Compare commits

..
Author SHA1 Message Date
mollusk 33e3998e7c Add orderly shutdown on window close 2026-06-16 17:37:00 -04:00
mollusk 44bad7b70b Fix jitter restart and WAV size overflow 2026-06-16 17:28:41 -04:00
mollusk 20643a24de Add audio controls and focused hotkeys 2026-06-16 17:23:38 -04:00
14 changed files with 1645 additions and 220 deletions
+542 -175
View File
File diff suppressed because it is too large Load Diff
+316
View File
@@ -0,0 +1,316 @@
//! Per-peer listener-side voice EQ.
//!
//! The EQ is deliberately small and local: three RBJ cookbook biquads at fixed
//! voice-oriented frequencies, with only gain exposed to the UI. State lives per
//! peer in the playout mixer so filter delay registers are continuous across 20ms
//! Opus frames; flat settings are treated as bypass so the default path is cheap
//! and sample-exact.
use serde::{Deserialize, Serialize};
const DEFAULT_SAMPLE_RATE: f32 = 48_000.0;
const LOW_SHELF_HZ: f32 = 160.0;
const MID_PEAK_HZ: f32 = 2_400.0;
const HIGH_SHELF_HZ: f32 = 6_500.0;
const MID_Q: f32 = 1.0;
const SHELF_Q: f32 = std::f32::consts::FRAC_1_SQRT_2;
const FLAT_EPSILON_DB: f32 = 0.001;
/// UI and config clamp for each band. Wide enough to be useful for voice, narrow
/// enough that a peer cannot accidentally make the listener-side limiter do all
/// the work.
pub const EQ_GAIN_DB_MIN: f32 = -12.0;
pub const EQ_GAIN_DB_MAX: f32 = 12.0;
/// Persisted per-peer EQ gains, in decibels. `Default` is flat/bypassed.
#[derive(Debug, Clone, Copy, Serialize, Deserialize, PartialEq)]
pub struct EqSettings {
#[serde(default)]
pub low_gain_db: f32,
#[serde(default)]
pub mid_gain_db: f32,
#[serde(default)]
pub high_gain_db: f32,
}
impl Default for EqSettings {
fn default() -> Self {
Self {
low_gain_db: 0.0,
mid_gain_db: 0.0,
high_gain_db: 0.0,
}
}
}
impl EqSettings {
pub fn flat() -> Self {
Self::default()
}
/// Clamp all public gains to the supported UI/DSP range.
pub fn clamped(self) -> Self {
Self {
low_gain_db: self.low_gain_db.clamp(EQ_GAIN_DB_MIN, EQ_GAIN_DB_MAX),
mid_gain_db: self.mid_gain_db.clamp(EQ_GAIN_DB_MIN, EQ_GAIN_DB_MAX),
high_gain_db: self.high_gain_db.clamp(EQ_GAIN_DB_MIN, EQ_GAIN_DB_MAX),
}
}
/// True when the EQ should be bypassed entirely.
pub fn is_flat(self) -> bool {
self.low_gain_db.abs() <= FLAT_EPSILON_DB
&& self.mid_gain_db.abs() <= FLAT_EPSILON_DB
&& self.high_gain_db.abs() <= FLAT_EPSILON_DB
}
}
/// A stateful three-band EQ. One instance belongs to one decoded peer stream.
pub struct Eq {
settings: EqSettings,
low: Biquad,
mid: Biquad,
high: Biquad,
}
impl Eq {
/// Build an EQ at the application's audio rate (48 kHz).
pub fn new(settings: EqSettings) -> Self {
Self::with_sample_rate(settings, DEFAULT_SAMPLE_RATE)
}
fn with_sample_rate(settings: EqSettings, sample_rate: f32) -> Self {
let settings = settings.clamped();
Self {
settings,
low: Biquad::low_shelf(sample_rate, LOW_SHELF_HZ, settings.low_gain_db, SHELF_Q),
mid: Biquad::peaking(sample_rate, MID_PEAK_HZ, settings.mid_gain_db, MID_Q),
high: Biquad::high_shelf(sample_rate, HIGH_SHELF_HZ, settings.high_gain_db, SHELF_Q),
}
}
pub fn settings(&self) -> EqSettings {
self.settings
}
/// Process one mono PCM frame in place. Flat settings are sample-exact bypass.
