Harden Nostr envelope migration compatibility

bitchat/Services/NostrTransport.swift

@@ -12,8 +12,11 @@ final class NostrTransport: Transport, @unchecked Sendable {
         let favoriteStatusForPeerID: @MainActor (PeerID) -> FavoritesPersistenceService.FavoriteRelationship?
         let currentIdentity: @MainActor () throws -> NostrIdentity?
         let registerPendingPrivateEnvelope: @MainActor (String) -> Void
-        let sendEvent: @MainActor (NostrEvent) -> Void
-        let now: @MainActor () -> Date
+        let sendPrivateEnvelopeBatch: @MainActor (
+            [NostrEvent],
+            @escaping @MainActor () -> Void
+        ) -> Bool
+        let envelopeRetryQueue: NostrPrivateEnvelopeRetryQueue
         /// Emits whether a relay that carries private messages is up
         /// (fail-closed behind Tor). A connected geohash/custom relay alone
         /// doesn't count: DM sends target the default relay set and would
@@ -23,6 +26,7 @@ final class NostrTransport: Transport, @unchecked Sendable {
         /// serialize behind each other; `live` passes the process-wide one.
         let ackPacer: AckPacer
 
+        @MainActor
         init(
             notificationCenter: NotificationCenter,
             loadFavorites: @escaping @MainActor () -> [Data: FavoritesPersistenceService.FavoriteRelationship],
@@ -30,11 +34,14 @@ final class NostrTransport: Transport, @unchecked Sendable {
             favoriteStatusForPeerID: @escaping @MainActor (PeerID) -> FavoritesPersistenceService.FavoriteRelationship?,
             currentIdentity: @escaping @MainActor () throws -> NostrIdentity?,
             registerPendingPrivateEnvelope: @escaping @MainActor (String) -> Void,
-            sendEvent: @escaping @MainActor (NostrEvent) -> Void,
+            sendPrivateEnvelopeBatch: @escaping @MainActor (
+                [NostrEvent],
+                @escaping @MainActor () -> Void
+            ) -> Bool,
             scheduleAfter: @escaping @Sendable (TimeInterval, @escaping @Sendable () -> Void) -> Void,
             relayConnectivity: @escaping @MainActor () -> AnyPublisher<Bool, Never>,
             ackPacer: AckPacer? = nil,
-            now: @escaping @MainActor () -> Date = Date.init
+            envelopeRetryQueue: NostrPrivateEnvelopeRetryQueue? = nil
         ) {
             self.notificationCenter = notificationCenter
             self.loadFavorites = loadFavorites
@@ -42,8 +49,12 @@ final class NostrTransport: Transport, @unchecked Sendable {
             self.favoriteStatusForPeerID = favoriteStatusForPeerID
             self.currentIdentity = currentIdentity
             self.registerPendingPrivateEnvelope = registerPendingPrivateEnvelope
-            self.sendEvent = sendEvent
-            self.now = now
+            self.sendPrivateEnvelopeBatch = sendPrivateEnvelopeBatch
+            self.envelopeRetryQueue = envelopeRetryQueue ?? NostrPrivateEnvelopeRetryQueue(
+                sendPrivateEnvelopeBatch: sendPrivateEnvelopeBatch,
+                registerPendingPrivateEnvelope: registerPendingPrivateEnvelope,
+                scheduleAfter: scheduleAfter
+            )
             self.relayConnectivity = relayConnectivity
             // Default pacer drives its throttle through the same injected
             // scheduler, so tests that step scheduleAfter manually keep
@@ -60,12 +71,18 @@ final class NostrTransport: Transport, @unchecked Sendable {
                 favoriteStatusForPeerID: { FavoritesPersistenceService.shared.getFavoriteStatus(forPeerID: $0) },
                 currentIdentity: { try idBridge.getCurrentNostrIdentity() },
                 registerPendingPrivateEnvelope: { NostrRelayManager.registerPendingPrivateEnvelope(id: $0) },
-                sendEvent: { NostrRelayManager.shared.sendEvent($0) },
+                sendPrivateEnvelopeBatch: { events, terminalFailure in
+                    NostrRelayManager.shared.sendPrivateEnvelopeBatch(
+                        events,
+                        terminalFailure: terminalFailure
+                    )
+                },
                 scheduleAfter: { delay, action in
                     DispatchQueue.main.asyncAfter(deadline: .now() + delay, execute: action)
                 },
                 relayConnectivity: { NostrRelayManager.shared.$isDMRelayConnected.eraseToAnyPublisher() },
-                ackPacer: NostrTransport.sharedAckPacer
+                ackPacer: NostrTransport.sharedAckPacer,
+                envelopeRetryQueue: NostrTransport.sharedEnvelopeRetryQueue
             )
         }
     }
@@ -131,7 +148,37 @@ final class NostrTransport: Transport, @unchecked Sendable {
         }
     }
     static let sharedAckPacer = AckPacer()
+    // Geohash acknowledgements use short-lived NostrTransport instances, so
+    // the retry owner must be process-wide. A per-transport cap would still be
+    // globally unbounded under outage as throwaway instances accumulated.
+    @MainActor
+    private static let sharedEnvelopeRetryQueue = NostrPrivateEnvelopeRetryQueue(
+        sendPrivateEnvelopeBatch: { events, terminalFailure in
+            NostrRelayManager.shared.sendPrivateEnvelopeBatch(
+                events,
+                terminalFailure: terminalFailure
+            )
+        },
+        registerPendingPrivateEnvelope: {
+            NostrRelayManager.registerPendingPrivateEnvelope(id: $0)
+        },
+        scheduleAfter: { delay, action in
+            DispatchQueue.main.asyncAfter(deadline: .now() + delay, execute: action)
+        }
+    )
+
+    @MainActor
+    static func resetControlRetriesForPanicWipe() {
+        sharedEnvelopeRetryQueue.removeAll()
+    }
+
+    private enum PrivateEnvelopeFailurePolicy {
+        case userMessage(messageID: String)
+        case retry(retryKey: String)
+    }
+
     private let dependencies: Dependencies
+    private let envelopeRetryQueue: NostrPrivateEnvelopeRetryQueue
     private var favoriteStatusObserver: NSObjectProtocol?
 
