WIP

submodules/AccountContext/Sources/PresentationCallManager.swift

@@ -280,6 +280,8 @@ public protocol PresentationGroupCall: class {
     var internalId: CallSessionInternalId { get }
     var peerId: PeerId { get }
     
+    var isVideo: Bool { get }
+    
     var audioOutputState: Signal<([AudioSessionOutput], AudioSessionOutput?), NoError> { get }
     
     var canBeRemoved: Signal<Bool, NoError> { get }
@@ -296,6 +298,8 @@ public protocol PresentationGroupCall: class {
     
     func toggleIsMuted()
     func setIsMuted(action: PresentationGroupCallMuteAction)
+    func requestVideo()
+    func disableVideo()
     func updateDefaultParticipantsAreMuted(isMuted: Bool)
     func setVolume(peerId: PeerId, volume: Double)
     func setFullSizeVideo(peerId: PeerId?)

submodules/TelegramCallsUI/Sources/PresentationGroupCall.swift

@@ -312,6 +312,8 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
     public let peerId: PeerId
     public let peer: Peer?
     
+    public private(set) var isVideo: Bool
+    
     private let temporaryJoinTimestamp: Int32
     
     private var internalState: InternalState = .requesting
@@ -319,6 +321,8 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
     private var callContext: OngoingGroupCallContext?
     private var ssrcMapping: [UInt32: PeerId] = [:]
     
+    private var requestedSsrcs = Set<UInt32>()
+    
     private var summaryInfoState = Promise<SummaryInfoState?>(nil)
     private var summaryParticipantsState = Promise<SummaryParticipantsState?>(nil)
     
@@ -433,6 +437,7 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
     }
     private let memberEventsPipeDisposable = MetaDisposable()
     
+    private let joinDisposable = MetaDisposable()
     private let requestDisposable = MetaDisposable()
     private var groupCallParticipantUpdatesDisposable: Disposable?
     
@@ -461,6 +466,10 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
         return self.incomingVideoSourcePromise.get()
     }
     
+    private var missingSsrcs = Set<UInt32>()
+    private let missingSsrcsDisposable = MetaDisposable()
+    private var isRequestingMissingSsrcs: Bool = false
+    
     init(
         accountContext: AccountContext,
         audioSession: ManagedAudioSession,
@@ -484,7 +493,7 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
         
         self.temporaryJoinTimestamp = Int32(CFAbsoluteTimeGetCurrent() + NSTimeIntervalSince1970)
         
-        self.videoCapturer = OngoingCallVideoCapturer()
+        self.isVideo = false
         
         var didReceiveAudioOutputs = false
         
@@ -623,9 +632,6 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
                 if !removedSsrc.isEmpty {
                     strongSelf.callContext?.removeSsrcs(ssrcs: removedSsrc)
                 }
-                if !addedParticipants.isEmpty {
-                    strongSelf.callContext?.addParticipants(participants: addedParticipants)
-                }
             }
         })
         
@@ -745,6 +751,7 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
         self.audioSessionActiveDisposable?.dispose()
         self.summaryStateDisposable?.dispose()
         self.audioSessionDisposable?.dispose()
+        self.joinDisposable.dispose()
         self.requestDisposable.dispose()
         self.groupCallParticipantUpdatesDisposable?.dispose()
         self.leaveDisposable.dispose()
@@ -756,6 +763,7 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
         self.participantsContextStateDisposable.dispose()
         self.myAudioLevelDisposable.dispose()
         self.memberEventsPipeDisposable.dispose()
+        self.missingSsrcsDisposable.dispose()
         
         self.myAudioLevelTimer?.invalidate()
         self.typingDisposable.dispose()
@@ -802,7 +810,14 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
             break
         default:
             if case let .active(callInfo) = internalState {
-                let callContext = OngoingGroupCallContext(video: self.videoCapturer)
+                let callContext = OngoingGroupCallContext(video: self.videoCapturer, participantDescriptionsRequired: { [weak self] ssrcs in
+                    Queue.mainQueue().async {
+                        guard let strongSelf = self else {
+                            return
+                        }
+                        strongSelf.maybeRequestParticipants(ssrcs: ssrcs)
+                    }
+                })
                 self.incomingVideoSourcePromise.set(callContext.videoSources
                 |> deliverOnMainQueue
                 |> map { [weak self] sources -> [PeerId: UInt32] in
@@ -818,8 +833,16 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
                     return result
                 })
                 self.callContext = callContext
-                self.requestDisposable.set((callContext.joinPayload
-                |> take(1)
+                self.joinDisposable.set((callContext.joinPayload
+                |> distinctUntilChanged(isEqual: { lhs, rhs in
+                    if lhs.0 != rhs.0 {
+                        return false
+                    }
+                    if lhs.1 != rhs.1 {
+                        return false
+                    }
+                    return true
+                })
                 |> deliverOnMainQueue).start(next: { [weak self] joinPayload, ssrc in
                     guard let strongSelf = self else {
                         return
@@ -963,14 +986,10 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
                 }
                 
