一个 Activity 为 FirstActivity finish()后，进入新的 Activity 为 SecondActivity ，生命周期里的源码到底是个怎样的实现流程呢。下面我们一起来剖析一下。
在 Android 10.0的源码中：
> Activity.java
/**
* Call this when your activity is done and should be closed. The
* ActivityResult is propagated back to whoever launched you via
* onActivityResult().
*/
public void finish() {
finish(DONT_FINISH_TASK_WITH_ACTIVITY);
}
重载了带参数的 finish() 方法。参数是 DONT_FINISH_TASK_WITH_ACTIVITY ，含义也很直白，不会销毁 Activity 所在的任务栈。
> Activity.java
/**
* Finishes the current activity and specifies whether to remove the task associated with this
* activity.
*/
private void finish(int finishTask) {
// mParent 一般为 null，在 ActivityGroup 中会使用到
if (mParent == null) {
int resultCode;
Intent resultData;
synchronized (this) {
resultCode = mResultCode;
resultData = mResultData;
}
if (false) Log.v(TAG, "Finishing self: token=" + mToken);
try {
if (resultData != null) {
resultData.prepareToLeaveProcess(this);
}
// Binder 调用 AMS.finishActivity()
if (ActivityTaskManager.getService()
.finishActivity(mToken, resultCode, resultData, finishTask)) {
mFinished = true;
}
} catch (RemoteException e) {
// Empty
}
} else {
mParent.finishFromChild(this);
}
// Activity was launched when user tapped a link in the Autofill Save UI - Save UI must
// be restored now.
if (mIntent != null && mIntent.hasExtra(AutofillManager.EXTRA_RESTORE_SESSION_TOKEN)) {
getAutofillManager().onPendingSaveUi(AutofillManager.PENDING_UI_OPERATION_RESTORE,
mIntent.getIBinderExtra(AutofillManager.EXTRA_RESTORE_SESSION_TOKEN));
}
}
这里的 mParent 大多数情况下都是 null ，不需要考虑 else 分支的情况。其中 Binder 调用了 ATMS.finishActivity() 方法。
Google 将 Android 的版本更新到10.0版本（sdk-29）之后，Acitivity的启动流程相比之前的版本发生了些变化，首先最大的变化是我们常见的用来管理 Activity 的AMS，变成了现在的ATMS（ActivityTaskManagerService）。
> IActivityTaskManager.aidl
boolean finishActivity(in IBinder token, int code, in Intent data, int finishTask);
>ActivityTaskManagerService.java
/**
* This is the internal entry point for handling Activity.finish().
*
* @param token The Binder token referencing the Activity we want to finish.
* @param resultCode Result code, if any, from this Activity.
* @param resultData Result data (Intent), if any, from this Activity.
* @param finishTask Whether to finish the task associated with this Activity.
*
* @return Returns true if the activity successfully finished, or false if it is still running.
*/
@Override
public final boolean finishActivity(IBinder token, int resultCode, Intent resultData,
int finishTask) {
// Refuse possible leaked file descriptors
if (resultData != null && resultData.hasFileDescriptors()) {
throw new IllegalArgumentException("File descriptors passed in Intent");
}
synchronized (mGlobalLock) {
ActivityRecord r = ActivityRecord.isInStackLocked(token);
if (r == null) {
return true;
}
// Keep track of the root activity of the task before we finish it
final TaskRecord tr = r.getTaskRecord();
ActivityRecord rootR = tr.getRootActivity();
if (rootR == null) {
Slog.w(TAG, "Finishing task with all activities already finished");
}
// Do not allow task to finish if last task in lockTask mode. Launchable priv-apps can
// finish.
if (getLockTaskController().activityBlockedFromFinish(r)) {
return false;
}
// TODO: There is a dup. of this block of code in ActivityStack.navigateUpToLocked
// We should consolidate.
if (mController != null) {
// Find the first activity that is not finishing.
final ActivityRecord next = r.getActivityStack().topRunningActivityLocked(token, 0);
if (next != null) {
// ask watcher if this is allowed
boolean resumeOK = true;
try {
resumeOK = mController.activityResuming(next.packageName);
} catch (RemoteException e) {
mController = null;
Watchdog.getInstance().setActivityController(null);
}
if (!resumeOK) {
Slog.i(TAG, "Not finishing activity because controller resumed");
return false;
}
}
}
// note down that the process has finished an activity and is in background activity
// starts grace period
if (r.app != null) {
r.app.setLastActivityFinishTimeIfNeeded(SystemClock.uptimeMillis());
}
final long origId = Binder.clearCallingIdentity();
try {
boolean res;
final boolean finishWithRootActivity =
finishTask == Activity.FINISH_TASK_WITH_ROOT_ACTIVITY;
if (finishTask == Activity.FINISH_TASK_WITH_ACTIVITY
|| (finishWithRootActivity && r == rootR)) {
// If requested, remove the task that is associated to this activity only if it
// was the root activity in the task. The result code and data is ignored
// because we don't support returning them across task boundaries. Also, to
// keep backwards compatibility we remove the task from recents when finishing
// task with root activity.
res = mStackSupervisor.removeTaskByIdLocked(tr.taskId, false,
finishWithRootActivity, "finish-activity");
if (!res) {
Slog.i(TAG, "Removing task failed to finish activity");
}
// Explicitly dismissing the activity so reset its relaunch flag.
r.mRelaunchReason = RELAUNCH_REASON_NONE;
} else {
res = tr.getStack().requestFinishActivityLocked(token, resultCode,
resultData, "app-request", true);
if (!res) {
Slog.i(TAG, "Failed to finish by app-request");
}
}
return res;
} finally {
Binder.restoreCallingIdentity(origId);
}
}
}
> ActivityManagerService.java
/**
* This is the internal entry point for handling Activity.finish().
*
* @param token The Binder token referencing the Activity we want to finish.
* @param resultCode Result code, if any, from this Activity.
* @param resultData Result data (Intent), if any, from this Activity.
* @param finishTask Whether to finish the task associated with this Activity.
*
* @return Returns true if the activity successfully finished, or false if it is still running.
*/
@Override
public final boolean finishActivity(IBinder token, int resultCode, Intent resultData,
int finishTask) {
// token 持有 ActivityRecord 的弱引用
return mActivityTaskManager.finishActivity(token, resultCode, resultData, finishTask);
}
注意方法参数中的 token 对象是 ActivityRecord 的静态内部类，它持有外部 ActivityRecord 的弱引用。继承自 IApplicationToken.Stub ，是一个 Binder 对象。ActivityRecord 就是对当前 Activity 的具体描述，包含了 Activity 的所有信息。
传入的 finishTask() 方法的参数是 DONT_FINISH_TASK_WITH_ACTIVITY，所以接着会调用 ActivityStack.requestFinishActivityLocked() 方法。
> ActivityRecord.java
static class Token extends IApplicationToken.Stub {
private final WeakReference weakActivity; //持有外部 ActivityRecord 的弱引用
private final String name;
Token(ActivityRecord activity, Intent intent) {
weakActivity = new WeakReference<>(activity);
name = intent.getComponent().flattenToShortString();
}
private static @Nullable ActivityRecord tokenToActivityRecordLocked(Token token) {
if (token == null) {
return null;
}
ActivityRecord r = token.weakActivity.get();
if (r == null || r.getActivityStack() == null) {
return null;
}
return r;
}
@Override
public String toString() {
StringBuilder sb = new StringBuilder(128);
sb.append("Token{");
sb.append(Integer.toHexString(System.identityHashCode(this)));
sb.append(' ');
sb.append(weakActivity.get());
sb.append('}');
return sb.toString();
}
@Override
public String getName() {
return name;
}
}
> ActivityStack.java
/**
* @return Returns true if the activity is being finished, false if for
* some reason it is being left as-is.
*/
final boolean requestFinishActivityLocked(IBinder token, int resultCode,
Intent resultData, String reason, boolean oomAdj) {
ActivityRecord r = isInStackLocked(token);
if (DEBUG_RESULTS || DEBUG_STATES) Slog.v(TAG_STATES,
"Finishing activity token=" + token + " r="
+ ", result=" + resultCode + ", data=" + resultData
+ ", reason=" + reason);
if (r == null) {
return false;
}
finishActivityLocked(r, resultCode, resultData, reason, oomAdj);
return true;
}
最后调用的是一个重载的 finishActivityLocked() 方法。
> ActivityStack.java
/**
* @return Returns true if this activity has been removed from the history
* list, or false if it is still in the list and will be removed later.
