fix container can't be started when task status is created

[ningmingxiao]

nerdctl start container failed if get staus timeout

nerdctl can't start this container anymore.

reproduce

1. nerdctl run -d busybox sleep 10
   c1a3a060f42feabdd2fd7440c05c5905a918ec1cad2ebf9e34e8ca1fa9921a80

2. wait container exited
   and add some sleep here

func Start(ctx context.Context, container containerd.Container, isAttach bool, isInteractive bool, client *containerd.Client, detachKeys string, checkpointDir string, cfg *config.Config, nerdctlCmd string, nerdctlArgs []string) (err error) {
...
	task, err := taskutil.NewTask(ctx, client, container, taskutil.TaskOptions{
		AttachStreamOpt: attachStreamOpt,
		IsInteractive:   isInteractive,
		IsTerminal:      isTerminal,
		IsDetach:        true,
		Con:             con,
		LogURI:          logURI,
		DetachKeys:      detachKeys,
		Namespace:       namespace,
		DetachC:         detachC,
		CheckpointDir:   checkpointDir,
	})
	if err != nil {
		return err
	}
	time.Sleep(time.Second*1000)
	statusC, err := task.Wait(ctx)
	if err != nil {
		return err
	}

3 ./nerdctl start c1a and then kill "nerdctl start" (will find runc init is running)

4. delete sleep and rebuild nerdctl
5. ./nerdctl start c1a
   FATA[0000] 1 errors:
   task c1a3a060f42feabdd2fd7440c05c5905a918ec1cad2ebf9e34e8ca1fa9921a80: already exists

@AkihiroSuda

[AkihiroSuda]

Can we have a test

[ningmingxiao]

I haven't figured out how to add test cases yet.

[AkihiroSuda]

?

[ningmingxiao]

here if old task is not deleted, new task can't be created. and default log level is debug, I think we can use error level @AkihiroSuda