DNF90 服务端与启动器完整源码（含 GM 工具功能增强）

服务端（Go）： - GM 塞背包/带属性装备发放：在线角色推送 op14 增量物品行，背包实时刷新，无需回角色选择界面 - GM 改金币/改等级：去掉在线拒绝，落库后推送 op14 钱包/op37 等级状态，在线实时刷新 - 等级端点统一切换到新版 GMSetCharacterLevel（自带在线刷新） - 无限/锁死疲劳模式 + 燃烧疲劳：每房间消耗 2 点、打怪与通关经验双倍 - 燃烧疲劳开关时给所有在线角色推送系统提示（op12 mode 0） - 深渊史诗掉率：GM 全局倍率 + 城镇增益药剂窗口倍率，真实接入掉落结算启动器（Python/JS）： - 标题栏自绘、窗口 1204x717、禁止拖拽改大小、仅最小化/关闭 - GM 工具：疲劳与燃烧、深渊史诗掉率、拍卖行上架（单一定价对齐） - 拍卖行上架前端对齐服务端单一价格语义，修复 400 报错
2026-11-04 14:31:55 +08:00
commit 645b64fe21
@@ -0,0 +1,223 @@
+package eventlog
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"log/slog"
+	"sync"
+	"time"
+)
+
+var (
+	// ErrPublishWorkerConfig 表示发布 worker 配置缺少必要字段。
+	ErrPublishWorkerConfig = errors.New("eventlog publish worker config is invalid")
+
+	// ErrPublishWorkerStopping 表示 worker 正在停机，不能再次启动。
+	ErrPublishWorkerStopping = errors.New("eventlog publish worker is stopping")
+)
+
+// PublishWorkerOptions 描述后台发布 worker 的日志、发布器和重试节奏。
+type PublishWorkerOptions struct {
+	Name           string
+	Log            *Log
+	Publisher      Publisher
+	Interval       time.Duration
+	Limit          int
+	RetryDelay     time.Duration
+	ClaimTimeout   time.Duration
+	MaxAttempts    int
+	ExcludeStreams []string
+	Logger         *slog.Logger
+}
+
+// PublishWorker 周期性领取 eventlog 待发布事件并调用 Publisher。
+type PublishWorker struct {
+	name           string
+	log            *Log
+	publisher      Publisher
+	interval       time.Duration
+	limit          int
+	retryDelay     time.Duration
+	claimTimeout   time.Duration
+	maxAttempts    int
+	excludeStreams []string
+	logger         *slog.Logger
+
+	mu       sync.Mutex
+	running  bool
+	stopping bool
+	cancel   context.CancelFunc
+	done     chan struct{}
+	stopped  chan struct{}
+}
+
+// NewPublishWorker 创建后台发布 worker 并补齐默认名称、间隔和重试延迟。
+func NewPublishWorker(options PublishWorkerOptions) (*PublishWorker, error) {
+	name := options.Name
+	if name == "" {
+		name = "eventlog-publisher"
+	}
+	if options.Log == nil {
+		return nil, fmt.Errorf("%w: log is required", ErrPublishWorkerConfig)
+	}
+	if options.Publisher == nil {
+		return nil, fmt.Errorf("%w: publisher is required", ErrPublishWorkerConfig)
+	}
+	interval := options.Interval
+	if interval <= 0 {
+		interval = time.Second
+	}
+	retryDelay := options.RetryDelay
+	if retryDelay <= 0 {
+		retryDelay = time.Second
+	}
+	return &PublishWorker{
+		name:           name,
+		log:            options.Log,
+		publisher:      options.Publisher,
+		interval:       interval,
+		limit:          options.Limit,
+		retryDelay:     retryDelay,
+		claimTimeout:   options.ClaimTimeout,
+		maxAttempts:    options.MaxAttempts,
+		excludeStreams: normalizeStreamList(options.ExcludeStreams),
+		logger:         options.Logger,
+	}, nil
+}
+
+// Name 返回 worker 名称，空实例返回默认名称。
+func (w *PublishWorker) Name() string {
+	if w == nil || w.name == "" {
+		return "eventlog-publisher"
+	}
+	return w.name
+}
+
+// Start 启动后台发布循环；重复启动保持幂等。
