ios: integrating webrtc calls with callkit (#686)

* ios: integrating webrtc calls with callkit

* accept call via chat item (e.g. when DND is on, and callkit blocks the call); refactor

* fix remote video, support logging from ios

* use callkit depending on CallController setting

* call sound

* update incoming call view

* fixing audio encryption

* refactor encryption webrtc fix

* log ontrack success/error

* accept / ignore call via notification

* remove unused imports

* remove unused file

* remove comments

apps/android/app/src/main/assets/www/call.js

@@ -96,7 +96,8 @@ const processCommand = (function () {
         const pc = new RTCPeerConnection(config.peerConnectionConfig);
         const remoteStream = new MediaStream();
         const localCamera = VideoCamera.User;
-        const localStream = await navigator.mediaDevices.getUserMedia(callMediaConstraints(mediaType, localCamera));
+        const constraints = callMediaConstraints(mediaType, localCamera);
+        const localStream = await navigator.mediaDevices.getUserMedia(constraints);
         const iceCandidates = getIceCandidates(pc, config);
         const call = { connection: pc, iceCandidates, localMedia: mediaType, localCamera, localStream, remoteStream, aesKey, useWorker };
         await setupMediaStreams(call);
@@ -116,8 +117,10 @@ const processCommand = (function () {
             });
             if (pc.connectionState == "disconnected" || pc.connectionState == "failed") {
                 pc.removeEventListener("connectionstatechange", connectionStateChange);
+                if (activeCall) {
+                    setTimeout(() => sendMessageToNative({ resp: { type: "ended" } }), 0);
+                }
                 endCall();
-                setTimeout(() => sendMessageToNative({ resp: { type: "ended" } }), 0);
             }
             else if (pc.connectionState == "connected") {
                 const stats = (await pc.getStats());
@@ -133,7 +136,7 @@ const processCommand = (function () {
                                 remoteCandidate: stats.get(iceCandidatePair.remoteCandidateId),
                             },
                         };
-                        setTimeout(() => sendMessageToNative({ resp }), 0);
+                        setTimeout(() => sendMessageToNative({ resp }), 500);
                         break;
                     }
                 }
@@ -256,19 +259,9 @@ const processCommand = (function () {
                     if (!activeCall || !pc) {
                         resp = { type: "error", message: "camera: call not started" };
                     }
-                    else if (activeCall.localMedia == CallMediaType.Audio) {
-                        resp = { type: "error", message: "camera: no video" };
-                    }
                     else {
-                        try {
-                            if (command.camera != activeCall.localCamera) {
-                                await replaceCamera(activeCall, command.camera);
-                            }
-                            resp = { type: "ok" };
-                        }
-                        catch (e) {
-                            resp = { type: "error", message: `camera: ${e.message}` };
-                        }
+                        await replaceMedia(activeCall, command.camera);
+                        resp = { type: "ok" };
                     }
                     break;
                 case "end":
@@ -281,7 +274,7 @@ const processCommand = (function () {
             }
         }
         catch (e) {
-            resp = { type: "error", message: e.message };
+            resp = { type: "error", message: `${command.type}: ${e.message}` };
         }
         const apiResp = { corrId, resp, command };
         sendMessageToNative(apiResp);
@@ -323,6 +316,8 @@ const processCommand = (function () {
             if (call.useWorker && !call.worker) {
                 const workerCode = `const callCrypto = (${callCryptoFunction.toString()})(); (${workerFunction.toString()})()`;
                 call.worker = new Worker(URL.createObjectURL(new Blob([workerCode], { type: "text/javascript" })));
+                call.worker.onerror = ({ error, filename, lineno, message }) => console.log(JSON.stringify({ error, filename, lineno, message }));
+                call.worker.onmessage = ({ data }) => console.log(JSON.stringify({ message: data }));
             }
         }
     }
@@ -346,14 +341,20 @@ const processCommand = (function () {
         // Pull tracks from remote stream as they arrive add them to remoteStream video
         const pc = call.connection;
         pc.ontrack = (event) => {
-            if (call.aesKey && call.key) {
-                console.log("set up decryption for receiving");
-                setupPeerTransform(TransformOperation.Decrypt, event.receiver, call.worker, call.aesKey, call.key);
-            }
-            for (const stream of event.streams) {
-                for (const track of stream.getTracks()) {
-                    call.remoteStream.addTrack(track);
+            try {
+                if (call.aesKey && call.key) {
+                    console.log("set up decryption for receiving");
+                    setupPeerTransform(TransformOperation.Decrypt, event.receiver, call.worker, call.aesKey, call.key);
                 }
+                for (const stream of event.streams) {
+                    for (const track of stream.getTracks()) {
+                        call.remoteStream.addTrack(track);
+                    }
+                }
+                console.log(`ontrack success`);
+            }
+            catch (e) {
+                console.log(`ontrack error: ${e.message}`);
             }
         };
     }
@@ -385,7 +386,7 @@ const processCommand = (function () {
             }
         }
     }
-    async function replaceCamera(call, camera) {
+    async function replaceMedia(call, camera) {
         const videos = getVideoElements();
         if (!videos)
             throw Error("no video elements");
@@ -401,6 +402,8 @@ const processCommand = (function () {
         videos.local.srcObject = localStream;
     }
     function replaceTracks(pc, tracks) {
+        if (!tracks.length)
+            return;
         const sender = pc.getSenders().find((s) => { var _a; return ((_a = s.track) === null || _a === void 0 ? void 0 : _a.kind) === tracks[0].kind; });
         if (sender)
             for (const t of tracks)
@@ -494,8 +497,8 @@ function callCryptoFunction() {
             const initial = data.subarray(0, n);
             const plaintext = data.subarray(n, data.byteLength);
             try {
-                const ciphertext = await crypto.subtle.encrypt({ name: "AES-GCM", iv: iv.buffer }, key, plaintext);
-                frame.data = concatN(initial, new Uint8Array(ciphertext), iv).buffer;
+                const ciphertext = new Uint8Array(plaintext.length ? await crypto.subtle.encrypt({ name: "AES-GCM", iv: iv.buffer }, key, plaintext) : 0);
+                frame.data = concatN(initial, ciphertext, iv).buffer;
                 controller.enqueue(frame);
             }
             catch (e) {
@@ -512,8 +515,8 @@ function callCryptoFunction() {
             const ciphertext = data.subarray(n, data.byteLength - IV_LENGTH);
             const iv = data.subarray(data.byteLength - IV_LENGTH, data.byteLength);
             try {
-                const plaintext = await crypto.subtle.decrypt({ name: "AES-GCM", iv }, key, ciphertext);
-                frame.data = concatN(initial, new Uint8Array(plaintext)).buffer;
+                const plaintext = new Uint8Array(ciphertext.length ? await crypto.subtle.decrypt({ name: "AES-GCM", iv }, key, ciphertext) : 0);
+                frame.data = concatN(initial, plaintext).buffer;
                 controller.enqueue(frame);
             }
             catch (e) {
@@ -619,9 +622,15 @@ function workerFunction() {
     // encryption using RTCRtpScriptTransform.
     if ("RTCTransformEvent" in self) {
         self.addEventListener("rtctransform", async ({ transformer }) => {
-            const { operation, aesKey } = transformer.options;
-            const { readable, writable } = transformer;
-            await setupTransform({ operation, aesKey, readable, writable });
+            try {
+                const { operation, aesKey } = transformer.options;
+                const { readable, writable } = transformer;
+                await setupTransform({ operation, aesKey, readable, writable });
+                self.postMessage({ result: "setupTransform success" });
+            }
+            catch (e) {
+                self.postMessage({ message: `setupTransform error: ${e.message}` });
+            }
         });
     }
     async function setupTransform({ operation, aesKey, readable, writable }) {

apps/android/app/src/main/assets/www/style.css

@@ -1,12 +1,10 @@
-video::-webkit-media-controls {
-  display: none;
-}
 html,
 body {
   padding: 0;
   margin: 0;
   background-color: black;
 }
+
 #remote-video-stream {
   position: absolute;
   width: 100%;
@@ -24,3 +22,20 @@ body {
   top: 0;
   right: 0;
 }
+
+*::-webkit-media-controls {
+  display: none !important;
+  -webkit-appearance: none !important;
+}
+*::-webkit-media-controls-panel {
+  display: none !important;
+  -webkit-appearance: none !important;
+}
+*::-webkit-media-controls-play-button {
+  display: none !important;
+  -webkit-appearance: none !important;
+}
+*::-webkit-media-controls-start-playback-button {
+  display: none !important;
+  -webkit-appearance: none !important;
+}

apps/android/app/src/main/java/chat/simplex/app/views/chat/ChatView.kt

@@ -209,18 +209,18 @@ fun ChatInfoToolbar(chat: Chat, back: () -> Unit, info: () -> Unit, startCall: (
     ) {
       val cInfo = chat.chatInfo
       toolbarButton(Icons.Outlined.ArrowBackIos, R.string.back, onClick = back)
-//      if (cInfo is ChatInfo.Direct) {
-//        Box(Modifier.fillMaxWidth(), contentAlignment = Alignment.CenterEnd) {
-//          Box(Modifier.width(85.dp), contentAlignment = Alignment.CenterStart) {
-//            toolbarButton(Icons.Outlined.Phone, R.string.icon_descr_audio_call) {
-//              startCall(CallMediaType.Audio)
-//            }
-//          }
-//          toolbarButton(Icons.Outlined.Videocam, R.string.icon_descr_video_call) {
-//            startCall(CallMediaType.Video)
-//          }
-//        }
-//      }
+      if (cInfo is ChatInfo.Direct) {
+        Box(Modifier.fillMaxWidth(), contentAlignment = Alignment.CenterEnd) {
+          Box(Modifier.width(85.dp), contentAlignment = Alignment.CenterStart) {
+            toolbarButton(Icons.Outlined.Phone, R.string.icon_descr_audio_call) {
+              startCall(CallMediaType.Audio)
+            }
+          }
+          toolbarButton(Icons.Outlined.Videocam, R.string.icon_descr_video_call) {
+            startCall(CallMediaType.Video)
+          }
+        }
+      }
       Row(
         Modifier
           .padding(horizontal = 80.dp)

apps/ios/Shared/AppDelegate.swift

@@ -70,7 +70,7 @@ class AppDelegate: NSObject, UIApplicationDelegate {
                 // TODO check if app in background
                 logger.debug("AppDelegate: didReceiveRemoteNotification: checkMessages")
                 // TODO remove
-                NtfManager.shared.notifyCheckingMessages()
+                // NtfManager.shared.notifyCheckingMessages()
                 receiveMessages(completionHandler)
             } else if let smpQueue = ntfData["checkMessage"] as? String {
                 // TODO check if app in background

apps/ios/Shared/ContentView.swift

@@ -10,6 +10,7 @@ import SwiftUI
 struct ContentView: View {
     @EnvironmentObject var chatModel: ChatModel
     @ObservedObject var alertManager = AlertManager.shared
+    @ObservedObject var callController = CallController.shared
     @State private var showNotificationAlert = false
 
     var body: some View {
@@ -17,11 +18,17 @@ struct ContentView: View {
             if let step = chatModel.onboardingStage {
                 if case .onboardingComplete = step,
                    let user = chatModel.currentUser {
-                    ChatListView(user: user)
-                    .onAppear {
-                        NtfManager.shared.requestAuthorization(onDeny: {
-                            alertManager.showAlert(notificationAlert())
-                        })
+                    ZStack(alignment: .top) {
+                        ChatListView(user: user)
+                        .onAppear {
+                            NtfManager.shared.requestAuthorization(onDeny: {
+                                alertManager.showAlert(notificationAlert())
+                            })
+                        }
+                        if chatModel.showCallView, let call = chatModel.activeCall {
+                            ActiveCallView(call: call)
+                        }
+                        IncomingCallView()
                     }
                 } else {
                     OnboardingView(onboarding: step)

apps/ios/Shared/Model/ChatModel.swift

@@ -9,6 +9,7 @@
 import Foundation
 import Combine
 import SwiftUI
+import WebKit
 
 final class ChatModel: ObservableObject {
     @Published var onboardingStage: OnboardingStage?
@@ -28,10 +29,10 @@ final class ChatModel: ObservableObject {
     @Published var tokenStatus = NtfTknStatus.new
     // current WebRTC call
     @Published var callInvitations: Dictionary<ChatId, CallInvitation> = [:]
-    @Published var activeCallInvitation: ContactRef?
     @Published var activeCall: Call?
     @Published var callCommand: WCallCommand?
     @Published var showCallView = false
+    var callWebView: WKWebView?
 
     var messageDelivery: Dictionary<Int64, () -> Void> = [:]
 

apps/ios/Shared/Model/NtfManager.swift

@@ -37,25 +37,12 @@ class NtfManager: NSObject, UNUserNotificationCenterDelegate, ObservableObject {
             Task { await acceptContactRequest(contactRequest) }
         } else if content.categoryIdentifier == ntfCategoryCallInvitation && (action == ntfActionAcceptCall || action == ntfActionRejectCall),
                   let chatId = content.userInfo["chatId"] as? String,
-                  case let .direct(contact) = chatModel.getChat(chatId)?.chatInfo,
                   let invitation = chatModel.callInvitations.removeValue(forKey: chatId) {
+            let cc = CallController.shared
             if action == ntfActionAcceptCall {
-                chatModel.activeCallInvitation = nil
-                chatModel.activeCall = Call(contact: contact, callState: .invitationReceived, localMedia: invitation.peerMedia)
-                chatModel.showCallView = true
-                chatModel.callCommand = .start(media: invitation.peerMedia, aesKey: invitation.sharedKey)
+                cc.answerCall(invitation: invitation)
             } else {
-                Task {
-                    do {
-                        try await apiRejectCall(contact)
-                        if chatModel.activeCall?.contact.id == chatId {
-                            DispatchQueue.main.async {
-                                chatModel.callCommand = .end
-                                chatModel.activeCall = nil
-                            }
-                        }
-                    }
-                }
+                cc.endCall(invitation: invitation)
             }
         } else {
             chatModel.chatId = content.targetContentIdentifier
@@ -89,6 +76,7 @@ class NtfManager: NSObject, UNUserNotificationCenterDelegate, ObservableObject {
             // this notification is deliverd from the notifications server
             // when the app is in foreground it does not need to be shown
             case ntfCategoryCheckMessage: return []
+            case ntfCategoryCallInvitation: return []
             default: return [.sound, .banner, .list]
             }
         } else {
@@ -136,11 +124,12 @@ class NtfManager: NSObject, UNUserNotificationCenterDelegate, ObservableObject {
                 actions: [
                     UNNotificationAction(
                         identifier: ntfActionAcceptCall,
-                        title: NSLocalizedString("Answer", comment: "accept incoming call via notification")
+                        title: NSLocalizedString("Accept", comment: "accept incoming call via notification"),
+                        options: .foreground
                     ),
                     UNNotificationAction(
                         identifier: ntfActionRejectCall,
-                        title: NSLocalizedString("Ignore", comment: "ignore incoming call via notification")
+                        title: NSLocalizedString("Reject", comment: "reject incoming call via notification")
                     )
                 ],
                 intentIdentifiers: [],
@@ -194,9 +183,9 @@ class NtfManager: NSObject, UNUserNotificationCenterDelegate, ObservableObject {
         addNotification(createMessageReceivedNtf(cInfo, cItem))
     }
 
-    func notifyCallInvitation(_ contact: Contact, _ invitation: CallInvitation) {
+    func notifyCallInvitation(_ invitation: CallInvitation) {
         logger.debug("NtfManager.notifyCallInvitation")
-        addNotification(createCallInvitationNtf(contact, invitation))
+        addNotification(createCallInvitationNtf(invitation))
     }
 
     // TODO remove

apps/ios/Shared/Model/Shared/CallTypes.swift

@@ -24,25 +24,23 @@ struct WebRTCExtraInfo: Codable {
 }
 
 struct CallInvitation {
+    var contact: Contact
+    var callkitUUID: UUID?
     var peerMedia: CallMediaType
     var sharedKey: String?
     var callTypeText: LocalizedStringKey {
         get {
             switch peerMedia {
-            case .video: return sharedKey == nil ? "video call (not e2e encrypted)." : "**e2e encrypted** video call."
-            case .audio: return sharedKey == nil ? "audio call (not e2e encrypted)." : "**e2e encrypted** audio call."
+            case .video: return sharedKey == nil ? "video call (not e2e encrypted)" : "**e2e encrypted** video call"
+            case .audio: return sharedKey == nil ? "audio call (not e2e encrypted)" : "**e2e encrypted** audio call"
             }
         }
     }
-    var callTitle: LocalizedStringKey {
-        get {
-            switch peerMedia {
-            case .video: return "Incoming video call"
-            case .audio: return "Incoming audio call"
-            }
-        }
-    }
-    var encryptionText: LocalizedStringKey { get { sharedKey == nil ? "no e2e encryption" : "with e2e encryption" } }
+
+    static let sampleData = CallInvitation(
+        contact: Contact.sampleData,
+        peerMedia: .audio
+    )
 }
 
 struct CallType: Codable {

apps/ios/Shared/Model/Shared/ChatTypes.swift

@@ -879,13 +879,13 @@ enum CICallStatus: String, Decodable {
     func text(_ sec: Int) -> String {
         switch self {
         case .pending: return NSLocalizedString("calling…", comment: "call status")
-        case .missed: return NSLocalizedString("missed", comment: "call status")
-        case .rejected: return NSLocalizedString("rejected", comment: "call status")
-        case .accepted: return NSLocalizedString("accepted", comment: "call status")
-        case .negotiated: return NSLocalizedString("connecting…", comment: "call status")
-        case .progress: return NSLocalizedString("in progress", comment: "call status")
-        case .ended: return String.localizedStringWithFormat(NSLocalizedString("ended %@", comment: "call status"), CICallStatus.durationText(sec))
-        case .error: return NSLocalizedString("error", comment: "call status")
+        case .missed: return NSLocalizedString("missed call", comment: "call status")
+        case .rejected: return NSLocalizedString("rejected call", comment: "call status")
+        case .accepted: return NSLocalizedString("accepted call", comment: "call status")
+        case .negotiated: return NSLocalizedString("connecting call…", comment: "call status")
+        case .progress: return NSLocalizedString("call in progress", comment: "call status")
+        case .ended: return String.localizedStringWithFormat(NSLocalizedString("ended call %@", comment: "call status"), CICallStatus.durationText(sec))
+        case .error: return NSLocalizedString("call error", comment: "call status")
         }
     }
 

apps/ios/Shared/Model/Shared/Notifications.swift

@@ -50,16 +50,16 @@ func createMessageReceivedNtf(_ cInfo: ChatInfo, _ cItem: ChatItem) -> UNMutable
     )
 }
 
-func createCallInvitationNtf(_ contact: Contact, _ invitation: CallInvitation) -> UNMutableNotificationContent {
+func createCallInvitationNtf(_ invitation: CallInvitation) -> UNMutableNotificationContent {
     let text = invitation.peerMedia == .video
                 ? NSLocalizedString("Incoming video call", comment: "notification")
                 : NSLocalizedString("Incoming audio call", comment: "notification")
     return createNotification(
         categoryIdentifier: ntfCategoryCallInvitation,
-        title: "\(contact.chatViewName):",
+        title: "\(invitation.contact.chatViewName):",
         body: text,
         targetContentIdentifier: nil,
-        userInfo: ["chatId": contact.id]
+        userInfo: ["chatId": invitation.contact.id]
     )
 }
 

apps/ios/Shared/Model/SimpleXAPI.swift

@@ -11,6 +11,7 @@ import UIKit
 import Dispatch
 import BackgroundTasks
 import SwiftUI
+import CallKit
 
 private var chatController: chat_ctrl?
 
@@ -629,22 +630,41 @@ func processReceivedMsg(_ res: ChatResponse) {
                 removeFile(fileName)
             }
         case let .callInvitation(contact, callType, sharedKey):
-            let invitation = CallInvitation(peerMedia: callType.media, sharedKey: sharedKey)
+            let uuid = UUID()
+            var invitation = CallInvitation(contact: contact, callkitUUID: uuid, peerMedia: callType.media, sharedKey: sharedKey)
             m.callInvitations[contact.id] = invitation
-            if (m.activeCallInvitation == nil) {
-                m.activeCallInvitation = ContactRef(contactId: contact.apiId, localDisplayName: contact.localDisplayName)
+            CallController.shared.reportNewIncomingCall(invitation: invitation) { error in
+                if let error = error {
+                    invitation.callkitUUID = nil
+                    m.callInvitations[contact.id] = invitation
+                    logger.error("reportNewIncomingCall error: \(error.localizedDescription)")
+                } else {
+                    logger.debug("reportNewIncomingCall success")
+                }
             }
-            NtfManager.shared.notifyCallInvitation(contact, invitation)
+
+// This will be called from notification service extension
+//            CXProvider.reportNewIncomingVoIPPushPayload([
+//                "displayName": contact.displayName,
+//                "contactId": contact.id,
+//                "uuid": invitation.callkitUUID
+//            ]) { error in
+//                if let error = error {
+//                    logger.error("reportNewIncomingVoIPPushPayload error \(error.localizedDescription)")
+//                } else {
+//                    logger.debug("reportNewIncomingVoIPPushPayload success for \(contact.id)")
+//                }
+//            }
         case let .callOffer(contact, callType, offer, sharedKey, _):
-            // TODO askConfirmation?
-            // TODO check encryption is compatible
             withCall(contact) { call in
-                m.activeCall = call.copy(callState: .offerReceived, peerMedia: callType.media, sharedKey: sharedKey)
+                call.callState = .offerReceived
+                call.peerMedia = callType.media
+                call.sharedKey = sharedKey
                 m.callCommand = .offer(offer: offer.rtcSession, iceCandidates: offer.rtcIceCandidates, media: callType.media, aesKey: sharedKey, useWorker: true)
             }
         case let .callAnswer(contact, answer):
             withCall(contact) { call in
-                m.activeCall = call.copy(callState: .negotiated)
+                call.callState = .answerReceived
                 m.callCommand = .answer(answer: answer.rtcSession, iceCandidates: answer.rtcIceCandidates)
             }
         case let .callExtraInfo(contact, extraInfo):
@@ -652,9 +672,12 @@ func processReceivedMsg(_ res: ChatResponse) {
                 m.callCommand = .ice(iceCandidates: extraInfo.rtcIceCandidates)
             }
         case let .callEnded(contact):
-            m.activeCallInvitation = nil
-            withCall(contact) { _ in
+            if let invitation = m.callInvitations.removeValue(forKey: contact.id) {
+                CallController.shared.reportCallRemoteEnded(invitation: invitation)
+            }
+            withCall(contact) { call in
                 m.callCommand = .end
+                CallController.shared.reportCallRemoteEnded(call: call)
             }
         default:
             logger.debug("unsupported event: \(res.responseType)")

apps/ios/Shared/Views/Call/ActiveCallView.swift

@@ -7,160 +7,211 @@
 //
 
 import SwiftUI
+import WebKit
 
 struct ActiveCallView: View {
-    @EnvironmentObject var chatModel: ChatModel
-    @Environment(\.dismiss) private var dismiss
-    @State private var coordinator: WebRTCCoordinator? = nil
-    @State private var webViewReady: Bool = false
+    @EnvironmentObject var m: ChatModel
+    @ObservedObject var call: Call
+    @State private var rtcWebView: WKWebView? = nil
     @State private var webViewMsg: WVAPIMessage? = nil
 
     var body: some View {
         ZStack(alignment: .bottom) {
-            WebRTCView(coordinator: $coordinator, webViewReady: $webViewReady, webViewMsg: $webViewMsg)
+            WebRTCView(rtcWebView: $rtcWebView, webViewMsg: $webViewMsg)
                 .onAppear() { sendCommandToWebView() }
-                .onChange(of: chatModel.callCommand) { _ in sendCommandToWebView() }
-                .onChange(of: webViewReady) { _ in sendCommandToWebView() }
+                .onChange(of: m.callCommand) { _ in sendCommandToWebView() }
+                .onChange(of: rtcWebView) { _ in sendCommandToWebView() }
                 .onChange(of: webViewMsg) { _ in processWebViewMessage() }
                 .background(.black)
-            ActiveCallOverlay(call: chatModel.activeCall, dismiss: { dismiss() })
+            if let call = m.activeCall, let webView = rtcWebView  {
+                ActiveCallOverlay(call: call, webView: webView)
+            }
         }
         .preferredColorScheme(.dark)
     }
 
     private func sendCommandToWebView() {
-        if chatModel.activeCall != nil && webViewReady,
-           let cmd = chatModel.callCommand,
-           let c = coordinator {
-            chatModel.callCommand = nil
-            logger.debug("ActiveCallView: command \(cmd.cmdType)")
-            c.sendCommand(command: cmd)
+        if m.activeCall != nil,
+           let wv = rtcWebView,
+           let cmd = m.callCommand {
+            m.callCommand = nil
+            sendCallCommand(wv, cmd)
         }
     }
 
