Question

I have an iPhone app that freezes sometimes when saving CoreData and then doesn’t relaunch. I did have a second thread that uses the database, but I think I have followed the pattern to make a separate context for that thread. Here is the crash report from the relaunch. Any ideas?

I have tried changing it to only run with one thread and here is the latest freeze point after entering the background.

#0  0x30851b98 in fsync
#1  0x3094e694 in _sqlite3_purgeEligiblePagerCacheMemory
#2  0x3094e6b8 in _sqlite3_purgeEligiblePagerCacheMemory
#3  0x30945372 in sqlite3_compileoption_get
#4  0x30957f06 in sqlite3_extended_errcode
#5  0x3095dc20 in sqlite3_extended_errcode
#6  0x3095dd8e in sqlite3_extended_errcode
#7  0x309646f8 in sqlite3_clear_bindings
#8  0x3098845a in sqlite3_open16
#9  0x3094495a in sqlite3_step
#10 0x31a1dc20 in _execute
#11 0x31acc6e8 in -[NSSQLiteConnection commitTransaction]
#12 0x31aca646 in -[NSSQLiteConnection endPrimaryKeyGeneration]
#13 0x31abeab4 in -[NSSQLCore prepareForSave:]
#14 0x31a4acd0 in -[NSSQLCore saveChanges:]
#15 0x31a1591e in -[NSSQLCore executeRequest:withContext:error:]
#16 0x31a1538a in -[NSPersistentStoreCoordinator executeRequest:withContext:error:]
#17 0x31a48544 in -[NSManagedObjectContext save:]
#18 0x000080aa in -[KPersistence saveManagedObjects:] at KPersistence.m:242
#19 0x00004320 in -[KinKastAppDelegate applicationDidEnterBackground:] at KinKastAppDelegate.m:126

Here is my implementation of saveManagedObjects

-(BOOL)saveManagedObjects:(NSError **)error
{
    [persistentStoreCoordinator lock];
    BOOL success = YES;
    if (managedObjectContext != nil) {
        if ([managedObjectContext hasChanges] && ![managedObjectContext save:error]) {
            VLog(@"Unresolved error %@, %@", *error, [*error userInfo]);
            success = NO;
        } 
    }
    [persistentStoreCoordinator unlock];
    return success;
}

Answer 1

So I now believe the problem was that I was sometimes writing an NSNumber with float value to an integer field. This caused the copy in memory to be different than the copy read from the database, and this caused the infinite merge loop. But if you restarted the app it worked fine because the value was simply an integer from then on.