*/
final boolean finishActivityLocked(ActivityRecord r, int resultCode, Intent resultData,
String reason, boolean oomAdj, boolean pauseImmediately) {
if (r.finishing) { //重复 finishing 的情况
Slog.w(TAG, "Duplicate finish request for " + r);
return false;
}
mWindowManager.deferSurfaceLayout();
try {
// 标记 r.finishing = true，
// 前面会做重复 finish 的检测就是依赖这个值
r.makeFinishingLocked();
final TaskRecord task = r.getTaskRecord();
EventLog.writeEvent(EventLogTags.AM_FINISH_ACTIVITY,
r.mUserId, System.identityHashCode(r),
task.taskId, r.shortComponentName, reason);
final ArrayList activities = task.mActivities;
final int index = activities.indexOf(r);
if (index < (activities.size() - 1)) {
task.setFrontOfTask();
if ((r.intent.getFlags() & Intent.FLAG_ACTIVITY_CLEAR_WHEN_TASK_RESET) != 0) {
// If the caller asked that this activity (and all above it)
// be cleared when the task is reset, don't lose that information,
// but propagate it up to the next activity.
ActivityRecord next = activities.get(index+1);
next.intent.addFlags(Intent.FLAG_ACTIVITY_CLEAR_WHEN_TASK_RESET);
}
}
// 暂停事件分发
r.pauseKeyDispatchingLocked();
adjustFocusedActivityStack(r, "finishActivity");
// 处理 activity result
finishActivityResultsLocked(r, resultCode, resultData);
final boolean endTask = index <= 0 && !task.isClearingToReuseTask();
final int transit = endTask ? TRANSIT_TASK_CLOSE : TRANSIT_ACTIVITY_CLOSE;
// mResumedActivity 就是当前 Activity，会进入此分支
if (mResumedActivity == r) {
if (DEBUG_VISIBILITY || DEBUG_TRANSITION) Slog.v(TAG_TRANSITION,
"Prepare close transition: finishing " + r);
if (endTask) {
mService.getTaskChangeNotificationController().notifyTaskRemovalStarted(
task.getTaskInfo());
}
getDisplay().mDisplayContent.prepareAppTransition(transit, false);
// Tell window manager to prepare for this one to be removed.
r.setVisibility(false);
if (mPausingActivity == null) {
if (DEBUG_PAUSE) Slog.v(TAG_PAUSE, "Finish needs to pause: " + r);
if (DEBUG_USER_LEAVING) Slog.v(TAG_USER_LEAVING,
"finish() => pause with userLeaving=false");
// 开始 pause mResumedActivity
startPausingLocked(false, false, null, pauseImmediately);
}
if (endTask) {
mService.getLockTaskController().clearLockedTask(task);
}
} else if (!r.isState(PAUSING)) {
// If the activity is PAUSING, we will complete the finish once
// it is done pausing; else we can just directly finish it here.
if (DEBUG_PAUSE) Slog.v(TAG_PAUSE, "Finish not pausing: " + r);
if (r.visible) {
prepareActivityHideTransitionAnimation(r, transit);
}
final int finishMode = (r.visible || r.nowVisible) ? FINISH_AFTER_VISIBLE
: FINISH_AFTER_PAUSE;
final boolean removedActivity = finishCurrentActivityLocked(r, finishMode, oomAdj,
"finishActivityLocked") == null;
// The following code is an optimization. When the last non-task overlay activity
// is removed from the task, we remove the entire task from the stack. However,
// since that is done after the scheduled destroy callback from the activity, that
// call to change the visibility of the task overlay activities would be out of
// sync with the activitiy visibility being set for this finishing activity above.
// In this case, we can set the visibility of all the task overlay activities when
// we detect the last one is finishing to keep them in sync.
if (task.onlyHasTaskOverlayActivities(true /* excludeFinishing */)) {
for (ActivityRecord taskOverlay : task.mActivities) {
if (!taskOverlay.mTaskOverlay) {
continue;
}
prepareActivityHideTransitionAnimation(taskOverlay, transit);
}
}
return removedActivity;
} else {
if (DEBUG_PAUSE) Slog.v(TAG_PAUSE, "Finish waiting for pause of: " + r);
}
return false;
} finally {
mWindowManager.continueSurfaceLayout();
}
}
void makeFinishingLocked() {
if (finishing) {
return;
}
finishing = true;
if (stopped) {
clearOptionsLocked();
}
if (mAtmService != null) {
mAtmService.getTaskChangeNotificationController().notifyTaskStackChanged();
}
}
调用 finish 之后肯定是要先 pause 当前 Activity，没毛病。接着看 startPausingLocked() 方法。
ActivityStack.java
/**
* Start pausing the currently resumed activity. It is an error to call this if there
* is already an activity being paused or there is no resumed activity.
*
* @param userLeaving True if this should result in an onUserLeaving to the current activity.
* @param uiSleeping True if this is happening with the user interface going to sleep (the
* screen turning off).
* @param resuming The activity we are currently trying to resume or null if this is not being
* called as part of resuming the top activity, so we shouldn't try to instigate
* a resume here if not null.
* @param pauseImmediately True if the caller does not want to wait for the activity callback to
* complete pausing.
* @return Returns true if an activity now is in the PAUSING state, and we are waiting for
* it to tell us when it is done.
*/
final boolean startPausingLocked(boolean userLeaving, boolean uiSleeping,
ActivityRecord resuming, boolean pauseImmediately) {
if (mPausingActivity != null) {
Slog.wtf(TAG, "Going to pause when pause is already pending for " + mPausingActivity
+ " state=" + mPausingActivity.getState());
if (!shouldSleepActivities()) {
// Avoid recursion among check for sleep and complete pause during sleeping.
// Because activity will be paused immediately after resume, just let pause
// be completed by the order of activity paused from clients.
completePauseLocked(false, resuming);
}
}
ActivityRecord prev = mResumedActivity;
if (prev == null) {
// 没有 onResume 的 Activity，不能执行 pause
if (resuming == null) {
Slog.wtf(TAG, "Trying to pause when nothing is resumed");
mRootActivityContainer.resumeFocusedStacksTopActivities();
}
return false;
}
if (prev == resuming) {
Slog.wtf(TAG, "Trying to pause activity that is in process of being resumed");
return false;
}
if (DEBUG_STATES) Slog.v(TAG_STATES, "Moving to PAUSING: " + prev);
else if (DEBUG_PAUSE) Slog.v(TAG_PAUSE, "Start pausing: " + prev);
mPausingActivity = prev;
mLastPausedActivity = prev;
mLastNoHistoryActivity = (prev.intent.getFlags() & Intent.FLAG_ACTIVITY_NO_HISTORY) != 0
|| (prev.info.flags & ActivityInfo.FLAG_NO_HISTORY) != 0 ? prev : null;
// 设置当前 Activity 状态为 PAUSING
prev.setState(PAUSING, "startPausingLocked");
prev.getTaskRecord().touchActiveTime();
clearLaunchTime(prev);
mService.updateCpuStats();
if (prev.attachedToProcess()) {
if (DEBUG_PAUSE) Slog.v(TAG_PAUSE, "Enqueueing pending pause: " + prev);
try {
EventLogTags.writeAmPauseActivity(prev.mUserId, System.identityHashCode(prev),
prev.shortComponentName, "userLeaving=" + userLeaving);
// 1. 通过 ClientLifecycleManager 分发生命周期事件
// 最终会向 H 发送 EXECUTE_TRANSACTION 事件
mService.getLifecycleManager().scheduleTransaction(prev.app.getThread(),
prev.appToken, PauseActivityItem.obtain(prev.finishing, userLeaving,
prev.configChangeFlags, pauseImmediately));
} catch (Exception e) {
// Ignore exception, if process died other code will cleanup.