pub fn process_frame(&mut self, frame: &mut [i16]) {
if self.settings.is_flat() {
return;
}
for sample in frame {
let x = *sample as f32;
let y = self.high.process(self.mid.process(self.low.process(x)));
*sample = y.round().clamp(i16::MIN as f32, i16::MAX as f32) as i16;
}
}
}
#[derive(Debug, Clone, Copy)]
struct Coeffs {
b0: f32,
b1: f32,
b2: f32,
a1: f32,
a2: f32,
}
impl Coeffs {
fn normalized(b0: f32, b1: f32, b2: f32, a0: f32, a1: f32, a2: f32) -> Self {
let inv_a0 = 1.0 / a0;
Self {
b0: b0 * inv_a0,
b1: b1 * inv_a0,
b2: b2 * inv_a0,
a1: a1 * inv_a0,
a2: a2 * inv_a0,
}
}
fn all_finite(self) -> bool {
self.b0.is_finite()
&& self.b1.is_finite()
&& self.b2.is_finite()
&& self.a1.is_finite()
&& self.a2.is_finite()
}
}
/// Direct Form II transposed biquad. The two delay registers are the state that
/// must survive across frames.
struct Biquad {
coeffs: Coeffs,
z1: f32,
z2: f32,
}
impl Biquad {
fn new(coeffs: Coeffs) -> Self {
debug_assert!(coeffs.all_finite());
Self {
coeffs,
z1: 0.0,
z2: 0.0,
}
}
fn low_shelf(sample_rate: f32, freq: f32, gain_db: f32, q: f32) -> Self {
let (a, cos_w0, alpha) = rbj_terms(sample_rate, freq, gain_db, q);
let sqrt_a = a.sqrt();
let b0 = a * ((a + 1.0) - (a - 1.0) * cos_w0 + 2.0 * sqrt_a * alpha);
let b1 = 2.0 * a * ((a - 1.0) - (a + 1.0) * cos_w0);
let b2 = a * ((a + 1.0) - (a - 1.0) * cos_w0 - 2.0 * sqrt_a * alpha);
let a0 = (a + 1.0) + (a - 1.0) * cos_w0 + 2.0 * sqrt_a * alpha;
let a1 = -2.0 * ((a - 1.0) + (a + 1.0) * cos_w0);
let a2 = (a + 1.0) + (a - 1.0) * cos_w0 - 2.0 * sqrt_a * alpha;
Self::new(Coeffs::normalized(b0, b1, b2, a0, a1, a2))
}
fn peaking(sample_rate: f32, freq: f32, gain_db: f32, q: f32) -> Self {
let (a, cos_w0, alpha) = rbj_terms(sample_rate, freq, gain_db, q);
let b0 = 1.0 + alpha * a;
let b1 = -2.0 * cos_w0;
let b2 = 1.0 - alpha * a;
let a0 = 1.0 + alpha / a;
let a1 = -2.0 * cos_w0;
let a2 = 1.0 - alpha / a;
Self::new(Coeffs::normalized(b0, b1, b2, a0, a1, a2))
}
fn high_shelf(sample_rate: f32, freq: f32, gain_db: f32, q: f32) -> Self {
let (a, cos_w0, alpha) = rbj_terms(sample_rate, freq, gain_db, q);
let sqrt_a = a.sqrt();
let b0 = a * ((a + 1.0) + (a - 1.0) * cos_w0 + 2.0 * sqrt_a * alpha);
let b1 = -2.0 * a * ((a - 1.0) + (a + 1.0) * cos_w0);
let b2 = a * ((a + 1.0) + (a - 1.0) * cos_w0 - 2.0 * sqrt_a * alpha);
let a0 = (a + 1.0) - (a - 1.0) * cos_w0 + 2.0 * sqrt_a * alpha;
let a1 = 2.0 * ((a - 1.0) - (a + 1.0) * cos_w0);
let a2 = (a + 1.0) - (a - 1.0) * cos_w0 - 2.0 * sqrt_a * alpha;
Self::new(Coeffs::normalized(b0, b1, b2, a0, a1, a2))
}
fn process(&mut self, x: f32) -> f32 {
let y = self.coeffs.b0 * x + self.z1;
self.z1 = self.coeffs.b1 * x - self.coeffs.a1 * y + self.z2;
self.z2 = self.coeffs.b2 * x - self.coeffs.a2 * y;
// Avoid carrying denormal-sized state forever on long quiet tails.