     // Reachability Cache (thread-safe)
@@ -148,7 +195,9 @@ final class NostrTransport: Transport, @unchecked Sendable {
         idBridge: NostrIdentityBridge,
         dependencies: Dependencies? = nil
     ) {
-        self.dependencies = dependencies ?? .live(idBridge: idBridge)
+        let resolvedDependencies = dependencies ?? .live(idBridge: idBridge)
+        self.dependencies = resolvedDependencies
+        self.envelopeRetryQueue = resolvedDependencies.envelopeRetryQueue
         
         setupObservers()
         
@@ -175,6 +224,18 @@ final class NostrTransport: Transport, @unchecked Sendable {
         }
     }
 
+    #if DEBUG
+    @MainActor
+    func debugEnqueueControlRetry(key: String, events: [NostrEvent]) {
+        envelopeRetryQueue.enqueue(key: key, events: events, registerPending: false)
+    }
+
+    @MainActor
+    var debugControlRetryCount: Int {
+        envelopeRetryQueue.debugPendingCount
+    }
+    #endif
+
     private func setupObservers() {
         favoriteStatusObserver = dependencies.notificationCenter.addObserver(
             forName: .favoriteStatusChanged,
@@ -264,7 +325,12 @@ final class NostrTransport: Transport, @unchecked Sendable {
                 SecureLogger.error("NostrTransport: failed to embed PM packet", category: .session)
                 return
             }
-            sendPrivateEnvelope(content: embedded, recipientHex: recipientHex, senderIdentity: senderIdentity)
+            sendPrivateEnvelope(
+                content: embedded,
+                recipientHex: recipientHex,
+                senderIdentity: senderIdentity,
+                failurePolicy: .userMessage(messageID: messageID)
+            )
         }
     }
 