Slog.w(TAG, "Exception thrown during pause", e);
mPausingActivity = null;
mLastPausedActivity = null;
mLastNoHistoryActivity = null;
}
} else {
mPausingActivity = null;
mLastPausedActivity = null;
mLastNoHistoryActivity = null;
}
// If we are not going to sleep, we want to ensure the device is
// awake until the next activity is started.
if (!uiSleeping && !mService.isSleepingOrShuttingDownLocked()) {
mStackSupervisor.acquireLaunchWakelock();
}
// mPausingActivity 在前面已经赋值，就是当前 Activity
if (mPausingActivity != null) {
// Have the window manager pause its key dispatching until the new
// activity has started. If we're pausing the activity just because
// the screen is being turned off and the UI is sleeping, don't interrupt
// key dispatch; the same activity will pick it up again on wakeup.
if (!uiSleeping) {
prev.pauseKeyDispatchingLocked();
} else if (DEBUG_PAUSE) {
Slog.v(TAG_PAUSE, "Key dispatch not paused for screen off");
}
if (pauseImmediately) { // 这里是 false，进入 else 分支
// If the caller said they don't want to wait for the pause, then complete
// the pause now.
completePauseLocked(false, resuming);
return false;
} else {
// 2. 发送一个延时 500ms 的消息，等待 pause 流程一点时间，最终会回调 activityPausedLocked() 方法
schedulePauseTimeout(prev);
return true;
}
} else {
// This activity failed to schedule the
// pause, so just treat it as being paused now.
if (DEBUG_PAUSE) Slog.v(TAG_PAUSE, "Activity not running, resuming next.");
if (resuming == null) {
mRootActivityContainer.resumeFocusedStacksTopActivities();
}
return false;
}
}
/**
* Schedule a pause timeout in case the app doesn't respond. We don't give it much time because
* this directly impacts the responsiveness seen by the user.
*/
private void schedulePauseTimeout(ActivityRecord r) {
final Message msg = mHandler.obtainMessage(PAUSE_TIMEOUT_MSG);
msg.obj = r;
r.pauseTime = SystemClock.uptimeMillis();
mHandler.sendMessageDelayed(msg, PAUSE_TIMEOUT);
if (DEBUG_PAUSE) Slog.v(TAG_PAUSE, "Waiting for pause to complete...");
}
这里面有两步重点操作。第一步是注释 1 处通过 ClientLifecycleManager 分发生命周期流程。第二步是发送一个延时 500ms 的消息，等待一下 onPause 流程。但是如果第一步中在 500ms 内已经完成了流程，则会取消这个消息。所以这两步的最终逻辑其实是一致的。这里就直接看第一步。
mService.getLifecycleManager().scheduleTransaction(prev.app.getThread(),
prev.appToken, PauseActivityItem.obtain(prev.finishing, userLeaving,
prev.configChangeFlags, pauseImmediately));
ClientLifecycleManager 它会向主线程的 Handler H 发送 EXECUTE_TRANSACTION 事件，调用 XXXActivityItem 的 execute() 和 postExecute() 方法。execute() 方法中会 Binder 调用
ActivityThread 中对应的 handleXXXActivity() 方法。在这里就是 handlePauseActivity() 方法，其中会通过 Instrumentation.callActivityOnPause(r.activity) 方法回调 Activity.onPause()。
> Instrumentation.java
/**
* Perform calling of an activity's {@link Activity#onPause} method. The
* default implementation simply calls through to that method.
*
* @param activity The activity being paused.
*/
public void callActivityOnPause(Activity activity) {
activity.performPause();
}
到这里，onPause() 方法就被执行了。但是流程没有结束，接着就该显示下一个 Activity 了。前面刚刚说过会调用 PauseActivityItem 的 execute() 和 postExecute() 方法。execute() 方法回调了当前 Activity.onPause()，而 postExecute() 方法就是去寻找要显示的 Activity 。
> PauseActivityItem.java
@Override
public void postExecute(ClientTransactionHandler client, IBinder token,
PendingTransactionActions pendingActions) {
if (mDontReport) {
return;
}
try {
// TODO(lifecycler): Use interface callback instead of AMS.
ActivityTaskManager.getService().activityPaused(token);
} catch (RemoteException ex) {
throw ex.rethrowFromSystemServer();
}
}
Binder 调用了 ActivityTaskManager.activityPaused() 方法。
> IActivityTaskManager.aidl
void activityPaused(in IBinder token);
> ActivityTaskManagerService.java
@Override
public final void activityPaused(IBinder token) {
final long origId = Binder.clearCallingIdentity();
synchronized (mGlobalLock) {
ActivityStack stack = ActivityRecord.getStackLocked(token);
if (stack != null) {
stack.activityPausedLocked(token, false);
}
}
Binder.restoreCallingIdentity(origId);
}
调用了 ActivityStack.activityPausedLocked() 方法。
> ActivityStack.java
final void activityPausedLocked(IBinder token, boolean timeout) {
if (DEBUG_PAUSE) Slog.v(TAG_PAUSE,
"Activity paused: token=" + token + ", timeout=" + timeout);
final ActivityRecord r = isInStackLocked(token);
if (r != null) {
mHandler.removeMessages(PAUSE_TIMEOUT_MSG, r); //标记1
if (mPausingActivity == r) {
if (DEBUG_STATES) Slog.v(TAG_STATES, "Moving to PAUSED: " + r
+ (timeout ? " (due to timeout)" : " (pause complete)"));
mService.mWindowManager.deferSurfaceLayout();
try {
//标记2
completePauseLocked(true /* resumeNext */, null /* resumingActivity */);
} finally {
mService.mWindowManager.continueSurfaceLayout();
}
return;
} else {
EventLog.writeEvent(EventLogTags.AM_FAILED_TO_PAUSE,
r.mUserId, System.identityHashCode(r), r.shortComponentName,
mPausingActivity != null
? mPausingActivity.shortComponentName : "(none)");
if (r.isState(PAUSING)) {
r.setState(PAUSED, "activityPausedLocked");
if (r.finishing) {
if (DEBUG_PAUSE) Slog.v(TAG,
"Executing finish of failed to pause activity: " + r);
finishCurrentActivityLocked(r, FINISH_AFTER_VISIBLE, false,
"activityPausedLocked");
}
}
}
}
mRootActivityContainer.ensureActivitiesVisible(null, 0, !PRESERVE_WINDOWS);
}
上面有这么一行代码 mHandler.removeMessages(PAUSE_TIMEOUT_MSG, r) ，移除的就是之前延迟 500ms 的消息。接着看 completePauseLocked() 方法。
> ActivityStack.java
private void completePauseLocked(boolean resumeNext, ActivityRecord resuming) {
ActivityRecord prev = mPausingActivity;
if (DEBUG_PAUSE) Slog.v(TAG_PAUSE, "Complete pause: " + prev);
if (prev != null) {
prev.setWillCloseOrEnterPip(false);
final boolean wasStopping = prev.isState(STOPPING);
// 设置状态为 PAUSED
prev.setState(PAUSED, "completePausedLocked");
if (prev.finishing) { // 1. finishing 为 true，进入此分支
if (DEBUG_PAUSE) Slog.v(TAG_PAUSE, "Executing finish of activity: " + prev);
prev = finishCurrentActivityLocked(prev, FINISH_AFTER_VISIBLE, false,
"completePausedLocked");
} else if (prev.hasProcess()) {
if (DEBUG_PAUSE) Slog.v(TAG_PAUSE, "Enqueue pending stop if needed: " + prev
+ " wasStopping=" + wasStopping + " visible=" + prev.visible);
if (prev.deferRelaunchUntilPaused) {
// Complete the deferred relaunch that was waiting for pause to complete.
if (DEBUG_PAUSE) Slog.v(TAG_PAUSE, "Re-launching after pause: " + prev);
prev.relaunchActivityLocked(false /* andResume */,
prev.preserveWindowOnDeferredRelaunch);
} else if (wasStopping) {
// We are also stopping, the stop request must have gone soon after the pause.
// We can't clobber it, because the stop confirmation will not be handled.
// We don't need to schedule another stop, we only need to let it happen.
prev.setState(STOPPING, "completePausedLocked");
} else if (!prev.visible || shouldSleepOrShutDownActivities()) {
// Clear out any deferred client hide we might currently have.
prev.setDeferHidingClient(false);
// If we were visible then resumeTopActivities will release resources before
// stopping.
addToStopping(prev, true /* scheduleIdle */, false /* idleDelayed */,
"completePauseLocked");
}
} else {
if (DEBUG_PAUSE) Slog.v(TAG_PAUSE, "App died during pause, not stopping: " + prev);
prev = null;
}
// It is possible the activity was freezing the screen before it was paused.