if self.z1.abs() < 1.0e-20 {
self.z1 = 0.0;
}
if self.z2.abs() < 1.0e-20 {
self.z2 = 0.0;
}
y
}
}
fn rbj_terms(sample_rate: f32, freq: f32, gain_db: f32, q: f32) -> (f32, f32, f32) {
let sr = sample_rate.max(1.0);
let f = freq.clamp(1.0, sr * 0.49);
let w0 = 2.0 * std::f32::consts::PI * f / sr;
let a = 10.0f32.powf(gain_db / 40.0);
let alpha = w0.sin() / (2.0 * q.max(0.001));
(a, w0.cos(), alpha)
}
#[cfg(test)]
mod tests {
use super::*;
fn sine(freq: f32, len: usize, amp: f32) -> Vec<i16> {
(0..len)
.map(|n| {
let t = n as f32 / DEFAULT_SAMPLE_RATE;
(amp * (2.0 * std::f32::consts::PI * freq * t).sin()).round() as i16
})
.collect()
}
fn rms(frame: &[i16]) -> f32 {
let sum: f32 = frame.iter().map(|&s| (s as f32).powi(2)).sum();
(sum / frame.len().max(1) as f32).sqrt()
}
#[test]
fn flat_eq_is_sample_exact_identity() {
let mut eq = Eq::new(EqSettings::flat());
let mut frame: Vec<i16> = (-480..480).map(|n| (n * 31) as i16).collect();
let original = frame.clone();
eq.process_frame(&mut frame);
assert_eq!(frame, original);
}
#[test]
fn low_shelf_boost_raises_low_frequency_energy() {
let mut eq = Eq::new(EqSettings {
low_gain_db: 9.0,
..EqSettings::flat()
});
let mut low = sine(100.0, 48_000, 3_000.0);
let before = rms(&low);
eq.process_frame(&mut low);
let after = rms(&low);
assert!(after > before * 1.6, "low shelf should boost low RMS: {before} -> {after}");
}
#[test]
fn high_shelf_boost_raises_high_frequency_energy() {
let mut eq = Eq::new(EqSettings {
high_gain_db: 9.0,
..EqSettings::flat()
});
let mut high = sine(8_000.0, 48_000, 3_000.0);
let before = rms(&high);
eq.process_frame(&mut high);
let after = rms(&high);
assert!(after > before * 1.6, "high shelf should boost high RMS: {before} -> {after}");
}
#[test]
fn coefficients_are_finite_across_supported_gain_range() {
for gain in [EQ_GAIN_DB_MIN, -6.0, 0.0, 6.0, EQ_GAIN_DB_MAX] {
for b in [
Biquad::low_shelf(DEFAULT_SAMPLE_RATE, LOW_SHELF_HZ, gain, SHELF_Q),
Biquad::peaking(DEFAULT_SAMPLE_RATE, MID_PEAK_HZ, gain, MID_Q),
Biquad::high_shelf(DEFAULT_SAMPLE_RATE, HIGH_SHELF_HZ, gain, SHELF_Q),
] {
assert!(b.coeffs.all_finite(), "coefficients must be finite at {gain} dB");
}
}
}
#[test]
fn hot_signal_does_not_nan_or_wrap() {
let mut eq = Eq::new(EqSettings {
low_gain_db: 12.0,
mid_gain_db: 12.0,
high_gain_db: 12.0,
});
let mut frame = sine(1_000.0, 48_000, 30_000.0);
eq.process_frame(&mut frame);
let peak = frame
.iter()
.map(|&s| i32::from(s).abs())
.max()
.unwrap_or(0);
assert!(peak > 1_000, "processed signal should retain audible energy");
assert!(
frame.iter().any(|&s| s > 0) && frame.iter().any(|&s| s < 0),
"a boosted sine should retain both polarities"
);
}
#[test]
fn settings_are_clamped() {
let s = EqSettings {
low_gain_db: -99.0,
mid_gain_db: 2.0,
high_gain_db: 99.0,
}
.clamped();
assert_eq!(s.low_gain_db, EQ_GAIN_DB_MIN);
assert_eq!(s.mid_gain_db, 2.0);
assert_eq!(s.high_gain_db, EQ_GAIN_DB_MAX);
}
}
+11 -4
View File
@@ -1,17 +1,22 @@
use std::sync::mpsc::{Sender, Receiver};
use std::sync::mpsc::{Receiver, Sender};
use std::sync::Arc;
use std::sync::atomic::AtomicUsize;
use thiserror::Error;
/// Target depth of the playback ring buffer, in samples (48kHz mono).