                 self.ssrcMapping.removeAll()
-                var addedParticipants: [(UInt32, String?)] = []
                 for participant in initialState.participants {
                     self.ssrcMapping[participant.ssrc] = participant.peer.id
-                    if participant.peer.id != self.accountContext.account.peerId {
-                        addedParticipants.append((participant.ssrc, participant.jsonParams))
-                    }
                 }
-                self.callContext?.setJoinResponse(payload: clientParams, participants: addedParticipants)
+                self.callContext?.setJoinResponse(payload: clientParams, participants: [])
                 
                 let accountContext = self.accountContext
                 let peerId = self.peerId
@@ -1135,6 +1154,65 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
         }
     }
     
+    private func maybeRequestParticipants(ssrcs: Set<UInt32>) {
+        var missingSsrcs = ssrcs
+        
+        var addedParticipants: [(UInt32, String?)] = []
+        
+        if let membersValue = self.membersValue {
+            for participant in membersValue.participants {
+                if missingSsrcs.contains(participant.ssrc) {
+                    missingSsrcs.remove(participant.ssrc)
+                    
+                    addedParticipants.append((participant.ssrc, participant.jsonParams))
+                }
+            }
+        }
+        
+        if !addedParticipants.isEmpty {
+            self.callContext?.addParticipants(participants: addedParticipants)
+        }
+        
+        if !missingSsrcs.isEmpty {
+            self.missingSsrcs.formUnion(missingSsrcs)
+            self.maybeRequestMissingSsrcs()
+        }
+    }
+    
+    private func maybeRequestMissingSsrcs() {
+        if self.isRequestingMissingSsrcs {
+            return
+        }
+        if self.missingSsrcs.isEmpty {
+            return
+        }
+        if case let .estabilished(callInfo, _, ssrc, _) = self.internalState {
+            self.isRequestingMissingSsrcs = true
+            
+            let requestedSsrcs = self.missingSsrcs
+            self.missingSsrcsDisposable.set((getGroupCallParticipants(account: self.account, callId: callInfo.id, accessHash: callInfo.accessHash, ssrcs: Array(requestedSsrcs), offset: "", limit: 100)
+            |> deliverOnMainQueue).start(next: { [weak self] state in
+                guard let strongSelf = self else {
+                    return
+                }
+                strongSelf.isRequestingMissingSsrcs = false
+                strongSelf.missingSsrcs.subtract(requestedSsrcs)
+                
+                var addedParticipants: [(UInt32, String?)] = []
+                
+                for participant in state.participants {
+                    addedParticipants.append((participant.ssrc, participant.jsonParams))
+                }
+                
+                if !addedParticipants.isEmpty {
+                    strongSelf.callContext?.addParticipants(participants: addedParticipants)
+                }
+                
+                strongSelf.maybeRequestMissingSsrcs()
+            }))
+        }
+    }
+    
     private func startCheckingCallIfNeeded() {
         if self.checkCallDisposable != nil {
             return
@@ -1271,6 +1349,25 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
         }
     }
     
+    public func requestVideo() {
+        if self.videoCapturer == nil {
+            let videoCapturer = OngoingCallVideoCapturer()
+            self.videoCapturer = videoCapturer
+        }
+        self.isVideo = true
+        if let videoCapturer = self.videoCapturer {
+            self.callContext?.requestVideo(videoCapturer)
+        }
+    }
+    
+    public func disableVideo() {
+        self.isVideo = false
+        if let _ = self.videoCapturer {
+            self.videoCapturer = nil
+            self.callContext?.disableVideo()
+        }
+    }
+    
     public func setVolume(peerId: PeerId, volume: Double) {
         for (ssrc, id) in self.ssrcMapping {
             if id == peerId {
@@ -1422,6 +1519,7 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
             }
         }
         
+        self.joinDisposable.set(nil)
         self.requestDisposable.set((currentOrRequestedCall
         |> deliverOnMainQueue).start(next: { [weak self] value in
             guard let strongSelf = self else {

submodules/TelegramCallsUI/Sources/VoiceChatController.swift

@@ -567,6 +567,7 @@ public final class VoiceChatController: ViewController {
         private let bottomPanelBackgroundNode: ASDisplayNode
         private let bottomCornersNode: ASImageNode
         fileprivate let audioOutputNode: CallControllerButtonItemNode
+        fileprivate let cameraButtonNode: CallControllerButtonItemNode
         fileprivate let leaveNode: CallControllerButtonItemNode
         fileprivate let actionButton: VoiceChatActionButton
         private let leftBorderNode: ASDisplayNode
@@ -699,6 +700,7 @@ public final class VoiceChatController: ViewController {
             self.bottomCornersNode.image = cornersImage(top: false, bottom: true, dark: false)
             
             self.audioOutputNode = CallControllerButtonItemNode()
+            self.cameraButtonNode = CallControllerButtonItemNode()
             self.leaveNode = CallControllerButtonItemNode()
             self.actionButton = VoiceChatActionButton()
 