// In that case go ahead and remove the freeze this activity has on the screen
// since it is no longer visible.
if (prev != null) {
prev.stopFreezingScreenLocked(true /*force*/);
}
mPausingActivity = null;
}
if (resumeNext) {
// 当前获取焦点的 ActivityStack
final ActivityStack topStack = mRootActivityContainer.getTopDisplayFocusedStack();
if (!topStack.shouldSleepOrShutDownActivities()) {
// 2. 恢复要显示的 activity
mRootActivityContainer.resumeFocusedStacksTopActivities(topStack, prev, null);
} else {
checkReadyForSleep();
ActivityRecord top = topStack.topRunningActivityLocked();
if (top == null || (prev != null && top != prev)) {
// If there are no more activities available to run, do resume anyway to start
// something. Also if the top activity on the stack is not the just paused
// activity, we need to go ahead and resume it to ensure we complete an
// in-flight app switch.
mRootActivityContainer.resumeFocusedStacksTopActivities();
}
}
}
if (prev != null) {
prev.resumeKeyDispatchingLocked();
if (prev.hasProcess() && prev.cpuTimeAtResume > 0) {
final long diff = prev.app.getCpuTime() - prev.cpuTimeAtResume;
if (diff > 0) {
final Runnable r = PooledLambda.obtainRunnable(
ActivityManagerInternal::updateForegroundTimeIfOnBattery,
mService.mAmInternal, prev.info.packageName,
prev.info.applicationInfo.uid,
diff);
mService.mH.post(r);
}
}
prev.cpuTimeAtResume = 0; // reset it
}
// Notify when the task stack has changed, but only if visibilities changed (not just
// focus). Also if there is an active pinned stack - we always want to notify it about
// task stack changes, because its positioning may depend on it.
if (mStackSupervisor.mAppVisibilitiesChangedSinceLastPause
|| (getDisplay() != null && getDisplay().hasPinnedStack())) {
mService.getTaskChangeNotificationController().notifyTaskStackChanged();
mStackSupervisor.mAppVisibilitiesChangedSinceLastPause = false;
}
mRootActivityContainer.ensureActivitiesVisible(resuming, 0, !PRESERVE_WINDOWS);
}
这里分了两步走。注释1 处判断了 finishing 状态，还记得 finishing 在何处被赋值为 true 的吗？在 Activity.finish() -> AMS.finishActivity() -> ActivityStack.requestFinishActivityLocked() -> ActivityStack.finishActivityLocked() 方法中。所以接着调用的是 finishCurrentActivityLocked() 方法。注释2 处就是来显示应该显示的 Activity ，就不再追进去细看了。
再跟到 finishCurrentActivityLocked() 方法中，看这名字，肯定是要 stop/destroy 没跑了。
> ActivityStack.java
/*
* 把前面带过来的参数标出来
* prev, FINISH_AFTER_VISIBLE, false,"completedPausedLocked"
*/
final ActivityRecord finishCurrentActivityLocked(ActivityRecord r, int mode, boolean oomAdj,
String reason) {
// First things first: if this activity is currently visible,
// and the resumed activity is not yet visible, then hold off on
// finishing until the resumed one becomes visible.
// The activity that we are finishing may be over the lock screen. In this case, we do not
// want to consider activities that cannot be shown on the lock screen as running and should
// proceed with finishing the activity if there is no valid next top running activity.
// Note that if this finishing activity is floating task, we don't need to wait the
// next activity resume and can destroy it directly.
final ActivityDisplay display = getDisplay();
// 获取将要显示的栈顶 Activity
final ActivityRecord next = display.topRunningActivity(true /* considerKeyguardState */);
final boolean isFloating = r.getConfiguration().windowConfiguration.tasksAreFloating();
// 1. mode 是 FINISH_AFTER_VISIBLE，进入此分支
if (mode == FINISH_AFTER_VISIBLE && (r.visible || r.nowVisible)
&& next != null && !next.nowVisible && !isFloating) {
if (!mStackSupervisor.mStoppingActivities.contains(r)) {
// 加入到 mStackSupervisor.mStoppingActivities
addToStopping(r, false /* scheduleIdle */, false /* idleDelayed */,
"finishCurrentActivityLocked");
}
if (DEBUG_STATES) Slog.v(TAG_STATES,
"Moving to STOPPING: "+ r + " (finish requested)");
// 设置状态为 STOPPING
r.setState(STOPPING, "finishCurrentActivityLocked");
if (oomAdj) {
mService.updateOomAdj();
}
return r;
}
// make sure the record is cleaned out of other places.
mStackSupervisor.mStoppingActivities.remove(r);
mStackSupervisor.mGoingToSleepActivities.remove(r);
final ActivityState prevState = r.getState();
if (DEBUG_STATES) Slog.v(TAG_STATES, "Moving to FINISHING: " + r);
r.setState(FINISHING, "finishCurrentActivityLocked");
// Don't destroy activity immediately if the display contains home stack, although there is
// no next activity at the moment but another home activity should be started later. Keep
// this activity alive until next home activity is resumed then user won't see a temporary
// black screen.
final boolean noRunningStack = next == null && display.topRunningActivity() == null
&& display.getHomeStack() == null;
final boolean noFocusedStack = r.getActivityStack() != display.getFocusedStack();
final boolean finishingInNonFocusedStackOrNoRunning = mode == FINISH_AFTER_VISIBLE
&& prevState == PAUSED && (noFocusedStack || noRunningStack);
// 下面的代码会执行 destroy 流程
if (mode == FINISH_IMMEDIATELY
|| (prevState == PAUSED
&& (mode == FINISH_AFTER_PAUSE || inPinnedWindowingMode()))
|| finishingInNonFocusedStackOrNoRunning
|| prevState == STOPPING
|| prevState == STOPPED
|| prevState == ActivityState.INITIALIZING) {
r.makeFinishingLocked();
boolean activityRemoved = destroyActivityLocked(r, true, "finish-imm:" + reason);
if (finishingInNonFocusedStackOrNoRunning) {
// Finishing activity that was in paused state and it was in not currently focused
// stack, need to make something visible in its place. Also if the display does not
// have running activity, the configuration may need to be updated for restoring
// original orientation of the display.
mRootActivityContainer.ensureVisibilityAndConfig(next, mDisplayId,
false /* markFrozenIfConfigChanged */, true /* deferResume */);
}
if (activityRemoved) {
mRootActivityContainer.resumeFocusedStacksTopActivities();
}
if (DEBUG_CONTAINERS) Slog.d(TAG_CONTAINERS,
"destroyActivityLocked: finishCurrentActivityLocked r=" + r +
" destroy returned removed=" + activityRemoved);
return activityRemoved ? null : r;
}
// Need to go through the full pause cycle to get this
// activity into the stopped state and then finish it.
if (DEBUG_ALL) Slog.v(TAG, "Enqueueing pending finish: " + r);
mStackSupervisor.mFinishingActivities.add(r);
r.resumeKeyDispatchingLocked();
mRootActivityContainer.resumeFocusedStacksTopActivities();
// If activity was not paused at this point - explicitly pause it to start finishing
// process. Finishing will be completed once it reports pause back.
if (r.isState(RESUMED) && mPausingActivity != null) {
startPausingLocked(false /* userLeaving */, false /* uiSleeping */, next /* resuming */,
false /* dontWait */);
}
return r;
}
注释 1 处 mode 的值是 FINISH_AFTER_VISIBLE ，并且现在新的 Activity 还没有 onResume，所以 r.visible || r.nowVisible 和 next != null && !next.nowVisible 都是成立的，并不会进入后面的 destroy 流程。虽然看到这还没得到想要的答案，但是起码是符合预期的。如果在这就直接 destroy 了，延迟 10s 才 onDestroy 的问题就无疾而终了。
对于这些暂时还不销毁的 Activity 都执行了 addToStopping(r, false, false) 方法。我们继续追进去。
> ActivityStack.java
private void addToStopping(ActivityRecord r, boolean scheduleIdle, boolean idleDelayed,
String reason) {
if (!mStackSupervisor.mStoppingActivities.contains(r)) {
EventLog.writeEvent(EventLogTags.AM_ADD_TO_STOPPING, r.mUserId,
System.identityHashCode(r), r.shortComponentName, reason);
mStackSupervisor.mStoppingActivities.add(r);
}
// If we already have a few activities waiting to stop, then give up
// on things going idle and start clearing them out. Or if r is the
// last of activity of the last task the stack will be empty and must
// be cleared immediately.
