit_result "queued"
-      exit 0
-    fi
   else
-    enqueue_job "${SCRIPT_TYPE}" "${PROJECT_DIR}"
-    emit_result "queued"
-    exit 0
+    claim_or_enqueue "${SCRIPT_TYPE}" "${PROJECT_DIR}"
   fi
 fi
 # ──────────────────────────────────────────────────────────────────────────────

package/dist/scripts/night-watch-helpers.sh

@@ -1029,55 +1029,13 @@ find_eligible_board_issue() {
 
 # ── Global Job Queue Gate ─────────────────────────────────────────────────────
 
-# Get the path to the queue lock file
-get_queue_lock_path() {
-  local queue_home="${NIGHT_WATCH_HOME:-${HOME}/.night-watch}"
-  printf "%s/%s" "${queue_home}" "queue.lock"
-}
-
-# Try to acquire the global queue gate using flock.
-# Uses a shared lock file (~/.night-watch/queue.lock).
-# Returns 0 if acquired (gate is free), 1 if busy.
-# Holds the lock fd open for the caller via GLOBAL_GATE_FD variable.
-acquire_global_gate() {
-  local lock_path
-  lock_path=$(get_queue_lock_path)
-
-  # Ensure the .night-watch directory exists
-  mkdir -p "$(dirname "${lock_path}")"
-
-  # Open the lock file as fd 200
-  exec 200>"${lock_path}"
-
-  # Try non-blocking flock
-  if flock --nonblock 200; then
-    GLOBAL_GATE_FD=200
-    return 0
-  else
-    GLOBAL_GATE_FD=""
-    return 1
-  fi
-}
-
-# Release the global queue gate
-release_global_gate() {
-  if [ -n "${GLOBAL_GATE_FD:-}" ]; then
-    flock --unlock "${GLOBAL_GATE_FD}" 2>/dev/null || true
-    exec 200>&-
-    GLOBAL_GATE_FD=""
-  fi
-}
-
 __night_watch_queue_cleanup() {
   local exit_code="${1:-0}"
-
   if [ "${NW_QUEUE_CLEANUP_ARMED:-0}" = "1" ]; then
     NW_QUEUE_CLEANUP_ARMED=0
     complete_queued_job
     dispatch_next_queued_job
-    release_global_gate
   fi
-
   return "${exit_code}"
 }
 
@@ -1090,6 +1048,35 @@ arm_global_queue_cleanup() {
   append_exit_trap "__night_watch_queue_cleanup \$?"
 }
 
+# Atomically claim a queue slot or enqueue for later dispatch.
+# Uses DB transaction (via `queue claim` CLI) for atomicity — no flock needed.
+# Sets NW_QUEUE_ENTRY_ID on success and arms the cleanup trap.
+# Calls enqueue_job and exits 0 if no slot is available.
+claim_or_enqueue() {
+  local script_type="${1:?script_type required}"
+  local project_dir="${2:?project_dir required}"
+  local provider_key
+  provider_key=$(resolve_provider_key "${project_dir}" "${script_type}")
+
+  local cli_bin
+  cli_bin=$(resolve_night_watch_cli) || {
+    log "ERROR: Cannot resolve night-watch CLI for claim"
+    return 1
+  }
+
+  local claim_id
+  if claim_id=$("${cli_bin}" queue claim "${script_type}" "${project_dir}" --provider-key "${provider_key}" 2>/dev/null); then
+    NW_QUEUE_ENTRY_ID="${claim_id}"
+    export NW_QUEUE_ENTRY_ID
+    arm_global_queue_cleanup
+    return 0
+  else
+    enqueue_job "${script_type}" "${project_dir}"
+    emit_result "queued"
+    exit 0
+  fi
+}
+
 # Enqueue the current job to the SQLite queue.
 # Usage: enqueue_job <job_type> <project_dir>
 # Stores project path/name, job type, and relevant NW_* env vars for later dispatch.
@@ -1115,7 +1102,12 @@ enqueue_job() {
 
   log "QUEUE: Enqueueing ${job_type} for ${project_name} (gate busy)"
 
-  "${cli_bin}" queue enqueue "${job_type}" "${project_dir}" --env "${env_json}" >> "${LOG_FILE:-/dev/null}" 2>&1 || {
+  local provider_key_arg=()
+  if [ -n "${NW_PROVIDER_KEY:-}" ]; then
+    provider_key_arg=(--provider-key "${NW_PROVIDER_KEY}")
+  fi
+
+  "${cli_bin}" queue enqueue "${job_type}" "${project_dir}" --env "${env_json}" "${provider_key_arg[@]}" >> "${LOG_FILE:-/dev/null}" 2>&1 || {
     log "WARN: Failed to enqueue job"
     return 1
   }
@@ -1123,6 +1115,18 @@ enqueue_job() {
   return 0
 }
 