/// Playback output channel count. Capture/encode/network remain mono; only the
/// listener-side playout bus is stereo.
pub const PLAYBACK_CHANNELS: usize = 2;
/// Target depth of the playback ring buffer, in interleaved samples (48kHz
/// stereo).
///
/// The playout chain is paced to keep the ring near this level: production is
/// driven by how fast PipeWire actually drains the ring (the hardware clock),
/// not by a fixed software timer — which is what eliminates the producer/
/// consumer beat that otherwise churns ~20% of audio into drops + silence.
/// 2880 = 60ms = 3×20ms frames, comfortably above the 2048-sample max quantum
/// 5760 = 60ms = 3×20ms stereo frames, comfortably above the 2048-frame max quantum
/// so a single hardware pull can never empty the ring before the mixer refills.
pub const PLAYBACK_TARGET_SAMPLES: usize = 2880;
pub const PLAYBACK_TARGET_SAMPLES: usize = 2880 * PLAYBACK_CHANNELS;
#[derive(Error, Debug)]
pub enum AudioError {
@@ -52,9 +57,11 @@ pub trait AudioBackend: Send + Sync {
}
pub mod echo_cancel;
pub mod eq;
pub mod gate;
pub mod limiter;
pub mod multitrack;
pub mod pan;
pub mod pipewire_impl;
pub mod pw_cli;
pub mod recorder;
+77
View File
@@ -0,0 +1,77 @@
//! Listener-side stereo pan law.
//!
//! Capture, Opus, and the network stay mono. These helpers are used only after a
//! peer has been decoded locally, just before the playout mix is written to the
//! stereo playback bus.
/// Clamp and compute constant-power pan gains for `pan` in `[-1.0, 1.0]`.
///
/// - `-1.0` is hard left `(1, 0)`
/// - `0.0` is center `(sqrt(1/2), sqrt(1/2))`
/// - `1.0` is hard right `(0, 1)`
pub fn pan_gains(pan: f32) -> (f32, f32) {
let pan = pan.clamp(-1.0, 1.0);
let theta = (pan + 1.0) * std::f32::consts::FRAC_PI_4;
(theta.cos(), theta.sin())
}
/// Gains used by the legacy-compatible playback mixer.
///
/// The pure law above is constant-power. The existing application, however, was
/// mono and users heard the full old mono signal in both ears. Scaling by sqrt(2)
/// makes `pan = 0` exactly dual-mono `(1, 1)`, preserving the default sound while
/// still following the same equal-power curve as a peer is moved away from center.
