ADHOC
/
replicant-packages_apps_Email
3
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Shut down Email process when sync is totally blocked 

* When sync threads get blocked by SSL/Socket code, we detect this
  and shutdown the ClientConnectionManager, hoping this will solve
  the problem.
* Apparently, this doesn't always work.  The result is that syncs
  are frozen until the device is rebooted (certainly) or until the
  Email process is stopped and restarted (hopefully)
* This CL kills the Email process after an unsuccessful attempt to
  clear blocked threads using ClientConnectionManager.shutdown()

Bug: 2615293
Change-Id: I8445b278817e508a855adeb88d6b2f969d28858b
This commit is contained in:
Marc Blank 2010-05-06 11:10:16 -07:00
parent 8f255ddeee
commit 3558e5f227
1 changed files with 18 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
src/com/android/exchange/SyncManager.java
View File

@ -76,6 +76,7 @@ import android.os.Debug;
import android.os.Handler;
import android.os.IBinder;
import android.os.PowerManager;
import android.os.Process;
import android.os.RemoteCallbackList;
import android.os.RemoteException;
import android.os.PowerManager.WakeLock;
@ -174,6 +175,8 @@ public class SyncManager extends Service implements Runnable {
// Service had a fatal error; can't run
public static final int PING_STATUS_UNABLE = 3;
private static final int MAX_CLIENT_CONNECTION_MANAGER_SHUTDOWNS = 1;
// We synchronize on this for all actions affecting the service and error maps
private static final Object sSyncLock = new Object();
// All threads can use this lock to wait for connectivity
@ -215,6 +218,8 @@ public class SyncManager extends Service implements Runnable {
private static String sDeviceId = null;
// ConnectionManager that all EAS threads can use
private static ClientConnectionManager sClientConnectionManager = null;
// Count of ClientConnectionManager shutdowns
private static volatile int sClientConnectionManagerShutdownCount = 0;
private boolean mStop = false;
@ -1162,6 +1167,12 @@ public class SyncManager extends Service implements Runnable {
static public synchronized ClientConnectionManager getClientConnectionManager() {
if (sClientConnectionManager == null) {
// After two tries, kill the process. Most likely, this will happen in the background
// The service will restart itself after about 5 seconds
if (sClientConnectionManagerShutdownCount > MAX_CLIENT_CONNECTION_MANAGER_SHUTDOWNS) {
alwaysLog("Shutting down process to unblock threads");
Process.killProcess(Process.myPid());
}
// Create a registry for our three schemes; http and https will use built-in factories
SchemeRegistry registry = new SchemeRegistry();
registry.register(new Scheme("http",
@ -1185,7 +1196,9 @@ public class SyncManager extends Service implements Runnable {
static private synchronized void shutdownConnectionManager() {
if (sClientConnectionManager != null) {
alwaysLog("Shutting down ClientConnectionManager");
sClientConnectionManager.shutdown();
sClientConnectionManagerShutdownCount++;
sClientConnectionManager = null;
}
}
@ -1465,7 +1478,7 @@ public class SyncManager extends Service implements Runnable {
}
// Shutdown the connection manager; this should close all of our
// sockets and generate IOExceptions all around.
syncManager.shutdownConnectionManager();
SyncManager.shutdownConnectionManager();
}
}
}}).start();
@ -2295,6 +2308,10 @@ public class SyncManager extends Service implements Runnable {
// TODO Handle this case
}
errorMap.remove(mailboxId);
// If we've had a successful sync, clear the shutdown count
synchronized (SyncManager.class) {
sClientConnectionManagerShutdownCount = 0;
}
break;
// I/O errors get retried at increasing intervals
case AbstractSyncService.EXIT_IO_ERROR: