android - 某些设备上的 ANR 内部功能

Question

我有一个使用表面 View 来显示 UI 的应用程序。该应用程序可稳定运行约 18,000 名用户，但有 3 台设备在返回表面 View Activity (sv Activity ->常规 Activity ->返回 sv Activity )时出现 ANR。

这 3 个设备是:

• 昂达平板电脑(Allwinner A31 芯片组)
• Sero 8(瑞芯微芯片组)
• 配备 Intel Atom 的 Acer 10"平板电脑

我尝试重新创建 ANR 但失败了。根据我的用户的说法，该应用程序可以正常运行数小时，除了上面列出的设备外没有任何问题。

来自 Android 4.2 的 ANR 堆栈跟踪:

DALVIK THREADS:
(mutexes: tll=0 tsl=0 tscl=0 ghl=0)

"main" prio=5 tid=1 WAIT
| group="main" sCount=1 dsCount=0 obj=0x41c899a0 self=0x41a6c010
| sysTid=5497 nice=0 sched=0/0 cgrp=apps handle=1074877404
| state=S schedstat=( 0 0 0 ) utm=541 stm=129 core=2
at java.lang.Object.wait(Native Method)
- waiting on <0x41c89da0> (a java.lang.VMThread) held by tid=1 (main)
at java.lang.Thread.parkFor(Thread.java:1231)
at sun.misc.Unsafe.park(Unsafe.java:323)
at java.util.concurrent.locks.LockSupport.park(LockSupport.java:159)
at java.util.concurrent.locks.AbstractQueuedSynchronizer.parkAndCheckInterrupt(AbstractQueuedSynchronizer.java:810)
at java.util.concurrent.locks.AbstractQueuedSynchronizer.acquireQueued(AbstractQueuedSynchronizer.java:843)
at java.util.concurrent.locks.AbstractQueuedSynchronizer.acquire(AbstractQueuedSynchronizer.java:1173)
at java.util.concurrent.locks.ReentrantLock$NonfairSync.lock(ReentrantLock.java:183)
at java.util.concurrent.locks.ReentrantLock.lock(ReentrantLock.java:259)
at android.view.SurfaceView.updateWindow(SurfaceView.java:597)
at android.view.SurfaceView.onWindowVisibilityChanged(SurfaceView.java:329)
at android.view.View.dispatchWindowVisibilityChanged(View.java:7544)
at android.view.ViewGroup.dispatchWindowVisibilityChanged(ViewGroup.java:1039)
at android.view.ViewGroup.dispatchWindowVisibilityChanged(ViewGroup.java:1039)
at android.view.ViewGroup.dispatchWindowVisibilityChanged(ViewGroup.java:1039)
at android.view.ViewGroup.dispatchWindowVisibilityChanged(ViewGroup.java:1039)
at android.view.ViewRootImpl.performTraversals(ViewRootImpl.java:1224)
at android.view.ViewRootImpl.doTraversal(ViewRootImpl.java:1002)
at android.view.ViewRootImpl$TraversalRunnable.run(ViewRootImpl.java:4400)
at android.view.Choreographer$CallbackRecord.run(Choreographer.java:749)
at android.view.Choreographer.doCallbacks(Choreographer.java:562)
at android.view.Choreographer.doFrame(Choreographer.java:532)
at android.view.Choreographer$FrameDisplayEventReceiver.run(Choreographer.java:735)
at android.os.Handler.handleCallback(Handler.java:725)
at android.os.Handler.dispatchMessage(Handler.java:92)
at android.os.Looper.loop(Looper.java:137)
at android.app.ActivityThread.main(ActivityThread.java:5041)
at java.lang.reflect.Method.invokeNative(Native Method)
at java.lang.reflect.Method.invoke(Method.java:511)
at com.android.internal.os.ZygoteInit$MethodAndArgsCaller.run(ZygoteInit.java:817)
at com.android.internal.os.ZygoteInit.main(ZygoteInit.java:584)
at dalvik.system.NativeStart.main(Native Method)

"SurfaceDraw" prio=5 tid=15 SUSPENDED
| group="main" sCount=1 dsCount=0 obj=0x42867d30 self=0x69a65b38
| sysTid=12101 nice=0 sched=0/0 cgrp=apps handle=1753253360
| state=S schedstat=( 0 0 0 ) utm=7914 stm=23 core=0
at android.graphics.Canvas.native_drawARGB(Native Method)
at android.graphics.Canvas.drawARGB(Canvas.java:801)
at com.davidgiga1993.mixingstationlibrary.surface.BaseSurface.b(BaseSurface.java:167)
at com.davidgiga1993.mixingstationlibrary.surface.k.run(DrawThread.java:27)