     private func processWebViewMessage() {
-        let m = chatModel
         if let msg = webViewMsg,
-           let call = chatModel.activeCall {
+           let call = m.activeCall,
+           let webView = rtcWebView {
             logger.debug("ActiveCallView: response \(msg.resp.respType)")
-            Task {
-                switch msg.resp {
-                case let .capabilities(capabilities):
-                    let callType = CallType(media: call.localMedia, capabilities: capabilities)
-                    try await apiSendCallInvitation(call.contact, callType)
-                    m.activeCall = call.copy(callState: .invitationSent, localCapabilities: capabilities)
-                case let .offer(offer, iceCandidates, capabilities):
-                    try await apiSendCallOffer(call.contact, offer, iceCandidates,
-                                               media: call.localMedia, capabilities: capabilities)
-                    m.activeCall = call.copy(callState: .offerSent, localCapabilities: capabilities)
-                case let .answer(answer, iceCandidates):
-                    try await apiSendCallAnswer(call.contact, answer, iceCandidates)
-                    m.activeCall = call.copy(callState: .negotiated)
-                case let .ice(iceCandidates):
-                    try await apiSendCallExtraInfo(call.contact, iceCandidates)
-                case let .connection(state):
-                    if let callStatus = WebRTCCallStatus.init(rawValue: state.connectionState),
-                       case .connected = callStatus {
-                        m.activeCall = call.copy(callState: .connected)
+            switch msg.resp {
+            case let .capabilities(capabilities):
+                let callType = CallType(media: call.localMedia, capabilities: capabilities)
+                Task {
+                    do {
+                        try await apiSendCallInvitation(call.contact, callType)
+                    } catch {
+                        logger.error("apiSendCallInvitation \(responseError(error))")
                     }
-                    try await apiCallStatus(call.contact, state.connectionState)
-                case let .connected(connectionInfo):
-                    m.activeCall = call.copy(callState: .connected, connectionInfo: connectionInfo)
-                case .ended:
-                    m.activeCall = nil
-                    m.activeCallInvitation = nil
-                    m.callCommand = nil
-                    m.showCallView = false
-                case .ok:
-                    switch msg.command {
-                    case let .media(media, enable):
-                        switch media {
-                        case .video: m.activeCall = call.copy(videoEnabled: enable)
-                        case .audio: m.activeCall = call.copy(audioEnabled: enable)
-                        }
-                    case let .camera(camera):
-                        m.activeCall = call.copy(localCamera: camera)
-                    case .end:
-                        m.activeCall = nil
-                        m.activeCallInvitation = nil
-                        m.callCommand = nil
-                        m.showCallView = false
-                    default: ()
+                    DispatchQueue.main.async {
+                        call.callState = .invitationSent
+                        call.localCapabilities = capabilities
                     }
-                case let .error(message):
-                    logger.debug("ActiveCallView: command error: \(message)")
-                case let .invalid(type):
-                    logger.debug("ActiveCallView: invalid response: \(type)")
                 }
+            case let .offer(offer, iceCandidates, capabilities):
+                Task {
+                    do {
+                        try await apiSendCallOffer(call.contact, offer, iceCandidates,
+                                               media: call.localMedia, capabilities: capabilities)
+                    } catch {
+                        logger.error("apiSendCallOffer \(responseError(error))")
+                    }
+                    DispatchQueue.main.async {
+                        call.callState = .offerSent
+                        call.localCapabilities = capabilities
+                    }
+                }
+            case let .answer(answer, iceCandidates):
+                Task {
+                    do {
+                        try await apiSendCallAnswer(call.contact, answer, iceCandidates)
+                    } catch {
+                        logger.error("apiSendCallAnswer \(responseError(error))")
+                    }
+                    DispatchQueue.main.async {
+                        call.callState = .negotiated
+                    }
+                }
+            case let .ice(iceCandidates):
+                Task {
+                    do {
+                        try await apiSendCallExtraInfo(call.contact, iceCandidates)
+                    } catch {
+                        logger.error("apiSendCallExtraInfo \(responseError(error))")
+                    }
+                }
+            case let .connection(state):
+                if let callStatus = WebRTCCallStatus.init(rawValue: state.connectionState),
+                   case .connected = callStatus {
+                    if case .outgoing = call.direction {
+                        CallController.shared.reportOutgoingCall(call: call, connectedAt: nil)
+                    }
+                    call.callState = .connected
+                    // CallKit doesn't work well with WKWebView
+                    // This is a hack to enable microphone in WKWebView after CallKit takes over it
+                    if CallController.useCallKit {
+                        DispatchQueue.main.asyncAfter(deadline: .now() + 2) {
+                            m.callCommand = .camera(camera: call.localCamera)
+                        }
+                    }
+                }
+                Task {
+                    do {
+                        try await apiCallStatus(call.contact, state.connectionState)
+                    } catch {
+                        logger.error("apiCallStatus \(responseError(error))")
+                    }
+                }
+            case let .connected(connectionInfo):
+                call.callState = .connected
+                call.connectionInfo = connectionInfo
+            case .ended:
+                closeCallView(webView)
+                call.callState = .ended
+                if let uuid = call.callkitUUID {
+                    CallController.shared.endCall(callUUID: uuid)
+                }
+            case .ok:
+                switch msg.command {
+                case .answer:
+                    call.callState = .negotiated
+                case let .camera(camera):
+                    call.localCamera = camera
+                    Task {
+                        // This disables microphone if it was disabled before flipping the camera
+                        await webView.setMicrophoneCaptureState(call.audioEnabled ? .active : .muted)
+                        // This compensates for the bug on some devices when remote video does not appear
+                        // await webView.setCameraCaptureState(.muted)
+                        // await webView.setCameraCaptureState(call.videoEnabled ? .active : .muted)
+                    }
+                case .end:
+                    closeCallView(webView)
+                    m.activeCall = nil
+                default: ()
+                }
+            case let .error(message):
+                logger.debug("ActiveCallView: command error: \(message)")
+            case let .invalid(type):
+                logger.debug("ActiveCallView: invalid response: \(type)")
             }
         }
     }
+
+    private func closeCallView(_ webView: WKWebView) {
+        m.showCallView = false
+        Task {
+            await webView.setMicrophoneCaptureState(.muted)
+            await webView.setCameraCaptureState(.muted)
+        }
+    }
 }
 
 struct ActiveCallOverlay: View {
     @EnvironmentObject var chatModel: ChatModel
-    var call: Call?
-    var dismiss: () -> Void
+    @ObservedObject var call: Call
+    var webView: WKWebView
 
     var body: some View {
         VStack {
-            if let call = call {
-                switch call.localMedia {
-                case .video:
-                    callInfoView(call, .leading)
-                    .foregroundColor(.white)
-                    .opacity(0.8)
-                    .padding()
+            switch call.localMedia {
+            case .video:
+                callInfoView(call, .leading)
+                .foregroundColor(.white)
+                .opacity(0.8)
+                .padding()
 
+                Spacer()
+
+                HStack {
+                    toggleAudioButton()
                     Spacer()
-
-                    HStack {
-                        controlButton(call, call.audioEnabled ? "mic.fill" : "mic.slash") {
-                            chatModel.callCommand = .media(media: .audio, enable: !call.audioEnabled)
-                        }
-                        Spacer()
+                    Color.clear.frame(width: 40, height: 40)
+                    Spacer()
+                    endCallButton()
+                    Spacer()
+                    if call.videoEnabled {
+                        flipCameraButton()
+                    } else {
                         Color.clear.frame(width: 40, height: 40)
-                        Spacer()
-                        callButton("phone.down.fill", size: 60) { dismiss() }
-                            .foregroundColor(.red)
-                        Spacer()
-                        controlButton(call, "arrow.triangle.2.circlepath") {
-                            chatModel.callCommand = .camera(camera: call.localCamera == .user ? .environment : .user)
-                        }
-                        Spacer()
-                        controlButton(call, call.videoEnabled ? "video.fill" : "video.slash") {
-                            chatModel.callCommand = .media(media: .video, enable: !call.videoEnabled)
-                        }
                     }
-                    .padding(.horizontal, 20)
-                    .padding(.bottom, 16)
-                    .frame(maxWidth: .infinity, alignment: .center)
-
-                case .audio:
-                    VStack {
-                        ProfileImage(imageStr: call.contact.profile.image)
-                            .scaledToFit()
-                            .frame(width: 192, height: 192)
-                        callInfoView(call, .center)
-                    }
-                    .foregroundColor(.white)
-                    .opacity(0.8)
-                    .padding()
-                    .frame(maxHeight: .infinity)
-
                     Spacer()
-
-                    ZStack(alignment: .bottom) {
-                        controlButton(call, call.audioEnabled ? "mic.fill" : "mic.slash") {
-                            chatModel.callCommand = .media(media: .audio, enable: !call.audioEnabled)
-                        }
-                        .frame(maxWidth: .infinity, alignment: .leading)
-                        callButton("phone.down.fill", size: 60) { dismiss() }
-                            .foregroundColor(.red)
-                    }
-                    .padding(.bottom, 60)
-                    .padding(.horizontal, 48)
+                    toggleVideoButton()
                 }
+                .padding(.horizontal, 20)
+                .padding(.bottom, 16)
+                .frame(maxWidth: .infinity, alignment: .center)
+
+            case .audio:
+                VStack {
+                    ProfileImage(imageStr: call.contact.profile.image)
+                        .scaledToFit()
+                        .frame(width: 192, height: 192)
+                    callInfoView(call, .center)
+                }
+                .foregroundColor(.white)
+                .opacity(0.8)
+                .padding()
+                .frame(maxHeight: .infinity)
+
+                Spacer()
+
+                ZStack(alignment: .bottom) {
+                    toggleAudioButton()
+                        .frame(maxWidth: .infinity, alignment: .leading)
+                    endCallButton()
+                }
+                .padding(.bottom, 60)
+                .padding(.horizontal, 48)
             }
         }
         .frame(maxWidth: .infinity)
@@ -186,6 +237,57 @@ struct ActiveCallOverlay: View {
         }
     }
 
+    private func endCallButton() -> some View {
+        let cc = CallController.shared
+        return callButton("phone.down.fill", size: 60) {
+            if let uuid = call.callkitUUID {
+                cc.endCall(callUUID: uuid)
+            } else {
+                cc.endCall(call: call) {}
+            }
+        }
+        .foregroundColor(.red)
+    }
+
+    private func toggleAudioButton() -> some View {
+        controlButton(call, call.audioEnabled ? "mic.fill" : "mic.slash") {
+            Task {
+                await webView.setMicrophoneCaptureState(call.audioEnabled ? .muted : .active)
+                DispatchQueue.main.async {
+                    call.audioEnabled = !call.audioEnabled
+                }
+            }
+        }
+    }
+
+    private func toggleVideoButton() -> some View {
+        controlButton(call, call.videoEnabled ? "video.fill" : "video.slash") {
+            Task {
+                await webView.setCameraCaptureState(call.videoEnabled ? .muted : .active)
+                DispatchQueue.main.async {
+                    call.videoEnabled = !call.videoEnabled
+                }
+            }
+        }
+    }
+
+    @ViewBuilder private func flipCameraButton() -> some View {
+        let cmd = WCallCommand.camera(camera: call.localCamera == .user ? .environment : .user)
+        controlButton(call, "arrow.triangle.2.circlepath") {
+            if call.audioEnabled {
+                chatModel.callCommand = cmd
+            } else {
+                Task {
+                    // Microphone has to be enabled before flipping the camera to avoid prompt for user permission when getUserMedia is called in webview
+                    await webView.setMicrophoneCaptureState(.active)
+                    DispatchQueue.main.async {
+                        chatModel.callCommand = cmd
+                    }
+                }
+            }
+        }
+    }
+
     @ViewBuilder private func controlButton(_ call: Call, _ imageName: String, _ perform: @escaping () -> Void) -> some View {
         if call.hasMedia {
             callButton(imageName, size: 40, perform)
@@ -211,9 +313,9 @@ struct ActiveCallOverlay: View {
 struct ActiveCallOverlay_Previews: PreviewProvider {
     static var previews: some View {
         Group{
-            ActiveCallOverlay(call: Call(contact: Contact.sampleData, callState: .offerSent, localMedia: .video), dismiss: {})
+            ActiveCallOverlay(call: Call(direction: .incoming, contact: Contact.sampleData, callkitUUID: UUID(), callState: .offerSent, localMedia: .video), webView: WKWebView())
                 .background(.black)
-            ActiveCallOverlay(call: Call(contact: Contact.sampleData, callState: .offerSent, localMedia: .audio), dismiss: {})
+            ActiveCallOverlay(call: Call(direction: .incoming, contact: Contact.sampleData, callkitUUID: UUID(), callState: .offerSent, localMedia: .audio), webView: WKWebView())
                 .background(.black)
         }
     }