@@ -290,7 +356,14 @@ final class NostrTransport: Transport, @unchecked Sendable {
                 SecureLogger.error("NostrTransport: failed to embed favorite notification", category: .session)
                 return
             }
-            sendPrivateEnvelope(content: embedded, recipientHex: recipientHex, senderIdentity: senderIdentity)
+            sendPrivateEnvelope(
+                content: embedded,
+                recipientHex: recipientHex,
+                senderIdentity: senderIdentity,
+                failurePolicy: .retry(
+                    retryKey: privateEnvelopeRetryKey(content: embedded, recipientHex: recipientHex)
+                )
+            )
         }
     }
 
@@ -322,7 +395,13 @@ extension NostrTransport {
                 SecureLogger.error("NostrTransport: failed to embed geohash PM packet", category: .session)
                 return
             }
-            sendPrivateEnvelope(content: embedded, recipientHex: recipientHex, senderIdentity: identity, registerPending: true)
+            sendPrivateEnvelope(
+                content: embedded,
+                recipientHex: recipientHex,
+                senderIdentity: identity,
+                registerPending: true,
+                failurePolicy: .userMessage(messageID: messageID)
+            )
         }
     }
 }
@@ -345,21 +424,89 @@ extension NostrTransport {
 
     /// Creates and sends a BitChat private-envelope event over Nostr.
     @MainActor
-    private func sendPrivateEnvelope(content: String, recipientHex: String, senderIdentity: NostrIdentity, registerPending: Bool = false) {
+    private func sendPrivateEnvelope(
+        content: String,
+        recipientHex: String,
+        senderIdentity: NostrIdentity,
+        registerPending: Bool = false,
+        failurePolicy: PrivateEnvelopeFailurePolicy
+    ) {
         guard let events = try? NostrProtocol.createPrivateEnvelopePublicationBatch(
             content: content,
             recipientPubkey: recipientHex,
-            senderIdentity: senderIdentity,
-            now: dependencies.now()
+            senderIdentity: senderIdentity
         ) else {
             SecureLogger.error("NostrTransport: failed to build Nostr private-envelope batch", category: .session)
             return
         }
+        let accepted = dependencies.sendPrivateEnvelopeBatch(events) { [self] in
+            handlePrivateEnvelopeFailure(
+                events: events,
+                registerPending: registerPending,
+                policy: failurePolicy
+            )
+        }
+        guard accepted else {
+            SecureLogger.error(
+                "NostrTransport: private-envelope migration pair was not accepted for relay delivery",
+                category: .session
+            )
+            handlePrivateEnvelopeFailure(
+                events: events,
+                registerPending: registerPending,
+                policy: failurePolicy
+            )
+            return
+        }
+        registerPendingPrivateEnvelopesIfNeeded(events, registerPending: registerPending)
+    }
+
+    @MainActor
+    private func handlePrivateEnvelopeFailure(
+        events: [NostrEvent],
+        registerPending: Bool,
+        policy: PrivateEnvelopeFailurePolicy
+    ) {
+        switch policy {
+        case .userMessage(let messageID):
+            deliverTransportEvent(.messageDeliveryStatusUpdated(
+                messageID: messageID,
+                status: .failed(reason: String(
+                    localized: "content.delivery.reason.not_delivered",
+                    comment: "Failure reason shown when a private message could not enter the relay delivery queue"
+                ))
+            ))
+        case .retry(let retryKey):
+            envelopeRetryQueue.enqueue(
+                key: retryKey,
+                events: events,
+                registerPending: registerPending
+            )
+        }
+    }
+
+    @MainActor
+    private func registerPendingPrivateEnvelopesIfNeeded(
+        _ events: [NostrEvent],
+        registerPending: Bool
+    ) {
+        guard registerPending else { return }
         for event in events {
-            if registerPending {
-                dependencies.registerPendingPrivateEnvelope(event.id)
-            }
-            dependencies.sendEvent(event)
+            dependencies.registerPendingPrivateEnvelope(event.id)
+        }
+    }
+
+    @MainActor
+    private func privateEnvelopeRetryKey(content: String, recipientHex: String) -> String {
+        "\(recipientHex.lowercased()):\(Data(content.utf8).sha256Fingerprint())"
+    }
+
+    @MainActor
+    private func deliverTransportEvent(_ event: TransportEvent) {
+        if let eventDelegate {
+            eventDelegate.didReceiveTransportEvent(event)
+        } else {
+            delegate?.receiveTransportEvent(event)
         }
     }
 