//对 mStoppingActivities 的存储容量做了限制。超出限制可能会提前触发销毁流程
boolean forceIdle = mStackSupervisor.mStoppingActivities.size() > MAX_STOPPING_TO_FORCE
|| (r.frontOfTask && mTaskHistory.size() <= 1);
if (scheduleIdle || forceIdle) {
if (DEBUG_PAUSE) Slog.v(TAG_PAUSE, "Scheduling idle now: forceIdle="
+ forceIdle + "immediate=" + !idleDelayed);
if (!idleDelayed) {
mStackSupervisor.scheduleIdleLocked();
} else {
mStackSupervisor.scheduleIdleTimeoutLocked(r);
}
} else {
checkReadyForSleep();
}
}
这些在等待销毁的 Activity 被保存在了 ActivityStackSupervisor 的 mStoppingActivities 集合中，它是一个 ArrayList
整个 finish 流程就到此为止了。前一个 Activity 被保存在了ActivityStackSupervisor.mStoppingActivities 集合中，新的 Activity 被显示出来了。
问题似乎进入了困境，什么时候回调 onStop/onDestroy 呢？其实这个才是根本问题。上面撸了一遍 finish() 并看不到本质，但是可以帮助我们形成一个完整的流程，这个一直是看 AOSP 最大的意义，帮助我们把零碎的上层知识形成一个完整的闭环。
回到正题来，在 Activity 跳转过程中，为了保证流畅的用户体验，只要前一个 Activity 与用户不可交互，即 onPause() 被回调之后，下一个 Activity 就要开始自己的生命周期流程了。所以
onStop/onDestroy 的调用时间是不确定的，可能整整过了 10s 才回调。那么，到底是由谁来驱动 onStop/onDestroy 的执行呢？我们来看看下一个 Activity 的 onResume 过程。
直接看 ActivityThread.handleResumeActivity() 方法，相信大家对生命周期的调用流程也很熟悉了。
> ActivityThread.java
@Override
public void handleResumeActivity(IBinder token, boolean finalStateRequest, boolean isForward,
String reason) {
// If we are getting ready to gc after going to the background, well
// we are back active so skip it.
unscheduleGcIdler();
mSomeActivitiesChanged = true;
// TODO Push resumeArgs into the activity for consideration
// 回调 onResume
final ActivityClientRecord r = performResumeActivity(token, finalStateRequest, reason);
if (r == null) {
// We didn't actually resume the activity, so skipping any follow-up actions.
return;
}
if (mActivitiesToBeDestroyed.containsKey(token)) {
// Although the activity is resumed, it is going to be destroyed. So the following
// UI operations are unnecessary and also prevents exception because its token may
// be gone that window manager cannot recognize it. All necessary cleanup actions
// performed below will be done while handling destruction.
return;
}
final Activity a = r.activity;
if (localLOGV) {
Slog.v(TAG, "Resume " + r + " started activity: " + a.mStartedActivity
+ ", hideForNow: " + r.hideForNow + ", finished: " + a.mFinished);
}
final int forwardBit = isForward
? WindowManager.LayoutParams.SOFT_INPUT_IS_FORWARD_NAVIGATION : 0;
// If the window hasn't yet been added to the window manager,
// and this guy didn't finish itself or start another activity,
// then go ahead and add the window.
boolean willBeVisible = !a.mStartedActivity;
if (!willBeVisible) {
try {
willBeVisible = ActivityTaskManager.getService().willActivityBeVisible(
a.getActivityToken());
} catch (RemoteException e) {
throw e.rethrowFromSystemServer();
}
}
if (r.window == null && !a.mFinished && willBeVisible) {
r.window = r.activity.getWindow();
View decor = r.window.getDecorView();
decor.setVisibility(View.INVISIBLE);
ViewManager wm = a.getWindowManager();
WindowManager.LayoutParams l = r.window.getAttributes();
a.mDecor = decor;
l.type = WindowManager.LayoutParams.TYPE_BASE_APPLICATION;
l.softInputMode |= forwardBit;
if (r.mPreserveWindow) {
a.mWindowAdded = true;
r.mPreserveWindow = false;
// Normally the ViewRoot sets up callbacks with the Activity
// in addView->ViewRootImpl#setView. If we are instead reusing
// the decor view we have to notify the view root that the
// callbacks may have changed.
ViewRootImpl impl = decor.getViewRootImpl();
if (impl != null) {
impl.notifyChildRebuilt();
}
}
if (a.mVisibleFromClient) {
if (!a.mWindowAdded) {
a.mWindowAdded = true;
// 添加 decorView 到 WindowManager
wm.addView(decor, l);
} else {
// The activity will get a callback for this {@link LayoutParams} change
// earlier. However, at that time the decor will not be set (this is set
// in this method), so no action will be taken. This call ensures the
// callback occurs with the decor set.
a.onWindowAttributesChanged(l);
}
}
// If the window has already been added, but during resume
// we started another activity, then don't yet make the
// window visible.
} else if (!willBeVisible) {
if (localLOGV) Slog.v(TAG, "Launch " + r + " mStartedActivity set");
r.hideForNow = true;
}
// Get rid of anything left hanging around.
cleanUpPendingRemoveWindows(r, false /* force */);
// The window is now visible if it has been added, we are not
// simply finishing, and we are not starting another activity.
if (!r.activity.mFinished && willBeVisible && r.activity.mDecor != null && !r.hideForNow) {
if (r.newConfig != null) {
performConfigurationChangedForActivity(r, r.newConfig);
if (DEBUG_CONFIGURATION) {
Slog.v(TAG, "Resuming activity " + r.activityInfo.name + " with newConfig "
+ r.activity.mCurrentConfig);
}
r.newConfig = null;
}
if (localLOGV) Slog.v(TAG, "Resuming " + r + " with isForward=" + isForward);
WindowManager.LayoutParams l = r.window.getAttributes();
if ((l.softInputMode
& WindowManager.LayoutParams.SOFT_INPUT_IS_FORWARD_NAVIGATION)
!= forwardBit) {
l.softInputMode = (l.softInputMode
& (~WindowManager.LayoutParams.SOFT_INPUT_IS_FORWARD_NAVIGATION))
| forwardBit;
if (r.activity.mVisibleFromClient) {
ViewManager wm = a.getWindowManager();
View decor = r.window.getDecorView();
wm.updateViewLayout(decor, l);
}
}
r.activity.mVisibleFromServer = true;
mNumVisibleActivities++;
if (r.activity.mVisibleFromClient) {
r.activity.makeVisible();
}
}
r.nextIdle = mNewActivities;
mNewActivities = r;
if (localLOGV) Slog.v(TAG, "Scheduling idle handler for " + r);
// 主线程空闲时会执行 Idler
Looper.myQueue().addIdleHandler(new Idler());
}
handleResumeActivity() 方法是整个 UI 显示流程的重中之重，它首先会回调 Activity.onResume() , 然后将 DecorView 添加到 Window 上，其中又包括了创建 ViewRootImpl，创建 Choreographer，与 WMS 进行 Binder 通信，注册 vsync 信号，执行measure/draw/layout。
在完成最终的界面绘制和显示之后，有这么一句代码 Looper.myQueue().addIdleHandler(new Idler())。IdleHandler 不知道大家是否熟悉，它提供了一种机制，当主线程消息队列空闲时，会执行 IdleHandler 的回调方法。至于怎么算 “空闲”，我们可以看一下 MessageQueue.next() 方法。
>MessageQueue.java
Message next() {
// Return here if the message loop has already quit and been disposed.