+# Resolve the provider bucket key for a given project and job type.
+# Uses the night-watch CLI to compute the canonical bucket key (e.g. claude-native, codex).
+# Prints the key to stdout, or an empty string if it cannot be resolved.
+# Usage: resolve_provider_key <project_dir> <job_type>
+resolve_provider_key() {
+  local project_dir="${1:?project_dir required}"
+  local job_type="${2:?job_type required}"
+  local cli_bin
+  cli_bin=$(resolve_night_watch_cli) || { printf ""; return 0; }
+  "${cli_bin}" queue resolve-key --project "${project_dir}" --job-type "${job_type}" 2>/dev/null || printf ""
+}
+
 # Dispatch the next queued job after the current job completes.
 # Picks highest-priority pending job, marks it dispatched, and spawns it.
 dispatch_next_queued_job() {
@@ -1143,16 +1147,6 @@ dispatch_next_queued_job() {
   "${cli_bin}" queue dispatch --log "${LOG_FILE:-/dev/null}" 2>/dev/null || true
 }
 
-queue_can_start_now() {
-  local cli_bin
-  cli_bin=$(resolve_night_watch_cli) || {
-    log "WARN: Cannot resolve night-watch CLI for queue slot check"
-    return 0
-  }
-
-  "${cli_bin}" queue can-start >/dev/null 2>&1
-}
-
 complete_queued_job() {
   local queue_entry_id="${NW_QUEUE_ENTRY_ID:-}"
   if [ -z "${queue_entry_id}" ]; then

package/dist/scripts/night-watch-plan-cron.sh

@@ -17,10 +17,12 @@ PROJECT_DIR="${1:?Usage: $0 /path/to/project}"
 PROJECT_NAME=$(basename "${PROJECT_DIR}")
 LOG_DIR="${PROJECT_DIR}/logs"
 LOG_FILE="${LOG_DIR}/plan.log"
+LOCK_FILE=""
 MAX_RUNTIME="${NW_PLAN_MAX_RUNTIME:-1800}"
 MAX_LOG_SIZE="524288"  # 512 KB
 PROVIDER_CMD="${NW_PROVIDER_CMD:-claude}"
 PROVIDER_LABEL="${NW_PROVIDER_LABEL:-}"
+SCRIPT_TYPE="planner"
 SCRIPT_START_TIME=$(date +%s)
 
 mkdir -p "${LOG_DIR}"
@@ -53,6 +55,8 @@ if ! ensure_provider_on_path "${PROVIDER_CMD}"; then
   exit 1
 fi
 
+PROJECT_RUNTIME_KEY=$(project_runtime_key "${PROJECT_DIR}")
+LOCK_FILE="/tmp/night-watch-plan-${PROJECT_RUNTIME_KEY}.lock"
 PROVIDER_MODEL_DISPLAY=$(resolve_provider_model_display "${PROVIDER_CMD}" "${PROVIDER_LABEL}")
 PRD_DIR="${NW_PRD_DIR:-docs/PRDs}"
 PLAN_TASK="${NW_PLAN_TASK:-}"
@@ -62,6 +66,25 @@ log_separator
 log "RUN-START: planner invoked project=${PROJECT_DIR} provider=${PROVIDER_CMD} dry_run=${NW_DRY_RUN:-0}"
 log "CONFIG: max_runtime=${MAX_RUNTIME}s prd_dir=${PRD_DIR}"
 
+if ! acquire_lock "${LOCK_FILE}"; then
+  exit 0
+fi
+# ── Global Job Queue Gate ────────────────────────────────────────────────────
+# Atomically claim a DB slot or enqueue for later dispatch — no flock needed.
+if [ "${NW_QUEUE_ENABLED:-0}" = "1" ]; then
+  if [ "${NW_QUEUE_DISPATCHED:-0}" = "1" ]; then
+    arm_global_queue_cleanup
+  else
+    claim_or_enqueue "${SCRIPT_TYPE}" "${PROJECT_DIR}"
+  fi
+fi
+# ──────────────────────────────────────────────────────────────────────────────
+cleanup_on_exit() {
+  rm -f "${LOCK_FILE}"
+}
+
+trap cleanup_on_exit EXIT
+
 # Dry-run mode
 if [ "${NW_DRY_RUN:-0}" = "1" ]; then
   echo "=== Dry Run: PRD Planner ==="