@@ -1094,6 +1096,7 @@ public final class VoiceChatController: ViewController {
             self.bottomPanelNode.addSubnode(self.bottomCornersNode)
             self.bottomPanelNode.addSubnode(self.bottomPanelBackgroundNode)
             self.bottomPanelNode.addSubnode(self.audioOutputNode)
+            self.bottomPanelNode.addSubnode(self.cameraButtonNode)
             self.bottomPanelNode.addSubnode(self.leaveNode)
             self.bottomPanelNode.addSubnode(self.actionButton)
             
@@ -1293,6 +1296,8 @@ public final class VoiceChatController: ViewController {
             
             self.audioOutputNode.addTarget(self, action: #selector(self.audioOutputPressed), forControlEvents: .touchUpInside)
             
+            self.cameraButtonNode.addTarget(self, action: #selector(self.cameraPressed), forControlEvents: .touchUpInside)
+            
             self.optionsButton.contextAction = { [weak self, weak optionsButton] sourceNode, gesture in
                 guard let strongSelf = self, let controller = strongSelf.controller, let strongOptionsButton = optionsButton else {
                     return
@@ -1705,6 +1710,14 @@ public final class VoiceChatController: ViewController {
             }
         }
         
+        @objc private func cameraPressed() {
+            if self.call.isVideo {
+                self.call.disableVideo()
+            } else {
+                self.call.requestVideo()
+            }
+        }
+        
         private func updateFloatingHeaderOffset(offset: CGFloat, transition: ContainedViewLayoutTransition, completion: (() -> Void)? = nil) {
             guard let (layout, _) = self.validLayout else {
                 return
@@ -1960,6 +1973,10 @@ public final class VoiceChatController: ViewController {
             let sideButtonSize = CGSize(width: 60.0, height: 60.0)
             self.audioOutputNode.update(size: sideButtonSize, content: CallControllerButtonItemNode.Content(appearance: soundAppearance, image: soundImage), text: soundTitle, transition: .animated(duration: 0.3, curve: .linear))
             
+            let cameraButtonSize = CGSize(width: 40.0, height: 40.0)
+            
+            self.cameraButtonNode.update(size: cameraButtonSize, content: CallControllerButtonItemNode.Content(appearance: CallControllerButtonItemNode.Content.Appearance.blurred(isFilled: false), image: .camera), text: " ", transition: .animated(duration: 0.3, curve: .linear))
+            
             self.leaveNode.update(size: sideButtonSize, content: CallControllerButtonItemNode.Content(appearance: .color(.custom(0xff3b30, 0.3)), image: .end), text: self.presentationData.strings.VoiceChat_Leave, transition: .immediate)
         }
         
@@ -2039,6 +2056,7 @@ public final class VoiceChatController: ViewController {
             transition.updateFrame(node: self.bottomPanelNode, frame: bottomPanelFrame)
             
             let sideButtonSize = CGSize(width: 60.0, height: 60.0)
+            let cameraButtonSize = CGSize(width: 40.0, height: 40.0)
             let centralButtonSize = CGSize(width: 440.0, height: 440.0)
                         
             let actionButtonFrame = CGRect(origin: CGPoint(x: floorToScreenPixels((size.width - centralButtonSize.width) / 2.0), y: floorToScreenPixels((bottomAreaHeight - centralButtonSize.height) / 2.0)), size: centralButtonSize)
@@ -2112,8 +2130,14 @@ public final class VoiceChatController: ViewController {
             let sideButtonOrigin = max(sideButtonMinimalInset, floor((size.width - 144.0) / 2.0) - sideButtonOffset - sideButtonSize.width)
             
