以下是谷歌在Android 12上的行为变化:

To provide a streamlined experience for short-running foreground services on Android 12, the system can delay the display of foreground service notifications by 10 seconds for certain foreground services. This change gives short-lived tasks a chance to complete before their notifications appear.

解决方案:在onCreate()中为你使用的服务调用start前台()

我在@humazed答案中添加了一些代码。所以没有初始通知。这可能是个变通办法，但对我来说很管用。

@Override
 public void onCreate() {
        super.onCreate();

        if (Build.VERSION.SDK_INT >= 26) {
            String CHANNEL_ID = "my_channel_01";
            NotificationChannel channel = new NotificationChannel(CHANNEL_ID,
                    "Channel human readable title",
                    NotificationManager.IMPORTANCE_DEFAULT);

            ((NotificationManager) getSystemService(Context.NOTIFICATION_SERVICE)).createNotificationChannel(channel);

            Notification notification = new NotificationCompat.Builder(this, CHANNEL_ID)
                    .setContentTitle("")
                    .setContentText("")
                    .setColor(ContextCompat.getColor(this, R.color.transparentColor))
                    .setSmallIcon(ContextCompat.getColor(this, R.color.transparentColor)).build();

            startForeground(1, notification);
        }
}

我在小图标和通知上添加了transparentColor。 它会起作用的。

服务

class TestService : Service() {

    override fun onCreate() {
        super.onCreate()
        Log.d(TAG, "onCreate")

        val nBuilder = NotificationCompat.Builder(this, "all")
            .setSmallIcon(R.drawable.ic_launcher_foreground)
            .setContentTitle("TestService")
            .setPriority(NotificationCompat.PRIORITY_DEFAULT)
        startForeground(1337, nBuilder.build())
    }

    override fun onStartCommand(intent: Intent?, flags: Int, startId: Int): Int {
        val rtn = super.onStartCommand(intent, flags, startId)

        if (intent?.action == STOP_ACTION) {
            Log.d(TAG, "onStartCommand -> STOP")
            stopForeground(true)
            stopSelf()
        } else {
            Log.d(TAG, "onStartCommand -> START")
        }

        return rtn
    }

    override fun onDestroy() {
        Log.d(TAG, "onDestroy")
        super.onDestroy()
    }

    override fun onBind(intent: Intent?): IBinder? = null

    companion object {

        private val TAG = "TestService"
        private val STOP_ACTION = "ly.zen.test.TestService.ACTION_STOP"

        fun start(context: Context) {
            ContextCompat.startForegroundService(context, Intent(context, TestService::class.java))
        }

        fun stop(context: Context) {
            val intent = Intent(context, TestService::class.java)
            intent.action = STOP_ACTION
            ContextCompat.startForegroundService(context, intent)
        }

    }

}

测试人员

val nChannel = NotificationChannel("all", "All", NotificationManager.IMPORTANCE_NONE)
val nManager = getSystemService(Context.NOTIFICATION_SERVICE) as NotificationManager
nManager.createNotificationChannel(nChannel)

start_test_service.setOnClickListener {
    TestService.start(this@MainActivity)
    TestService.stop(this@MainActivity)
}

结果

D/TestService: onCreate
D/TestService: onStartCommand -> START
D/TestService: onStartCommand -> STOP
D/TestService: onDestroy

好吧，我注意到的一些东西可能也会对其他人有所帮助。这是严格的测试，看看我是否能找出如何解决我所看到的事件。为了简单起见，假设我有一个从演示者调用这个的方法。

context.startForegroundService(new Intent(context, TaskQueueExecutorService.class));

try {
    Thread.sleep(10000);
} catch (InterruptedException e) {
    e.printStackTrace();
}       

这将导致同样的错误。在方法完成之前，服务不会启动，因此服务中没有onCreate()。

So even if you update the UI off the main thread, IF you have anything that might hold up that method after it, it won't start on time and give you the dreaded Foreground Error. In my case we were loading some things onto a queue and each called startForegroundService, but some logic was involved with each in the background. So if the logic took too long to finish that method since they were called back to back, crash time. The old startService just ignored it and went on it's way and since we called it each time, the next round would finish up.