@@ -377,7 +524,14 @@ extension NostrTransport {
                     SecureLogger.error("NostrTransport: failed to embed READ ack", category: .session)
                     return
                 }
-                sendPrivateEnvelope(content: ack, recipientHex: recipientHex, senderIdentity: senderIdentity)
+                sendPrivateEnvelope(
+                    content: ack,
+                    recipientHex: recipientHex,
+                    senderIdentity: senderIdentity,
+                    failurePolicy: .retry(
+                        retryKey: privateEnvelopeRetryKey(content: ack, recipientHex: recipientHex)
+                    )
+                )
 
             case .deliveredDirect(let messageID, let peerID):
                 guard let recipientNpub = resolveRecipientNpub(for: peerID),
@@ -388,17 +542,40 @@ extension NostrTransport {
                     SecureLogger.error("NostrTransport: failed to embed DELIVERED ack", category: .session)
                     return
                 }
-                sendPrivateEnvelope(content: ack, recipientHex: recipientHex, senderIdentity: senderIdentity)
+                sendPrivateEnvelope(
+                    content: ack,
+                    recipientHex: recipientHex,
+                    senderIdentity: senderIdentity,
+                    failurePolicy: .retry(
+                        retryKey: privateEnvelopeRetryKey(content: ack, recipientHex: recipientHex)
+                    )
+                )
 
             case .deliveredGeohash(let messageID, let recipientHex, let identity):
                 SecureLogger.debug("GeoDM: send DELIVERED mid=\(messageID.prefix(8))…", category: .session)
                 guard let embedded = NostrEmbeddedBitChat.encodeAckForNostrNoRecipient(type: .delivered, messageID: messageID, senderPeerID: senderPeerID) else { return }
-                sendPrivateEnvelope(content: embedded, recipientHex: recipientHex, senderIdentity: identity, registerPending: true)
+                sendPrivateEnvelope(
+                    content: embedded,
+                    recipientHex: recipientHex,
+                    senderIdentity: identity,
+                    registerPending: true,
+                    failurePolicy: .retry(
+                        retryKey: privateEnvelopeRetryKey(content: embedded, recipientHex: recipientHex)
+                    )
+                )
 
