[WIP] Call tones

submodules/TelegramCallsUI/Sources/GroupCallNavigationAccessoryPanel.swift

@@ -110,6 +110,9 @@ public final class GroupCallNavigationAccessoryPanel: ASDisplayNode {
     private let joinButtonTitleNode: ImmediateTextNode
     private let joinButtonBackgroundNode: ASImageNode
     
+    private var previewImageNode: ASImageNode?
+    private var previewImage: UIImage?
+    
     private var audioLevelView: VoiceBlobView?
     
     private let micButton: HighlightTrackingButtonNode
@@ -536,8 +539,18 @@ public final class GroupCallNavigationAccessoryPanel: ASDisplayNode {
                     guard let self, let data else {
                         return
                     }
-                    let _ = self
-                    let _ = data
+                    
+                    var image: UIImage?
+                    for i in 0 ..< 100 {
+                        image = UIImage(data: data.subdata(in: i ..< data.count))
+                        if image != nil {
+                            break
+                        }
+                    }
+                    self.previewImage = image
+                    if let (size, leftInset, rightInset) = self.validLayout {
+                        self.updateLayout(size: size, leftInset: leftInset, rightInset: rightInset, transition: .animated(duration: 0.2, curve: .easeInOut))
+                    }
                 })
             }
         }
@@ -668,6 +681,26 @@ public final class GroupCallNavigationAccessoryPanel: ASDisplayNode {
         staticTransition.updateFrame(node: self.joinButtonBackgroundNode, frame: CGRect(origin: CGPoint(), size: joinButtonFrame.size))
         staticTransition.updateFrame(node: self.joinButtonTitleNode, frame: CGRect(origin: CGPoint(x: floorToScreenPixels((joinButtonFrame.width - joinButtonTitleSize.width) / 2.0), y: floorToScreenPixels((joinButtonFrame.height - joinButtonTitleSize.height) / 2.0)), size: joinButtonTitleSize))
         
+        if let previewImage = self.previewImage {
+            let previewImageNode: ASImageNode
+            if let current = self.previewImageNode {
+                previewImageNode = current
+            } else {
+                previewImageNode = ASImageNode()
+                previewImageNode.clipsToBounds = true
+                previewImageNode.cornerRadius = 8.0
+                previewImageNode.contentMode = .scaleAspectFill
+                self.previewImageNode = previewImageNode
+                self.addSubnode(previewImageNode)
+            }
+            previewImageNode.image = previewImage
+            let previewSize = CGSize(width: 40.0, height: 40.0)
+            previewImageNode.frame = CGRect(origin: CGPoint(x: joinButtonFrame.minX - previewSize.width - 8.0, y: joinButtonFrame.minY + floor((joinButtonFrame.height - previewSize.height) / 2.0)), size: previewSize)
+        } else if let previewImageNode = self.previewImageNode {
+            self.previewImageNode = nil
+            previewImageNode.removeFromSupernode()
+        }
+        
         let micButtonSize = CGSize(width: 36.0, height: 36.0)
         let micButtonFrame = CGRect(origin: CGPoint(x: size.width - rightInset - 7.0 - micButtonSize.width, y: floor((panelHeight - micButtonSize.height) / 2.0)), size: micButtonSize)
         staticTransition.updateFrame(node: self.micButton, frame: micButtonFrame)

submodules/TelegramCallsUI/Sources/PresentationCallToneData.swift

@@ -4,12 +4,12 @@ import AVFoundation
 private func loadToneData(name: String, addSilenceDuration: Double = 0.0) -> Data? {
     let outputSettings: [String: Any] = [
         AVFormatIDKey: kAudioFormatLinearPCM as NSNumber,
-        AVSampleRateKey: 44100.0 as NSNumber,
+        AVSampleRateKey: 48000.0 as NSNumber,
         AVLinearPCMBitDepthKey: 16 as NSNumber,
         AVLinearPCMIsNonInterleaved: false as NSNumber,
         AVLinearPCMIsFloatKey: false as NSNumber,
         AVLinearPCMIsBigEndianKey: false as NSNumber,
-        AVNumberOfChannelsKey: 2 as NSNumber
+        AVNumberOfChannelsKey: 1 as NSNumber
     ]
     