             if self.audioOutputNode.supernode === self.bottomPanelNode {
-                transition.updateFrame(node: self.audioOutputNode, frame: CGRect(origin: CGPoint(x: sideButtonOrigin, y: floor((bottomAreaHeight - sideButtonSize.height) / 2.0)), size: sideButtonSize))
+                let cameraButtonDistance: CGFloat = 4.0
+                
+                let audioOutputFrame = CGRect(origin: CGPoint(x: sideButtonOrigin, y: floor((bottomAreaHeight - sideButtonSize.height - cameraButtonDistance - cameraButtonSize.height) / 2.0) + cameraButtonDistance + cameraButtonSize.height), size: sideButtonSize)
+                
+                transition.updateFrame(node: self.audioOutputNode, frame: audioOutputFrame)
                 transition.updateFrame(node: self.leaveNode, frame: CGRect(origin: CGPoint(x: size.width - sideButtonOrigin - sideButtonSize.width, y: floor((bottomAreaHeight - sideButtonSize.height) / 2.0)), size: sideButtonSize))
+                
+                transition.updateFrame(node: self.cameraButtonNode, frame: CGRect(origin: CGPoint(x: floor(audioOutputFrame.midX - cameraButtonSize.width / 2.0), y: audioOutputFrame.minY - cameraButtonDistance - cameraButtonSize.height), size: cameraButtonSize))
             }
             if isFirstTime {
                 while !self.enqueuedTransitions.isEmpty {
@@ -2139,10 +2163,13 @@ public final class VoiceChatController: ViewController {
                 if self.actionButton.supernode !== self.bottomPanelNode {
                     self.actionButton.ignoreHierarchyChanges = true
                     self.audioOutputNode.isHidden = false
+                    self.cameraButtonNode.isHidden = false
                     self.leaveNode.isHidden = false
                     self.audioOutputNode.layer.removeAllAnimations()
+                    self.cameraButtonNode.layer.removeAllAnimations()
                     self.leaveNode.layer.removeAllAnimations()
                     self.bottomPanelNode.addSubnode(self.audioOutputNode)
+                    self.bottomPanelNode.addSubnode(self.cameraButtonNode)
                     self.bottomPanelNode.addSubnode(self.leaveNode)
                     self.bottomPanelNode.addSubnode(self.actionButton)
                     self.containerLayoutUpdated(layout, navigationHeight :navigationHeight, transition: .immediate)
@@ -2386,7 +2413,7 @@ public final class VoiceChatController: ViewController {
         override func gestureRecognizerShouldBegin(_ gestureRecognizer: UIGestureRecognizer) -> Bool {
             if gestureRecognizer is DirectionalPanGestureRecognizer {
                 let location = gestureRecognizer.location(in: self.bottomPanelNode.view)
-                if self.audioOutputNode.frame.contains(location) || self.leaveNode.frame.contains(location) {
+                if self.audioOutputNode.frame.contains(location) || (!self.cameraButtonNode.isHidden && self.cameraButtonNode.frame.contains(location)) || self.leaveNode.frame.contains(location) {
                     return false
                 }
             }

submodules/TelegramCore/Sources/GroupCalls.swift

@@ -190,8 +190,10 @@ public enum GetGroupCallParticipantsError {
     case generic
 }
 
-public func getGroupCallParticipants(account: Account, callId: Int64, accessHash: Int64, offset: String, limit: Int32) -> Signal<GroupCallParticipantsContext.State, GetGroupCallParticipantsError> {
-    return account.network.request(Api.functions.phone.getGroupParticipants(call: .inputGroupCall(id: callId, accessHash: accessHash), ids: [], sources: [], offset: offset, limit: limit))
+public func getGroupCallParticipants(account: Account, callId: Int64, accessHash: Int64, ssrcs: [UInt32] = [], offset: String, limit: Int32) -> Signal<GroupCallParticipantsContext.State, GetGroupCallParticipantsError> {
+    return account.network.request(Api.functions.phone.getGroupParticipants(call: .inputGroupCall(id: callId, accessHash: accessHash), ids: [], sources: ssrcs.map {
+        Int32(bitPattern: $0)
+    }, offset: offset, limit: limit))
     |> mapError { _ -> GetGroupCallParticipantsError in
         return .generic
     }

submodules/TelegramVoip/Sources/GroupCallContext.swift

@@ -50,11 +50,12 @@ public final class OngoingGroupCallContext {
         
         let videoSources = ValuePromise<Set<UInt32>>(Set(), ignoreRepeated: true)
         
-        init(queue: Queue, inputDeviceId: String, outputDeviceId: String, video: OngoingCallVideoCapturer?) {
+        init(queue: Queue, inputDeviceId: String, outputDeviceId: String, video: OngoingCallVideoCapturer?, participantDescriptionsRequired: @escaping (Set<UInt32>) -> Void) {
             self.queue = queue
             
             var networkStateUpdatedImpl: ((GroupCallNetworkState) -> Void)?
             var audioLevelsUpdatedImpl: (([NSNumber]) -> Void)?
+            var participantDescriptionsRequiredImpl: (([NSNumber]) -> Void)?
             
             let videoSources = self.videoSources
             self.context = GroupCallThreadLocalContext(
@@ -70,6 +71,9 @@ public final class OngoingGroupCallContext {
                 videoCapturer: video?.impl,
                 incomingVideoSourcesUpdated: { ssrcs in
                     videoSources.set(Set(ssrcs.map { $0.uint32Value }))
+                },
+                participantDescriptionsRequired: { ssrcs in
+                    participantDescriptionsRequired(Set(ssrcs.map { $0.uint32Value }))
                 }
             )
             
@@ -167,6 +171,30 @@ public final class OngoingGroupCallContext {
             self.context.setIsMuted(isMuted)
         }
         
