我在iOS 8.1上遇到CoreData并发问题。
我正在为崩溃获得以下堆栈跟踪:
NSInternalInconsistencyException - Invalid rowCache row is nil
0 CoreFoundation 0x0000000183b6659c __exceptionPreprocess + 132
1 libobjc.A.dylib 0x00000001942640e4 objc_exception_throw + 56
2 CoreData 0x000000018385b8b8 -[NSSQLCore _newRowCacheRowForToManyUpdatesForRelationship:rowCacheOriginal:originalSnapshot:value:added:deleted:sourceRowPK:properties:sourceObject:newIndexes:reorderedIndexes:] + 6668
3 CoreData 0x00000001838fbea0 -[NSSQLCore recordToManyChangesForObject:inRow:usingTimestamp:inserted:] + 2604
4 CoreData 0x0000000183857638 -[NSSQLCore prepareForSave:] + 1052
5 CoreData 0x00000001838569b4 -[NSSQLCore saveChanges:] + 520
6 CoreData 0x000000018381f078 -[NSSQLCore executeRequest:withContext:error:] + 716
7 CoreData 0x00000001838e6254 __65-[NSPersistentStoreCoordinator executeRequest:withContext:error:]_block_invoke + 4048
8 CoreData 0x00000001838ed654 gutsOfBlockToNSPersistentStoreCoordinatorPerform + 176
9 libdispatch.dylib 0x00000001948a936c _dispatch_client_callout + 12
10 libdispatch.dylib 0x00000001948b26e8 _dispatch_barrier_sync_f_invoke + 72
11 CoreData 0x00000001838e0cb4 _perform + 176
12 CoreData 0x000000018381ec34 -[NSPersistentStoreCoordinator executeRequest:withContext:error:] + 296
13 CoreData 0x0000000183845400 -[NSManagedObjectContext save:] + 1280
这种崩溃发生在几个地方(大约20个),但在我的数据导入功能中最为突出,它导入1000个记录并保存每100个。
关于我的CoreData设置的一些注意事项:
performBlock:
发生所有这些保存的代码具有以下基本模式:
[backgroundContext performBlock:^{
...
NSArray *result = [backgroundContext fetch...];
...
if ([backgroundContext save:&error]) { // <-- App is crashing here
}
}];
据我了解,NSPersistentStore
背后的backgroundContext
不应该出现任何并发问题,但这就是崩溃告诉我的内容。
此外,这仅发生在我的用户群的不到0.02%。这是非常罕见的(我无法重现)但是用户不能够从中恢复而不删除并重新安装应用程序。它将一直为他们打开并崩溃。
请注意,这仅适用于64位iOS 8.1.X - iOS 7.X和8.0.X不会出现此行为,我在iPhone 5s之前的任何版本都没有看到它。
澄清:删除持久性商店可以解决所有用户的此问题。这似乎表明它不是并发问题。
创建商店&amp;上下文。
[_persistenStoreCoordinator addPersistentStoreWithType:NSSQLiteStoreType
configuration:nil
URL:[DatabaseManager storeURL]
options:@{NSMigratePersistentStoresAutomaticallyOption:@YES,
NSInferMappingModelAutomaticallyOption:@YES}
error:&error];
创建上下文
_backgroundContext = [[NSManagedObjectContext alloc] initWithConcurrencyType:NSPrivateQueueConcurrencyType];
[_backgroundContext setMergePolicy:NSMergeByPropertyObjectTrumpMergePolicy];
[_backgroundContext setPersistentStoreCoordinator:_persistenStoreCoordinator];
if ([_backgroundContext respondsToSelector:@selector(setName:)]) {
[_backgroundContext setName:@"DatabaseManager.BackgroundQueue"];
}
_mainContext = [[NSManagedObjectContext alloc] initWithConcurrencyType:NSMainQueueConcurrencyType];
[_mainContext setMergePolicy:NSMergeByPropertyObjectTrumpMergePolicy];
[_mainContext setPersistentStoreCoordinator:_persistenStoreCoordinator];
if ([_mainContext respondsToSelector:@selector(setName:)]) {
[_mainContext setName:@"DatabaseManager.MainQueue"];
}
[[NSNotificationCenter defaultCenter] addObserver:self
selector:@selector(mainContextDidSave:)
name:NSManagedObjectContextDidSaveNotification
object:_mainContext];
[[NSNotificationCenter defaultCenter] addObserver:self
selector:@selector(backgroundContextDidSave:)
name:NSManagedObjectContextDidSaveNotification
object:_backgroundContext];
倾听变化
- (void) mainContextDidSave:(NSNotification *)notification
{
[_backgroundContext performBlock:^{
[self->_backgroundContext mergeChangesFromContextDidSaveNotification:notification];
}];
}
- (void) backgroundContextDidSave:(NSNotification*)notification
{
[_mainContext performBlock:^{
NSArray* updated = [notification.userInfo valueForKey:NSUpdatedObjectsKey];
// Fault all objects that will be updated.