     let nsName: NSString = name as NSString
@@ -63,9 +63,9 @@ private func loadToneData(name: String, addSilenceDuration: Double = 0.0) -> Dat
     }
     
     if !addSilenceDuration.isZero {
-        let sampleRate = 44100
+        let sampleRate = 48000
         let numberOfSamples = Int(Double(sampleRate) * addSilenceDuration)
-        let numberOfChannels = 2
+        let numberOfChannels = 1
         let numberOfBytes = numberOfSamples * 2 * numberOfChannels
         
         data.append(Data(count: numberOfBytes))

submodules/TelegramCallsUI/Sources/PresentationGroupCall.swift

@@ -2484,11 +2484,19 @@ public final class PresentationGroupCallImpl: PresentationGroupCall {
     }
 
     private func beginTone(tone: PresentationCallTone) {
+        if self.isStream {
+            switch tone {
+            case .groupJoined, .groupLeft:
+                return
+            default:
+                break
+            }
+        }
         if let toneData = presentationCallToneData(tone) {
             self.genericCallContext?.setTone(tone: OngoingGroupCallContext.Tone(
                 samples: toneData,
                 sampleRate: 44100,
-                loopCount: 1000
+                loopCount: tone.loopCount ?? 100000
             ))
         }
         

submodules/TelegramVoip/Sources/GroupCallContext.swift

@@ -415,6 +415,7 @@ public final class OngoingGroupCallContext {
     private final class Impl {
         let queue: Queue
         let context: GroupCallThreadLocalContext
+        let audioDevice: SharedCallAudioDevice?
         
         let sessionId = UInt32.random(in: 0 ..< UInt32(Int32.max))
         
@@ -433,6 +434,12 @@ public final class OngoingGroupCallContext {
         init(queue: Queue, inputDeviceId: String, outputDeviceId: String, audioSessionActive: Signal<Bool, NoError>, video: OngoingCallVideoCapturer?, requestMediaChannelDescriptions: @escaping (Set<UInt32>, @escaping ([MediaChannelDescription]) -> Void) -> Disposable, rejoinNeeded: @escaping () -> Void, outgoingAudioBitrateKbit: Int32?, videoContentType: VideoContentType, enableNoiseSuppression: Bool, disableAudioInput: Bool, preferX264: Bool, logPath: String) {
             self.queue = queue
             
+            #if DEBUG
+            self.audioDevice = SharedCallAudioDevice(disableRecording: disableAudioInput)
+            #else
+            self.audioDevice = nil
+            #endif
+            
             var networkStateUpdatedImpl: ((GroupCallNetworkState) -> Void)?
             var audioLevelsUpdatedImpl: (([NSNumber]) -> Void)?
             
@@ -538,7 +545,8 @@ public final class OngoingGroupCallContext {
                 enableNoiseSuppression: enableNoiseSuppression,
                 disableAudioInput: disableAudioInput,
                 preferX264: preferX264,
-                logPath: logPath
+                logPath: logPath,
+                audioDevice: self.audioDevice
             )
             
             let queue = self.queue
@@ -592,6 +600,7 @@ public final class OngoingGroupCallContext {
                     return
                 }
                 #if os(iOS)
+                self.audioDevice?.setManualAudioSessionIsActive(isActive)
                 self.context.setManualAudioSessionIsActive(isActive)
                 #endif
             }))
@@ -898,7 +907,7 @@ public final class OngoingGroupCallContext {
         }
         
         func setTone(tone: Tone?) {
-            self.context.setTone(tone.flatMap { tone in
+            self.audioDevice?.setTone(tone.flatMap { tone in
                 CallAudioTone(samples: tone.samples, sampleRate: tone.sampleRate, loopCount: tone.loopCount)
             })
         }

submodules/TelegramVoip/Sources/OngoingCallContext.swift

@@ -722,7 +722,7 @@ public final class OngoingCallContext {
         let impl: SharedCallAudioDevice
         
         public static func create() -> AudioDevice? {
-            return AudioDevice(impl: SharedCallAudioDevice())
+            return AudioDevice(impl: SharedCallAudioDevice(disableRecording: false))
         }
         
         private init(impl: SharedCallAudioDevice) {

submodules/TgVoipWebrtc/PublicHeaders/TgVoipWebrtc/OngoingCallThreadLocalContext.h