+        func requestVideo(_ capturer: OngoingCallVideoCapturer?) {
+            let queue = self.queue
+            self.context.requestVideo(capturer?.impl, completion: { [weak self] payload, ssrc in
+                queue.async {
+                    guard let strongSelf = self else {
+                        return
+                    }
+                    strongSelf.joinPayload.set(.single((payload, ssrc)))
+                }
+            })
+        }
+        
+        public func disableVideo() {
+            let queue = self.queue
+            self.context.disableVideo({ [weak self] payload, ssrc in
+                queue.async {
+                    guard let strongSelf = self else {
+                        return
+                    }
+                    strongSelf.joinPayload.set(.single((payload, ssrc)))
+                }
+            })
+        }
+        
         func switchAudioInput(_ deviceId: String) {
             self.context.switchAudioInput(deviceId)
         }
@@ -278,10 +306,10 @@ public final class OngoingGroupCallContext {
         }
     }
     
-    public init(inputDeviceId: String = "", outputDeviceId: String = "", video: OngoingCallVideoCapturer?) {
+    public init(inputDeviceId: String = "", outputDeviceId: String = "", video: OngoingCallVideoCapturer?, participantDescriptionsRequired: @escaping (Set<UInt32>) -> Void) {
         let queue = self.queue
         self.impl = QueueLocalObject(queue: queue, generate: {
-            return Impl(queue: queue, inputDeviceId: inputDeviceId, outputDeviceId: outputDeviceId, video: video)
+            return Impl(queue: queue, inputDeviceId: inputDeviceId, outputDeviceId: outputDeviceId, video: video, participantDescriptionsRequired: participantDescriptionsRequired)
         })
     }
     
@@ -291,6 +319,18 @@ public final class OngoingGroupCallContext {
         }
     }
     
+    public func requestVideo(_ capturer: OngoingCallVideoCapturer?) {
+        self.impl.with { impl in
+            impl.requestVideo(capturer)
+        }
+    }
+    
+    public func disableVideo() {
+        self.impl.with { impl in
+            impl.disableVideo()
+        }
+    }
+    
     public func switchAudioInput(_ deviceId: String) {
         self.impl.with { impl in
             impl.switchAudioInput(deviceId)

submodules/TgVoipWebrtc/PublicHeaders/TgVoipWebrtc/OngoingCallThreadLocalContext.h

@@ -168,7 +168,7 @@ typedef NS_ENUM(int32_t, GroupCallNetworkState) {
 
 @interface GroupCallThreadLocalContext : NSObject
 
-- (instancetype _Nonnull)initWithQueue:(id<OngoingCallThreadLocalContextQueueWebrtc> _Nonnull)queue networkStateUpdated:(void (^ _Nonnull)(GroupCallNetworkState))networkStateUpdated audioLevelsUpdated:(void (^ _Nonnull)(NSArray<NSNumber *> * _Nonnull))audioLevelsUpdated inputDeviceId:(NSString * _Nonnull)inputDeviceId outputDeviceId:(NSString * _Nonnull)outputDeviceId videoCapturer:(OngoingCallThreadLocalContextVideoCapturer * _Nullable)videoCapturer incomingVideoSourcesUpdated:(void (^ _Nonnull)(NSArray<NSNumber *> * _Nonnull))incomingVideoSourcesUpdated;
+- (instancetype _Nonnull)initWithQueue:(id<OngoingCallThreadLocalContextQueueWebrtc> _Nonnull)queue networkStateUpdated:(void (^ _Nonnull)(GroupCallNetworkState))networkStateUpdated audioLevelsUpdated:(void (^ _Nonnull)(NSArray<NSNumber *> * _Nonnull))audioLevelsUpdated inputDeviceId:(NSString * _Nonnull)inputDeviceId outputDeviceId:(NSString * _Nonnull)outputDeviceId videoCapturer:(OngoingCallThreadLocalContextVideoCapturer * _Nullable)videoCapturer incomingVideoSourcesUpdated:(void (^ _Nonnull)(NSArray<NSNumber *> * _Nonnull))incomingVideoSourcesUpdated participantDescriptionsRequired:(void (^ _Nonnull)(NSArray<NSNumber *> * _Nonnull))participantDescriptionsRequired;
 
 - (void)stop;
 
@@ -177,6 +177,8 @@ typedef NS_ENUM(int32_t, GroupCallNetworkState) {
 - (void)removeSsrcs:(NSArray<NSNumber *> * _Nonnull)ssrcs;
 - (void)addParticipants:(NSArray<OngoingGroupCallParticipantDescription *> * _Nonnull)participants;
 - (void)setIsMuted:(bool)isMuted;
+- (void)requestVideo:(OngoingCallThreadLocalContextVideoCapturer * _Nullable)videoCapturer completion:(void (^ _Nonnull)(NSString * _Nonnull, uint32_t))completion;
+- (void)disableVideo:(void (^ _Nonnull)(NSString * _Nonnull, uint32_t))completion;
 