for (NSManagedObject* obj in updated) {
NSManagedObject* mainThreadObject = [self->_mainContext objectWithID:obj.objectID];
[mainThreadObject willAccessValueForKey:nil];
}
[self->_mainContext mergeChangesFromContextDidSaveNotification:notification];
}];
}
如何在每个上下文上执行代码。我把一个街区传到这里:
- (void)executeBackgroundOperation:(void (^)(NSManagedObjectContext *))operation {
NSAssert(operation, @"No Background operation to perform");
[_backgroundContext performBlock:^{
operation(self->_backgroundContext);
}];
}
- (void)executeMainThreadOperation:(void (^)(NSManagedObjectContext *))operation {
NSAssert(operation, @"No Main Thread operation to perform");
[_mainContext performBlock:^{
operation(self->_mainContext);
}];
}
崩溃报告中唯一包含lock
的其他线程是两个Javascript线程,如下所示:
0 libsystem_kernel.dylib 0x3a77cb38 __psynch_cvwait + 24
1 libsystem_pthread.dylib 0x3a7fa2dd pthread_cond_wait + 38
2 libc++.1.dylib 0x39a11e91 _ZNSt3__118condition_variable4waitERNS_11unique_lockINS_5mutexEEE + 34
3 JavaScriptCore 0x2dcd4cb5 _ZN3JSC8GCThread16waitForNextPhaseEv + 102
4 JavaScriptCore 0x2dcd4d19 _ZN3JSC8GCThread12gcThreadMainEv + 50
5 JavaScriptCore 0x2db09597 _ZN3WTFL19wtfThreadEntryPointEPv + 12
6 libsystem_pthread.dylib 0x3a7f9e93 _pthread_body + 136
7 libsystem_pthread.dylib 0x3a7f9e07 _pthread_start + 116
8 libsystem_pthread.dylib 0x3a7f7b90 thread_start + 6
答案 0 :(得分:3)
您认为这可能不是并发问题是正确的。删除和安装应用程序通常可以在迁移未正确完成时修复核心数据错误。
如果您发布了应用程序,更改了核心数据模型,然后发布了更新(没有迁移),这可能是您崩溃的原因。
按照本教程查看是否可以解决您的问题。 http://www.raywenderlich.com/27657/how-to-perform-a-lightweight-core-data-migration
答案 1 :(得分:0)
@StephenFurlani,
我不是Core-data的专家。但我喜欢解决问题。我认为app将managedObject从backgroundContext合并到MainContext时崩溃了(正如你提到的代码):
- (void) backgroundContextDidSave:(NSNotification*)notification
{
[_mainContext performBlock:^{
NSArray* updated = [notification.userInfo valueForKey:NSUpdatedObjectsKey];
// Fault all objects that will be updated.
for (NSManagedObject* obj in updated) {
NSManagedObject* mainThreadObject = [self->_mainContext objectWithID:obj.objectID];
[mainThreadObject willAccessValueForKey:nil];
}
[self->_mainContext mergeChangesFromContextDidSaveNotification:notification];
}];
}
以下是一些参考链接:
解决方案1 :请尝试以下代码:
NSMangedObjectContext *temporaryContext = [[NSManagedObjectContext alloc] initWithConcurrencyType:NSPrivateQueueConcurrencyType];
temporaryContext.parentContext = mainMOC;
[temporaryContext performBlock:^{
// do something that takes some time asynchronously using the temp context
// push to parent
NSError *error;
if (![temporaryContext save:&error])
{
// handle error
}
// save parent to disk asynchronously
[mainMOC performBlock:^{
NSError *error;
if (![mainMOC save:&error])
{
// handle error
}
}];
}];
解决方案2:从backgroundContext的NSManagedObject刷新值到mainContext,然后调用mainContext的processPendingChanges或mergeChangesFromContextDidSaveNotification。 尝试使用此代码而不是代码(backgroundContextDidSave方法):
- (void) backgroundContextDidSave:(NSNotification*)notification
{
[_mainContext performBlock:^{
NSArray* updated = [notification.userInfo valueForKey:NSUpdatedObjectsKey];
// Fault all objects that will be updated.
for (NSManagedObject* obj in updated) {
NSManagedObject* mainThreadObject = [self->_mainContext objectWithID:obj.objectID];
if (mainThreadObject) {
[self->_mainContext refreshObject: mainThreadObject mergeChanges:NO];
}
}
[self->_mainContext processPendingChanges];
}];
}
非常好的帮助 - 完整链接: CORE DATA with multiple managed object contexts
答案 2 :(得分:0)
我遇到了类似的崩溃,不一样,但我也在使用Stack#3设置。在从后台保存期间,崩溃发生在与您相同的行上。我发现原因是当CoreData操作旋转多个线程时,线程以错误的顺序完成,但有时只是。对我们来说这是一个问题,因为我们在模型中有一些依赖操作和对象关系。
我通过管理自己的线程和我自己的队列进行后台下载来修复它。通过这种方式,我可以处理订单并始终知道我有一个FIFO类型的流程。这也修复了保存的崩溃。
如果您想尝试一下,我在下面添加了一些代码。
- (void) keepAlive
{
[self performSelector:@selector(keepAlive) withObject:nil afterDelay:60];
}
- (void)backgroundThreadMain
{
// Add selector to prevent CFRunLoopRunInMode from returning immediately
[self performSelector:@selector(keepAlive) withObject:nil afterDelay:60];
BOOL done = NO;
do
{
// Start the run loop but return after each source is handled.
SInt32 result = CFRunLoopRunInMode(kCFRunLoopDefaultMode, 10, YES);
// If a source explicitly stopped the run loop, or if there are no
// sources or timers, go ahead and exit.
if ((result == kCFRunLoopRunStopped) || (result == kCFRunLoopRunFinished))
done = YES;
}
while (!done);
}
self.backgroundThread = [[NSThread alloc] initWithTarget:self selector:@selector(backgroundThreadMain) object:nil];
[self.backgroundThread start];
然后,只要我需要运行一段代码,我就会使用对该线程的静态引用。
[[CoreDataControl coreDataControlManager] performSelector:@selector(saveSite:) onThread:[CoreDataControl coreDataControlManager].backgroundThread withObject:site waitUntilDone:NO];
最坏的情况是,您可以将保存包装在try catch块中,如果抛出异常,则在两个上下文中调用reset。这将清除可能导致错误的任何rowcache。