[GH-ISSUE #1660] iOS app crashes when clearing notifications (mutation during enumeration) #1157

New issue

Open

opened 2026-05-07 00:30:41 +02:00 by BreizhHardware · 1 comment

BreizhHardware commented

2026-05-07 00:30:41 +02:00

Owner

Originally created by @jflammia on GitHub (Mar 17, 2026).
Original GitHub issue: https://github.com/binwiederhier/ntfy/issues/1660

Bug Description

The ntfy iOS app crashes with SIGABRT when attempting to clear/delete all notifications for a subscription. The app becomes unusable once enough notifications accumulate (~870 in my case) because the only recovery is to delete and reinstall the app.

Crash Analysis

Exception: EXC_CRASH / SIGABRT (Abort trap: 6)
Device: iPhone 15 Pro (iPhone16,1), iOS 26.3.1
App Version: 1.3 (build 4)

Root Cause

The crash occurs in Store.swift at delete(allNotificationsFor:) (line ~151 in ntfy-ios repo):

func delete(allNotificationsFor subscription: Subscription) {
    guard let notifications = subscription.notifications else { return }
    do {
        notifications.forEach { notification in
            context.delete(notification as! Notification)  // Mutates collection during enumeration
        }
        try context.save()
    } catch let error {
        Log.w(Store.tag, "Cannot delete notification(s)", error)
        rollbackAndRefresh()
    }
}

The forEach loop calls context.delete() on each notification while iterating over subscription.notifications (an NSSet). Core Data's _processRecentChanges: fires on the run loop observer during this iteration and tries to enumerate the same collection, causing a mutation-during-enumeration crash.

Stack Trace

Main thread (triggered):

__pthread_kill → abort() → __abort_message → demangling_terminate_handler()
→ _objc_terminate() → __cxa_rethrow → objc_exception_rethrow
→ -[NSManagedObjectContext _processRecentChanges:]
→ _performRunLoopAction
→ __CFRUNLOOP_IS_CALLING_OUT_TO_AN_OBSERVER_CALLBACK_FUNCTION__

Last exception backtrace:

__exceptionPreprocess → objc_exception_throw
→ +[__NSFastEnumerationEnumerator allocWithZone:]   ← collection mutated during enumeration
→ -[NSManagedObjectContext _processRecentChanges:]

Background SQLite thread was inside _prefetchObjectsForDeletePropagation at crash time, confirming Core Data was processing the deletes concurrently.

Suggested Fix

Option A — Copy before iterating (minimal fix):

func delete(allNotificationsFor subscription: Subscription) {
    guard let notifications = subscription.notifications else { return }
    let snapshot = Array(notifications)  // Copy to avoid mutation during enumeration
    do {
        snapshot.forEach { notification in
            context.delete(notification as! Notification)
        }
        try context.save()
    } catch let error {
        Log.w(Store.tag, "Cannot delete notification(s)", error)
        rollbackAndRefresh()
    }
}

Option B — NSBatchDeleteRequest (better for large datasets):

func delete(allNotificationsFor subscription: Subscription) {
    let fetchRequest: NSFetchRequest<NSFetchRequestResult> = Notification.fetchRequest()
    fetchRequest.predicate = NSPredicate(format: "subscription == %@", subscription)
    let batchDelete = NSBatchDeleteRequest(fetchRequest: fetchRequest)
    batchDelete.resultType = .resultTypeObjectIDs
    
    do {
        let result = try context.execute(batchDelete) as? NSBatchDeleteResult
        if let objectIDs = result?.result as? [NSManagedObjectID] {
            let changes = [NSDeletedObjectsKey: objectIDs]
            NSManagedObjectContext.mergeChanges(fromRemoteContextSave: changes, into: [context])
        }
    } catch let error {
        Log.w(Store.tag, "Cannot batch delete notification(s)", error)
    }
}

Option B is preferred as it doesn't load objects into memory and handles any number of notifications efficiently.

The same pattern should also be applied to delete(notifications:) which has the same mutation-during-enumeration vulnerability.

Steps to Reproduce

Subscribe to a topic on a self-hosted ntfy server
Accumulate 500+ notifications over time
Attempt to clear/delete all notifications in the iOS app
App crashes immediately

Environment

Self-hosted ntfy server v2.17.0
ntfy iOS app v1.3 (build 4) from App Store
iPhone 15 Pro, iOS 26.3.1

Workaround

Delete and reinstall the ntfy iOS app to clear the local Core Data database.