这让我想知道，如果我从后台线程调用服务，它是否可以在启动时完全绑定并立即运行，因此我开始试验。即使这样不会立即启动，它也不会崩溃。

new Handler(Looper.getMainLooper()).post(new Runnable() {
        public void run() {
               context.startForegroundService(new Intent(context, 
           TaskQueueExecutorService.class));
               try {
                   Thread.sleep(10000);
               } catch (InterruptedException e) {
                  e.printStackTrace();
              }       
        }
});

我不会假装知道为什么它没有崩溃，尽管我怀疑这迫使它等待，直到主线程能够及时处理它。我知道将它绑定到主线程并不理想，但由于我的使用是在后台调用它，所以我并不真正关心它是否等待完成而不是崩溃。

我一直在研究这个问题，这是我目前为止的发现。如果我们有类似这样的代码，就会发生崩溃:

MyForegroundService.java

public class MyForegroundService extends Service {
    @Override
    public void onCreate() {
        super.onCreate();
        startForeground(...);
    }
}

MainActivity.java

Intent serviceIntent = new Intent(this, MyForegroundService.class);
startForegroundService(serviceIntent);
...
stopService(serviceIntent);

异常在以下代码块中抛出:

ActiveServices.java

private final void bringDownServiceLocked(ServiceRecord r) {
    ...
    if (r.fgRequired) {
        Slog.w(TAG_SERVICE, "Bringing down service while still waiting for start foreground: "
                  + r);
        r.fgRequired = false;
        r.fgWaiting = false;
        mAm.mAppOpsService.finishOperation(AppOpsManager.getToken(mAm.mAppOpsService),
                    AppOpsManager.OP_START_FOREGROUND, r.appInfo.uid, r.packageName);
        mAm.mHandler.removeMessages(
                    ActivityManagerService.SERVICE_FOREGROUND_TIMEOUT_MSG, r);
        if (r.app != null) {
            Message msg = mAm.mHandler.obtainMessage(
                ActivityManagerService.SERVICE_FOREGROUND_CRASH_MSG);
            msg.obj = r.app;
            msg.getData().putCharSequence(
                ActivityManagerService.SERVICE_RECORD_KEY, r.toString());
            mAm.mHandler.sendMessage(msg);
         }
    }
    ...
}

这个方法在MyForegroundService的onCreate()之前执行，因为Android在主线程处理程序上安排了服务的创建，但在BinderThread上调用了bringdownservicelock，这是一个竞争条件。这意味着MyForegroundService没有机会调用start前台，这将导致崩溃。

为了解决这个问题，我们必须确保在MyForegroundService的onCreate()之前不调用bringdownservicellocked。

public class MyForegroundService extends Service {

    private static final String ACTION_STOP = "com.example.MyForegroundService.ACTION_STOP";

    private final BroadcastReceiver stopReceiver = new BroadcastReceiver() {
        @Override
        public void onReceive(Context context, Intent intent) {
            context.removeStickyBroadcast(intent);
            stopForeground(true);
            stopSelf();
        }
    };

    @Override
    public void onCreate() {
        super.onCreate();
        startForeground(...);
        registerReceiver(
            stopReceiver, new IntentFilter(ACTION_STOP));
    }

    @Override
    public void onDestroy() {
        super.onDestroy();
        unregisterReceiver(stopReceiver);
    }

    public static void stop(Context context) {
        context.sendStickyBroadcast(new Intent(ACTION_STOP));
    }
}

通过使用粘性广播，我们确保广播不会丢失，stopReceiver在MyForegroundService的onCreate()中注册后立即接收到停止意图。此时我们已经调用了startForeground(…)。我们还必须删除这个粘滞的广播，以防止下一次stopReceiver被通知。

请注意，sendStickyBroadcast方法已弃用，我仅将其用作解决此问题的临时方法。

提醒一下，我在这上面浪费了太多时间。我一直得到这个异常，即使我调用start前台(..)作为onCreate(..)的第一件事。 最后，我发现这个问题是由于使用NOTIFICATION_ID = 0引起的。使用任何其他值似乎都可以解决这个问题。