apps/ios/Shared/Views/Call/CallController.swift

@@ -0,0 +1,215 @@
+//
+//  CallController.swift
+//  SimpleX (iOS)
+//
+//  Created by Evgeny on 21/05/2022.
+//  Copyright © 2022 SimpleX Chat. All rights reserved.
+//
+
+import Foundation
+import CallKit
+import AVFoundation
+
+class CallController: NSObject, CXProviderDelegate, ObservableObject {
+    static let useCallKit = false
+    static let shared = CallController()
+    private let provider = CXProvider(configuration: CallController.configuration)
+    private let controller = CXCallController()
+    private let callManager = CallManager()
+    @Published var activeCallInvitation: CallInvitation?
+
+// PKPushRegistry will be used from notification service extension
+//    let registry = PKPushRegistry(queue: nil)
+
+    static let configuration: CXProviderConfiguration = {
+        let configuration = CXProviderConfiguration()
+        configuration.supportsVideo = true
+        configuration.supportedHandleTypes = [.generic]
+        configuration.includesCallsInRecents = true // TODO disable or add option
+        configuration.maximumCallsPerCallGroup = 1
+        return configuration
+    }()
+
+    override init() {
+        super.init()
+        self.provider.setDelegate(self, queue: nil)
+//        self.registry.delegate = self
+//        self.registry.desiredPushTypes = [.voIP]
+    }
+
+    func providerDidReset(_ provider: CXProvider) {
+    }
+
+    func provider(_ provider: CXProvider, perform action: CXStartCallAction) {
+        logger.debug("CallController.provider CXStartCallAction")
+        if callManager.startOutgoingCall(callUUID: action.callUUID) {
+            action.fulfill()
+            provider.reportOutgoingCall(with: action.callUUID, startedConnectingAt: nil)
+        } else {
+            action.fail()
+        }
+    }
+
+    func provider(_ provider: CXProvider, perform action: CXAnswerCallAction) {
+        logger.debug("CallController.provider CXAnswerCallAction")
+        if callManager.answerIncomingCall(callUUID: action.callUUID) {
+            action.fulfill()
+        } else {
+            action.fail()
+        }
+    }
+
+    func provider(_ provider: CXProvider, perform action: CXEndCallAction) {
+        logger.debug("CallController.provider CXEndCallAction")
+        callManager.endCall(callUUID: action.callUUID) { ok in
+            if ok {
+                action.fulfill()
+            } else {
+                action.fail()
+            }
+        }
+    }
+
+    func provider(_ provider: CXProvider, timedOutPerforming action: CXAction) {
+        print("timed out", #function)
+        action.fulfill()
+    }
+
+    func provider(_ provider: CXProvider, didActivate audioSession: AVAudioSession) {
+        print("received", #function)
+//        do {
+//            try audioSession.setCategory(.playAndRecord, mode: .voiceChat, options: .mixWithOthers)
+//            logger.debug("audioSession category set")
+//            try audioSession.setActive(true)
+//            logger.debug("audioSession activated")
+//        } catch {
+//            print(error)
+//            logger.error("failed activating audio session")
+//        }
+    }
+
+    func provider(_ provider: CXProvider, didDeactivate audioSession: AVAudioSession) {
+        print("received", #function)
+    }
+
+//    func pushRegistry(_ registry: PKPushRegistry, didUpdate pushCredentials: PKPushCredentials, for type: PKPushType) {
+//
+//    }
+
+// This will be needed when we have notification service extension
+//    func pushRegistry(_ registry: PKPushRegistry, didReceiveIncomingPushWith payload: PKPushPayload, for type: PKPushType, completion: @escaping () -> Void) {
+//        if type == .voIP {
+//           // Extract the call information from the push notification payload
+//            if let displayName = payload.dictionaryPayload["displayName"] as? String,
+//               let contactId = payload.dictionaryPayload["contactId"] as? String,
+//               let uuidStr = payload.dictionaryPayload["uuid"] as? String,
+//               let uuid = UUID(uuidString: uuidStr) {
+//                let callUpdate = CXCallUpdate()
+//                callUpdate.remoteHandle = CXHandle(type: .phoneNumber, value: displayName)
+//                provider.reportNewIncomingCall(with: uuid, update: callUpdate, completion: { error in
+//                    if error != nil {
+//                        let m = ChatModel.shared
+//                        m.callInvitations.removeValue(forKey: contactId)
+//                    }
+//                    // Tell PushKit that the notification is handled.
+//                    completion()
+//                })
+//            }
+//        }
+//    }
+
+    func reportNewIncomingCall(invitation: CallInvitation, completion: @escaping (Error?) -> Void) {
+        logger.debug("CallController.reportNewIncomingCall")
+        if CallController.useCallKit, let uuid = invitation.callkitUUID {
+            let update = CXCallUpdate()
+            update.remoteHandle = CXHandle(type: .generic, value: invitation.contact.displayName)
+            update.hasVideo = invitation.peerMedia == .video
+            provider.reportNewIncomingCall(with: uuid, update: update, completion: completion)
+        } else {
+            NtfManager.shared.notifyCallInvitation(invitation)
+            activeCallInvitation = invitation
+        }
+    }
+
+    func reportOutgoingCall(call: Call, connectedAt dateConnected: Date?) {
+        if CallController.useCallKit, let uuid = call.callkitUUID {
+            provider.reportOutgoingCall(with: uuid, connectedAt: dateConnected)
+        }
+    }
+
+    func reportCallRemoteEnded(invitation: CallInvitation) {
+        if CallController.useCallKit, let uuid = invitation.callkitUUID {
+            provider.reportCall(with: uuid, endedAt: nil, reason: .remoteEnded)
+        } else if invitation.contact.id == activeCallInvitation?.contact.id {
+            activeCallInvitation = nil
+        }
+    }
+
+    func reportCallRemoteEnded(call: Call) {
+        if CallController.useCallKit, let uuid = call.callkitUUID {
+            provider.reportCall(with: uuid, endedAt: nil, reason: .remoteEnded)
+        }
+    }
+
+    func startCall(_ contact: Contact, _ media: CallMediaType) {
+        logger.debug("CallController.startCall")
+        let uuid = callManager.newOutgoingCall(contact, media)
+        if CallController.useCallKit {
+            let handle = CXHandle(type: .generic, value: contact.displayName)
+            let action = CXStartCallAction(call: uuid, handle: handle)
+            action.isVideo = media == .video
+            requestTransaction(with: action)
+        } else if callManager.startOutgoingCall(callUUID: uuid) {
+            logger.debug("CallController.startCall: call started")
+        } else {
+            logger.error("CallController.startCall: no active call")
+        }
+    }
+
+    func answerCall(invitation: CallInvitation) {
+        callManager.answerIncomingCall(invitation: invitation)
+        if invitation.contact.id == self.activeCallInvitation?.contact.id {
+            self.activeCallInvitation = nil
+        }
+    }
+
+    func endCall(callUUID: UUID) {
+        if CallController.useCallKit {
+            requestTransaction(with: CXEndCallAction(call: callUUID))
+        } else {
+            callManager.endCall(callUUID: callUUID) { ok in
+                if ok {
+                    logger.debug("CallController.endCall: call ended")
+                } else {
+                    logger.error("CallController.endCall: no actove call pr call invitation to end")
+                }
+            }
+        }
+    }
+
+    func endCall(invitation: CallInvitation) {
+        callManager.endCall(invitation: invitation) {
+            if invitation.contact.id == self.activeCallInvitation?.contact.id {
+                DispatchQueue.main.async {
+                    self.activeCallInvitation = nil
+                }
+            }
+        }
+    }
+
+    func endCall(call: Call, completed: @escaping () -> Void) {
+        callManager.endCall(call: call, completed: completed)
+    }
+
+    private func requestTransaction(with action: CXAction) {
+        let t = CXTransaction()
+        t.addAction(action)
+        controller.request(t) { error in
+            if let error = error {
+                logger.error("CallController.requestTransaction error requesting transaction: \(error.localizedDescription)")
+            } else {
+                logger.debug("CallController.requestTransaction requested transaction successfully")
+            }
+        }
+    }
+}

apps/ios/Shared/Views/Call/CallManager.swift

@@ -0,0 +1,104 @@
+//
+//  CallManager.swift
+//  SimpleX (iOS)
+//
+//  Created by Evgeny on 22/05/2022.
+//  Copyright © 2022 SimpleX Chat. All rights reserved.
+//
+
+import Foundation
+
+class CallManager {
+    func newOutgoingCall(_ contact: Contact, _ media: CallMediaType) -> UUID {
+        let uuid = UUID()
+        ChatModel.shared.activeCall = Call(direction: .outgoing, contact: contact, callkitUUID: uuid, callState: .waitCapabilities, localMedia: media)
+        return uuid
+    }
+
+    func startOutgoingCall(callUUID: UUID) -> Bool {
+        let m = ChatModel.shared
+        if let call = m.activeCall, call.callkitUUID == callUUID {
+            m.showCallView = true
+            m.callCommand = .capabilities(useWorker: true)
+            return true
+        }
+        return false
+    }
+
+    func answerIncomingCall(callUUID: UUID) -> Bool {
+        if let invitation = getCallInvitation(callUUID) {
+            answerIncomingCall(invitation: invitation)
+            return true
+        }
+        return false
+    }
+
+    func answerIncomingCall(invitation: CallInvitation) {
+        let m = ChatModel.shared
+        m.callInvitations.removeValue(forKey: invitation.contact.id)
+        m.activeCall = Call(
+            direction: .incoming,
+            contact: invitation.contact,
+            callkitUUID: invitation.callkitUUID,
+            callState: .invitationAccepted,
+            localMedia: invitation.peerMedia,
+            sharedKey: invitation.sharedKey
+        )
+        m.showCallView = true
+        m.callCommand = .start(media: invitation.peerMedia, aesKey: invitation.sharedKey, useWorker: true)
+    }
+
+    func endCall(callUUID: UUID, completed: @escaping (Bool) -> Void) {
+        if let call = ChatModel.shared.activeCall, call.callkitUUID == callUUID {
+            endCall(call: call) { completed(true) }
+        } else if let invitation = getCallInvitation(callUUID) {
+            endCall(invitation: invitation) { completed(true) }
+        } else {
+            completed(false)
+        }
+    }
+
+    func endCall(call: Call, completed: @escaping () -> Void) {
+        let m = ChatModel.shared
+        if case .ended = call.callState {
+            logger.debug("CallController.provider CXEndCallAction: call ended")
+            m.activeCall = nil
+            m.showCallView = false
+            completed()
+        } else {
+            logger.debug("CallController.provider CXEndCallAction: ending call...")
+            m.callCommand = .end
+            m.showCallView = false
+            Task {
+                do {
+                    try await apiEndCall(call.contact)
+                } catch {
+                    logger.error("CallController.provider apiEndCall error: \(responseError(error))")
+                }
+                DispatchQueue.main.async {
+                    m.activeCall = nil
+                    completed()
+                }
+            }
+        }
+    }
+
+    func endCall(invitation: CallInvitation, completed: @escaping () -> Void) {
+        ChatModel.shared.callInvitations.removeValue(forKey: invitation.contact.id)
+        Task {
+            do {
+                try await apiRejectCall(invitation.contact)
+            } catch {
+                logger.error("CallController.provider apiRejectCall error: \(responseError(error))")
+            }
+            completed()
+        }
+    }
+
+    private func getCallInvitation(_ callUUID: UUID) -> CallInvitation? {
+        if let (_, invitation) = ChatModel.shared.callInvitations.first(where: { (_, inv) in inv.callkitUUID == callUUID }) {
+            return invitation
+        }
+        return nil
+    }
+}

apps/ios/Shared/Views/Call/IncomingCallView.swift

@@ -0,0 +1,86 @@
+//
+//  IncomingCallView.swift
+//  SimpleX (iOS)
+//
+//  Created by Evgeny on 24/05/2022.
+//  Copyright © 2022 SimpleX Chat. All rights reserved.
+//
+
+import SwiftUI
+
+struct IncomingCallView: View {
+    @EnvironmentObject var m: ChatModel
+    @ObservedObject var cc = CallController.shared
+
+    var body: some View {
+        let sp = SoundPlayer.shared
+        if let invitation = cc.activeCallInvitation {
+            if m.showCallView {
+                incomingCall(invitation)
+            } else {
+                incomingCall(invitation)
+                .onAppear { sp.startRingtone() }
+                .onDisappear { sp.stopRingtone() }
+            }
+        }
+    }
+
+    private func incomingCall(_ invitation: CallInvitation) -> some View {
+        VStack(alignment: .leading, spacing: 6) {
+            HStack {
+                Image(systemName: invitation.peerMedia == .video ? "video.fill" : "phone.fill").foregroundColor(.green)
+                Text(invitation.callTypeText)
+            }
+            HStack {
+                ProfilePreview(profileOf: invitation.contact, color: .white)
+                Spacer()
+
+                callButton("Reject", "phone.down.fill", .red) {
+                    cc.endCall(invitation: invitation)
+                }
+
+                callButton("Ignore", "multiply", .accentColor) {
+                    cc.activeCallInvitation = nil
+                }
+
+                callButton("Accept", "checkmark", .green) {
+                    if let call = m.activeCall {
+                        cc.endCall(call: call) {
+                            DispatchQueue.main.async {
+                                cc.answerCall(invitation: invitation)
+                            }
+                        }
+                    } else {
+                        cc.answerCall(invitation: invitation)
+                    }
+                }
+            }
+        }
+        .padding(.horizontal, 16)
+        .padding(.vertical, 12)
+        .frame(maxWidth: .infinity)
+        .background(Color(uiColor: .tertiarySystemGroupedBackground))
+    }
+
+    private func callButton(_ text: LocalizedStringKey, _ image: String, _ color: Color, action: @escaping () -> Void) -> some View {
+        Button(action: action, label: {
+            VStack(spacing: 2) {
+                Image(systemName: image)
+                    .scaleEffect(1.24)
+                    .foregroundColor(color)
+                    .frame(width: 24, height: 24)
+                Text(text)
+                    .font(.caption)
+                    .foregroundColor(.secondary)
+            }
+            .frame(minWidth: 44)
+        })
+    }
+}
+
+struct IncomingCallView_Previews: PreviewProvider {
+    static var previews: some View {
+        CallController.shared.activeCallInvitation = CallInvitation.sampleData
+        return IncomingCallView()
+    }
+}

apps/ios/Shared/Views/Call/SoundPlayer.swift

@@ -0,0 +1,45 @@
+//
+//  SoundPlayer.swift
+//  SimpleX (iOS)
+//
+//  Created by Evgeny on 24/05/2022.
+//  Copyright © 2022 SimpleX Chat. All rights reserved.
+//
+
+import Foundation
+import AVFoundation
+
+class SoundPlayer {
+    static let shared = SoundPlayer()
+    private var audioPlayer: AVAudioPlayer?
+
+    func startRingtone() {
+        audioPlayer?.stop()
+        logger.debug("startRingtone")
+        guard let path = Bundle.main.path(forResource: "ringtone2", ofType: "m4a", inDirectory: "sounds") else {
+            logger.debug("startRingtone: file not found")
+            return
+        }
+        do {
+            let player = try AVAudioPlayer(contentsOf: URL(fileURLWithPath: path))
+            if player.prepareToPlay() {
+                audioPlayer = player
+            }
+        } catch {
+            logger.debug("startRingtone: AVAudioPlayer error \(error.localizedDescription)")
+        }
+
+        Task {
+            while let player = audioPlayer {
+                player.play()
+                AudioServicesPlayAlertSound(kSystemSoundID_Vibrate)
+                _ = try? await Task.sleep(nanoseconds: UInt64(player.duration * 1_000_000_000))
+            }
+        }
+    }
+
+    func stopRingtone() {
+        audioPlayer?.stop()
+        audioPlayer = nil
+    }
+}

apps/ios/Shared/Views/Call/WebRTC.swift

@@ -9,70 +9,39 @@
 import Foundation
 import SwiftUI
 
-class Call: Equatable {
+class Call: ObservableObject, Equatable {
     static func == (lhs: Call, rhs: Call) -> Bool {
         lhs.contact.apiId == rhs.contact.apiId
     }
 