"AsyncTask #3" prio=5 tid=14 WAIT
| group="main" sCount=1 dsCount=0 obj=0x421734d0 self=0x690d59d8
| sysTid=5820 nice=0 sched=0/0 cgrp=apps handle=1762495280
| state=S schedstat=( 0 0 0 ) utm=0 stm=0 core=0
at java.lang.Object.wait(Native Method)
- waiting on <0x421735f0> (a java.lang.VMThread) held by tid=14 (AsyncTask #3)
at java.lang.Thread.parkFor(Thread.java:1231)
at sun.misc.Unsafe.park(Unsafe.java:323)
at java.util.concurrent.locks.LockSupport.park(LockSupport.java:159)
at java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await(AbstractQueuedSynchronizer.java:2019)
at java.util.concurrent.LinkedBlockingQueue.take(LinkedBlockingQueue.java:413)
at java.util.concurrent.ThreadPoolExecutor.getTask(ThreadPoolExecutor.java:1013)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1073)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:573)
at java.lang.Thread.run(Thread.java:856)

"Binder_3" prio=5 tid=13 NATIVE
| group="main" sCount=1 dsCount=0 obj=0x4215a380 self=0x684f9540
| sysTid=5691 nice=0 sched=0/0 cgrp=apps handle=1693280728
| state=S schedstat=( 0 0 0 ) utm=3 stm=0 core=3
#00 pc 00016fe4 /system/lib/libc.so (__ioctl+8)
#01 pc 0002a75d /system/lib/libc.so (ioctl+16)
#02 pc 00016ba1 /system/lib/libbinder.so (android::IPCThreadState::talkWithDriver(bool)+132)
#03 pc 00017363 /system/lib/libbinder.so (android::IPCThreadState::joinThreadPool(bool)+154)
#04 pc 0001b15d /system/lib/libbinder.so
#05 pc 00011267 /system/lib/libutils.so (android::Thread::_threadLoop(void*)+114)
#06 pc 00046887 /system/lib/libandroid_runtime.so (android::AndroidRuntime::javaThreadShell(void*)+66)
#07 pc 00010dcd /system/lib/libutils.so
#08 pc 0000e3d8 /system/lib/libc.so (__thread_entry+72)
#09 pc 0000dac4 /system/lib/libc.so (pthread_create+160)
at dalvik.system.NativeStart.run(Native Method)

"AsyncTask #2" prio=5 tid=12 WAIT
| group="main" sCount=1 dsCount=0 obj=0x42145450 self=0x64f1aac8
| sysTid=5525 nice=0 sched=0/0 cgrp=apps handle=1693560600
| state=S schedstat=( 0 0 0 ) utm=0 stm=0 core=3
at java.lang.Object.wait(Native Method)
- waiting on <0x421455c8> (a java.lang.VMThread) held by tid=12 (AsyncTask #2)
at java.lang.Thread.parkFor(Thread.java:1231)
at sun.misc.Unsafe.park(Unsafe.java:323)
at java.util.concurrent.locks.LockSupport.park(LockSupport.java:159)
at java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await(AbstractQueuedSynchronizer.java:2019)
at java.util.concurrent.LinkedBlockingQueue.take(LinkedBlockingQueue.java:413)
at java.util.concurrent.ThreadPoolExecutor.getTask(ThreadPoolExecutor.java:1013)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1073)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:573)
at java.lang.Thread.run(Thread.java:856)

"AsyncTask #1" prio=5 tid=11 WAIT
| group="main" sCount=1 dsCount=0 obj=0x42141d40 self=0x64ef35e8
| sysTid=5524 nice=0 sched=0/0 cgrp=apps handle=1693143224
| state=S schedstat=( 0 0 0 ) utm=0 stm=0 core=2
at java.lang.Object.wait(Native Method)
- waiting on <0x42141ed8> (a java.lang.VMThread) held by tid=11 (AsyncTask #1)
at java.lang.Thread.parkFor(Thread.java:1231)
at sun.misc.Unsafe.park(Unsafe.java:323)
at java.util.concurrent.locks.LockSupport.park(LockSupport.java:159)
at java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await(AbstractQueuedSynchronizer.java:2019)
at java.util.concurrent.LinkedBlockingQueue.take(LinkedBlockingQueue.java:413)
at java.util.concurrent.ThreadPoolExecutor.getTask(ThreadPoolExecutor.java:1013)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1073)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:573)
at java.lang.Thread.run(Thread.java:856)

根据我的解释，ANR 发生在 surfaceCreated 被调用之前。

这里是surfaceview和绘图线程的源代码:

public class BaseSurface extends SurfaceView implements SurfaceHolder.Callback
{
    protected SurfaceHolder holder;
    private DrawThread drawThread;

    @Override
    public void surfaceChanged(SurfaceHolder holder, int format, int width, int height)
    {
        Log.d("Surface", "Changed");
    }

    @Override
    public void surfaceCreated(SurfaceHolder holder)
    {
        synchronized (holder)
        {
            this.holder = holder;

            if (drawThread != null)
            {
                drawThread.Active = false;
                try
                {
                    drawThread.join();
                }
                catch (InterruptedException e)
                {
                }
            }

            drawThread = new DrawThread(this);
            drawThread.Active = true;
            drawThread.start();
        }
    }

    @Override
    public void surfaceDestroyed(SurfaceHolder holder)
    {
        synchronized (holder)
        {

            drawThread.Active = false;
            boolean retry = true;
            while (retry)
            {
                try
                {
                    drawThread.join();
                    retry = false;
                }
                catch (InterruptedException e)
                {
                }
            }

            drawThread = null;
            this.holder = null;
        }
    }

    public void Update()
    {
        if (holder == null)
            return;

