lesavka/server/src/audio.rs

//  server/src/audio.rs
#![forbid(unsafe_code)]

use anyhow::Context;
use futures_util::Stream;
use gstreamer as gst;
use gstreamer_app as gst_app;
use gst::prelude::*;
use gst::ElementFactory;
use gst::MessageView::*;
use lesavka_common::lesavka::AudioPacket;
use tokio_stream::wrappers::ReceiverStream;
use tonic::Status;
use tracing::{debug, error, warn};

/// “Speaker” stream coming **from** the remote host (UAC2‑gadget playback
/// endpoint) **towards** the client.
pub struct AudioStream {
    _pipeline: gst::Pipeline,
    inner:     ReceiverStream<Result<AudioPacket, Status>>,
}

impl Stream for AudioStream {
    type Item = Result<AudioPacket, Status>;
    fn poll_next(
        mut self: std::pin::Pin<&mut Self>,
        cx: &mut std::task::Context<'_>,
    ) -> std::task::Poll<Option<Self::Item>> {
        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<AudioStream> {
    //   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");

    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),
    })
}

fn build_pipeline_desc(dev: &str) -> anyhow::Result<String> {
    // choose the first encoder that exists on the system
    let enc = ["voaacenc", "avenc_aac", "fdkaacenc"]
        .into_iter()
        .find(|e| ElementFactory::find(e).is_some())      // cheap run‑time probe
        .ok_or_else(|| anyhow::anyhow!("no AAC encoder plugin available"))?;

    Ok(format!(
        // 48 kHz stereo, floats, gadget is master → provide‑clock=false
        "alsasrc device=\"{dev}\" provide-clock=false do-timestamp=true ! \
         audioconvert ! audioresample ! \
         audio/x-raw,format=F32LE,channels=2,rate=48000 ! \
         {enc} bitrate=192000 ! aacparse ! queue ! \
         appsink name=asink emit-signals=true max-buffers=64 drop=true"
    ))
}