+    var direction: CallDirection
     var contact: Contact
-    var callState: CallState
+    var callkitUUID: UUID?
     var localMedia: CallMediaType
-    var localCapabilities: CallCapabilities?
-    var peerMedia: CallMediaType?
-    var sharedKey: String?
-    var audioEnabled: Bool
-    var videoEnabled: Bool
-    var localCamera: VideoCamera
-    var connectionInfo: ConnectionInfo?
+    @Published var callState: CallState
+    @Published var localCapabilities: CallCapabilities?
+    @Published var peerMedia: CallMediaType?
+    @Published var sharedKey: String?
+    @Published var audioEnabled = true
+    @Published var videoEnabled: Bool
+    @Published var localCamera = VideoCamera.user
+    @Published var connectionInfo: ConnectionInfo?
 
     init(
+        direction: CallDirection,
         contact: Contact,
+        callkitUUID: UUID?,
         callState: CallState,
         localMedia: CallMediaType,
-        localCapabilities: CallCapabilities? = nil,
-        peerMedia: CallMediaType? = nil,
-        sharedKey: String? = nil,
-        audioEnabled: Bool? = nil,
-        videoEnabled: Bool? = nil,
-        localCamera: VideoCamera = .user,
-        connectionInfo: ConnectionInfo? = nil
+        sharedKey: String? = nil
     ) {
+        self.direction = direction
         self.contact = contact
+        self.callkitUUID = callkitUUID
         self.callState = callState
         self.localMedia = localMedia
-        self.localCapabilities = localCapabilities
-        self.peerMedia = peerMedia
         self.sharedKey = sharedKey
-        self.audioEnabled = audioEnabled ?? true
-        self.videoEnabled = videoEnabled ?? (localMedia == .video)
-        self.localCamera = localCamera
-        self.connectionInfo = connectionInfo
-    }
-
-    func copy(
-        contact: Contact? = nil,
-        callState: CallState? = nil,
-        localMedia: CallMediaType? = nil,
-        localCapabilities: CallCapabilities? = nil,
-        peerMedia: CallMediaType? = nil,
-        sharedKey: String? = nil,
-        audioEnabled: Bool? = nil,
-        videoEnabled: Bool? = nil,
-        localCamera: VideoCamera? = nil,
-        connectionInfo: ConnectionInfo? = nil
-    ) -> Call {
-        Call (
-            contact: contact ?? self.contact,
-            callState: callState ?? self.callState,
-            localMedia: localMedia ?? self.localMedia,
-            localCapabilities: localCapabilities ?? self.localCapabilities,
-            peerMedia: peerMedia ?? self.peerMedia,
-            sharedKey: sharedKey ?? self.sharedKey,
-            audioEnabled: audioEnabled ?? self.audioEnabled,
-            videoEnabled: videoEnabled ?? self.videoEnabled,
-            localCamera: localCamera ?? self.localCamera,
-            connectionInfo: connectionInfo ?? self.connectionInfo
-        )
+        self.videoEnabled = localMedia == .video
     }
 
     var encrypted: Bool { get { localEncrypted && sharedKey != nil } }
@@ -82,7 +51,7 @@ class Call: Equatable {
             switch callState {
             case .waitCapabilities: return ""
             case .invitationSent: return localEncrypted ? "e2e encrypted" : "no e2e encryption"
-            case .invitationReceived: return sharedKey == nil ? "contact has no e2e encryption" : "contact has e2e encryption"
+            case .invitationAccepted: return sharedKey == nil ? "contact has no e2e encryption" : "contact has e2e encryption"
             default: return !localEncrypted ? "no e2e encryption" : sharedKey == nil ? "contact has no e2e encryption" : "e2e encrypted" 
             }
         }
@@ -90,24 +59,33 @@ class Call: Equatable {
     var hasMedia: Bool { get { callState == .offerSent || callState == .negotiated || callState == .connected } }
 }
 
+enum CallDirection {
+    case incoming
+    case outgoing
+}
+
 enum CallState {
-    case waitCapabilities
-    case invitationSent
-    case invitationReceived
-    case offerSent
-    case offerReceived
-    case negotiated
+    case waitCapabilities   // outgoing call started
+    case invitationSent     // outgoing call - sent invitation
+    case invitationAccepted // incoming call started
+    case offerSent          // incoming - webrtc started and offer sent
+    case offerReceived      // outgoing - webrtc offer received via API
+    case answerReceived     // incoming - webrtc answer received via API
+    case negotiated         // outgoing - webrtc offer processed and answer sent, incoming - webrtc answer processed
     case connected
+    case ended
 
     var text: LocalizedStringKey {
         switch self {
         case .waitCapabilities: return "starting…"
         case .invitationSent: return "waiting for answer…"
-        case .invitationReceived: return "starting…"
+        case .invitationAccepted: return "starting…"
         case .offerSent: return "waiting for confirmation…"
         case .offerReceived: return "received answer…"
+        case .answerReceived: return "received confirmation…"
         case .negotiated: return "connecting…"
         case .connected: return "connected"
+        case .ended: return "ended"
         }
     }
 }
@@ -275,16 +253,16 @@ enum WCallResponse: Equatable, Decodable {
     var respType: String {
         get {
             switch self {
-            case .capabilities: return("capabilities")
-            case .offer: return("offer")
-            case .answer: return("answer")
-            case .ice: return("ice")
-            case .connection: return("connection")
-            case .connected: return("connected")
-            case .ended: return("ended")
-            case .ok: return("ok")
-            case .error: return("error")
-            case .invalid: return("invalid")
+            case .capabilities: return "capabilities"
+            case .offer: return "offer"
+            case .answer: return "answer"
+            case .ice: return "ice"
+            case .connection: return "connection"
+            case .connected: return "connected"
+            case .ended: return "ended"
+            case .ok: return "ok"
+            case .error: return "error"
+            case .invalid: return "invalid"
             }
         }
     }

apps/ios/Shared/Views/Call/WebRTCView.swift

@@ -9,20 +9,24 @@
 import SwiftUI
 import WebKit
 
-class WebRTCCoordinator: NSObject, WKNavigationDelegate, WKScriptMessageHandler {
-    var webViewReady: Binding<Bool>
+class WebRTCCoordinator: NSObject, WKNavigationDelegate, WKScriptMessageHandler, WKUIDelegate {
+    var rtcWebView: Binding<WKWebView?>
     var webViewMsg: Binding<WVAPIMessage?>
-    private var webView: WKWebView?
 
-    internal init(webViewReady: Binding<Bool>, webViewMsg: Binding<WVAPIMessage?>) {
-        self.webViewReady = webViewReady
+    internal init(rtcWebView: Binding<WKWebView?>, webViewMsg: Binding<WVAPIMessage?>) {
+        self.rtcWebView = rtcWebView
         self.webViewMsg = webViewMsg
     }
 
     func webView(_ webView: WKWebView, didFinish navigation: WKNavigation!) {
         webView.allowsBackForwardNavigationGestures = false
-        self.webView = webView
-        webViewReady.wrappedValue = true
+        self.rtcWebView.wrappedValue = webView
+        ChatModel.shared.callWebView = webView
+    }
+
+    func webView(_ webView: WKWebView, decideMediaCapturePermissionsFor origin : WKSecurityOrigin, initiatedBy frame: WKFrameInfo, type: WKMediaCaptureType) async -> WKPermissionDecision {
+        print("webView", #function)
+        return .grant
     }
 
     // receive message from WKWebView
@@ -31,34 +35,37 @@ class WebRTCCoordinator: NSObject, WKNavigationDelegate, WKScriptMessageHandler
         didReceive message: WKScriptMessage
     ) {
         logger.debug("WebRTCCoordinator.userContentController")
-        if let msgStr = message.body as? String,
-           let msg: WVAPIMessage = decodeJSON(msgStr) {
-            webViewMsg.wrappedValue = msg
-            if case .invalid = msg.resp {
-                logger.error("WebRTCCoordinator.userContentController: invalid message \(String(describing: message.body))")
+        switch message.name {
+        case "webrtc":
+            if let msgStr = message.body as? String,
+               let msg: WVAPIMessage = decodeJSON(msgStr) {
+                // this is the binding that communicates messages from webview to swift view
+                webViewMsg.wrappedValue = msg
+                if case .invalid = msg.resp {
+                    logger.error("WebRTCCoordinator.userContentController: invalid message \(String(describing: message.body))")
+                }
+            } else {
+                logger.error("WebRTCCoordinator.userContentController: message parsing error \(String(describing: message.body))")
             }
-        } else {
-            logger.error("WebRTCCoordinator.userContentController: message parsing error \(String(describing: message.body))")
-        }
-    }
-
-    func sendCommand(command: WCallCommand) {
-        if let webView = webView {
-            logger.debug("WebRTCCoordinator.sendCommand")
-            let apiCmd = encodeJSON(WVAPICall(command: command))
-            let js = "processCommand(\(apiCmd))"
-            webView.evaluateJavaScript(js)
+        case "logger":
+            if let msgStr = message.body as? String {
+                logger.error("WebRTCCoordinator console.log: \(msgStr)")
+            } else {
+                logger.error("WebRTCCoordinator console.log: \(String(describing: message.body))")
+            }
+        default:
+            logger.error("WebRTCCoordinator.userContentController: invalid message.name \(message.name)")
         }
     }
 }
 
 struct WebRTCView: UIViewRepresentable {
-    @Binding var coordinator: WebRTCCoordinator?
-    @Binding var webViewReady: Bool
+    @State private var coordinator: WebRTCCoordinator?
+    @Binding var rtcWebView: WKWebView?
     @Binding var webViewMsg: WVAPIMessage?
 
     func makeCoordinator() -> WebRTCCoordinator {
-        WebRTCCoordinator(webViewReady: $webViewReady, webViewMsg: $webViewMsg)
+        WebRTCCoordinator(rtcWebView: $rtcWebView, webViewMsg: $webViewMsg)
     }
 
     func makeUIView(context: Context) -> WKWebView {
@@ -72,10 +79,14 @@ struct WebRTCView: UIViewRepresentable {
         cfg.mediaTypesRequiringUserActionForPlayback = []
         cfg.allowsInlineMediaPlayback = true
 
-        let source = "sendMessageToNative = (msg) => webkit.messageHandlers.webrtc.postMessage(JSON.stringify(msg))"
-        let script = WKUserScript(source: source, injectionTime: .atDocumentEnd, forMainFrameOnly: false)
-        wkController.addUserScript(script)
-        wkController.add(wkCoordinator, name: "webrtc")
+        let addScript = { (handler: String, source: String) in
+            let script = WKUserScript(source: source, injectionTime: .atDocumentEnd, forMainFrameOnly: false)
+            wkController.addUserScript(script)
+            wkController.add(wkCoordinator, name: handler)
+        }
+
+        addScript("webrtc", "sendMessageToNative = (msg) => webkit.messageHandlers.webrtc.postMessage(JSON.stringify(msg))")
+        addScript("logger", "console.log = (arg) => webkit.messageHandlers.logger.postMessage(JSON.stringify(arg))")
 
         let wkWebView = WKWebView(frame: .zero, configuration: cfg)
         wkWebView.navigationDelegate = wkCoordinator
@@ -93,16 +104,22 @@ struct WebRTCView: UIViewRepresentable {
     }
 }
 