 - (void)setVolumeForSsrc:(uint32_t)ssrc volume:(double)volume;
 - (void)setFullSizeVideoSsrc:(uint32_t)ssrc;

submodules/TgVoipWebrtc/Sources/OngoingCallThreadLocalContext.mm

@@ -818,7 +818,7 @@ static void (*InternalVoipLoggingFunction)(NSString *) = NULL;
 
 @implementation GroupCallThreadLocalContext
 
-- (instancetype _Nonnull)initWithQueue:(id<OngoingCallThreadLocalContextQueueWebrtc> _Nonnull)queue networkStateUpdated:(void (^ _Nonnull)(GroupCallNetworkState))networkStateUpdated audioLevelsUpdated:(void (^ _Nonnull)(NSArray<NSNumber *> * _Nonnull))audioLevelsUpdated inputDeviceId:(NSString * _Nonnull)inputDeviceId outputDeviceId:(NSString * _Nonnull)outputDeviceId videoCapturer:(OngoingCallThreadLocalContextVideoCapturer * _Nullable)videoCapturer incomingVideoSourcesUpdated:(void (^ _Nonnull)(NSArray<NSNumber *> * _Nonnull))incomingVideoSourcesUpdated {
+- (instancetype _Nonnull)initWithQueue:(id<OngoingCallThreadLocalContextQueueWebrtc> _Nonnull)queue networkStateUpdated:(void (^ _Nonnull)(GroupCallNetworkState))networkStateUpdated audioLevelsUpdated:(void (^ _Nonnull)(NSArray<NSNumber *> * _Nonnull))audioLevelsUpdated inputDeviceId:(NSString * _Nonnull)inputDeviceId outputDeviceId:(NSString * _Nonnull)outputDeviceId videoCapturer:(OngoingCallThreadLocalContextVideoCapturer * _Nullable)videoCapturer incomingVideoSourcesUpdated:(void (^ _Nonnull)(NSArray<NSNumber *> * _Nonnull))incomingVideoSourcesUpdated participantDescriptionsRequired:(void (^ _Nonnull)(NSArray<NSNumber *> * _Nonnull))participantDescriptionsRequired {
     self = [super init];
     if (self != nil) {
         _queue = queue;
@@ -855,6 +855,13 @@ static void (*InternalVoipLoggingFunction)(NSString *) = NULL;
                     [mappedSources addObject:@(it)];
                 }
                 incomingVideoSourcesUpdated(mappedSources);
+            },
+            .participantDescriptionsRequired = [participantDescriptionsRequired](std::vector<uint32_t> const &ssrcs) {
+                NSMutableArray<NSNumber *> *mappedSources = [[NSMutableArray alloc] init];
+                for (auto it : ssrcs) {
+                    [mappedSources addObject:@(it)];
+                }
+                participantDescriptionsRequired(mappedSources);
             }
         }));
     }
@@ -868,105 +875,106 @@ static void (*InternalVoipLoggingFunction)(NSString *) = NULL;
     }
 }
 