// This can happen if the application tries to restart a looper after quit
// which is not supported.
final long ptr = mPtr;
if (ptr == 0) {
return null;
}
int pendingIdleHandlerCount = -1; // -1 only during first iteration
int nextPollTimeoutMillis = 0;
for (;;) {
if (nextPollTimeoutMillis != 0) {
Binder.flushPendingCommands();
}
// 阻塞方法，主要是通过 native 层的 epoll 监听文件描述符的写入事件来实现的。
// 如果 nextPollTimeoutMillis = -1，一直阻塞不会超时。
// 如果 nextPollTimeoutMillis = 0，不会阻塞，立即返回。
// 如果 nextPollTimeoutMillis > 0，最长阻塞nextPollTimeoutMillis毫秒(超时)，如果期间有程序唤醒会立即返回。
nativePollOnce(ptr, nextPollTimeoutMillis);
synchronized (this) {
// Try to retrieve the next message. Return if found.
final long now = SystemClock.uptimeMillis();
Message prevMsg = null;
Message msg = mMessages;
if (msg != null && msg.target == null) {
// Stalled by a barrier. Find the next asynchronous message in the queue.
// msg.target == null表示此消息为消息屏障（通过postSyncBarrier方法发送来的）
// 如果发现了一个消息屏障，会循环找出第一个异步消息（如果有异步消息的话），所有同步消息都将忽略（平常发送的一般都是同步消息）
do {
prevMsg = msg;
msg = msg.next;
} while (msg != null && !msg.isAsynchronous());
}
if (msg != null) {
if (now < msg.when) {
// Next message is not ready. Set a timeout to wake up when it is ready.
// 消息触发时间未到，设置下一次轮询的超时时间
nextPollTimeoutMillis = (int) Math.min(msg.when - now, Integer.MAX_VALUE);
} else {
// Got a message.
// 得到 Message
mBlocked = false;
if (prevMsg != null) {
prevMsg.next = msg.next;
} else {
mMessages = msg.next;
}
msg.next = null;
if (DEBUG) Log.v(TAG, "Returning message: " + msg);
msg.markInUse(); // 标记 FLAG_IN_USE
return msg;
}
} else {
// No more messages.
nextPollTimeoutMillis = -1;
}
// Process the quit message now that all pending messages have been handled.
if (mQuitting) {
dispose();
return null;
}
// If first time idle, then get the number of idlers to run.
// Idle handles only run if the queue is empty or if the first message
// in the queue (possibly a barrier) is due to be handled in the future.
/*
* 两个条件：
* 1. pendingIdleHandlerCount = -1
* 2. 此次取到的 mMessage 为空或者需要延迟处理
*/
if (pendingIdleHandlerCount < 0
&& (mMessages == null || now < mMessages.when)) {
pendingIdleHandlerCount = mIdleHandlers.size();
}
if (pendingIdleHandlerCount <= 0) {
// No idle handlers to run. Loop and wait some more.
// 没有 idle handler 需要运行，继续循环
mBlocked = true;
continue;
}
if (mPendingIdleHandlers == null) {
mPendingIdleHandlers = new IdleHandler[Math.max(pendingIdleHandlerCount, 4)];
}
mPendingIdleHandlers = mIdleHandlers.toArray(mPendingIdleHandlers);
}
// Run the idle handlers.
// We only ever reach this code block during the first iteration.
// 下一次 next 时，pendingIdleHandlerCount 又会被置为 -1，不会导致死循环
for (int i = 0; i < pendingIdleHandlerCount; i++) {
final IdleHandler idler = mPendingIdleHandlers[i];
mPendingIdleHandlers[i] = null; // release the reference to the handler
boolean keep = false;
try {
// 执行 Idler
keep = idler.queueIdle();
} catch (Throwable t) {
Log.wtf(TAG, "IdleHandler threw exception", t);
}
if (!keep) {
synchronized (this) {
mIdleHandlers.remove(idler);
}
}
}
// Reset the idle handler count to 0 so we do not run them again.
// 将 pendingIdleHandlerCount 置零
pendingIdleHandlerCount = 0;
// While calling an idle handler, a new message could have been delivered
// so go back and look again for a pending message without waiting.
nextPollTimeoutMillis = 0;
}
}
在正常的消息处理机制之后，额外对 IdleHandler 进行了处理。当本次取到的 Message 为空或者需要延时处理的时候，就会去执行 mIdleHandlers 数组中的 IdleHandler 对象。其中还有一些关于 pendingIdleHandlerCount 的额外逻辑来防止循环处理。
所以，不出意外的话，当新的 Activity 完成页面绘制并显示之后，主线程就可以停下歇一歇，来执行 IdleHandler 了。再回来 handleResumeActivity() 中来，Looper.myQueue().addIdleHandler(new Idler()) ，这里的 Idler 是 IdleHandler 的一个具体实现类。
>ActivityThread.java
private class Idler implements MessageQueue.IdleHandler {
@Override
public final boolean queueIdle() {
ActivityClientRecord a = mNewActivities;
boolean stopProfiling = false;
if (mBoundApplication != null && mProfiler.profileFd != null
&& mProfiler.autoStopProfiler) {
stopProfiling = true;
}
if (a != null) {
mNewActivities = null;
IActivityTaskManager am = ActivityTaskManager.getService();
ActivityClientRecord prev;
do {
if (localLOGV) Slog.v(
TAG, "Reporting idle of " + a +
" finished=" +
(a.activity != null && a.activity.mFinished));
if (a.activity != null && !a.activity.mFinished) {
try {
// 调用 ActivityTaskManagerService.activityIdle()
am.activityIdle(a.token, a.createdConfig, stopProfiling);
a.createdConfig = null;
} catch (RemoteException ex) {
throw ex.rethrowFromSystemServer();
}
}
prev = a;
a = a.nextIdle;
prev.nextIdle = null;
} while (a != null);
}
if (stopProfiling) {
mProfiler.stopProfiling();
}
applyPendingProcessState();
return false;
}
}
Binder 调用了 ActivityTaskManagerService.activityIdle() 。
>ActivityTaskManagerService.java
@Override
public final void activityIdle(IBinder token, Configuration config, boolean stopProfiling) {
final long origId = Binder.clearCallingIdentity();
try {
WindowProcessController proc = null;
synchronized (mGlobalLock) {
ActivityStack stack = ActivityRecord.getStackLocked(token);
if (stack == null) {
return;
}
final ActivityRecord r = mStackSupervisor.activityIdleInternalLocked(token,
false /* fromTimeout */, false /* processPausingActivities */, config);
if (r != null) {
proc = r.app;
}
if (stopProfiling && proc != null) {
proc.clearProfilerIfNeeded();
}
}
} finally {
Binder.restoreCallingIdentity(origId);
}
}
调用了 ActivityStackSupervisor.activityIdleInternalLocked() 方法。
>ActivityStackSupervisor.java
final ActivityRecord activityIdleInternalLocked(final IBinder token, boolean fromTimeout,
boolean processPausingActivities, Configuration config) {
if (DEBUG_ALL) Slog.v(TAG, "Activity idle: " + token);
ArrayList finishes = null;
ArrayList startingUsers = null;
int NS = 0;
int NF = 0;
boolean booting = false;
boolean activityRemoved = false;
ActivityRecord r = ActivityRecord.forTokenLocked(token);
if (r != null) {
if (DEBUG_IDLE) Slog.d(TAG_IDLE, "activityIdleInternalLocked: Callers="
+ Debug.getCallers(4));
mHandler.removeMessages(IDLE_TIMEOUT_MSG, r);
r.finishLaunchTickingLocked();
if (fromTimeout) {
reportActivityLaunchedLocked(fromTimeout, r, INVALID_DELAY,
-1 /* launchState */);
}
// This is a hack to semi-deal with a race condition
// in the client where it can be constructed with a
// newer configuration from when we asked it to launch.
// We'll update with whatever configuration it now says
// it used to launch.
if (config != null) {
r.setLastReportedGlobalConfiguration(config);
}
// We are now idle. If someone is waiting for a thumbnail from
// us, we can now deliver.
r.idle = true;
//Slog.i(TAG, "IDLE: mBooted=" + mBooted + ", fromTimeout=" + fromTimeout);
// Check if able to finish booting when device is booting and all resumed activities
// are idle.
if ((mService.isBooting() && mRootActivityContainer.allResumedActivitiesIdle())
|| fromTimeout) {
booting = checkFinishBootingLocked();
}
// When activity is idle, we consider the relaunch must be successful, so let's clear
// the flag.
r.mRelaunchReason = RELAUNCH_REASON_NONE;
}
if (mRootActivityContainer.allResumedActivitiesIdle()) {
if (r != null) {
mService.scheduleAppGcsLocked();
}
if (mLaunchingActivityWakeLock.isHeld()) {
mHandler.removeMessages(LAUNCH_TIMEOUT_MSG);
if (VALIDATE_WAKE_LOCK_CALLER &&
Binder.getCallingUid() != Process.myUid()) {
throw new IllegalStateException("Calling must be system uid");
}
mLaunchingActivityWakeLock.release();
}
mRootActivityContainer.ensureActivitiesVisible(null, 0, !PRESERVE_WINDOWS);
}
// Atomically retrieve all of the other things to do.