+func sendCallCommand(_ webView: WKWebView, _ command: WCallCommand) {
+    logger.debug("sendCallCommand: \(command.cmdType)")
+    let apiCmd = encodeJSON(WVAPICall(command: command))
+    let js = "processCommand(\(apiCmd))"
+    webView.evaluateJavaScript(js)
+}
+
 struct CallViewDebug: View {
-    @State private var coordinator: WebRTCCoordinator? = nil
     @State private var commandStr = ""
-    @State private var webViewReady: Bool = false
+    @State private var rtcWebView: WKWebView? = nil
     @State private var webViewMsg: WVAPIMessage? = nil
     @FocusState private var keyboardVisible: Bool
 
     var body: some View {
         VStack(spacing: 30) {
-            WebRTCView(coordinator: $coordinator, webViewReady: $webViewReady, webViewMsg: $webViewMsg).frame(maxHeight: 260)
+            WebRTCView(rtcWebView: $rtcWebView, webViewMsg: $webViewMsg).frame(maxHeight: 260)
                 .onChange(of: webViewMsg) { _ in
                     if let resp = webViewMsg {
                         commandStr = encodeJSON(resp)
@@ -130,21 +147,21 @@ struct CallViewDebug: View {
                     commandStr = ""
                 }
                 Button("Send") {
-                    if let c = coordinator,
+                    if let wv = rtcWebView,
                        let command: WCallCommand = decodeJSON(commandStr) {
-                        c.sendCommand(command: command)
+                        sendCallCommand(wv, command)
                     }
                 }
             }
             HStack(spacing: 20) {
                 Button("Capabilities") {
-                    if let c = coordinator {
-                        c.sendCommand(command: .capabilities(useWorker: true))
+                    if let wv = rtcWebView {
+                        sendCallCommand(wv, .capabilities(useWorker: true))
                     }
                 }
                 Button("Start") {
-                    if let c = coordinator {
-                        c.sendCommand(command: .start(media: .video))
+                    if let wv = rtcWebView {
+                        sendCallCommand(wv, .start(media: .video))
                     }
                 }
                 Button("Accept") {

apps/ios/Shared/Views/Chat/ChatItem/CICallItemView.swift

@@ -26,7 +26,7 @@ struct CICallItemView: View {
                     acceptCallButton()
                 }
             case .missed: missedCallIcon(sent).foregroundColor(.red)
-            case .rejected: Image(systemName: "phone.down").foregroundColor(.secondary)
+            case .rejected: Image(systemName: "phone.down").foregroundColor(.red)
             case .accepted: connectingCallIcon()
             case .negotiated: connectingCallIcon()
             case .progress: Image(systemName: "phone.and.waveform.fill").foregroundColor(.green)
@@ -61,16 +61,9 @@ struct CICallItemView: View {
     @ViewBuilder private func acceptCallButton() -> some View {
         if case let .direct(contact) = chatInfo {
             Button {
-                if let invitation = m.callInvitations.removeValue(forKey: contact.id) {
-                    m.activeCallInvitation = nil
-                    m.activeCall = Call(
-                        contact: contact,
-                        callState: .invitationReceived,
-                        localMedia: invitation.peerMedia,
-                        sharedKey: invitation.sharedKey
-                    )
-                    m.showCallView = true
-                    m.callCommand = .start(media: invitation.peerMedia, aesKey: invitation.sharedKey, useWorker: true)
+                if let invitation = m.callInvitations[contact.id] {
+                    CallController.shared.answerCall(invitation: invitation)
+                    logger.debug("acceptCallButton call answered")
                 } else {
                     AlertManager.shared.showAlertMsg(title: "Call already ended!")
                 }

apps/ios/Shared/Views/Chat/ChatView.swift

@@ -105,27 +105,21 @@ struct ChatView: View {
                     ChatInfoView(chat: chat, showChatInfo: $showChatInfo)
                 }
             }
-//            ToolbarItem(placement: .navigationBarTrailing) {
-//                if case let .direct(contact) = cInfo {
-//                    HStack {
-//                        callButton(contact, .audio, imageName: "phone")
-//                        callButton(contact, .video, imageName: "video")
-//                    }
-//                }
-//            }
+            ToolbarItem(placement: .navigationBarTrailing) {
+                if case let .direct(contact) = cInfo {
+                    HStack {
+                        callButton(contact, .audio, imageName: "phone")
+                        callButton(contact, .video, imageName: "video")
+                    }
+                }
+            }
         }
         .navigationBarBackButtonHidden(true)
     }
 
     private func callButton(_ contact: Contact, _ media: CallMediaType, imageName: String) -> some View {
         Button {
-            chatModel.activeCall = Call(
-                contact: contact,
-                callState: .waitCapabilities,
-                localMedia: media
-            )
-            chatModel.showCallView = true
-            chatModel.callCommand = .capabilities(useWorker: true)
+            CallController.shared.startCall(contact, media)
         } label: {
             Image(systemName: imageName)
         }

apps/ios/Shared/Views/ChatList/ChatListView.swift

@@ -49,29 +49,6 @@ struct ChatListView: View {
                     NewChatButton()
                 }
             }
-            .fullScreenCover(isPresented: $chatModel.showCallView) {
-                ActiveCallView()
-            }
-            .onChange(of: chatModel.showCallView) { _ in
-                if (chatModel.showCallView) { return }
-                if let call = chatModel.activeCall {
-                    Task {
-                        do {
-                            try await apiEndCall(call.contact)
-                        } catch {
-                            logger.error("ChatListView apiEndCall error: \(error.localizedDescription)")
-                        }
-                    }
-                }
-                chatModel.callCommand = .end
-            }
-            .onChange(of: chatModel.activeCallInvitation) { _ in
-                if let contactRef = chatModel.activeCallInvitation,
-                   case let .direct(contact) = chatModel.getChat(contactRef.id)?.chatInfo,
-                   let invitation = chatModel.callInvitations[contactRef.id] {
-                    answerCallAlert(contact, invitation)
-                }
-            }
         }
         .navigationViewStyle(.stack)
 
@@ -95,34 +72,6 @@ struct ChatListView: View {
                 $0.chatInfo.chatViewName.localizedLowercase.contains(s)
             }
     }
-
-    private func answerCallAlert(_ contact: Contact, _ invitation: CallInvitation) {
-        return AlertManager.shared.showAlert(Alert(
-            title: Text(invitation.callTitle),
-            message: Text(contact.profile.displayName).bold() +
-                Text(" wants to connect with you via ") +
-                Text(invitation.callTypeText),
-            primaryButton: .default(Text("Answer")) {
-                if let activeCallInvitation = chatModel.activeCallInvitation {
-                    chatModel.callInvitations.removeValue(forKey: activeCallInvitation.id)
-                    chatModel.activeCallInvitation = nil
-                    chatModel.activeCall = Call(
-                        contact: contact,
-                        callState: .invitationReceived,
-                        localMedia: invitation.peerMedia,
-                        sharedKey: invitation.sharedKey
-                    )
-                    chatModel.showCallView = true
-                    chatModel.callCommand = .start(media: invitation.peerMedia, aesKey: invitation.sharedKey, useWorker: true)
-                } else {
-                    DispatchQueue.main.async {
-                        AlertManager.shared.showAlertMsg(title: "Call already ended!")
-                    }
-                }
-            },
-            secondaryButton: .cancel()
-        ))
-    }
 }
 
 struct ChatListView_Previews: PreviewProvider {

apps/ios/Shared/Views/UserSettings/SettingsView.swift

@@ -38,18 +38,7 @@ struct SettingsView: View {
                         UserProfile()
                             .navigationTitle("Your chat profile")
                     } label: {
-                        HStack {
-                            ProfileImage(imageStr: user.image)
-                                .frame(width: 44, height: 44)
-                                .padding(.trailing, 6)
-                                .padding(.vertical, 6)
-                            VStack(alignment: .leading) {
-                                Text(user.displayName)
-                                    .fontWeight(.bold)
-                                    .font(.title2)
-                                Text(user.fullName)
-                            }
-                        }
+                        ProfilePreview(profileOf: user)
                         .padding(.leading, -8)
                     }
                     NavigationLink {
@@ -242,6 +231,26 @@ struct SettingsView: View {
     }
 }
 
+struct ProfilePreview: View {
+    var profileOf: NamedChat
+    var color = Color(uiColor: .tertiarySystemGroupedBackground)
+
+    var body: some View {
+        HStack {
+            ProfileImage(imageStr: profileOf.image, color: color)
+                .frame(width: 44, height: 44)
+                .padding(.trailing, 6)
+                .padding(.vertical, 6)
+            VStack(alignment: .leading) {
+                Text(profileOf.displayName)
+                    .fontWeight(.bold)
+                    .font(.title2)
+                Text(profileOf.fullName)
+            }
+        }
+    }
+}
+
 struct SettingsView_Previews: PreviewProvider {
     static var previews: some View {
         let chatModel = ChatModel()

apps/ios/SimpleX Localizations/en.xcloc/Localized Contents/en.xliff

@@ -90,14 +90,14 @@
         <target>**Scan QR code**: to connect to your contact in person or via video call.</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
-      <trans-unit id="**e2e encrypted** audio call." xml:space="preserve">
-        <source>**e2e encrypted** audio call.</source>
-        <target>**e2e encrypted** audio call.</target>
+      <trans-unit id="**e2e encrypted** audio call" xml:space="preserve">
+        <source>**e2e encrypted** audio call</source>
+        <target>**e2e encrypted** audio call</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
-      <trans-unit id="**e2e encrypted** video call." xml:space="preserve">
-        <source>**e2e encrypted** video call.</source>
-        <target>**e2e encrypted** video call.</target>
+      <trans-unit id="**e2e encrypted** video call" xml:space="preserve">
+        <source>**e2e encrypted** video call</source>
+        <target>**e2e encrypted** video call</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
       <trans-unit id="*bold*" xml:space="preserve">
@@ -1118,9 +1118,9 @@ SimpleX servers cannot see your profile.</target>
         <target>accepted</target>
         <note>call status</note>
       </trans-unit>
-      <trans-unit id="audio call (not e2e encrypted)." xml:space="preserve">
-        <source>audio call (not e2e encrypted).</source>
-        <target>audio call (not e2e encrypted).</target>
+      <trans-unit id="audio call (not e2e encrypted)" xml:space="preserve">
+        <source>audio call (not e2e encrypted)</source>
+        <target>audio call (not e2e encrypted)</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
       <trans-unit id="bold" xml:space="preserve">
@@ -1279,9 +1279,9 @@ SimpleX servers cannot see your profile.</target>
         <target>via relay</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
-      <trans-unit id="video call (not e2e encrypted)." xml:space="preserve">
-        <source>video call (not e2e encrypted).</source>
-        <target>video call (not e2e encrypted).</target>
+      <trans-unit id="video call (not e2e encrypted)" xml:space="preserve">
+        <source>video call (not e2e encrypted)</source>
+        <target>video call (not e2e encrypted)</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
       <trans-unit id="waiting for answer…" xml:space="preserve">
@@ -1380,14 +1380,14 @@ SimpleX servers cannot see your profile.</target>
         <target>%@ wants to connect!</target>
         <note>notification title</note>
       </trans-unit>
-      <trans-unit id="**e2e encrypted** audio call." xml:space="preserve">
-        <source>**e2e encrypted** audio call.</source>
-        <target>**e2e encrypted** audio call.</target>
+      <trans-unit id="**e2e encrypted** audio call" xml:space="preserve">
+        <source>**e2e encrypted** audio call</source>
+        <target>**e2e encrypted** audio call</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
-      <trans-unit id="**e2e encrypted** video call." xml:space="preserve">
-        <source>**e2e encrypted** video call.</source>
-        <target>**e2e encrypted** video call.</target>
+      <trans-unit id="**e2e encrypted** video call" xml:space="preserve">
+        <source>**e2e encrypted** video call</source>
+        <target>**e2e encrypted** video call</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
       <trans-unit id="Accept contact request from %@?" xml:space="preserve">
@@ -1415,9 +1415,9 @@ SimpleX servers cannot see your profile.</target>
         <target>accepted</target>
         <note>call status</note>
       </trans-unit>
-      <trans-unit id="audio call (not e2e encrypted)." xml:space="preserve">
-        <source>audio call (not e2e encrypted).</source>
-        <target>audio call (not e2e encrypted).</target>
+      <trans-unit id="audio call (not e2e encrypted)" xml:space="preserve">
+        <source>audio call (not e2e encrypted)</source>
+        <target>audio call (not e2e encrypted)</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
       <trans-unit id="calling…" xml:space="preserve">
@@ -1491,9 +1491,9 @@ SimpleX servers cannot see your profile.</target>
         <target>via one-time link</target>
         <note>chat list item description</note>
       </trans-unit>
-      <trans-unit id="video call (not e2e encrypted)." xml:space="preserve">
-        <source>video call (not e2e encrypted).</source>
-        <target>video call (not e2e encrypted).</target>
+      <trans-unit id="video call (not e2e encrypted)" xml:space="preserve">
+        <source>video call (not e2e encrypted)</source>
+        <target>video call (not e2e encrypted)</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
       <trans-unit id="with e2e encryption" xml:space="preserve">

