Merge branch 'master' into rnnoise

.github/workflows/ci.yml

@@ -95,6 +95,8 @@ jobs:
           mkdir -p 16x16/devices 24x24/devices 32x32/devices 48x48/devices 64x64/devices 96x96/devices scalable/devices
           mkdir -p 16x16/status 24x24/status 32x32/status 48x48/status 64x64/status 96x96/status scalable/status
           cd ${GITHUB_WORKSPACE}
+          cp ci/gtk-for-windows/gtk-nsis-pack/share/icons/Adwaita/icon-theme.cache ${artifact_dir}/share/icons/Adwaita/icon-theme.cache
+          cp ci/gtk-for-windows/gtk-nsis-pack/share/icons/Adwaita/index.theme ${artifact_dir}/share/icons/Adwaita/index.theme
           cat "ci/used-icons.txt" | sed 's/\r$//' | xargs -I % cp ci/gtk-for-windows/gtk-nsis-pack/share/icons/Adwaita/16x16/%.symbolic.png ${artifact_dir}/share/icons/Adwaita/16x16/%.symbolic.png || :
           cat "ci/used-icons.txt" | sed 's/\r$//' | xargs -I % cp ci/gtk-for-windows/gtk-nsis-pack/share/icons/Adwaita/24x24/%.symbolic.png ${artifact_dir}/share/icons/Adwaita/24x24/%.symbolic.png || :
           cat "ci/used-icons.txt" | sed 's/\r$//' | xargs -I % cp ci/gtk-for-windows/gtk-nsis-pack/share/icons/Adwaita/32x32/%.symbolic.png ${artifact_dir}/share/icons/Adwaita/32x32/%.symbolic.png || :

src/abaddon.cpp

@@ -58,8 +58,8 @@ Abaddon::Abaddon()
     m_discord.signal_voice_connected().connect(sigc::mem_fun(*this, &Abaddon::OnVoiceConnected));
     m_discord.signal_voice_disconnected().connect(sigc::mem_fun(*this, &Abaddon::OnVoiceDisconnected));
     m_discord.signal_voice_speaking().connect([this](const VoiceSpeakingData &m) {
-        printf("%llu has ssrc %u\n", (uint64_t)m.UserID, m.SSRC);
-        m_audio->AddSSRC(m.SSRC);
+        spdlog::get("voice")->debug("{} SSRC: {}", m.UserID, m.SSRC);
+        m_audio.AddSSRC(m.SSRC);
     });
 #endif
 
@@ -244,8 +244,7 @@ int Abaddon::StartGTK() {
     }
 
 #ifdef WITH_VOICE
-    m_audio = std::make_unique<AudioManager>();
-    if (!m_audio->OK()) {
+    if (!m_audio.OK()) {
         Gtk::MessageDialog dlg(*m_main_window, "The audio engine could not be initialized!", false, Gtk::MESSAGE_ERROR, Gtk::BUTTONS_OK, true);
         dlg.set_position(Gtk::WIN_POS_CENTER);
         dlg.run();
@@ -341,8 +340,8 @@ void Abaddon::StartDiscord() {
 }
 
 void Abaddon::StopDiscord() {
-    if (m_discord.Stop())
-        SaveState();
+    if (m_discord.IsStarted()) SaveState();
+    m_discord.Stop();
     m_main_window->UpdateMenus();
 }
 
@@ -441,13 +440,13 @@ void Abaddon::DiscordOnThreadUpdate(const ThreadUpdateData &data) {
 