// 获取要 stop 的 Activity
final ArrayList stops = processStoppingActivitiesLocked(r,
true /* remove */, processPausingActivities);
NS = stops != null ? stops.size() : 0;
if ((NF = mFinishingActivities.size()) > 0) {
finishes = new ArrayList<>(mFinishingActivities);
mFinishingActivities.clear();
}
if (mStartingUsers.size() > 0) {
startingUsers = new ArrayList<>(mStartingUsers);
mStartingUsers.clear();
}
// Stop any activities that are scheduled to do so but have been
// waiting for the next one to start.
// 该 stop 的 stop 掉
for (int i = 0; i < NS; i++) {
r = stops.get(i);
final ActivityStack stack = r.getActivityStack();
if (stack != null) {
if (r.finishing) {
stack.finishCurrentActivityLocked(r, ActivityStack.FINISH_IMMEDIATELY, false,
"activityIdleInternalLocked");
} else {
stack.stopActivityLocked(r);
}
}
}
// Finish any activities that are scheduled to do so but have been
// waiting for the next one to start.
// 该 destroy 的 destroy 掉
for (int i = 0; i < NF; i++) {
r = finishes.get(i);
final ActivityStack stack = r.getActivityStack();
if (stack != null) {
activityRemoved |= stack.destroyActivityLocked(r, true, "finish-idle");
}
}
if (!booting) {
// Complete user switch
if (startingUsers != null) {
for (int i = 0; i < startingUsers.size(); i++) {
mService.mAmInternal.finishUserSwitch(startingUsers.get(i));
}
}
}
mService.mH.post(() -> mService.mAmInternal.trimApplications());
//dump();
//mWindowManager.dump();
if (activityRemoved) {
mRootActivityContainer.resumeFocusedStacksTopActivities();
}
return r;
}
stops 和 finishes 分别是要 stop 和 destroy 的两个 ActivityRecord 数组。stops 数组是通过 ActivityStackSuperVisor.processStoppingActivitiesLocked() 方法获取的，追进去看一下。
>ActivityStackSuperVisor.java
final ArrayList processStoppingActivitiesLocked(ActivityRecord idleActivity,
boolean remove, boolean processPausingActivities) {
ArrayList stops = null;
final boolean nowVisible = mRootActivityContainer.allResumedActivitiesVisible();
// 遍历 mStoppingActivities
for (int activityNdx = mStoppingActivities.size() - 1; activityNdx >= 0; --activityNdx) {
ActivityRecord s = mStoppingActivities.get(activityNdx);
final boolean animating = s.mAppWindowToken.isSelfAnimating();
if (DEBUG_STATES) Slog.v(TAG, "Stopping " + s + ": nowVisible=" + nowVisible
+ " animating=" + animating + " finishing=" + s.finishing);
if (nowVisible && s.finishing) {
// If this activity is finishing, it is sitting on top of
// everyone else but we now know it is no longer needed...
// so get rid of it. Otherwise, we need to go through the
// normal flow and hide it once we determine that it is
// hidden by the activities in front of it.
if (DEBUG_STATES) Slog.v(TAG, "Before stopping, can hide: " + s);
s.setVisibility(false);
}
if (remove) {
final ActivityStack stack = s.getActivityStack();
final boolean shouldSleepOrShutDown = stack != null
? stack.shouldSleepOrShutDownActivities()
: mService.isSleepingOrShuttingDownLocked();
if (!animating || shouldSleepOrShutDown) {
if (!processPausingActivities && s.isState(PAUSING)) {
// Defer processing pausing activities in this iteration and reschedule
// a delayed idle to reprocess it again
removeTimeoutsForActivityLocked(idleActivity);
scheduleIdleTimeoutLocked(idleActivity);
continue;
}
if (DEBUG_STATES) Slog.v(TAG, "Ready to stop: " + s);
if (stops == null) {
stops = new ArrayList<>();
}
stops.add(s);
mStoppingActivities.remove(activityNdx);
}
}
}
return stops;
}
中间的详细处理逻辑就不看了，我们只需要关注这里遍历的是 ActivityStackSuperVisor 中的 mStoppingActivities 集合 。在前面分析 finish() 流程到最后的 addToStopping() 方法时提到过，
这些在等待销毁的 Activity 被保存在了
ActivityStackSupervisor的mStoppingActivities集合中，它是一个ArrayList。
看到这里，终于打通了流程。再回头想一下文章开头的例子，由于人为的在 SecondActivity 不间断的向主线程塞消息，导致 Idler 迟迟无法被执行，按理说 onStop/onDestroy 也就不会被回调。可实际情况是这样吗？并不是，明明是过了 10s 被回调。这就说明了即使主线程迟迟没有机会执行 Idler，系统仍然提供了兜底机制，防止已经不需要的 Activity 长时间无法被回收，从而造成内存泄漏等问题。从实际现象就可以猜测到，这个兜底机制就是 onResume 之后 10s 主动去进行释放操作。
再回到之前显示待跳转 Activity 的 ActivityStackSuperVisor.resumeFocusedStackTopActivityLocked() 方法。我这里就不带着大家追进去了，直接给出调用链。
ASS.resumeFocusedStackTopActivityLocked() -> ActivityStack.resumeTopActivityUncheckedLocked() -> ActivityStack.resumeTopActivityInnerLocked() -> ActivityRecord.completeResumeLocked() -> ASS.scheduleIdleTimeoutLocked()
> ActivityStackSuperVisor.java
void scheduleIdleTimeoutLocked(ActivityRecord next) {
if (DEBUG_IDLE) Slog.d(TAG_IDLE,
"scheduleIdleTimeoutLocked: Callers=" + Debug.getCallers(4));
Message msg = mHandler.obtainMessage(IDLE_TIMEOUT_MSG, next);
mHandler.sendMessageDelayed(msg, IDLE_TIMEOUT);
}
IDLE_TIMEOUT 的值是 10，这里延迟 10s 发送了一个消息。这个消息是在 ActivityStackSupervisorHandler 中处理的。
private final class ActivityStackSupervisorHandler extends Handler {
public ActivityStackSupervisorHandler(Looper looper) {
super(looper);
}
void activityIdleInternal(ActivityRecord r, boolean processPausingActivities) {
synchronized (mService.mGlobalLock) {
//进入此处
activityIdleInternalLocked(r != null ? r.appToken : null, true /* fromTimeout */,
processPausingActivities, null);
}
}
@Override
public void handleMessage(Message msg) {
switch (msg.what) {
case REPORT_MULTI_WINDOW_MODE_CHANGED_MSG: {
synchronized (mService.mGlobalLock) {
for (int i = mMultiWindowModeChangedActivities.size() - 1; i >= 0; i--) {
final ActivityRecord r = mMultiWindowModeChangedActivities.remove(i);
r.updateMultiWindowMode();
}
}
} break;
case REPORT_PIP_MODE_CHANGED_MSG: {
synchronized (mService.mGlobalLock) {
for (int i = mPipModeChangedActivities.size() - 1; i >= 0; i--) {
final ActivityRecord r = mPipModeChangedActivities.remove(i);
r.updatePictureInPictureMode(mPipModeChangedTargetStackBounds,
false /* forceUpdate */);
}
}
} break;
case IDLE_TIMEOUT_MSG: { //进入此处
if (DEBUG_IDLE) Slog.d(TAG_IDLE,
"handleMessage: IDLE_TIMEOUT_MSG: r=" + msg.obj);
// We don't at this point know if the activity is fullscreen,
// so we need to be conservative and assume it isn't.