@@ -23,7 +23,7 @@
 
 @interface SharedCallAudioDevice : NSObject
 
-- (instancetype _Nonnull)init;
+- (instancetype _Nonnull)initWithDisableRecording:(bool)disableRecording;
 
 + (void)setupAudioSession;
 
@@ -397,7 +397,8 @@ typedef NS_ENUM(int32_t, OngoingGroupCallRequestedVideoQuality) {
     enableNoiseSuppression:(bool)enableNoiseSuppression
     disableAudioInput:(bool)disableAudioInput
     preferX264:(bool)preferX264
-    logPath:(NSString * _Nonnull)logPath;
+    logPath:(NSString * _Nonnull)logPath
+audioDevice:(SharedCallAudioDevice * _Nullable)audioDevice;
 
 - (void)stop;
 

submodules/TgVoipWebrtc/Sources/OngoingCallThreadLocalContext.mm

@@ -73,35 +73,31 @@ public:
     
 public:
     virtual rtc::scoped_refptr<webrtc::tgcalls_ios_adm::AudioDeviceModuleIOS> audioDeviceModule() = 0;
+    virtual void start() = 0;
 };
 
 }
 
 class SharedAudioDeviceModuleImpl: public tgcalls::SharedAudioDeviceModule {
 public:
-    SharedAudioDeviceModuleImpl() {
-        if (tgcalls::StaticThreads::getThreads()->getWorkerThread()->IsCurrent()) {
-            _audioDeviceModule = rtc::make_ref_counted<webrtc::tgcalls_ios_adm::AudioDeviceModuleIOS>(false, false, 1);
-            _audioDeviceModule->Init();
-            if (!_audioDeviceModule->Playing()) {
-                _audioDeviceModule->InitPlayout();
-            }
-        } else {
-            tgcalls::StaticThreads::getThreads()->getWorkerThread()->BlockingCall([&]() {
-                _audioDeviceModule = rtc::make_ref_counted<webrtc::tgcalls_ios_adm::AudioDeviceModuleIOS>(false, false, 1);
-                _audioDeviceModule->Init();
-                if (!_audioDeviceModule->Playing()) {
-                    _audioDeviceModule->InitPlayout();
-                }
-            });
-        }
+    SharedAudioDeviceModuleImpl(bool disableAudioInput) {
+        RTC_DCHECK(tgcalls::StaticThreads::getThreads()->getWorkerThread()->IsCurrent());
+        _audioDeviceModule = rtc::make_ref_counted<webrtc::tgcalls_ios_adm::AudioDeviceModuleIOS>(false, disableAudioInput, disableAudioInput ? 2 : 1);
     }
     
     virtual ~SharedAudioDeviceModuleImpl() override {
         if (tgcalls::StaticThreads::getThreads()->getWorkerThread()->IsCurrent()) {
+            if (_audioDeviceModule->Playing()) {
+                _audioDeviceModule->StopPlayout();
+                _audioDeviceModule->StopRecording();
+            }
             _audioDeviceModule = nullptr;
         } else {
             tgcalls::StaticThreads::getThreads()->getWorkerThread()->BlockingCall([&]() {
+                if (_audioDeviceModule->Playing()) {
+                    _audioDeviceModule->StopPlayout();
+                    _audioDeviceModule->StopRecording();
+                }
                 _audioDeviceModule = nullptr;
             });
         }
@@ -112,6 +108,18 @@ public:
         return _audioDeviceModule;
     }
     
+    virtual void start() override {
+        RTC_DCHECK(tgcalls::StaticThreads::getThreads()->getWorkerThread()->IsCurrent());
+        
+        _audioDeviceModule->Init();
+        if (!_audioDeviceModule->Playing()) {
+            _audioDeviceModule->InitPlayout();
+            //_audioDeviceModule->InitRecording();
+            _audioDeviceModule->InternalStartPlayout();
+            //_audioDeviceModule->InternalStartRecording();
+        }
+    }
+    
 private:
     rtc::scoped_refptr<webrtc::tgcalls_ios_adm::AudioDeviceModuleIOS> _audioDeviceModule;
 };
@@ -120,11 +128,11 @@ private:
     std::shared_ptr<tgcalls::ThreadLocalObject<tgcalls::SharedAudioDeviceModule>> _audioDeviceModule;
 }
 