 #ifdef WITH_VOICE
 void Abaddon::OnVoiceConnected() {
-    m_audio->StartCaptureDevice();
+    m_audio.StartCaptureDevice();
     ShowVoiceWindow();
 }
 
 void Abaddon::OnVoiceDisconnected() {
-    m_audio->StopCaptureDevice();
-    m_audio->RemoveAllSSRCs();
+    m_audio.StopCaptureDevice();
+    m_audio.RemoveAllSSRCs();
     if (m_voice_window != nullptr) {
         m_voice_window->close();
     }
@@ -461,25 +460,25 @@ void Abaddon::ShowVoiceWindow() {
 
     wnd->signal_mute().connect([this](bool is_mute) {
         m_discord.SetVoiceMuted(is_mute);
-        m_audio->SetCapture(!is_mute);
+        m_audio.SetCapture(!is_mute);
     });
 
     wnd->signal_deafen().connect([this](bool is_deaf) {
         m_discord.SetVoiceDeafened(is_deaf);
-        m_audio->SetPlayback(!is_deaf);
+        m_audio.SetPlayback(!is_deaf);
     });
 
     wnd->signal_gate().connect([this](double gate) {
-        m_audio->SetCaptureGate(gate);
+        m_audio.SetCaptureGate(gate);
     });
 
     wnd->signal_gain().connect([this](double gain) {
-        m_audio->SetCaptureGain(gain);
+        m_audio.SetCaptureGain(gain);
     });
 
     wnd->signal_mute_user_cs().connect([this](Snowflake id, bool is_mute) {
         if (const auto ssrc = m_discord.GetSSRCOfUser(id); ssrc.has_value()) {
-            m_audio->SetMuteSSRC(*ssrc, is_mute);
+            m_audio.SetMuteSSRC(*ssrc, is_mute);
         }
     });
 
@@ -1080,7 +1079,7 @@ EmojiResource &Abaddon::GetEmojis() {
 
 #ifdef WITH_VOICE
 AudioManager &Abaddon::GetAudio() {
-    return *m_audio;
+    return m_audio;
 }
 #endif
 

src/abaddon.hpp

@@ -12,6 +12,7 @@
 #include "imgmanager.hpp"
 #include "emojis.hpp"
 #include "notifications/notifications.hpp"
+#include "audio/manager.hpp"
 
 #define APP_TITLE "Abaddon"
 
@@ -173,7 +174,7 @@ private:
     EmojiResource m_emojis;
 
 #ifdef WITH_VOICE
-    std::unique_ptr<AudioManager> m_audio;
+    AudioManager m_audio;
     Gtk::Window *m_voice_window = nullptr;
 #endif
 

src/audio/manager.cpp

@@ -50,6 +50,15 @@ void capture_data_callback(ma_device *pDevice, void *pOutput, const void *pInput
     if (mgr == nullptr) return;
 
     mgr->OnCapturedPCM(static_cast<const int16_t *>(pInput), frameCount);
+
+    /*
+     * You can simply increment it by 480 in UDPSocket::SendEncrypted but this is wrong
+     * The timestamp is supposed to be strictly linear eg. if there's discontinuous
+     * transmission for 1 second then the timestamp should be 48000 greater than the
+     * last packet. So it's incremented here because this is fired 100x per second
+     * and is always called in sync with UDPSocket::SendEncrypted
+     */
+    mgr->m_rtp_timestamp += 480;
 }
 
 AudioManager::AudioManager() {
@@ -484,6 +493,10 @@ AudioDevices &AudioManager::GetDevices() {
     return m_devices;
 }
 
+uint32_t AudioManager::GetRTPTimestamp() const noexcept {
+    return m_rtp_timestamp;
+}
+
 AudioManager::type_signal_opus_packet AudioManager::signal_opus_packet() {
     return m_signal_opus_packet;
 }

src/audio/manager.hpp

@@ -65,6 +65,8 @@ public:
 
     [[nodiscard]] AudioDevices &GetDevices();
 
+    [[nodiscard]] uint32_t GetRTPTimestamp() const noexcept;
+
 private:
     void OnCapturedPCM(const int16_t *pcm, ma_uint32 frames);
 
@@ -116,6 +118,7 @@ private:
     AudioDevices m_devices;
 