             case .readGeohash(let messageID, let recipientHex, let identity):
                 SecureLogger.debug("GeoDM: send READ mid=\(messageID.prefix(8))…", category: .session)
                 guard let embedded = NostrEmbeddedBitChat.encodeAckForNostrNoRecipient(type: .readReceipt, messageID: messageID, senderPeerID: senderPeerID) else { return }
-                sendPrivateEnvelope(content: embedded, recipientHex: recipientHex, senderIdentity: identity, registerPending: true)
+                sendPrivateEnvelope(
+                    content: embedded,
+                    recipientHex: recipientHex,
+                    senderIdentity: identity,
+                    registerPending: true,
+                    failurePolicy: .retry(
+                        retryKey: privateEnvelopeRetryKey(content: embedded, recipientHex: recipientHex)
+                    )
+                )
             }
         }
     }
@@ -418,3 +595,122 @@ extension NostrTransport {
         return nil
     }
 }
+
+/// Bounded retry owner for non-user private control payloads. It is separate
+/// from `NostrTransport` so a scheduled retry from a short-lived geohash
+/// transport remains valid after that transport deinitializes. Scheduler
+/// callbacks retain this queue, never the transport; an evicted key simply
+/// becomes a harmless no-op when its already-scheduled callback fires.
+@MainActor
+final class NostrPrivateEnvelopeRetryQueue {
+    private struct PendingRetry {
+        let events: [NostrEvent]
+        let registerPending: Bool
+        var attempt: Int
+        var isScheduled: Bool
+    }
+
+    private let sendPrivateEnvelopeBatch: @MainActor (
+        [NostrEvent],
+        @escaping @MainActor () -> Void
+    ) -> Bool
+    private let registerPendingPrivateEnvelope: @MainActor (String) -> Void
+    private let scheduleAfter: @Sendable (
+        TimeInterval,
+        @escaping @Sendable () -> Void
+    ) -> Void
+    private var pending: [String: PendingRetry] = [:]
+    private var insertionOrder: [String] = []
+
+    init(
+        sendPrivateEnvelopeBatch: @escaping @MainActor (
+            [NostrEvent],
+            @escaping @MainActor () -> Void
+        ) -> Bool,
+        registerPendingPrivateEnvelope: @escaping @MainActor (String) -> Void,
+        scheduleAfter: @escaping @Sendable (
+            TimeInterval,
+            @escaping @Sendable () -> Void
+        ) -> Void
+    ) {
+        self.sendPrivateEnvelopeBatch = sendPrivateEnvelopeBatch
+        self.registerPendingPrivateEnvelope = registerPendingPrivateEnvelope
+        self.scheduleAfter = scheduleAfter
+    }
+
+    func enqueue(key: String, events: [NostrEvent], registerPending: Bool) {
+        guard pending[key] == nil else { return }
+        if pending.count >= TransportConfig.nostrPrivateEnvelopeRetryQueueCap,
+           let evictedKey = insertionOrder.first {
+            insertionOrder.removeFirst()
+            pending.removeValue(forKey: evictedKey)
+            // These are control payloads, never user-authored messages. Keep
+            // the bounded-loss decision explicit rather than silently growing
+            // memory during a prolonged outage.
+            SecureLogger.warning(
+                "📮 Private control retry queue full — evicted oldest whole migration pair",
+                category: .session
+            )
+        }
+        pending[key] = PendingRetry(
+            events: events,
+            registerPending: registerPending,
+            attempt: 0,
+            isScheduled: false
+        )
+        insertionOrder.append(key)
+        schedule(key: key)
+    }
+
+    private func schedule(key: String) {
+        guard var item = pending[key], !item.isScheduled else { return }
+        item.isScheduled = true
+        pending[key] = item
+        let exponent = min(item.attempt, 5)
+        let delay = min(2.0 * pow(2.0, Double(exponent)), 60.0)
+        scheduleAfter(delay) { [self] in
+            Task { @MainActor [self] in
+                self.retry(key: key)
+            }
+        }
+    }
+
+    private func retry(key: String) {
+        guard var item = pending[key] else { return }
+        item.isScheduled = false
+        pending[key] = item
+
+        let accepted = sendPrivateEnvelopeBatch(item.events) { [self] in
+            self.enqueue(
+                key: key,
+                events: item.events,
+                registerPending: item.registerPending
+            )
+        }
+        if accepted {
+            remove(key: key)
+            if item.registerPending {
+                for event in item.events {
+                    registerPendingPrivateEnvelope(event.id)
+                }
+            }
+        } else {
+            item.attempt += 1
+            pending[key] = item
+            schedule(key: key)
+        }
+    }
+
+    private func remove(key: String) {
+        pending.removeValue(forKey: key)
+        insertionOrder.removeAll { $0 == key }
+    }
+
+    func removeAll() {
+        pending.removeAll()
+        insertionOrder.removeAll()
+    }
+
+    var debugPendingCount: Int { pending.count }
+    func debugContains(key: String) -> Bool { pending[key] != nil }
+}