fix(web,agent): P1 正确性与鉴权加固，P2 抽 fsutil 并补测试（v1.2.8）

internal/scheduler/scheduler.go

@@ -2,6 +2,7 @@ package scheduler
 
 import (
 	"context"
+	"log"
 	"time"
 )
 
@@ -15,8 +16,17 @@ type Scheduler struct {
 	jobs []Job
 }
 
+// New 会丢掉间隔非正的任务：time.NewTicker 对非正间隔直接 panic，这类任务应由调用方单独启动。
 func New(jobs ...Job) *Scheduler {
-	return &Scheduler{jobs: jobs}
+	valid := make([]Job, 0, len(jobs))
+	for _, job := range jobs {
+		if job.Interval <= 0 {
+			log.Printf("scheduler: 任务 %s 间隔为 %v，已跳过", job.Name, job.Interval)
+			continue
+		}
+		valid = append(valid, job)
+	}
+	return &Scheduler{jobs: valid}
 }
 
 func (s *Scheduler) Start(ctx context.Context) {
@@ -28,13 +38,22 @@ func (s *Scheduler) Start(ctx context.Context) {
 func (s *Scheduler) runJob(ctx context.Context, job Job) {
 	ticker := time.NewTicker(job.Interval)
 	defer ticker.Stop()
-	job.Run(ctx)
+	// 单个任务 panic 不能带走整个进程：包一层 recover，记录后继续下一轮
+	run := func() {
+		defer func() {
+			if r := recover(); r != nil {
+				log.Printf("scheduler: 任务 %s panic: %v", job.Name, r)
+			}
+		}()
+		job.Run(ctx)
+	}
+	run()
 	for {
 		select {
 		case <-ctx.Done():
 			return
 		case <-ticker.C:
-			job.Run(ctx)
+			run()
 		}
 	}
 }

internal/scheduler/scheduler_test.go

@@ -0,0 +1,40 @@
+package scheduler
+
+import (
+	"context"
+	"sync/atomic"
+	"testing"
+	"time"
+)
+
+// TestNewSkipsNonPositiveIntervals：time.NewTicker 对非正间隔会 panic，这类任务必须在建表时丢掉。
+func TestNewSkipsNonPositiveIntervals(t *testing.T) {
+	s := New(
+		Job{Name: "zero", Interval: 0, Run: func(context.Context) {}},
+		Job{Name: "negative", Interval: -time.Second, Run: func(context.Context) {}},
+		Job{Name: "ok", Interval: time.Hour, Run: func(context.Context) {}},
+	)
+	if len(s.jobs) != 1 || s.jobs[0].Name != "ok" {
+		t.Fatalf("非正间隔的任务应被跳过，实际保留 %d 个", len(s.jobs))
+	}
+}
+
+// TestRunJobRecoversFromPanic：单个任务 panic 不能带走整个进程，且要能继续下一轮。
+func TestRunJobRecoversFromPanic(t *testing.T) {
+	var calls int32
+	s := New(Job{Name: "panic", Interval: 5 * time.Millisecond, Run: func(context.Context) {
+		atomic.AddInt32(&calls, 1)
+		panic("boom")
+	}})
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	s.Start(ctx)
+
+	deadline := time.Now().Add(2 * time.Second)
+	for time.Now().Before(deadline) && atomic.LoadInt32(&calls) < 3 {
+		time.Sleep(5 * time.Millisecond)
+	}
+	if got := atomic.LoadInt32(&calls); got < 3 {
+		t.Fatalf("panic 后应继续后续轮次，实际只执行了 %d 次", got)
+	}
+}