     DenoiseState *m_rnnoise;
+    std::atomic<uint32_t> m_rtp_timestamp = 0;
 
 public:
     using type_signal_opus_packet = sigc::signal<void(int payload_size)>;

src/discord/discord.cpp

@@ -2813,7 +2813,6 @@ void DiscordClient::SetVoiceState(Snowflake user_id, const VoiceState &state) {
         spdlog::get("discord")->error("SetVoiceState called with missing channel ID");
         return;
     }
-    spdlog::get("discord")->debug("SetVoiceState: {} -> {}", user_id, *state.ChannelID);
 
     auto flags = VoiceStateFlags::Clear;
     if (state.IsSelfMuted) flags |= VoiceStateFlags::SelfMute;

src/discord/snowflake.hpp

@@ -2,6 +2,8 @@
 #include <cstdint>
 #include <nlohmann/json.hpp>
 #include <glibmm/ustring.h>
+#include <spdlog/fmt/fmt.h>
+#include <spdlog/fmt/ostr.h>
 
 struct Snowflake {
     Snowflake();
@@ -39,6 +41,13 @@ private:
     unsigned long long m_num;
 };
 
+template<>
+struct fmt::formatter<Snowflake> : fmt::formatter<std::string> {
+    auto format(Snowflake id, format_context &ctx) -> decltype(ctx.out()) {
+        return format_to(ctx.out(), "[id: {}]", static_cast<uint64_t>(id));
+    }
+};
+
 namespace std {
 template<>
 struct hash<Snowflake> {

src/discord/voiceclient.cpp

@@ -49,17 +49,18 @@ void UDPSocket::SetSSRC(uint32_t ssrc) {
 
 void UDPSocket::SendEncrypted(const uint8_t *data, size_t len) {
     m_sequence++;
-    m_timestamp += 480; // this is important
+
+    const uint32_t timestamp = Abaddon::Get().GetAudio().GetRTPTimestamp();
 
     std::vector<uint8_t> rtp(12 + len + crypto_secretbox_MACBYTES, 0);
     rtp[0] = 0x80; // ver 2
     rtp[1] = 0x78; // payload type 0x78
     rtp[2] = (m_sequence >> 8) & 0xFF;
     rtp[3] = (m_sequence >> 0) & 0xFF;
-    rtp[4] = (m_timestamp >> 24) & 0xFF;
-    rtp[5] = (m_timestamp >> 16) & 0xFF;
-    rtp[6] = (m_timestamp >> 8) & 0xFF;
-    rtp[7] = (m_timestamp >> 0) & 0xFF;
+    rtp[4] = (timestamp >> 24) & 0xFF;
+    rtp[5] = (timestamp >> 16) & 0xFF;
+    rtp[6] = (timestamp >> 8) & 0xFF;
+    rtp[7] = (timestamp >> 0) & 0xFF;
     rtp[8] = (m_ssrc >> 24) & 0xFF;
     rtp[9] = (m_ssrc >> 16) & 0xFF;
     rtp[10] = (m_ssrc >> 8) & 0xFF;
@@ -382,7 +383,7 @@ void DiscordVoiceClient::Discovery() {
         const auto response = m_udp.Receive();
         if (response.size() >= 74 && response[0] == 0x00 && response[1] == 0x02) {
             const char *ip = reinterpret_cast<const char *>(response.data() + 8);
-            uint16_t port = (response[73] << 8) | response[74];
+            uint16_t port = (response[72] << 8) | response[73];
             m_log->info("Discovered IP and port: {}:{}", ip, port);
             SelectProtocol(ip, port);
             break;

src/discord/voiceclient.hpp

@@ -171,7 +171,6 @@ private:
     uint32_t m_ssrc;
 
     uint16_t m_sequence = 0;
-    uint32_t m_timestamp = 0;
 
 public:
     using type_signal_data = sigc::signal<void, std::vector<uint8_t>>;