        Canvas canvas = holder.lockCanvas();
        if (canvas != null)
        {
            synchronized (holder)
            {
                //drawing the ui...
                holder.unlockCanvasAndPost(canvas);
            }
        }
    }
}

public class DrawThread extends Thread
{
    public boolean Active = false;
    private BaseSurface surface;

    private long frameStartTime;
    public float FPS = 38f;// = 26fps; 27f = 37fps
    private int sleepTime;

    public DrawThread(BaseSurface surface)
    {
        super("SurfaceDraw");
        this.surface = surface;
    }

    @Override
    public void run()
    {
        while (Active)
        {
            frameStartTime = SystemClock.uptimeMillis();
            surface.Update();
            try
            {
                sleepTime = (int) (FPS - (SystemClock.uptimeMillis() - frameStartTime));
                if (sleepTime > 0 && sleepTime < 1000)
                {
                    Thread.sleep(sleepTime);
                }
            }
            catch (InterruptedException ex)
            {
                Log.d("DrawThread", "Interrupred");
            }
        }
        Log.d("DrawThread", "Finished");
    }
}

过去几天我已经搜索了很多，但没有找到任何线索，说明为什么会这样。我发现的唯一类似问题在这里:https://groups.google.com/forum/#!msg/android-developers/0VuqnrYe7b0/Yw1mHodmrwoJ但他没有发布任何解决方案，他的问题与特定设备无关。

是否有其他人在使用某些特定设备之前遇到过这些问题并且知道该问题的解决方案？

编辑:

我找到了一种(随机)重现 ANR 的方法。导致 ANR 的真正问题发生在表面 active 关闭时。这是“好”关闭和“坏”关闭的堆栈跟踪:

好

04-24 14:54:10.798: D/DrawThread(1526): Finished
04-24 14:54:10.798: D/Surface(1526): surfaceDestroyed

不好

04-24 14:54:16.851: D/DrawThread(1526): Finished
04-24 14:54:16.851: D/Surface(1526): surfaceDestroyed
04-24 14:54:16.860: E/SurfaceHolder(1526): Exception locking surface
04-24 14:54:16.860: E/SurfaceHolder(1526): java.lang.IllegalArgumentException
04-24 14:54:16.860: E/SurfaceHolder(1526):  at android.view.Surface.lockCanvasNative(Native Method)
04-24 14:54:16.860: E/SurfaceHolder(1526):  at android.view.Surface.lockCanvas(Surface.java:76)
04-24 14:54:16.860: E/SurfaceHolder(1526):  at android.view.SurfaceView$4.internalLockCanvas(SurfaceView.java:744)
04-24 14:54:16.860: E/SurfaceHolder(1526):  at android.view.SurfaceView$4.lockCanvas(SurfaceView.java:720)
04-24 14:54:16.860: E/SurfaceHolder(1526):  at com.davidgiga1993.mixingstationlibrary.surface.BaseSurface.Update(BaseSurface.java:169)
04-24 14:54:16.860: E/SurfaceHolder(1526):  at com.davidgiga1993.mixingstationlibrary.surface.DrawThread.run(DrawThread.java:27)
04-24 14:54:16.860: D/Surface(1526): surfaceCreated

为什么会调用异常？绘制线程在表面被破坏之前停止并且没有任何东西再接触表面 View 。另外，为什么要在该异常之后调用表面的 surfaceCreated？该 Activity 此时甚至不再可见。

我也尝试删除所有同步块(synchronized block)，但它们并没有改变行为。

Answer 1

查看 android-developers 链接中的 ANR 跟踪，他们运行的是 Android 4.2，当 SurfaceView 试图锁定其 Surface 时，他们的主线程停止运行。我认为问题在于渲染线程调用了 lockCanvas()，它锁定了 Surface(使用 ReentrantLock)，然后发生了导致 SurfaceView 需要更新的事情(例如它的大小或位置变了)。您可以在(大概)调用 lockCanvas() 的线程的跟踪中看到它正在积极运行(“Thread-3899”在 NATIVE 中，state=R)在一些看起来很复杂的 Skia 代码。因此，要么 Skia 代码永远循环，要么只需要很长时间才能完成。

在您的情况下，渲染线程 (SurfaceDraw) 被挂起，可能是因为它完成了它正在做的事情并从 native 代码返回到 VM。你的是一个简单的 drawARGB() 调用，所以我不确定为什么会花这么长时间。有可能是其他原因使它停滞不前，这正是 ANR 快照机制最终 catch 进度时发生的情况。

在调用 lockCanvas() 之前获取 SurfaceHolder 的锁可能是明智的，以确保您不会在持有 Canvas 锁的情况下阻塞等待它。

(FWIW，在 SurfaceHolder 实例上同步让我有点紧张，因为你不知道 SurfaceView 中的某些东西是否会出于其自身的邪恶目的锁定它。不过不要认为这就是这里的问题。)