claude-evolve 1.3.1 → 1.3.2

package/bin/claude-evolve-run

@@ -29,6 +29,10 @@ while [[ $# -gt 0 ]]; do
     force_sequential="true"
     shift
     ;;
+  --keep-awake|--caffeinate)
+    use_caffeinate="true"
+    shift
+    ;;
   --help)
     cat <<EOF
 claude-evolve run - Execute evolution candidates
@@ -40,6 +44,7 @@ OPTIONS:
   --timeout <sec>  Kill evaluator after specified seconds (default: no timeout)
   --parallel       Force parallel execution mode
   --sequential     Force sequential execution mode
+  --keep-awake     Prevent system sleep during execution (macOS only)
   --help           Show this help message
 
 DESCRIPTION:
@@ -59,6 +64,13 @@ EOF
   esac
 done
 
+# Check if caffeinate should be used
+if [[ "$use_caffeinate" == "true" ]] && command -v caffeinate >/dev/null 2>&1; then
+  echo "[INFO] Using caffeinate to prevent system sleep"
+  # Re-run this script with caffeinate
+  exec caffeinate -dims "$0" "$@"
+fi
+
 # Determine execution mode
 use_parallel=false
 if [[ "$force_parallel" == "true" ]]; then

package/bin/claude-evolve-run-parallel

@@ -68,8 +68,16 @@ start_worker() {
   echo "[DISPATCHER] Starting worker..."
   $worker_cmd &
   local pid=$!
-  worker_pids+=($pid)
-  echo "[DISPATCHER] Worker $pid started"
+  
+  # Verify worker started successfully
+  sleep 0.1
+  if kill -0 "$pid" 2>/dev/null; then
+    worker_pids+=($pid)
+    echo "[DISPATCHER] Worker $pid started"
+  else
+    echo "[ERROR] Worker failed to start" >&2
+    ((consecutive_failures++))
+  fi
 }
 
 # Check worker exit status and handle accordingly
@@ -144,11 +152,75 @@ shutdown_workers() {
   exit 0
 }
 
-trap shutdown_workers INT TERM
+# Better signal handling with logging
+handle_signal() {
+  local signal="$1"
+  echo "[DISPATCHER] Received signal: $signal" >&2
+  echo "[DISPATCHER] Active workers: ${#worker_pids[@]}" >&2
+  
+  # For expensive workers, give option to continue
+  if [[ ${#worker_pids[@]} -gt 0 ]]; then
+    echo "[DISPATCHER] Warning: ${#worker_pids[@]} expensive workers are still running!" >&2
+    echo "[DISPATCHER] Press Ctrl+C again within 5 seconds to force shutdown, or wait..." >&2
+    
+    # Give 5 seconds to reconsider
+    local count=5
+    while [[ $count -gt 0 ]]; do
+      sleep 1
+      ((count--))
+      # Check if we got another signal
+      if [[ -f /tmp/evolve-force-shutdown-$$ ]]; then
+        echo "[DISPATCHER] Force shutdown requested" >&2
+        rm -f /tmp/evolve-force-shutdown-$$
+        shutdown_workers
+        exit 1
+      fi
+    done
+    
+    echo "[DISPATCHER] Continuing with active workers..." >&2
+    return
+  fi
+  
+  shutdown_workers
+}
+
+# Set up signal handlers
+trap 'handle_signal INT' INT
+trap 'handle_signal TERM' TERM
+trap 'echo "[DISPATCHER] Exiting with code $?" >&2' EXIT
+
+# Check for stuck "running" candidates from previous runs
+check_stuck_candidates() {
+  if read_csv_with_lock csv_content; then
+    local stuck_count=$(echo "$csv_content" | awk -F',' 'NR>1 && $5 == "running" { count++ } END { print count+0 }')
+    if [[ $stuck_count -gt 0 ]]; then
+      echo "[DISPATCHER] Found $stuck_count candidates stuck in 'running' status"
+      echo "[DISPATCHER] Resetting them to 'pending' for retry..."
+      
+      # Reset stuck candidates
+      if acquire_csv_lock; then
+        awk -F',' -v OFS=',' '
+          NR==1 { print }
+          NR>1 { 
+            if ($5 == "running") $5 = "pending"
+            print 
+          }
+        ' "$FULL_CSV_PATH" > "${FULL_CSV_PATH}.tmp" && mv -f "${FULL_CSV_PATH}.tmp" "$FULL_CSV_PATH"
+        release_csv_lock
+      fi
+    fi
+  fi
+}
 
 # Main dispatcher loop
 echo "[DISPATCHER] Starting main dispatch loop"
 
+# Check for stuck candidates from previous runs
+check_stuck_candidates
+
+# Set error handling
+set +e  # Don't exit on error in the main loop
+
 while true; do
   # Check if too many consecutive failures
   if [[ $consecutive_failures -ge $MAX_FAILURES ]]; then
@@ -164,7 +236,7 @@ while true; do
   fi
   
   # Count pending work
-  pending_count=$(count_pending)
+  pending_count=$(count_pending || echo "0")
   active_workers=${#worker_pids[@]}
   
   echo "[DISPATCHER] Status: $pending_count pending, $active_workers active workers"
@@ -180,6 +252,7 @@ while true; do
   if [[ $pending_count -eq 0 && $active_workers -eq 0 ]]; then
     echo "[DISPATCHER] No pending candidates found. Evolution complete."
     echo "[DISPATCHER] Run 'claude-evolve ideate' to generate more candidates."
+    echo "[DISPATCHER] Exiting main loop: no work remaining" >&2
     break
   fi
   
@@ -210,10 +283,31 @@ while true; do
         handle_worker_exit "$pid" "$exit_code"
       fi
     done
+    
+    # Safety check - if we have workers but the array is corrupted
+    if [[ ${#worker_pids[@]} -eq 0 ]] && jobs -r | grep -q .; then
+      echo "[DISPATCHER] Warning: Lost track of workers but jobs still running!" >&2
+      echo "[DISPATCHER] Attempting to recover..." >&2
+      
+      # Try to recover PIDs from jobs
+      while read -r job_info; do
+        if [[ $job_info =~ \[([0-9]+)\][[:space:]]+([0-9]+) ]]; then
+          local recovered_pid="${BASH_REMATCH[2]}"
+          echo "[DISPATCHER] Recovered worker PID: $recovered_pid" >&2
+          worker_pids+=($recovered_pid)
+        fi
+      done < <(jobs -l)
+    fi
   else
     # No active workers but we might be waiting for ideation or have pending work
     sleep 1
   fi
 done
 
-echo "[DISPATCHER] Evolution run complete"
+echo "[DISPATCHER] Evolution run complete"
+
+# Final cleanup check
+if [[ ${#worker_pids[@]} -gt 0 ]]; then
+  echo "[DISPATCHER] Warning: ${#worker_pids[@]} workers still active at exit"
+  shutdown_workers
+fi

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-evolve",
-  "version": "1.3.1",
+  "version": "1.3.2",
   "bin": {
     "claude-evolve": "./bin/claude-evolve",
     "claude-evolve-main": "./bin/claude-evolve-main",