iOS Crash: NSInvalidArgumentException during BLE write with active scanning (EXC_BAD_ACCESS)

[Thilakforabm]

Hi there! Firstly, thank you for this excellent plugin – it's been invaluable for our BLE project.

I'm experiencing intermittent iOS crashes that occur during BLE write operations while scanning is active. I wanted to report this and see if others have encountered similar issues.
Environment

Plugin version: @capacitor-community/bluetooth-le@7.1.1
Platform: iOS
Ionic/Capacitor: React + Capacitor
Scan pattern: 15-second scan cycles with requestLEScan (stop and restart every 15s)

The Issue
After running the app for a while (typically 10-30 minutes), the app crashes during a BleClient.write() operation with:
EXC_BAD_ACCESS (code=1, address=0x10)

Sometimes I also see:
*** Terminating app due to uncaught exception 'NSInvalidArgumentException',
reason: '-[NSTaggedPointerString objectForKey:]: unrecognized selector sent to instance 0x8000000000000000'

When It Happens

✅ Occurs during or right after a BLE write operation
✅ Happens while requestLEScan is actively running
✅ Random/intermittent – not tied to a specific device or characteristic
❌ Does not happen when scanning is stopped before writes

My Analysis
Based on the crash logs and assembly dump, I believe there's a race condition in requestLEScan:
In Plugin.swift line ~163:
swift}, { (device, advertisementData, rssi) in
self.deviceMap[device.getId()] = device // ⚠️ Overwrites existing Device
let data = self.getScanResult(device, advertisementData, rssi)
self.notifyListeners("onScanResult", data: data)
}
The race condition:

Device A is discovered → deviceMap["ABC"] = Device1
I call write() on Device A → operation starts, holds reference to Device1
Any device advertises (Device A, B, C, etc.) → New scan callback fires
Callback executes self.deviceMap[device.getId()] = device → Creates new Device object
If this happens to Device A → deviceMap["ABC"] = Device2 (new object!)
Device1 loses its strong reference → ARC deallocates it
Write callback tries to access Device1 → crash (accessing deallocated memory)

The critical issue: The scan callback creates a new Device object on every advertisement, even for devices already in the map. This means:

Every scan result (from any device) modifies deviceMap
If the scanned device happens to be one with an ongoing operation, that operation's Device object gets replaced
The original Device object is deallocated while still being used
This explains why the crash is random – it depends on the timing of advertisements from all nearby devices

Proposed Fix:
always use the existing device from the map when building scan results:
swift}, { (device, advertisementData, rssi) in
// Use existing device if available, otherwise add new one
let existingDevice = self.deviceMap[device.getId()]
let deviceToUse = existingDevice ?? device

if existingDevice == nil {
    self.deviceMap[device.getId()] = device
}

let data = self.getScanResult(deviceToUse, advertisementData, rssi)
self.notifyListeners("onScanResult", data: data)

}

This occurred more frequently for me because i write multiple messages almost every second to my peripheral while scanning

Questions
Is this a known issue or has anyone experienced similar crashes if not shall i raise an issue?
Would the proposed fix be appropriate, or is there a better approach?
Are there any workarounds you'd recommend while this is being investigated?

[peitschie]

I've converted this into an issue at #779

This looks like a crash in the iOS stack due to threaded access to the dictionary as Swift dictionaries are not thread-safe. We'll need some synchronisation/locking to stop the concurrent access while modifying.

[mklingmann]

I fixed it by replacing the deviceMap dict in Plugin.swift

private var deviceMap = [String: Device]()

with a thread safe dictionary (already in the source)

private var deviceMap = ThreadSafeDictionary<String, Device>()