+static void processJoinPayload(tgcalls::GroupJoinPayload &payload, void (^ _Nonnull completion)(NSString * _Nonnull, uint32_t)) {
+    NSMutableDictionary *dict = [[NSMutableDictionary alloc] init];
+    
+    int32_t signedSsrc = *(int32_t *)&payload.ssrc;
+    
+    dict[@"ssrc"] = @(signedSsrc);
+    dict[@"ufrag"] = [NSString stringWithUTF8String:payload.ufrag.c_str()];
+    dict[@"pwd"] = [NSString stringWithUTF8String:payload.pwd.c_str()];
+    
+    NSMutableArray *fingerprints = [[NSMutableArray alloc] init];
+    for (auto &fingerprint : payload.fingerprints) {
+        [fingerprints addObject:@{
+            @"hash": [NSString stringWithUTF8String:fingerprint.hash.c_str()],
+            @"fingerprint": [NSString stringWithUTF8String:fingerprint.fingerprint.c_str()],
+            @"setup": [NSString stringWithUTF8String:fingerprint.setup.c_str()]
+        }];
+    }
+    
+    dict[@"fingerprints"] = fingerprints;
+    
+    NSMutableArray *parsedVideoSsrcGroups = [[NSMutableArray alloc] init];
+    NSMutableArray *parsedVideoSources = [[NSMutableArray alloc] init];
+    for (auto &group : payload.videoSourceGroups) {
+        NSMutableDictionary *parsedGroup = [[NSMutableDictionary alloc] init];
+        parsedGroup[@"semantics"] = [NSString stringWithUTF8String:group.semantics.c_str()];
+        NSMutableArray *sources = [[NSMutableArray alloc] init];
+        for (auto &source : group.ssrcs) {
+            [sources addObject:@(source)];
+            if (![parsedVideoSources containsObject:@(source)]) {
+                [parsedVideoSources addObject:@(source)];
+            }
+        }
+        parsedGroup[@"sources"] = sources;
+        [parsedVideoSsrcGroups addObject:parsedGroup];
+    }
+    if (parsedVideoSsrcGroups.count != 0) {
+        dict[@"ssrc-groups"] = parsedVideoSsrcGroups;
+    }
+    
+    NSMutableArray *videoPayloadTypes = [[NSMutableArray alloc] init];
+    for (auto &payloadType : payload.videoPayloadTypes) {
+        NSMutableDictionary *parsedType = [[NSMutableDictionary alloc] init];
+        parsedType[@"id"] = @(payloadType.id);
+        NSString *name = [NSString stringWithUTF8String:payloadType.name.c_str()];
+        parsedType[@"name"] = name;
+        parsedType[@"clockrate"] = @(payloadType.clockrate);
+        if (![name isEqualToString:@"rtx"]) {
+            parsedType[@"channels"] = @(payloadType.channels);
+        }
+        
+        NSMutableDictionary *parsedParameters = [[NSMutableDictionary alloc] init];
+        for (auto &it : payloadType.parameters) {
+            NSString *key = [NSString stringWithUTF8String:it.first.c_str()];
+            NSString *value = [NSString stringWithUTF8String:it.second.c_str()];
+            parsedParameters[key] = value;
+        }
+        if (parsedParameters.count != 0) {
+            parsedType[@"parameters"] = parsedParameters;
+        }
+        
+        if (![name isEqualToString:@"rtx"]) {
+            NSMutableArray *parsedFbs = [[NSMutableArray alloc] init];
+            for (auto &it : payloadType.feedbackTypes) {
+                NSMutableDictionary *parsedFb = [[NSMutableDictionary alloc] init];
+                parsedFb[@"type"] = [NSString stringWithUTF8String:it.type.c_str()];
+                if (it.subtype.size() != 0) {
+                    parsedFb[@"subtype"] = [NSString stringWithUTF8String:it.subtype.c_str()];
+                }
+                [parsedFbs addObject:parsedFb];
+            }
+            parsedType[@"rtcp-fbs"] = parsedFbs;
+        }
+        
+        [videoPayloadTypes addObject:parsedType];
+    }
+    if (videoPayloadTypes.count != 0) {
+        dict[@"payload-types"] = videoPayloadTypes;
+    }
+    
+    NSMutableArray *parsedExtensions = [[NSMutableArray alloc] init];
+    for (auto &it : payload.videoExtensionMap) {
+        NSMutableDictionary *parsedExtension = [[NSMutableDictionary alloc] init];
+        parsedExtension[@"id"] = @(it.first);
+        parsedExtension[@"uri"] = [NSString stringWithUTF8String:it.second.c_str()];
+        [parsedExtensions addObject:parsedExtension];
+    }
+    if (parsedExtensions.count != 0) {
+        dict[@"rtp-hdrexts"] = parsedExtensions;
+    }
+    
+    NSData *data = [NSJSONSerialization dataWithJSONObject:dict options:0 error:nil];
+    NSString *string = [[NSString alloc] initWithData:data encoding:NSUTF8StringEncoding];
+    
+    completion(string, payload.ssrc);
+}
+
 - (void)emitJoinPayload:(void (^ _Nonnull)(NSString * _Nonnull, uint32_t))completion {
     if (_instance) {
         _instance->emitJoinPayload([completion](tgcalls::GroupJoinPayload payload) {
-            NSMutableDictionary *dict = [[NSMutableDictionary alloc] init];
-            
-            int32_t signedSsrc = *(int32_t *)&payload.ssrc;
-            
-            dict[@"ssrc"] = @(signedSsrc);
-            dict[@"ufrag"] = [NSString stringWithUTF8String:payload.ufrag.c_str()];
-            dict[@"pwd"] = [NSString stringWithUTF8String:payload.pwd.c_str()];
-            
-            NSMutableArray *fingerprints = [[NSMutableArray alloc] init];
-            for (auto &fingerprint : payload.fingerprints) {
-                [fingerprints addObject:@{
-                    @"hash": [NSString stringWithUTF8String:fingerprint.hash.c_str()],
-                    @"fingerprint": [NSString stringWithUTF8String:fingerprint.fingerprint.c_str()],
-                    @"setup": [NSString stringWithUTF8String:fingerprint.setup.c_str()]
