chore: 初始开源发布

iotlcm-bot：独立飞书应用机器人（读取 Worker 终态 Run JSON，渲染并发送全量同步结果卡片）。 - Go module：git.5iot.com/iotlcm/iotlcm-bot → git.5iot.com/iotlcm-public/iotlcm-bot - 飞书凭据占位：appId/appSecret/chatId → ${FEISHU_APP_ID}/${FEISHU_APP_SECRET}/${FEISHU_CHAT_ID} - 移除内部资产：CI 触发器 - Dockerfile 改用公共基础镜像（golang:1.25-alpine → scratch） - README 重写（命令/配置/挂载契约/常见问题），去掉内部环境路径 - 补 Apache-2.0 LICENSE，LF 行尾 + .gitattributes
2026-10-22 16:14:24 +08:00
commit 5732bf3384
33 changed files with 2676 additions and 0 deletions
@@ -0,0 +1,107 @@
+package notify
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+)
+
+func TestAlertStateSurvivesRestart(t *testing.T) {
+	dir := t.TempDir()
+	now := time.Now()
+
+	// 首次：没有任何记录 → 应该告警
+	first := loadAlertState(dir)
+	if !first.shouldAlert("dir gone", now, DegradedAlertInterval) {
+		t.Fatal("first alert must be sent")
+	}
+	first.LastDegradedAlert = now
+	first.LastDegradedReason = "dir gone"
+	if err := saveAlertState(dir, first); err != nil {
+		t.Fatal(err)
+	}
+
+	// 模拟进程退出后重启：重新从磁盘加载
+	reloaded := loadAlertState(dir)
+	if reloaded.shouldAlert("dir gone", now.Add(time.Second), DegradedAlertInterval) {
+		t.Fatal("after restart the same failure must NOT alert again immediately (this is the alert-flood bug)")
+	}
+	// 间隔到了才允许再报
+	if !reloaded.shouldAlert("dir gone", now.Add(DegradedAlertInterval+time.Second), DegradedAlertInterval) {
+		t.Fatal("alert must be sent again after the interval elapses")
+	}
+	// 故障原因变化时立即告警
+	if !reloaded.shouldAlert("permission denied", now.Add(time.Second), DegradedAlertInterval) {
+		t.Fatal("a changed failure reason must alert immediately")
+	}
+}
+
+func TestExitCountIsPersistedAndIncremented(t *testing.T) {
+	dir := t.TempDir()
+	st := loadAlertState(dir)
+	if st.ExitCount != 0 {
+		t.Fatalf("fresh state exitCount=%d want 0", st.ExitCount)
+	}
+	st.ExitCount++
+	if err := saveAlertState(dir, st); err != nil {
+		t.Fatal(err)
+	}
+	// 重启后计数必须延续，否则无法识别「反复重启」
+	if got := loadAlertState(dir).ExitCount; got != 1 {
+		t.Fatalf("exitCount after reload=%d want 1", got)
+	}
+}
+
+// 关键回归：挂载失效 → 退出 → 重启后，同一次故障只能推一条飞书。
+func TestWatchAlertsOnceAcrossRestarts(t *testing.T) {
+	shortenExitDelay(t)
+	if !ExitOnDegraded {
+		t.Skip("ExitOnDegraded disabled")
+	}
+	stateDir := t.TempDir()
+	archiveDir := filepath.Join(t.TempDir(), "gone")
+	fake := &fakeFeishu{}
+	newSvc := func() Service {
+		return Service{
+			archiveDir:   archiveDir,
+			environment:  "test",
+			stateDir:     stateDir,
+			pollInterval: 10 * time.Millisecond,
+			store:        newTestStore(stateDir),
+			client:       fake,
+			logger:       discardLogger(),
+		}
+	}
+	// 第一次运行：应告警并退出
+	if err := newSvc().Watch(context.Background()); err == nil {
+		t.Fatal("Watch must exit when the mount is degraded")
+	}
+	if fake.sent != 1 {
+		t.Fatalf("first run must send exactly one alert, sent=%d", fake.sent)
+	}
+	// 模拟被 restart 策略拉起：新进程、新内存，但 state 目录相同
+	if err := newSvc().Watch(context.Background()); err == nil {
+		t.Fatal("Watch must exit again")
+	}
+	if fake.sent != 1 {
+		t.Fatalf("restart must NOT resend the same alert (alert flood), sent=%d", fake.sent)
+	}
+	// 退出计数应累计到 2
+	if got := loadAlertState(stateDir).ExitCount; got != 2 {
+		t.Fatalf("exitCount=%d want 2", got)
+	}
+}
+
+func TestAlertStateMissingDirIsNotFatal(t *testing.T) {
+	// 状态目录不可写时不应影响告警逻辑（退化为每次都报，但至少不崩）
+	st := loadAlertState("")
+	if !st.shouldAlert("x", time.Now(), DegradedAlertInterval) {
+		t.Fatal("empty state must allow alerting")
+	}
+	if err := saveAlertState("", st); err != nil {
+		t.Fatalf("saving to empty state dir must be a no-op, got %v", err)
+	}
+	_ = os.Remove("nonexistent")
+}