+func (w *PublishWorker) Start(ctx context.Context) error {
+	if err := ctxErr(ctx); err != nil {
+		return err
+	}
+	if w == nil || w.log == nil || w.publisher == nil {
+		return ErrPublishWorkerConfig
+	}
+	w.mu.Lock()
+	if w.running {
+		w.mu.Unlock()
+		return nil
+	}
+	if w.stopping {
+		w.mu.Unlock()
+		return ErrPublishWorkerStopping
+	}
+	// Worker 使用独立后台 context，不继承启动请求的生命周期。
+	// 停机时必须走 Stop，确保当前发布循环有机会收敛。
+	runCtx, cancel := context.WithCancel(context.Background())
+	w.done = make(chan struct{})
+	w.stopped = make(chan struct{})
+	w.running = true
+	w.cancel = cancel
+	done := w.done
+	stopped := w.stopped
+	w.mu.Unlock()
+
+	go w.run(runCtx, done, stopped)
+	return nil
+}
+
+// Stop 请求后台发布循环退出并等待收敛。
+func (w *PublishWorker) Stop(ctx context.Context) error {
+	if ctx == nil {
+		ctx = context.Background()
+	}
+	if w == nil {
+		return nil
+	}
+	w.mu.Lock()
+	if !w.running && !w.stopping {
+		w.mu.Unlock()
+		return nil
+	}
+	done := w.done
+	stopped := w.stopped
+	cancel := w.cancel
+	shouldSignal := w.running
+	if shouldSignal {
+		w.cancel = nil
+		w.running = false
+		w.stopping = true
+	}
+	w.mu.Unlock()
+
+	if shouldSignal && done != nil {
+		close(done)
+	}
+	if stopped == nil {
+		return nil
+	}
+	// Stop 等待 run 退出；超时后取消后台 context，避免进程关停被永久卡住。
+	select {
+	case <-stopped:
+		if cancel != nil {
+			cancel()
+		}
+		w.clearStopped(stopped)
+		return nil
+	case <-ctx.Done():
+		if cancel != nil {
+			cancel()
+		}
+		return ctx.Err()
+	}
+}
+
+func (w *PublishWorker) run(ctx context.Context, done <-chan struct{}, stopped chan struct{}) {
+	defer func() {
+		close(stopped)
+		w.clearStopped(stopped)
+	}()
+	// 启动后先跑一轮，避免 outbox 事件必须等到下一个 tick 才发布。
+	w.publishOnce(ctx)
+	ticker := time.NewTicker(w.interval)
+	defer ticker.Stop()
+	for {
+		select {
+		case <-done:
+			return
+		case <-ticker.C:
+			w.publishOnce(ctx)
+		}
+	}
+}
+
+func (w *PublishWorker) clearStopped(stopped chan struct{}) {
+	w.mu.Lock()
+	defer w.mu.Unlock()
+	if w.stopped != stopped {
+		return
+	}
+	w.running = false
+	w.stopping = false
+	w.cancel = nil
+	w.done = nil
+	w.stopped = nil
+}
+
+func (w *PublishWorker) publishOnce(ctx context.Context) {
+	// PublishPending 内部负责 claim、发布、重试、死信；Worker 只负责节拍和日志。
+	stats, err := w.log.PublishPending(ctx, w.publisher, PublishOptions{
+		Limit:          w.limit,
+		RetryDelay:     w.retryDelay,
+		ClaimTimeout:   w.claimTimeout,
+		MaxAttempts:    w.maxAttempts,
+		ExcludeStreams: w.excludeStreams,
+	})
+	if err != nil && w.logger != nil {
+		w.logger.Error("eventlog publish failed", "worker", w.Name(), "fetched", stats.Fetched, "published", stats.Published, "failed", stats.Failed, "dead_lettered", stats.DeadLettered, "error", err)
+		return
+	}
+	if stats.Published > 0 && w.logger != nil {
+		w.logger.Info("eventlog published events", "worker", w.Name(), "published", stats.Published, "failed", stats.Failed, "dead_lettered", stats.DeadLettered)
+	}
+}