-- (instancetype _Nonnull)init {
+- (instancetype _Nonnull)initWithDisableRecording:(bool)disableRecording {
     self = [super init];
     if (self != nil) {
-        _audioDeviceModule.reset(new tgcalls::ThreadLocalObject<tgcalls::SharedAudioDeviceModule>(tgcalls::StaticThreads::getThreads()->getWorkerThread(), []() mutable {
-            return (tgcalls::SharedAudioDeviceModule *)(new SharedAudioDeviceModuleImpl());
+        _audioDeviceModule.reset(new tgcalls::ThreadLocalObject<tgcalls::SharedAudioDeviceModule>(tgcalls::StaticThreads::getThreads()->getWorkerThread(), [disableRecording]() mutable {
+            return (tgcalls::SharedAudioDeviceModule *)(new SharedAudioDeviceModuleImpl(disableRecording));
         }));
     }
     return self;
@@ -164,6 +172,12 @@ private:
         [[RTCAudioSession sharedInstance] audioSessionDidDeactivate:[AVAudioSession sharedInstance]];
     }
     [RTCAudioSession sharedInstance].isAudioEnabled = isAudioSessionActive;
+    
+    if (isAudioSessionActive) {
+        _audioDeviceModule->perform([](tgcalls::SharedAudioDeviceModule *audioDeviceModule) {
+            audioDeviceModule->start();
+        });
+    }
 }
 
 @end
@@ -1596,6 +1610,8 @@ private:
     
     rtc::scoped_refptr<webrtc::tgcalls_ios_adm::AudioDeviceModuleIOS> _currentAudioDeviceModule;
     rtc::Thread *_currentAudioDeviceModuleThread;
+    
+    SharedCallAudioDevice * _audioDevice;
 }
 
 @end
@@ -1617,7 +1633,8 @@ private:
     enableNoiseSuppression:(bool)enableNoiseSuppression
     disableAudioInput:(bool)disableAudioInput
     preferX264:(bool)preferX264
-    logPath:(NSString * _Nonnull)logPath {
+    logPath:(NSString * _Nonnull)logPath
+audioDevice:(SharedCallAudioDevice * _Nullable)audioDevice {
     self = [super init];
     if (self != nil) {
         _queue = queue;
@@ -1629,6 +1646,12 @@ private:
         _networkStateUpdated = [networkStateUpdated copy];
         _videoCapturer = videoCapturer;
         
+        _audioDevice = audioDevice;
+        std::shared_ptr<tgcalls::ThreadLocalObject<tgcalls::SharedAudioDeviceModule>> audioDeviceModule;
+        if (_audioDevice) {
+            audioDeviceModule = [_audioDevice getAudioDeviceModule];
+        }
+        
         tgcalls::VideoContentType _videoContentType;
         switch (videoContentType) {
             case OngoingGroupCallVideoContentTypeGeneric: {
@@ -1837,7 +1860,10 @@ private:
                 return std::make_shared<RequestMediaChannelDescriptionTaskImpl>(task);
             },
             .minOutgoingVideoBitrateKbit = minOutgoingVideoBitrateKbit,
-            .createAudioDeviceModule = [weakSelf, queue, disableAudioInput](webrtc::TaskQueueFactory *taskQueueFactory) -> rtc::scoped_refptr<webrtc::AudioDeviceModule> {
+            .createAudioDeviceModule = [weakSelf, queue, disableAudioInput, audioDeviceModule](webrtc::TaskQueueFactory *taskQueueFactory) -> rtc::scoped_refptr<webrtc::AudioDeviceModule> {
+                if (audioDeviceModule) {
+                    return audioDeviceModule->getSyncAssumingSameThread()->audioDeviceModule();
+                } else {
                     rtc::Thread *audioDeviceModuleThread = rtc::Thread::Current();
                     auto resultModule = rtc::make_ref_counted<webrtc::tgcalls_ios_adm::AudioDeviceModuleIOS>(false, disableAudioInput, disableAudioInput ? 2 : 1);
                     [queue dispatch:^{
@@ -1849,6 +1875,7 @@ private:
                     }];
                     return resultModule;
                 }
+            }
         }));
     }
     return self;