FetchedResultsController in TCA

[lightandshadow68]

Has anyone found a way to successfully leverage CoreData's FetchedResultsController in a dependency?

I've created a custom combine publisher to return an array of FetchedResultsSectionInfo, but occasionally get flagged with a data race using the thread sanitizer diagnostic in Xcode when models are saved.

What's odd is that when I print the thread in those methods, it's always "com.apple.root.user-initiated-qos.cooperative" Yet, the sanitizer report refers to other threads.

I'm open to completely scrapping Combine in this particular case, and moving to a AsyncThrowingSequence. I'm actually using .values to get the results back as such, but I think the problem has something to do with something intrinsic to how FetchedResultsController reloads after saves.

     WARNING: ThreadSanitizer: Swift access race (pid=48960)
  Modifying access of Swift variable at 0x00010c463c20 by thread T28:
    #0 CoreDataPersistence.ManagedObjectSectionsPublisher.(Inner in _132E40AC37633BC07BAC211BE31E5511).request(Combine.Subscribers.Demand) -> () <null> (<app-name>:arm64+0x10183ff9c)
    #1 protocol witness for Combine.Subscription.request(Combine.Subscribers.Demand) -> () in conformance CoreDataPersistence.ManagedObjectSectionsPublisher<A>.(Inner in _132E40AC37633BC07BAC211BE31E5511)<A1> : Combine.Subscription in CoreDataPersistence <null> (<app-name>:arm64+0x101841f08)
    #2 Combine.Publishers.ReceiveOn.(Inner in _1178A6B2012BC46DB46053E713A746B4).request(Combine.Subscribers.Demand) -> () <null> (Combine:arm64+0x48e0c)

  Previous read of size 8 at 0x00010c463c20 by thread T31:
    #0 CoreDataPersistence.ManagedObjectSectionsPublisher.(Inner in _132E40AC37633BC07BAC211BE31E5511).fulfillDemand() -> () <null> (<app-name>:arm64+0x10184066c)
    #1 CoreDataPersistence.ManagedObjectSectionsPublisher.(Inner in _132E40AC37633BC07BAC211BE31E5511).updateSections() -> () <null> (<app-name>:arm64+0x10183fbb4)
    #2 (1) suspend resume partial function for closure #1 @Sendable () async -> () in CoreDataPersistence.ManagedObjectSectionsPublisher.(Inner in _132E40AC37633BC07BAC211BE31E5511).controllerDidChangeContent(__C.NSFetchedResultsController<__C.NSFetchRequestResult>) -> () <null> (<app-name>:arm64+0x101841620)
    #3 swift::runJobInEstablishedExecutorContext(swift::Job*) <null> (libswift_Concurrency.dylib:arm64+0x38108)

  Location is heap block of size 64 at 0x00010c463c00 allocated by thread T12:
    #0 calloc <null> (libclang_rt.tsan_iossim_dynamic.dylib:arm64+0x4fc30)
    #1 _malloc_type_calloc_outlined <null> (libsystem_malloc.dylib:arm64+0xf16c)
    #2 CoreDataPersistence.ManagedObjectSectionsPublisher.receive<A where A1: Combine.Subscriber, A1.Failure == Swift.Error, A1.Input == Swift.Array<__C.NSFetchedResultsSectionInfo>>(subscriber: A1) -> () <null> (<app-name>:arm64+0x10183e1a0)
    #3 protocol witness for Combine.Publisher.receive<A where A1: Combine.Subscriber, A.Failure == A1.Failure, A.Output == A1.Input>(subscriber: A1) -> () in conformance CoreDataPersistence.ManagedObjectSectionsPublisher<A> : Combine.Publisher in CoreDataPersistence <null> (<app-name>:arm64+0x101842094)
    #4 Combine.Publishers.ReceiveOn.receive<A where A1: Combine.Subscriber, A.Failure == A1.Failure, A.Output == A1.Input>(subscriber: A1) -> () <null> (Combine:arm64+0x48a98)
    #5 swift::runJobInEstablishedExecutorContext(swift::Job*) <null> (libswift_Concurrency.dylib:arm64+0x38108)

  Thread T28 (tid=17129169, running) is a GCD worker thread

  Thread T31 (tid=17130304, running) is a GCD worker thread

  Thread T12 (tid=17128366, finished) is a GCD worker thread

SUMMARY: ThreadSanitizer: Swift access race