fix some potential audio issues, clean up the code a bit, and suppress some warnings (#737)

* fix some potential audio issues, clean up the code a bit, and surpress some warnings

* use assign to initialize the buffer

Author: xianshijing-lk

livekit/src/room/utils/utf8_chunk.rs

@@ -59,11 +59,11 @@ impl<'a> Iterator for Utf8AwareChunks<'a> {
 pub trait Utf8AwareChunkExt {
     /// Splits the bytes into chunks of the specified size, ensuring that
     /// UTF-8 character boundaries are respected.
-    fn utf8_aware_chunks(&self, chunk_size: usize) -> Utf8AwareChunks;
+    fn utf8_aware_chunks(&self, chunk_size: usize) -> Utf8AwareChunks<'_>;
 }
 
 impl Utf8AwareChunkExt for [u8] {
-    fn utf8_aware_chunks(&self, chunk_size: usize) -> Utf8AwareChunks {
+    fn utf8_aware_chunks(&self, chunk_size: usize) -> Utf8AwareChunks<'_> {
         Utf8AwareChunks::new(self, chunk_size)
     }
 }

webrtc-sys/build.rs

@@ -216,7 +216,8 @@ fn main() {
                 .file("src/objc_video_factory.mm")
                 .file("src/objc_video_frame_buffer.mm")
                 .flag("-stdlib=libc++")
-                .flag("-std=c++20");
+                .flag("-std=c++20")
+                .flag("-Wno-nullability-completeness");
         }
         "ios" => {
             println!("cargo:rustc-link-lib=framework=Foundation");

webrtc-sys/include/livekit/audio_track.h

@@ -17,6 +17,7 @@
 #pragma once
 
 #include <memory>
+#include <vector>
 
 #include "api/audio/audio_frame.h"
 #include "api/audio_options.h"
@@ -137,7 +138,7 @@ class AudioTrackSource {
     void (*on_complete_)(const SourceContext*) RTC_GUARDED_BY(mutex_);
 
     int missed_frames_ RTC_GUARDED_BY(mutex_) = 0;
-    int16_t* silence_buffer_ = nullptr;
+    std::vector<int16_t> silence_buffer_;
 
     int sample_rate_;
     int num_channels_;

webrtc-sys/src/audio_track.cpp

@@ -150,7 +150,7 @@ AudioTrackSource::InternalSource::InternalSource(
 
   int samples10ms = sample_rate / 100 * num_channels;
 
-  silence_buffer_ = new int16_t[samples10ms]();
+  silence_buffer_.assign(samples10ms, 0);
   queue_size_samples_ = queue_size_ms / 10 * samples10ms;
   notify_threshold_samples_ = queue_size_samples_;  // TODO: this is currently
                                                     // using x2 the queue size
@@ -164,18 +164,21 @@ AudioTrackSource::InternalSource::InternalSource(
       audio_queue_.get(),
       [this, samples10ms]() {
         webrtc::MutexLock lock(&mutex_);
+        constexpr int kBitsPerSample = sizeof(int16_t) * 8;
 
         if (buffer_.size() >= samples10ms) {
+          // Reset |missed_frames_| to 0 so that it won't keep sending silence to webrtc due to audio callback timing drifts.
+          missed_frames_ = 0;
           for (auto sink : sinks_)
-            sink->OnData(buffer_.data(), sizeof(int16_t) * 8, sample_rate_,
+            sink->OnData(buffer_.data(), kBitsPerSample, sample_rate_,
                          num_channels_, samples10ms / num_channels_);
 
           buffer_.erase(buffer_.begin(), buffer_.begin() + samples10ms);
         } else {
           missed_frames_++;
           if (missed_frames_ >= silence_frames_threshold) {
             for (auto sink : sinks_)
-              sink->OnData(silence_buffer_, sizeof(int16_t) * 8, sample_rate_,
+              sink->OnData(silence_buffer_.data(), kBitsPerSample, sample_rate_,
                            num_channels_, samples10ms / num_channels_);
           }
         }
@@ -192,7 +195,6 @@ AudioTrackSource::InternalSource::InternalSource(
 }
 
 AudioTrackSource::InternalSource::~InternalSource() {
-  delete[] silence_buffer_;
 }
 
 bool AudioTrackSource::InternalSource::capture_frame(