feat: nlog, config hot-reload, and concurrency fixes

internal/config/config.go

@@ -8,6 +8,8 @@ import (
 	"path/filepath"
 	"strconv"
 
+	"github.com/cedar2025/xboard-node/internal/nlog"
+	"golang.org/x/term"
 	"gopkg.in/yaml.v3"
 )
 
@@ -331,40 +333,47 @@ func (c *Config) ExpandNodes() []*Config {
 }
 
 func InitLogger(cfg LogConfig) {
-	var level slog.Level
+	var minLevel slog.Level
 	switch cfg.Level {
 	case "debug":
-		level = slog.LevelDebug
+		minLevel = slog.LevelDebug
 	case "warn":
-		level = slog.LevelWarn
+		minLevel = slog.LevelWarn
 	case "error":
-		level = slog.LevelError
+		minLevel = slog.LevelError
 	default:
-		level = slog.LevelInfo
+		minLevel = slog.LevelInfo
 	}
 
 	var w io.Writer
+	useColor := false
 	switch cfg.Output {
 	case "stdout", "":
 		w = os.Stdout
+		useColor = term.IsTerminal(int(os.Stdout.Fd()))
 	case "stderr":
 		w = os.Stderr
+		useColor = term.IsTerminal(int(os.Stderr.Fd()))
 	default:
 		dir := filepath.Dir(cfg.Output)
 		if err := os.MkdirAll(dir, 0o755); err != nil {
-			slog.Error("failed to create log dir, falling back to stdout", "error", err)
+			fmt.Fprintf(os.Stderr, "failed to create log dir, falling back to stdout: %v\n", err)
 			w = os.Stdout
+			useColor = term.IsTerminal(int(os.Stdout.Fd()))
 		} else {
 			f, err := os.OpenFile(cfg.Output, os.O_CREATE|os.O_WRONLY|os.O_APPEND, 0o644)
 			if err != nil {
-				slog.Error("failed to open log file, falling back to stdout", "error", err)
+				fmt.Fprintf(os.Stderr, "failed to open log file, falling back to stdout: %v\n", err)
 				w = os.Stdout
+				useColor = term.IsTerminal(int(os.Stdout.Fd()))
 			} else {
 				w = f
+				useColor = false
 			}
 		}
 	}
 
-	handler := slog.NewTextHandler(w, &slog.HandlerOptions{Level: level})
-	slog.SetDefault(slog.New(handler))
+	nlog.Init(w, minLevel, useColor)
+	// Application logging goes through nlog; silence slog.Default for stray library use.
+	slog.SetDefault(slog.New(slog.DiscardHandler))
 }

internal/config/watcher.go

@@ -0,0 +1,117 @@
+package config
+
+import (
+	"context"
+	"path/filepath"
+	"sync"
+	"sync/atomic"
+	"time"
+
+	"github.com/cedar2025/xboard-node/internal/nlog"
+	"github.com/fsnotify/fsnotify"
+)
+
+// Watcher reloads the config file on change (debounced) and calls onChange.
+type Watcher struct {
+	path     string
+	debounce time.Duration
+	onChange func(*Config)
+	watcher  *fsnotify.Watcher
+
+	stopOnce sync.Once
+	stopCh   chan struct{}
+	stopped  atomic.Bool
+}
+
+// WatchConfig watches path; invalid reloads are logged and ignored.
+// Stop via Stop() or cancel ctx.
+func WatchConfig(ctx context.Context, path string, onChange func(*Config)) (*Watcher, error) {
+	absPath, err := filepath.Abs(path)
+	if err != nil {
+		return nil, err
+	}
+
+	fsw, err := fsnotify.NewWatcher()
+	if err != nil {
+		return nil, err
+	}
+
+	// Watch parent dir so rename/atomic saves are visible.
+	dir := filepath.Dir(absPath)
+	if err := fsw.Add(dir); err != nil {
+		fsw.Close()
+		return nil, err
+	}
+
+	w := &Watcher{
+		path:     absPath,
+		debounce: 1 * time.Second,
+		onChange: onChange,
+		watcher:  fsw,
+		stopCh:   make(chan struct{}),
+	}
+
+	go w.loop(ctx)
+	nlog.Core().Info("config watcher started", "path", absPath)
+	return w, nil
+}
+
+func (w *Watcher) loop(ctx context.Context) {
+	var timer *time.Timer
+	defer func() {
+		if timer != nil {
+			timer.Stop()
+		}
+		w.watcher.Close()
+	}()
+
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-w.stopCh:
+			return
+
+		case ev, ok := <-w.watcher.Events:
+			if !ok {
+				return
+			}
+			if filepath.Clean(ev.Name) != w.path {
+				continue
+			}
+			if !ev.Has(fsnotify.Write) && !ev.Has(fsnotify.Create) {
+				continue
+			}
+
+			if timer == nil {
+				timer = time.AfterFunc(w.debounce, w.reload)
+			} else {
+				timer.Reset(w.debounce)
+			}
+
+		case err, ok := <-w.watcher.Errors:
+			if !ok {
+				return
+			}
+			nlog.Core().Warn("config watcher error", "error", err)
+		}
+	}
+}
+
+func (w *Watcher) reload() {
+	if w.stopped.Load() {
+		return
+	}
+	cfg, err := Load(w.path)
+	if err != nil {
+		nlog.Core().Error("config reload failed, keeping current config", "error", err)
+		return
+	}
+	nlog.Core().Info("config reloaded successfully")
+	w.onChange(cfg)
+}
+
+func (w *Watcher) Stop() {
+	w.stopped.Store(true)
+	w.stopOnce.Do(func() { close(w.stopCh) })
+}