apps/ios/SimpleX Localizations/ru.xcloc/Localized Contents/ru.xliff

@@ -90,14 +90,14 @@
         <target>**Сканировать QR код**: соединиться с вашим контактом при встрече или во время видеозвонка.</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
-      <trans-unit id="**e2e encrypted** audio call." xml:space="preserve">
-        <source>**e2e encrypted** audio call.</source>
-        <target>**e2e зашифрованный** аудиозвонок.</target>
+      <trans-unit id="**e2e encrypted** audio call" xml:space="preserve">
+        <source>**e2e encrypted** audio call</source>
+        <target>**e2e зашифрованный** аудиозвонок</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
-      <trans-unit id="**e2e encrypted** video call." xml:space="preserve">
-        <source>**e2e encrypted** video call.</source>
-        <target>**e2e зашифрованный** видеозвонок.</target>
+      <trans-unit id="**e2e encrypted** video call" xml:space="preserve">
+        <source>**e2e encrypted** video call</source>
+        <target>**e2e зашифрованный** видеозвонок</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
       <trans-unit id="*bold*" xml:space="preserve">
@@ -1118,9 +1118,9 @@ SimpleX серверы не могут получить доступ к ваше
         <target>принятый звонок</target>
         <note>call status</note>
       </trans-unit>
-      <trans-unit id="audio call (not e2e encrypted)." xml:space="preserve">
-        <source>audio call (not e2e encrypted).</source>
-        <target>аудиозвонок (не e2e зашифрованный).</target>
+      <trans-unit id="audio call (not e2e encrypted)" xml:space="preserve">
+        <source>audio call (not e2e encrypted)</source>
+        <target>аудиозвонок (не e2e зашифрованный)</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
       <trans-unit id="bold" xml:space="preserve">
@@ -1279,9 +1279,9 @@ SimpleX серверы не могут получить доступ к ваше
         <target>через relay сервер</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
-      <trans-unit id="video call (not e2e encrypted)." xml:space="preserve">
-        <source>video call (not e2e encrypted).</source>
-        <target>видеозвонок (не e2e зашифрованный).</target>
+      <trans-unit id="video call (not e2e encrypted)" xml:space="preserve">
+        <source>video call (not e2e encrypted)</source>
+        <target>видеозвонок (не e2e зашифрованный)</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
       <trans-unit id="waiting for answer…" xml:space="preserve">
@@ -1380,14 +1380,14 @@ SimpleX серверы не могут получить доступ к ваше
         <target>%@ хочет соединиться!</target>
         <note>notification title</note>
       </trans-unit>
-      <trans-unit id="**e2e encrypted** audio call." xml:space="preserve">
-        <source>**e2e encrypted** audio call.</source>
-        <target>**e2e зашифрованный** аудиозвонок.</target>
+      <trans-unit id="**e2e encrypted** audio call" xml:space="preserve">
+        <source>**e2e encrypted** audio call</source>
+        <target>**e2e зашифрованный** аудиозвонок</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
-      <trans-unit id="**e2e encrypted** video call." xml:space="preserve">
-        <source>**e2e encrypted** video call.</source>
-        <target>**e2e зашифрованный** видеозвонок.</target>
+      <trans-unit id="**e2e encrypted** video call" xml:space="preserve">
+        <source>**e2e encrypted** video call</source>
+        <target>**e2e зашифрованный** видеозвонок</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
       <trans-unit id="Accept contact request from %@?" xml:space="preserve">
@@ -1415,9 +1415,9 @@ SimpleX серверы не могут получить доступ к ваше
         <target>принятый звонок</target>
         <note>call status</note>
       </trans-unit>
-      <trans-unit id="audio call (not e2e encrypted)." xml:space="preserve">
-        <source>audio call (not e2e encrypted).</source>
-        <target>аудиозвонок (не e2e зашифрованный).</target>
+      <trans-unit id="audio call (not e2e encrypted)" xml:space="preserve">
+        <source>audio call (not e2e encrypted)</source>
+        <target>аудиозвонок (не e2e зашифрованный)</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
       <trans-unit id="calling…" xml:space="preserve">
@@ -1491,9 +1491,9 @@ SimpleX серверы не могут получить доступ к ваше
         <target>через одноразовую ссылку</target>
         <note>chat list item description</note>
       </trans-unit>
-      <trans-unit id="video call (not e2e encrypted)." xml:space="preserve">
-        <source>video call (not e2e encrypted).</source>
-        <target>видеозвонок (не e2e зашифрованный).</target>
+      <trans-unit id="video call (not e2e encrypted)" xml:space="preserve">
+        <source>video call (not e2e encrypted)</source>
+        <target>видеозвонок (не e2e зашифрованный)</target>
         <note>No comment provided by engineer.</note>
       </trans-unit>
       <trans-unit id="with e2e encryption" xml:space="preserve">

apps/ios/SimpleX NSE/ru.lproj/Localizable.strings

@@ -1,8 +1,8 @@
 /* No comment provided by engineer. */
-"**e2e encrypted** audio call." = "**e2e зашифрованный** аудиозвонок.";
+"**e2e encrypted** audio call" = "**e2e зашифрованный** аудиозвонок";
 
 /* No comment provided by engineer. */
-"**e2e encrypted** video call." = "**e2e зашифрованный** видеозвонок.";
+"**e2e encrypted** video call" = "**e2e зашифрованный** видеозвонок";
 
 /* notification title */
 "%@ is connected!" = "соединение с %@ установлено!";
@@ -17,7 +17,7 @@
 "accepted" = "принятый звонок";
 
 /* No comment provided by engineer. */
-"audio call (not e2e encrypted)." = "аудиозвонок (не e2e зашифрованный).";
+"audio call (not e2e encrypted)" = "аудиозвонок (не e2e зашифрованный)";
 
 /* call status */
 "calling…" = "входящий звонок…";
@@ -69,7 +69,7 @@
 "via one-time link" = "через одноразовую ссылку";
 
 /* No comment provided by engineer. */
-"video call (not e2e encrypted)." = "видеозвонок (не e2e зашифрованный).";
+"video call (not e2e encrypted)" = "видеозвонок (не e2e зашифрованный)";
 
 /* No comment provided by engineer. */
 "with e2e encryption" = "e2e зашифровано";

apps/ios/SimpleX--iOS--Info.plist

@@ -23,8 +23,10 @@
 	<false/>
 	<key>UIBackgroundModes</key>
 	<array>
+		<string>audio</string>
 		<string>fetch</string>
 		<string>remote-notification</string>
+		<string>voip</string>
 	</array>
 </dict>
 </plist>

apps/ios/SimpleX.xcodeproj/project.pbxproj

@@ -13,6 +13,7 @@
 		3CDBCF4227FAE51000354CDD /* ComposeLinkView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 3CDBCF4127FAE51000354CDD /* ComposeLinkView.swift */; };
 		3CDBCF4827FF621E00354CDD /* CILinkView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 3CDBCF4727FF621E00354CDD /* CILinkView.swift */; };
 		5C029EA82837DBB3004A9677 /* CICallItemView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C029EA72837DBB3004A9677 /* CICallItemView.swift */; };
+		5C029EAA283942EA004A9677 /* CallController.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C029EA9283942EA004A9677 /* CallController.swift */; };
 		5C063D2727A4564100AEC577 /* ChatPreviewView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C063D2627A4564100AEC577 /* ChatPreviewView.swift */; };
 		5C116CDC27AABE0400E66D01 /* ContactRequestView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C116CDB27AABE0400E66D01 /* ContactRequestView.swift */; };
 		5C13730B28156D2700F43030 /* ContactConnectionView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C13730A28156D2700F43030 /* ContactConnectionView.swift */; };
@@ -27,6 +28,10 @@
 		5C3A88CE27DF50170060F1C2 /* DetermineWidth.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C3A88CD27DF50170060F1C2 /* DetermineWidth.swift */; };
 		5C3A88D127DF57800060F1C2 /* FramedItemView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C3A88D027DF57800060F1C2 /* FramedItemView.swift */; };
 		5C5346A827B59A6A004DF848 /* ChatHelp.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C5346A727B59A6A004DF848 /* ChatHelp.swift */; };
+		5C55A91F283AD0E400C4E99E /* CallManager.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C55A91E283AD0E400C4E99E /* CallManager.swift */; };
+		5C55A921283CCCB700C4E99E /* IncomingCallView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C55A920283CCCB700C4E99E /* IncomingCallView.swift */; };
+		5C55A923283CEDE600C4E99E /* SoundPlayer.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C55A922283CEDE600C4E99E /* SoundPlayer.swift */; };
+		5C55A92E283D0FDE00C4E99E /* sounds in Resources */ = {isa = PBXBuildFile; fileRef = 5C55A92D283D0FDE00C4E99E /* sounds */; };
 		5C577F7D27C83AA10006112D /* MarkdownHelp.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C577F7C27C83AA10006112D /* MarkdownHelp.swift */; };
 		5C5E5D3B2824468B00B0488A /* ActiveCallView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C5E5D3A2824468B00B0488A /* ActiveCallView.swift */; };
 		5C5E5D3D282447AB00B0488A /* CallTypes.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C5E5D3C282447AB00B0488A /* CallTypes.swift */; };
@@ -130,6 +135,7 @@
 		3CDBCF4127FAE51000354CDD /* ComposeLinkView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ComposeLinkView.swift; sourceTree = "<group>"; };
 		3CDBCF4727FF621E00354CDD /* CILinkView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = CILinkView.swift; sourceTree = "<group>"; };
 		5C029EA72837DBB3004A9677 /* CICallItemView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = CICallItemView.swift; sourceTree = "<group>"; };
+		5C029EA9283942EA004A9677 /* CallController.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = CallController.swift; sourceTree = "<group>"; };
 		5C063D2627A4564100AEC577 /* ChatPreviewView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ChatPreviewView.swift; sourceTree = "<group>"; };
 		5C116CDB27AABE0400E66D01 /* ContactRequestView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ContactRequestView.swift; sourceTree = "<group>"; };
 		5C13730A28156D2700F43030 /* ContactConnectionView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ContactConnectionView.swift; sourceTree = "<group>"; };
@@ -146,6 +152,10 @@
 		5C3A88D027DF57800060F1C2 /* FramedItemView.swift */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.swift; path = FramedItemView.swift; sourceTree = "<group>"; };
 		5C422A7C27A9A6FA0097A1E1 /* SimpleX (iOS).entitlements */ = {isa = PBXFileReference; lastKnownFileType = text.plist.entitlements; path = "SimpleX (iOS).entitlements"; sourceTree = "<group>"; };
 		5C5346A727B59A6A004DF848 /* ChatHelp.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ChatHelp.swift; sourceTree = "<group>"; };
+		5C55A91E283AD0E400C4E99E /* CallManager.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = CallManager.swift; sourceTree = "<group>"; };
+		5C55A920283CCCB700C4E99E /* IncomingCallView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = IncomingCallView.swift; sourceTree = "<group>"; };
+		5C55A922283CEDE600C4E99E /* SoundPlayer.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = SoundPlayer.swift; sourceTree = "<group>"; };
+		5C55A92D283D0FDE00C4E99E /* sounds */ = {isa = PBXFileReference; lastKnownFileType = folder; path = sounds; sourceTree = "<group>"; };
 		5C577F7C27C83AA10006112D /* MarkdownHelp.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = MarkdownHelp.swift; sourceTree = "<group>"; };
 		5C5E5D3A2824468B00B0488A /* ActiveCallView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ActiveCallView.swift; sourceTree = "<group>"; };
 		5C5E5D3C282447AB00B0488A /* CallTypes.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = CallTypes.swift; sourceTree = "<group>"; };
@@ -271,6 +281,10 @@
 				3C714776281C081000CB4D4B /* WebRTCView.swift */,
 				5C9D13A2282187BB00AB8B43 /* WebRTC.swift */,
 				5C5E5D3A2824468B00B0488A /* ActiveCallView.swift */,
+				5C029EA9283942EA004A9677 /* CallController.swift */,
+				5C55A91E283AD0E400C4E99E /* CallManager.swift */,
+				5C55A920283CCCB700C4E99E /* IncomingCallView.swift */,
+				5C55A922283CEDE600C4E99E /* SoundPlayer.swift */,
 			);
 			path = Call;
 			sourceTree = "<group>";
@@ -362,6 +376,7 @@
 		5CA059BD279559F40002BEB4 = {
 			isa = PBXGroup;
 			children = (
+				5C55A92D283D0FDE00C4E99E /* sounds */,
 				3C714779281C0F6800CB4D4B /* www */,
 				5CC2C0FD2809BF11000C35E3 /* SimpleX--iOS--InfoPlist.strings */,
 				5CC2C0FA2809BF11000C35E3 /* Localizable.strings */,
@@ -625,6 +640,7 @@
 			isa = PBXResourcesBuildPhase;
 			buildActionMask = 2147483647;
 			files = (
+				5C55A92E283D0FDE00C4E99E /* sounds in Resources */,
 				3C71477A281C0F6800CB4D4B /* www in Resources */,
 				5CA059EF279559F40002BEB4 /* Assets.xcassets in Resources */,
 				5CC2C0FC2809BF11000C35E3 /* Localizable.strings in Resources */,
@@ -657,6 +673,7 @@
 			files = (
 				5C6AD81327A834E300348BD7 /* NewChatButton.swift in Sources */,
 				5CDCAD7F281894FB00503DA2 /* API.swift in Sources */,
+				5C55A923283CEDE600C4E99E /* SoundPlayer.swift in Sources */,
 				5CDCAD81281A7E2700503DA2 /* Notifications.swift in Sources */,
 				5CB924D727A8563F00ACCCDD /* SettingsView.swift in Sources */,
 				5CEACCE327DE9246000BD591 /* ComposeView.swift in Sources */,
@@ -667,6 +684,7 @@
 				5CDCAD5328186F9500503DA2 /* GroupDefaults.swift in Sources */,
 				5C13730B28156D2700F43030 /* ContactConnectionView.swift in Sources */,
 				5CE4407927ADB701007B033A /* EmojiItemView.swift in Sources */,
+				5C029EAA283942EA004A9677 /* CallController.swift in Sources */,
 				5C5346A827B59A6A004DF848 /* ChatHelp.swift in Sources */,
 				648010AB281ADD15009009B9 /* CIFileView.swift in Sources */,
 				3CDBCF4227FAE51000354CDD /* ComposeLinkView.swift in Sources */,
@@ -701,10 +719,12 @@
 				5C2E260F27A30FDC00F70299 /* ChatView.swift in Sources */,
 				5C2E260B27A30CFA00F70299 /* ChatListView.swift in Sources */,
 				5C971E2127AEBF8300C8A3CE /* ChatInfoImage.swift in Sources */,
+				5C55A921283CCCB700C4E99E /* IncomingCallView.swift in Sources */,
 				6454036F2822A9750090DDFF /* ComposeFileView.swift in Sources */,
 				5C5F2B6D27EBC3FE006A9D5F /* ImagePicker.swift in Sources */,
 				5C577F7D27C83AA10006112D /* MarkdownHelp.swift in Sources */,
 				5CA059EB279559F40002BEB4 /* SimpleXApp.swift in Sources */,
+				5C55A91F283AD0E400C4E99E /* CallManager.swift in Sources */,
 				5CCD403727A5F9A200368C90 /* ScanToConnectView.swift in Sources */,
 				649BCDA22805D6EF00C3A862 /* CIImageView.swift in Sources */,
 				5CCD403A27A5F9BE00368C90 /* CreateGroupView.swift in Sources */,

apps/ios/ru.lproj/Localizable.strings

@@ -38,10 +38,10 @@
 "**Create link / QR code** for your contact to use." = "**Создать ссылку / QR код** для вашего контакта.";
 
