// server/src/audio.rs #![forbid(unsafe_code)] use anyhow::{anyhow, Context}; use chrono::Local; use futures_util::Stream; use gstreamer as gst; use gstreamer_app as gst_app; use gst::prelude::*; use gst::{ElementFactory, MessageView}; use gst::MessageView::*; use lesavka_common::lesavka::AudioPacket; use tokio_stream::wrappers::ReceiverStream; use tonic::Status; use tracing::{debug, error, warn}; use std::time::{Duration, SystemTime, UNIX_EPOCH}; /// “Speaker” stream coming **from** the remote host (UAC2‑gadget playback /// endpoint) **towards** the client. pub struct AudioStream { _pipeline: gst::Pipeline, inner: ReceiverStream>, } impl Stream for AudioStream { type Item = Result; fn poll_next( mut self: std::pin::Pin<&mut Self>, cx: &mut std::task::Context<'_>, ) -> std::task::Poll> { Stream::poll_next(std::pin::Pin::new(&mut self.inner), cx) } } impl Drop for AudioStream { fn drop(&mut self) { let _ = self._pipeline.set_state(gst::State::Null); } } /*───────────────────────────────────────────────────────────────────────────*/ /* ear() - capture from ALSA (“speaker”) and push AAC AUs via gRPC */ /*───────────────────────────────────────────────────────────────────────────*/ pub async fn ear(alsa_dev: &str, id: u32) -> anyhow::Result { // NB: one *logical* speaker → id==0. A 2nd logical stream could be // added later (for multi‑channel) without changing the client. gst::init().context("gst init")?; /*──────────── pipeline description ──────────── * * ALSA (UAC2 gadget) AAC+ADTS AppSink * ┌───────────┐ raw 48 kHz ┌─────────┐ AU/ADTS ┌──────────┐ * │ alsasrc │────────────► voaacenc │────────► appsink │ * └───────────┘ └─────────┘ └──────────┘ */ let desc = build_pipeline_desc(alsa_dev)?; let pipeline: gst::Pipeline = gst::parse::launch(&desc)? .downcast() .expect("pipeline"); let sink: gst_app::AppSink = pipeline .by_name("asink") .expect("asink") .downcast() .expect("appsink"); if let Some(tap) = pipeline .by_name("debugtap") .and_then(|e| e.downcast::().ok()) { clip_tap(tap); } let (tx, rx) = tokio::sync::mpsc::channel(8192); let bus = pipeline.bus().expect("bus"); std::thread::spawn(move || { for msg in bus.iter_timed(gst::ClockTime::NONE) { match msg.view() { Error(e) => error!("💥 audio pipeline: {} ({})", e.error(), e.debug().unwrap_or_default()), Warning(w) => warn!("⚠️ audio pipeline: {} ({})", w.error(), w.debug().unwrap_or_default()), StateChanged(s) if s.current() == gst::State::Playing => debug!("🎶 audio pipeline PLAYING"), _ => {} } } }); /*──────────── callbacks ────────────*/ sink.set_callbacks( gst_app::AppSinkCallbacks::builder() .new_sample(move |s| { let sample = s.pull_sample().map_err(|_| gst::FlowError::Eos)?; let buffer = sample.buffer().ok_or(gst::FlowError::Error)?; let map = buffer.map_readable().map_err(|_| gst::FlowError::Error)?; static CNT: std::sync::atomic::AtomicU64 = std::sync::atomic::AtomicU64::new(0); let n = CNT.fetch_add(1, std::sync::atomic::Ordering::Relaxed); if n < 10 || n % 300 == 0 { debug!("🎧 ear #{n}: {} bytes", map.len()); } let pts_us = buffer .pts() .unwrap_or(gst::ClockTime::ZERO) .nseconds() / 1_000; // push non‑blocking; drop oldest on overflow if tx.try_send(Ok(AudioPacket { id, pts: pts_us, data: map.as_slice().to_vec(), })).is_err() { static DROPS: std::sync::atomic::AtomicU64 = std::sync::atomic::AtomicU64::new(0); let d = DROPS.fetch_add(1, std::sync::atomic::Ordering::Relaxed); if d % 300 == 0 { warn!("🎧💔 dropped {d} audio AUs (client too slow)"); } } Ok(gst::FlowSuccess::Ok) }) .build(), ); pipeline.set_state(gst::State::Playing) .context("starting audio pipeline")?; Ok(AudioStream { _pipeline: pipeline, inner: ReceiverStream::new(rx), }) } /*────────────────────────── build_pipeline_desc ───────────────────────────*/ fn build_pipeline_desc(dev: &str) -> anyhow::Result { let reg = gst::Registry::get(); // first available encoder let enc = ["fdkaacenc", "voaacenc", "avenc_aac"] .into_iter() .find(|&e| { reg.find_plugin(e).is_some() || reg .find_feature(e, ElementFactory::static_type()) .is_some() }) .ok_or_else(|| anyhow!("no AAC encoder plugin available"))?; Ok(format!( concat!( "alsasrc device=\"{dev}\" do-timestamp=true ! ", "audio/x-raw,format=S16LE,channels=2,rate=48000 ! ", "audioconvert ! audioresample ! {enc} bitrate=192000 ! ", "aacparse ! ", "capsfilter caps=audio/mpeg,stream-format=adts,channels=2,rate=48000 ! ", "tee name=t ", "t. ! queue ! appsink name=asink emit-signals=true ", "t. ! queue ! appsink name=debugtap emit-signals=true max-buffers=500 drop=true" ), dev = dev, enc = enc )) } /*────────────────────────────── clip_tap() ────────────────────────────────*/ /// Called once per pipeline; spawns a thread that writes a 1 s AAC file /// at the start of every wall‑clock minute **while log‑level == TRACE**. fn clip_tap(tap: gst_app::AppSink) { use gst::prelude::*; std::thread::spawn(move || { use std::fs::File; use std::io::Write; let mut collecting = Vec::with_capacity(200_000); // ~1 s let mut next_min_boundary = next_minute(); loop { match tap.pull_sample() { Ok(s) => { let buf = s.buffer().unwrap(); let map = buf.map_readable().unwrap(); collecting.extend_from_slice(map.as_slice()); // once per minute boundary & trace‑level if tracing::enabled!(tracing::Level::TRACE) && SystemTime::now() >= next_min_boundary { if !collecting.is_empty() { let ts = chrono::Local::now() .format("%Y%m%d-%H%M%S") .to_string(); let path = format!("/tmp/ear-{ts}.aac"); if let Ok(mut f) = File::create(&path) { let _ = f.write_all(&collecting); tracing::debug!("📼 wrote 1 s clip → {}", path); } } collecting.clear(); next_min_boundary = next_minute(); } if collecting.len() > 192_000 { // keep at most ~1 s collecting.truncate(192_000); } } Err(_) => break, // EOS } } }); fn next_minute() -> SystemTime { let now = SystemTime::now() .duration_since(UNIX_EPOCH) .unwrap(); let secs = now.as_secs(); let next = (secs / 60 + 1) * 60; UNIX_EPOCH + Duration::from_secs(next) } }