package/dist/scripts/night-watch-pr-reviewer-cron.sh

@@ -82,9 +82,6 @@ else
   LOCK_FILE="${GLOBAL_LOCK_FILE}"
 fi
 
-# ── Global Job Queue Gate ────────────────────────────────────────────────────
-# Acquire global gate before per-project lock to serialize jobs across projects.
-# When gate is busy, enqueue the job and exit cleanly.
 SCRIPT_TYPE="reviewer"
 
 emit_result() {
@@ -106,24 +103,12 @@ extract_review_score_from_text() {
 }
 
 # ── Global Job Queue Gate ────────────────────────────────────────────────────
-# Acquire global gate before per-project lock to serialize jobs across projects.
-# When gate is busy, enqueue the job and exit cleanly.
+# Atomically claim a DB slot or enqueue for later dispatch — no flock needed.
 if [ "${NW_QUEUE_ENABLED:-0}" = "1" ]; then
   if [ "${NW_QUEUE_DISPATCHED:-0}" = "1" ]; then
     arm_global_queue_cleanup
-  elif acquire_global_gate; then
-    if queue_can_start_now; then
-      arm_global_queue_cleanup
-    else
-      release_global_gate
-      enqueue_job "${SCRIPT_TYPE}" "${PROJECT_DIR}"
-      emit_result "queued"
-      exit 0
-    fi
   else
-    enqueue_job "${SCRIPT_TYPE}" "${PROJECT_DIR}"
-    emit_result "queued"
-    exit 0
+    claim_or_enqueue "${SCRIPT_TYPE}" "${PROJECT_DIR}"
   fi
 fi
 # ──────────────────────────────────────────────────────────────────────────────

package/dist/scripts/night-watch-qa-cron.sh

@@ -56,24 +56,12 @@ emit_result() {
 }
 
 # ── Global Job Queue Gate ────────────────────────────────────────────────────
-# Acquire global gate before per-project lock to serialize jobs across projects.
-# When gate is busy, enqueue the job and exit cleanly.
+# Atomically claim a DB slot or enqueue for later dispatch — no flock needed.
 if [ "${NW_QUEUE_ENABLED:-0}" = "1" ]; then
   if [ "${NW_QUEUE_DISPATCHED:-0}" = "1" ]; then
     arm_global_queue_cleanup
-  elif acquire_global_gate; then
-    if queue_can_start_now; then
-      arm_global_queue_cleanup
-    else
-      release_global_gate
-      enqueue_job "${SCRIPT_TYPE}" "${PROJECT_DIR}"
-      emit_result "queued"
-      exit 0
-    fi
   else
-    enqueue_job "${SCRIPT_TYPE}" "${PROJECT_DIR}"
-    emit_result "queued"
-    exit 0
+    claim_or_enqueue "${SCRIPT_TYPE}" "${PROJECT_DIR}"
   fi
 fi
 # ──────────────────────────────────────────────────────────────────────────────

package/dist/scripts/night-watch-slicer-cron.sh

@@ -61,23 +61,13 @@ log "CONFIG: max_runtime=${MAX_RUNTIME}s roadmap_path=${NW_ROADMAP_PATH:-ROADMAP
 if ! acquire_lock "${LOCK_FILE}"; then
   exit 0
 fi
-# ── Global Job Queue Gate ──────────────────────────────────────
+# ── Global Job Queue Gate ────────────────────────────────────────────────────
+# Atomically claim a DB slot or enqueue for later dispatch — no flock needed.
 if [ "${NW_QUEUE_ENABLED:-0}" = "1" ]; then
   if [ "${NW_QUEUE_DISPATCHED:-0}" = "1" ]; then
     arm_global_queue_cleanup
-  elif acquire_global_gate; then
-    if queue_can_start_now; then
-      arm_global_queue_cleanup
-    else
-      release_global_gate
-      enqueue_job "slicer" "${PROJECT_DIR}"
-      emit_result "queued"
-      exit 0
-    fi
   else
-    enqueue_job "slicer" "${PROJECT_DIR}"
-    emit_result "queued"
-    exit 0
+    claim_or_enqueue "slicer" "${PROJECT_DIR}"
   fi
 fi
 # ──────────────────────────────────────────────────────────────────────────────