Originally created by @jflammia on GitHub (Mar 17, 2026). Original GitHub issue: https://github.com/binwiederhier/ntfy/issues/1660 ## Bug Description The ntfy iOS app crashes with `SIGABRT` when attempting to clear/delete all notifications for a subscription. The app becomes unusable once enough notifications accumulate (~870 in my case) because the only recovery is to delete and reinstall the app. ## Crash Analysis **Exception:** `EXC_CRASH` / `SIGABRT` (Abort trap: 6) **Device:** iPhone 15 Pro (iPhone16,1), iOS 26.3.1 **App Version:** 1.3 (build 4) ### Root Cause The crash occurs in `Store.swift` at `delete(allNotificationsFor:)` (line ~151 in ntfy-ios repo): ```swift func delete(allNotificationsFor subscription: Subscription) { guard let notifications = subscription.notifications else { return } do { notifications.forEach { notification in context.delete(notification as! Notification) // Mutates collection during enumeration } try context.save() } catch let error { Log.w(Store.tag, "Cannot delete notification(s)", error) rollbackAndRefresh() } } ``` The `forEach` loop calls `context.delete()` on each notification while iterating over `subscription.notifications` (an `NSSet`). Core Data's `_processRecentChanges:` fires on the run loop observer during this iteration and tries to enumerate the same collection, causing a **mutation-during-enumeration crash**. ### Stack Trace **Main thread (triggered):** ``` __pthread_kill → abort() → __abort_message → demangling_terminate_handler() → _objc_terminate() → __cxa_rethrow → objc_exception_rethrow → -[NSManagedObjectContext _processRecentChanges:] → _performRunLoopAction → __CFRUNLOOP_IS_CALLING_OUT_TO_AN_OBSERVER_CALLBACK_FUNCTION__ ``` **Last exception backtrace:** ``` __exceptionPreprocess → objc_exception_throw → +[__NSFastEnumerationEnumerator allocWithZone:] ← collection mutated during enumeration → -[NSManagedObjectContext _processRecentChanges:] ``` **Background SQLite thread** was inside `_prefetchObjectsForDeletePropagation` at crash time, confirming Core Data was processing the deletes concurrently. ## Suggested Fix **Option A — Copy before iterating (minimal fix):** ```swift func delete(allNotificationsFor subscription: Subscription) { guard let notifications = subscription.notifications else { return } let snapshot = Array(notifications) // Copy to avoid mutation during enumeration do { snapshot.forEach { notification in context.delete(notification as! Notification) } try context.save() } catch let error { Log.w(Store.tag, "Cannot delete notification(s)", error) rollbackAndRefresh() } } ``` **Option B — NSBatchDeleteRequest (better for large datasets):** ```swift func delete(allNotificationsFor subscription: Subscription) { let fetchRequest: NSFetchRequest<NSFetchRequestResult> = Notification.fetchRequest() fetchRequest.predicate = NSPredicate(format: "subscription == %@", subscription) let batchDelete = NSBatchDeleteRequest(fetchRequest: fetchRequest) batchDelete.resultType = .resultTypeObjectIDs do { let result = try context.execute(batchDelete) as? NSBatchDeleteResult if let objectIDs = result?.result as? [NSManagedObjectID] { let changes = [NSDeletedObjectsKey: objectIDs] NSManagedObjectContext.mergeChanges(fromRemoteContextSave: changes, into: [context]) } } catch let error { Log.w(Store.tag, "Cannot batch delete notification(s)", error) } } ``` Option B is preferred as it doesn't load objects into memory and handles any number of notifications efficiently. The same pattern should also be applied to `delete(notifications:)` which has the same mutation-during-enumeration vulnerability. ## Steps to Reproduce 1. Subscribe to a topic on a self-hosted ntfy server 2. Accumulate 500+ notifications over time 3. Attempt to clear/delete all notifications in the iOS app 4. App crashes immediately ## Environment - Self-hosted ntfy server v2.17.0 - ntfy iOS app v1.3 (build 4) from App Store - iPhone 15 Pro, iOS 26.3.1 ## Workaround Delete and reinstall the ntfy iOS app to clear the local Core Data database.

BreizhHardware commented

2026-05-07 00:30:42 +02:00

Author

Owner

@RezzZ commented on GitHub (Mar 20, 2026):

app crashes for me too, but messages do get deleted and a restart of the app works for me. so no workaround required here

@RezzZ commented on GitHub (Mar 20, 2026): app crashes for me too, but messages do get deleted and a restart of the app works for me. so no workaround required here