activityIdleInternal((ActivityRecord) msg.obj,
true /* processPausingActivities */);
} break;
case IDLE_NOW_MSG: {
if (DEBUG_IDLE) Slog.d(TAG_IDLE, "handleMessage: IDLE_NOW_MSG: r=" + msg.obj);
activityIdleInternal((ActivityRecord) msg.obj,
false /* processPausingActivities */);
} break;
case RESUME_TOP_ACTIVITY_MSG: {
synchronized (mService.mGlobalLock) {
mRootActivityContainer.resumeFocusedStacksTopActivities();
}
} break;
case SLEEP_TIMEOUT_MSG: {
synchronized (mService.mGlobalLock) {
if (mService.isSleepingOrShuttingDownLocked()) {
Slog.w(TAG, "Sleep timeout! Sleeping now.");
checkReadyForSleepLocked(false /* allowDelay */);
}
}
} break;
case LAUNCH_TIMEOUT_MSG: {
synchronized (mService.mGlobalLock) {
if (mLaunchingActivityWakeLock.isHeld()) {
Slog.w(TAG, "Launch timeout has expired, giving up wake lock!");
if (VALIDATE_WAKE_LOCK_CALLER
&& Binder.getCallingUid() != Process.myUid()) {
throw new IllegalStateException("Calling must be system uid");
}
mLaunchingActivityWakeLock.release();
}
}
} break;
case LAUNCH_TASK_BEHIND_COMPLETE: {
synchronized (mService.mGlobalLock) {
ActivityRecord r = ActivityRecord.forTokenLocked((IBinder) msg.obj);
if (r != null) {
handleLaunchTaskBehindCompleteLocked(r);
}
}
} break;
case RESTART_ACTIVITY_PROCESS_TIMEOUT_MSG: {
final ActivityRecord r = (ActivityRecord) msg.obj;
String processName = null;
int uid = 0;
synchronized (mService.mGlobalLock) {
if (r.attachedToProcess()
&& r.isState(ActivityStack.ActivityState.RESTARTING_PROCESS)) {
processName = r.app.mName;
uid = r.app.mUid;
}
}
if (processName != null) {
mService.mAmInternal.killProcess(processName, uid,
"restartActivityProcessTimeout");
}
} break;
case REPORT_HOME_CHANGED_MSG: {
synchronized (mService.mGlobalLock) {
mHandler.removeMessages(REPORT_HOME_CHANGED_MSG);
// Start home activities on displays with no activities.
mRootActivityContainer.startHomeOnEmptyDisplays("homeChanged");
}
} break;
case TOP_RESUMED_STATE_LOSS_TIMEOUT_MSG: {
ActivityRecord r = (ActivityRecord) msg.obj;
Slog.w(TAG, "Activity top resumed state loss timeout for " + r);
synchronized (mService.mGlobalLock) {
if (r.hasProcess()) {
mService.logAppTooSlow(r.app, r.topResumedStateLossTime,
"top state loss for " + r);
}
}
handleTopResumedStateReleased(true /* timeout */);
} break;
}
}
}
忘记 activityIdleInternalLocked 方法的话可以 ctrl+F 向上搜索一下。如果 10s 内主线程执行了 Idler 的话，就会移除这个消息。
到这里，所有的问题就全部理清了。
Activity 的 onStop/onDestroy 是依赖 IdleHandler 来回调的，正常情况下当主线程空闲时会调用。但是由于某些特殊场景下的问题，导致主线程迟迟无法空闲，onStop/onDestroy 也会迟迟得不到调用。但这并不意味着 Activity 永远得不到回收，系统提供了一个兜底机制，当 onResume 回调 10s 之后，如果仍然没有得到调用，会主动触发。
虽然有兜底机制，但无论如何这肯定不是我们想看到的。如果我们项目中的 onStop/onDestroy 延迟了 10s 调用，该如何排查问题呢？可以利用 Looper.getMainLooper().setMessageLogging() 方法，打印出主线程消息队列中的消息。每处理一条消息，都会打印如下内容：
logging.println(">>>>> Dispatching to " + msg.target + " " + msg.callback + ": " + msg.what);
logging.println("<<<<< Finished to " + msg.target + " " + msg.callback);
此日志消息来源于
>Looper.java
public static void loop() {
final Looper me = myLooper();
if (me == null) {
throw new RuntimeException("No Looper; Looper.prepare() wasn't called on this thread.");
}
final MessageQueue queue = me.mQueue;
// Make sure the identity of this thread is that of the local process,
// and keep track of what that identity token actually is.
Binder.clearCallingIdentity();
final long ident = Binder.clearCallingIdentity();
// Allow overriding a threshold with a system prop. e.g.
// adb shell 'setprop log.looper.1000.main.slow 1 && stop && start'
final int thresholdOverride =
SystemProperties.getInt("log.looper."
+ Process.myUid() + "."
+ Thread.currentThread().getName()
+ ".slow", 0);
boolean slowDeliveryDetected = false;
for (;;) {
Message msg = queue.next(); // might block
if (msg == null) {
// No message indicates that the message queue is quitting.
return;
}
// This must be in a local variable, in case a UI event sets the logger
final Printer logging = me.mLogging;
if (logging != null) {
//获取主线程 Dispatching to 该 message 的日志
logging.println(">>>>> Dispatching to " + msg.target + " " +
msg.callback + ": " + msg.what);
}
// Make sure the observer won't change while processing a transaction.
final Observer observer = sObserver;
final long traceTag = me.mTraceTag;
long slowDispatchThresholdMs = me.mSlowDispatchThresholdMs;
long slowDeliveryThresholdMs = me.mSlowDeliveryThresholdMs;
if (thresholdOverride > 0) {
slowDispatchThresholdMs = thresholdOverride;
slowDeliveryThresholdMs = thresholdOverride;
}
final boolean logSlowDelivery = (slowDeliveryThresholdMs > 0) && (msg.when > 0);
final boolean logSlowDispatch = (slowDispatchThresholdMs > 0);
final boolean needStartTime = logSlowDelivery || logSlowDispatch;
final boolean needEndTime = logSlowDispatch;
if (traceTag != 0 && Trace.isTagEnabled(traceTag)) {
Trace.traceBegin(traceTag, msg.target.getTraceName(msg));
}
final long dispatchStart = needStartTime ? SystemClock.uptimeMillis() : 0;
final long dispatchEnd;
Object token = null;
if (observer != null) {
token = observer.messageDispatchStarting();
}
long origWorkSource = ThreadLocalWorkSource.setUid(msg.workSourceUid);
try {
msg.target.dispatchMessage(msg);
if (observer != null) {
observer.messageDispatched(token, msg);
}
dispatchEnd = needEndTime ? SystemClock.uptimeMillis() : 0;
} catch (Exception exception) {
if (observer != null) {
observer.dispatchingThrewException(token, msg, exception);
}
throw exception;
} finally {
ThreadLocalWorkSource.restore(origWorkSource);
if (traceTag != 0) {
Trace.traceEnd(traceTag);
}
}
if (logSlowDelivery) {
if (slowDeliveryDetected) {
if ((dispatchStart - msg.when) <= 10) {
Slog.w(TAG, "Drained");
slowDeliveryDetected = false;
}
} else {
if (showSlowLog(slowDeliveryThresholdMs, msg.when, dispatchStart, "delivery",
msg)) {
// Once we write a slow delivery log, suppress until the queue drains.
slowDeliveryDetected = true;
}
}
}
if (logSlowDispatch) {
showSlowLog(slowDispatchThresholdMs, dispatchStart, dispatchEnd, "dispatch", msg);
}
if (logging != null) {
//获取主线程 Finished to 该 message 的日志
logging.println("<<<<< Finished to " + msg.target + " " + msg.callback);
}
// Make sure that during the course of dispatching the
// identity of the thread wasn't corrupted.
final long newIdent = Binder.clearCallingIdentity();
if (ident != newIdent) {
Log.wtf(TAG, "Thread identity changed from 0x"
+ Long.toHexString(ident) + " to 0x"
+ Long.toHexString(newIdent) + " while dispatching to "
+ msg.target.getClass().getName() + " "
+ msg.callback + " what=" + msg.what);
}
msg.recycleUnchecked();
}
}
另外，由于 onStop/onDestroy 调用时机的不确定性，在做资源释放等操作的时候，一定要考虑好，以避免产生资源没有及时释放的情况。
本文参考链接：https://juejin.cn/post/6898588053451833351
最后，感兴趣的同学可以顺着目前最新的 Android 11.0(R) 的 AOSP 中，去走一遍这个流程看看。