 /* No comment provided by engineer. */
-"**e2e encrypted** audio call." = "**e2e зашифрованный** аудиозвонок.";
+"**e2e encrypted** audio call" = "**e2e зашифрованный** аудиозвонок";
 
 /* No comment provided by engineer. */
-"**e2e encrypted** video call." = "**e2e зашифрованный** видеозвонок.";
+"**e2e encrypted** video call" = "**e2e зашифрованный** видеозвонок";
 
 /* No comment provided by engineer. */
 "**Paste received link** or open it in the browser and tap **Open in mobile app**." = "**Вставить полученную ссылку**, или откройте её в браузере и нажмите **Open in mobile app**.";
@@ -122,7 +122,7 @@
 "Attach" = "Прикрепить";
 
 /* No comment provided by engineer. */
-"audio call (not e2e encrypted)." = "аудиозвонок (не e2e зашифрованный).";
+"audio call (not e2e encrypted)" = "аудиозвонок (не e2e зашифрованный)";
 
 /* No comment provided by engineer. */
 "bold" = "жирный";
@@ -678,7 +678,7 @@
 "via relay" = "через relay сервер";
 
 /* No comment provided by engineer. */
-"video call (not e2e encrypted)." = "видеозвонок (не e2e зашифрованный).";
+"video call (not e2e encrypted)" = "видеозвонок (не e2e зашифрованный)";
 
 /* No comment provided by engineer. */
 "waiting for answer…" = "ожидается ответ…";

packages/simplex-chat-webrtc/src/call.ts

@@ -289,7 +289,8 @@ const processCommand = (function () {
     const pc = new RTCPeerConnection(config.peerConnectionConfig)
     const remoteStream = new MediaStream()
     const localCamera = VideoCamera.User
-    const localStream = await navigator.mediaDevices.getUserMedia(callMediaConstraints(mediaType, localCamera))
+    const constraints = callMediaConstraints(mediaType, localCamera)
+    const localStream = await navigator.mediaDevices.getUserMedia(constraints)
     const iceCandidates = getIceCandidates(pc, config)
     const call = {connection: pc, iceCandidates, localMedia: mediaType, localCamera, localStream, remoteStream, aesKey, useWorker}
     await setupMediaStreams(call)
@@ -310,8 +311,10 @@ const processCommand = (function () {
       })
       if (pc.connectionState == "disconnected" || pc.connectionState == "failed") {
         pc.removeEventListener("connectionstatechange", connectionStateChange)
+        if (activeCall) {
+          setTimeout(() => sendMessageToNative({resp: {type: "ended"}}), 0)
+        }
         endCall()
-        setTimeout(() => sendMessageToNative({resp: {type: "ended"}}), 0)
       } else if (pc.connectionState == "connected") {
         const stats = (await pc.getStats()) as Map<string, any>
         for (const stat of stats.values()) {
@@ -326,7 +329,7 @@ const processCommand = (function () {
                 remoteCandidate: stats.get(iceCandidatePair.remoteCandidateId),
               },
             }
-            setTimeout(() => sendMessageToNative({resp}), 0)
+            setTimeout(() => sendMessageToNative({resp}), 500)
             break
           }
         }
@@ -442,17 +445,9 @@ const processCommand = (function () {
         case "camera":
           if (!activeCall || !pc) {
             resp = {type: "error", message: "camera: call not started"}
-          } else if (activeCall.localMedia == CallMediaType.Audio) {
-            resp = {type: "error", message: "camera: no video"}
           } else {
-            try {
-              if (command.camera != activeCall.localCamera) {
-                await replaceCamera(activeCall, command.camera)
-              }
-              resp = {type: "ok"}
-            } catch (e) {
-              resp = {type: "error", message: `camera: ${(e as Error).message}`}
-            }
+            await replaceMedia(activeCall, command.camera)
+            resp = {type: "ok"}
           }
           break
         case "end":
@@ -464,7 +459,7 @@ const processCommand = (function () {
           break
       }
     } catch (e) {
-      resp = {type: "error", message: (e as Error).message}
+      resp = {type: "error", message: `${command.type}: ${(e as Error).message}`}
     }
     const apiResp = {corrId, resp, command}
     sendMessageToNative(apiResp)
@@ -506,6 +501,9 @@ const processCommand = (function () {
       if (call.useWorker && !call.worker) {
         const workerCode = `const callCrypto = (${callCryptoFunction.toString()})(); (${workerFunction.toString()})()`
         call.worker = new Worker(URL.createObjectURL(new Blob([workerCode], {type: "text/javascript"})))
+        call.worker.onerror = ({error, filename, lineno, message}: ErrorEvent) =>
+          console.log(JSON.stringify({error, filename, lineno, message}))
+        call.worker.onmessage = ({data}) => console.log(JSON.stringify({message: data}))
       }
     }
   }
@@ -532,14 +530,19 @@ const processCommand = (function () {
     // Pull tracks from remote stream as they arrive add them to remoteStream video
     const pc = call.connection
     pc.ontrack = (event) => {
-      if (call.aesKey && call.key) {
-        console.log("set up decryption for receiving")
-        setupPeerTransform(TransformOperation.Decrypt, event.receiver as RTCRtpReceiverWithEncryption, call.worker, call.aesKey, call.key)
-      }
-      for (const stream of event.streams) {
-        for (const track of stream.getTracks()) {
-          call.remoteStream.addTrack(track)
+      try {
+        if (call.aesKey && call.key) {
+          console.log("set up decryption for receiving")
+          setupPeerTransform(TransformOperation.Decrypt, event.receiver as RTCRtpReceiverWithEncryption, call.worker, call.aesKey, call.key)
         }
+        for (const stream of event.streams) {
+          for (const track of stream.getTracks()) {
+            call.remoteStream.addTrack(track)
+          }
+        }
+        console.log(`ontrack success`)
+      } catch (e) {
+        console.log(`ontrack error: ${(e as Error).message}`)
       }
     }
   }
@@ -573,7 +576,7 @@ const processCommand = (function () {
     }
   }
 
-  async function replaceCamera(call: Call, camera: VideoCamera): Promise<void> {
+  async function replaceMedia(call: Call, camera: VideoCamera): Promise<void> {
     const videos = getVideoElements()
     if (!videos) throw Error("no video elements")
     const pc = call.connection
@@ -588,6 +591,7 @@ const processCommand = (function () {
   }
 
   function replaceTracks(pc: RTCPeerConnection, tracks: MediaStreamTrack[]) {
+    if (!tracks.length) return
     const sender = pc.getSenders().find((s) => s.track?.kind === tracks[0].kind)
     if (sender) for (const t of tracks) sender.replaceTrack(t)
   }
@@ -713,8 +717,10 @@ function callCryptoFunction(): CallCrypto {
       const initial = data.subarray(0, n)
       const plaintext = data.subarray(n, data.byteLength)
       try {
-        const ciphertext = await crypto.subtle.encrypt({name: "AES-GCM", iv: iv.buffer}, key, plaintext)
-        frame.data = concatN(initial, new Uint8Array(ciphertext), iv).buffer
+        const ciphertext = new Uint8Array(
+          plaintext.length ? await crypto.subtle.encrypt({name: "AES-GCM", iv: iv.buffer}, key, plaintext) : 0
+        )
+        frame.data = concatN(initial, ciphertext, iv).buffer
         controller.enqueue(frame)
       } catch (e) {
         console.log(`encryption error ${e}`)
@@ -731,8 +737,8 @@ function callCryptoFunction(): CallCrypto {
       const ciphertext = data.subarray(n, data.byteLength - IV_LENGTH)
       const iv = data.subarray(data.byteLength - IV_LENGTH, data.byteLength)
       try {
-        const plaintext = await crypto.subtle.decrypt({name: "AES-GCM", iv}, key, ciphertext)
-        frame.data = concatN(initial, new Uint8Array(plaintext)).buffer
+        const plaintext = new Uint8Array(ciphertext.length ? await crypto.subtle.decrypt({name: "AES-GCM", iv}, key, ciphertext) : 0)
+        frame.data = concatN(initial, plaintext).buffer
         controller.enqueue(frame)
       } catch (e) {
         console.log(`decryption error ${e}`)
@@ -864,9 +870,14 @@ function workerFunction() {
   // encryption using RTCRtpScriptTransform.
   if ("RTCTransformEvent" in self) {
     self.addEventListener("rtctransform", async ({transformer}: any) => {
-      const {operation, aesKey} = transformer.options
-      const {readable, writable} = transformer
-      await setupTransform({operation, aesKey, readable, writable})
+      try {
+        const {operation, aesKey} = transformer.options
+        const {readable, writable} = transformer
+        await setupTransform({operation, aesKey, readable, writable})
+        self.postMessage({result: "setupTransform success"})
+      } catch (e) {
+        self.postMessage({message: `setupTransform error: ${(e as Error).message}`})
+      }
     })
   }
 

packages/simplex-chat-webrtc/src/style.css

@@ -1,12 +1,10 @@
-video::-webkit-media-controls {
-  display: none;
-}
 html,
 body {
   padding: 0;
   margin: 0;
   background-color: black;
 }
+
 #remote-video-stream {
   position: absolute;
   width: 100%;
@@ -24,3 +22,20 @@ body {
   top: 0;
   right: 0;
 }
+
+*::-webkit-media-controls {
+  display: none !important;
+  -webkit-appearance: none !important;
+}
+*::-webkit-media-controls-panel {
+  display: none !important;
+  -webkit-appearance: none !important;
+}
+*::-webkit-media-controls-play-button {
+  display: none !important;
+  -webkit-appearance: none !important;
+}
+*::-webkit-media-controls-start-playback-button {
+  display: none !important;
+  -webkit-appearance: none !important;
+}

src/Simplex/Chat.hs

@@ -880,6 +880,7 @@ callStatusItemContent userId Contact {contactId} chatItemId receivedStatus = do
         (Just CISCallPending, WCSDisconnected) -> Just (CISCallMissed, 0)
         (Just CISCallEnded, _) -> Nothing -- if call already ended or failed -> no change
         (Just CISCallError, _) -> Nothing
+        (Just _, WCSConnecting) -> Just (CISCallNegotiated, 0)
         (Just _, WCSConnected) -> Just (CISCallProgress, 0) -- if call ended that was never connected, duration = 0
         (Just _, WCSDisconnected) -> Just (CISCallEnded, 0)
         (Just _, WCSFailed) -> Just (CISCallError, 0)

src/Simplex/Chat/Call.hs

@@ -202,16 +202,18 @@ instance ToJSON WebRTCExtraInfo where
   toJSON = J.genericToJSON J.defaultOptions
   toEncoding = J.genericToEncoding J.defaultOptions
 
-data WebRTCCallStatus = WCSConnected | WCSDisconnected | WCSFailed
+data WebRTCCallStatus = WCSConnecting | WCSConnected | WCSDisconnected | WCSFailed
   deriving (Show)
 
 instance StrEncoding WebRTCCallStatus where
   strEncode = \case
+    WCSConnecting -> "connecting"
     WCSConnected -> "connected"
     WCSDisconnected -> "disconnected"
     WCSFailed -> "failed"
   strP =
     A.takeTill (== ' ') >>= \case
+      "connecting" -> pure WCSConnecting
       "connected" -> pure WCSConnected
       "disconnected" -> pure WCSDisconnected
       "failed" -> pure WCSFailed