-                }];
-            }
-            
-            dict[@"fingerprints"] = fingerprints;
-            
-            NSMutableArray *parsedVideoSsrcGroups = [[NSMutableArray alloc] init];
-            NSMutableArray *parsedVideoSources = [[NSMutableArray alloc] init];
-            for (auto &group : payload.videoSourceGroups) {
-                NSMutableDictionary *parsedGroup = [[NSMutableDictionary alloc] init];
-                parsedGroup[@"semantics"] = [NSString stringWithUTF8String:group.semantics.c_str()];
-                NSMutableArray *sources = [[NSMutableArray alloc] init];
-                for (auto &source : group.ssrcs) {
-                    [sources addObject:@(source)];
-                    if (![parsedVideoSources containsObject:@(source)]) {
-                        [parsedVideoSources addObject:@(source)];
-                    }
-                }
-                parsedGroup[@"sources"] = sources;
-                [parsedVideoSsrcGroups addObject:parsedGroup];
-            }
-            if (parsedVideoSsrcGroups.count != 0) {
-                dict[@"ssrc-groups"] = parsedVideoSsrcGroups;
-            }
-            if (parsedVideoSources.count != 0) {
-                //dict[@"sources"] = parsedVideoSources;
-            }
-            
-            NSMutableArray *videoPayloadTypes = [[NSMutableArray alloc] init];
-            for (auto &payloadType : payload.videoPayloadTypes) {
-                NSMutableDictionary *parsedType = [[NSMutableDictionary alloc] init];
-                parsedType[@"id"] = @(payloadType.id);
-                NSString *name = [NSString stringWithUTF8String:payloadType.name.c_str()];
-                parsedType[@"name"] = name;
-                parsedType[@"clockrate"] = @(payloadType.clockrate);
-                if (![name isEqualToString:@"rtx"]) {
-                    parsedType[@"channels"] = @(payloadType.channels);
-                }
-                
-                NSMutableDictionary *parsedParameters = [[NSMutableDictionary alloc] init];
-                for (auto &it : payloadType.parameters) {
-                    NSString *key = [NSString stringWithUTF8String:it.first.c_str()];
-                    NSString *value = [NSString stringWithUTF8String:it.second.c_str()];
-                    parsedParameters[key] = value;
-                }
-                if (parsedParameters.count != 0) {
-                    parsedType[@"parameters"] = parsedParameters;
-                }
-                
-                if (![name isEqualToString:@"rtx"]) {
-                    NSMutableArray *parsedFbs = [[NSMutableArray alloc] init];
-                    for (auto &it : payloadType.feedbackTypes) {
-                        NSMutableDictionary *parsedFb = [[NSMutableDictionary alloc] init];
-                        parsedFb[@"type"] = [NSString stringWithUTF8String:it.type.c_str()];
-                        if (it.subtype.size() != 0) {
-                            parsedFb[@"subtype"] = [NSString stringWithUTF8String:it.subtype.c_str()];
-                        }
-                        [parsedFbs addObject:parsedFb];
-                    }
-                    parsedType[@"rtcp-fbs"] = parsedFbs;
-                }
-                
-                [videoPayloadTypes addObject:parsedType];
-            }
-            if (videoPayloadTypes.count != 0) {
-                dict[@"payload-types"] = videoPayloadTypes;
-            }
-            
-            NSMutableArray *parsedExtensions = [[NSMutableArray alloc] init];
-            for (auto &it : payload.videoExtensionMap) {
-                NSMutableDictionary *parsedExtension = [[NSMutableDictionary alloc] init];
-                parsedExtension[@"id"] = @(it.first);
-                parsedExtension[@"uri"] = [NSString stringWithUTF8String:it.second.c_str()];
-                [parsedExtensions addObject:parsedExtension];
-            }
-            if (parsedExtensions.count != 0) {
-                dict[@"rtp-hdrexts"] = parsedExtensions;
-            }
-            
-            NSData *data = [NSJSONSerialization dataWithJSONObject:dict options:0 error:nil];
-            NSString *string = [[NSString alloc] initWithData:data encoding:NSUTF8StringEncoding];
-            
-            completion(string, payload.ssrc);
+            processJoinPayload(payload, completion);
         });
     }
 }
@@ -1295,6 +1303,22 @@ static void (*InternalVoipLoggingFunction)(NSString *) = NULL;
     }
 }
 
+- (void)requestVideo:(OngoingCallThreadLocalContextVideoCapturer * _Nullable)videoCapturer completion:(void (^ _Nonnull)(NSString * _Nonnull, uint32_t))completion {
+    if (_instance) {
+        _instance->setVideoCapture([videoCapturer getInterface], [completion](auto payload){
+            processJoinPayload(payload, completion);
+        });
+    }
+}
+
+- (void)disableVideo:(void (^ _Nonnull)(NSString * _Nonnull, uint32_t))completion {
+    if (_instance) {
+        _instance->setVideoCapture(nullptr, [completion](auto payload){
+            processJoinPayload(payload, completion);
+        });
+    }
+}
+
 - (void)setVolumeForSsrc:(uint32_t)ssrc volume:(double)volume {
     if (_instance) {
         _instance->setVolume(ssrc, volume);

submodules/TgVoipWebrtc/tgcalls

@@ -1 +1 @@
-Subproject commit bca416c0b76eac786f01b34e9d09e92df7863168
+Subproject commit d8748598133e39c3fc8c52a77a443c2d1da25032