feat: add global process tracker for daemon cleanup

add process_tracker.go with daemon PID tracking
register/unregister daemons in StartDaemon/StopDaemon
track all spawned processes for later cleanup

test/cli/harness/node.go

@@ -274,6 +274,11 @@ func (n *Node) StartDaemonWithReq(req RunRequest, authorization string) *Node {
 	res := n.Runner.MustRun(newReq)
 
 	n.Daemon = res
+	
+	// Register the daemon process for cleanup tracking
+	if res.Cmd != nil && res.Cmd.Process != nil {
+		globalProcessTracker.RegisterProcess(res.Cmd.Process)
+	}
 
 	log.Debugf("node %d started, checking API", n.ID)
 	n.WaitOnAPI(authorization)
@@ -317,6 +322,10 @@ func (n *Node) StopDaemon() *Node {
 		log.Debugf("didn't stop node %d since no daemon present", n.ID)
 		return n
 	}
+	
+	// Store PID for cleanup tracking
+	pid := n.Daemon.Cmd.Process.Pid
+	
 	watch := make(chan struct{}, 1)
 	go func() {
 		_, _ = n.Daemon.Cmd.Process.Wait()
@@ -326,6 +335,7 @@ func (n *Node) StopDaemon() *Node {
 	// os.Interrupt does not support interrupts on Windows https://github.com/golang/go/issues/46345
 	if runtime.GOOS == "windows" {
 		if n.signalAndWait(watch, syscall.SIGKILL, 5*time.Second) {
+			globalProcessTracker.UnregisterProcess(pid)
 			return n
 		}
 		log.Panicf("timed out stopping node %d with peer ID %s", n.ID, n.PeerID())
@@ -333,18 +343,22 @@ func (n *Node) StopDaemon() *Node {
 
 	log.Debugf("signaling node %d with SIGTERM", n.ID)
 	if n.signalAndWait(watch, syscall.SIGTERM, 1*time.Second) {
+		globalProcessTracker.UnregisterProcess(pid)
 		return n
 	}
 	log.Debugf("signaling node %d with SIGTERM", n.ID)
 	if n.signalAndWait(watch, syscall.SIGTERM, 2*time.Second) {
+		globalProcessTracker.UnregisterProcess(pid)
 		return n
 	}
 	log.Debugf("signaling node %d with SIGQUIT", n.ID)
 	if n.signalAndWait(watch, syscall.SIGQUIT, 5*time.Second) {
+		globalProcessTracker.UnregisterProcess(pid)
 		return n
 	}
 	log.Debugf("signaling node %d with SIGKILL", n.ID)
 	if n.signalAndWait(watch, syscall.SIGKILL, 5*time.Second) {
+		globalProcessTracker.UnregisterProcess(pid)
 		return n
 	}
 	log.Panicf("timed out stopping node %d with peer ID %s", n.ID, n.PeerID())

test/cli/harness/process_tracker.go

@@ -0,0 +1,83 @@
+package harness
+
+import (
+	"os"
+	"sync"
+	"syscall"
+	"time"
+)
+
+// processTracker keeps track of all daemon processes started during tests
+type processTracker struct {
+	mu        sync.Mutex
+	processes map[int]*os.Process
+}
+
+// globalProcessTracker is a package-level tracker for all spawned daemons
+var globalProcessTracker = &processTracker{
+	processes: make(map[int]*os.Process),
+}
+
+// RegisterProcess adds a process to the tracker
+func (pt *processTracker) RegisterProcess(proc *os.Process) {
+	if proc == nil {
+		return
+	}
+	pt.mu.Lock()
+	defer pt.mu.Unlock()
+	pt.processes[proc.Pid] = proc
+	log.Debugf("registered daemon process PID %d", proc.Pid)
+}
+
+// UnregisterProcess removes a process from the tracker
+func (pt *processTracker) UnregisterProcess(pid int) {
+	pt.mu.Lock()
+	defer pt.mu.Unlock()
+	delete(pt.processes, pid)
+	log.Debugf("unregistered daemon process PID %d", pid)
+}
+
+// KillAll forcefully terminates all tracked processes
+func (pt *processTracker) KillAll() {
+	pt.mu.Lock()
+	defer pt.mu.Unlock()
+	
+	for pid, proc := range pt.processes {
+		log.Debugf("force killing daemon process PID %d", pid)
+		
+		// Try SIGTERM first
+		if err := proc.Signal(syscall.SIGTERM); err != nil {
+			if !os.IsProcessDone(err) {
+				log.Debugf("error sending SIGTERM to PID %d: %v", pid, err)
+			}
+		}
+		
+		// Give it a moment to terminate
+		time.Sleep(100 * time.Millisecond)
+		
+		// Force kill if still running
+		if err := proc.Kill(); err != nil {
+			if !os.IsProcessDone(err) {
+				log.Debugf("error killing PID %d: %v", pid, err)
+			}
+		}
+		
+		// Clean up entry
+		delete(pt.processes, pid)
+	}
+	
+	if len(pt.processes) > 0 {
+		log.Debugf("cleaned up %d daemon processes", len(pt.processes))
+	}
+}
+
+// IsProcessDone checks if an error indicates the process has already exited
+func IsProcessDone(err error) bool {
+	return err == os.ErrProcessDone
+}
+
+// CleanupDaemonProcesses kills all tracked daemon processes
+// This should be called in test cleanup or panic recovery
+func CleanupDaemonProcesses() {
+	globalProcessTracker.KillAll()
+}