pub fn playback_pan_gains(pan: f32) -> (f32, f32) {
let (left, right) = pan_gains(pan);
(left * std::f32::consts::SQRT_2, right * std::f32::consts::SQRT_2)
}
#[cfg(test)]
mod tests {
use super::*;
const EPS: f32 = 1.0e-6;
#[test]
fn hard_left_and_right_are_endpoints() {
assert_eq!(pan_gains(-1.0), (1.0, 0.0));
let (l, r) = pan_gains(1.0);
assert!(l.abs() < EPS, "left at hard-right should be zero-ish, got {l}");
assert!((r - 1.0).abs() < EPS, "right at hard-right should be one, got {r}");
}
#[test]
fn center_is_equal_and_power_preserving() {
let (l, r) = pan_gains(0.0);
assert!((l - r).abs() < EPS);
assert!((l - std::f32::consts::FRAC_1_SQRT_2).abs() < EPS);
assert!(((l * l + r * r) - 1.0).abs() < EPS);
}
#[test]
fn gains_move_monotonically() {
let pans = [-1.0, -0.5, 0.0, 0.5, 1.0];
let mut prev_l = f32::INFINITY;
let mut prev_r = f32::NEG_INFINITY;
for pan in pans {
let (l, r) = pan_gains(pan);
assert!(l <= prev_l + EPS, "left gain must not rise as pan moves right");
assert!(r >= prev_r - EPS, "right gain must not fall as pan moves right");
prev_l = l;
prev_r = r;
}
}
#[test]
fn playback_center_preserves_legacy_dual_mono() {
let (l, r) = playback_pan_gains(0.0);
assert!((l - 1.0).abs() < EPS);
assert!((r - 1.0).abs() < EPS);
}
#[test]
fn input_is_clamped() {
assert_eq!(pan_gains(-9.0), pan_gains(-1.0));
assert_eq!(pan_gains(9.0), pan_gains(1.0));
}
}
+23 -18
View File
@@ -283,8 +283,9 @@ fn run_playback(
let core = context.connect_rc(None)
.map_err(|e| AudioError::Init(e.to_string()))?;
// Ring buffer setup: 9600 samples (200ms capacity for mono 48kHz).
const RING_CAPACITY: usize = 9600;
// Ring buffer setup: 19200 interleaved samples (200ms capacity for stereo
// 48kHz).
const RING_CAPACITY: usize = 9600 * crate::audio::PLAYBACK_CHANNELS;
let rb = HeapRb::<i16>::new(RING_CAPACITY);
let (mut producer, consumer) = rb.split();
@@ -371,7 +372,7 @@ fn run_playback(
let data = &mut datas[0];
let mut total_size = 0;
if let Some(slice) = data.data() {
let stride = 2; // S16LE Mono = 2 bytes per frame
let stride = 2 * crate::audio::PLAYBACK_CHANNELS; // S16LE stereo
// Fill exactly what the graph asked for this cycle (with
// a safe fallback), never the whole mapped slice — that
// over-pull past the ring depth was the original crackle.
@@ -383,17 +384,20 @@ fn run_playback(
user_data.callback_count.fetch_add(1, Ordering::Relaxed);
let mut starved = 0u64;
for i in 0..n_frames {
let val = match user_data.consumer.try_pop() {
Some(v) => v,
None => {
starved += 1;
0
}
};
let bytes = val.to_le_bytes();
let start = i * stride;
slice[start] = bytes[0];
slice[start + 1] = bytes[1];
for ch in 0..crate::audio::PLAYBACK_CHANNELS {
let val = match user_data.consumer.try_pop() {
Some(v) => v,
None => {
starved += 1;
0
}
};
let bytes = val.to_le_bytes();
let offset = start + ch * 2;
slice[offset] = bytes[0];
slice[offset + 1] = bytes[1];
}
}
if starved > 0 {
// One wait-free atomic add per quantum — RT-safe.
@@ -403,7 +407,8 @@ fn run_playback(
// actually pulled (excluding underruns, which removed
// nothing) so the mixer paces against true ring depth.
// Wait-free fetch_sub, RT-safe.
let popped = n_frames - starved as usize;
let requested_samples = n_frames * crate::audio::PLAYBACK_CHANNELS;
let popped = requested_samples - starved as usize;
if popped > 0 {
user_data.fill_gauge.fetch_sub(popped, Ordering::Relaxed);
}
@@ -411,7 +416,7 @@ fn run_playback(
}
let chunk = data.chunk_mut();
*chunk.offset_mut() = 0;
*chunk.stride_mut() = 2;
*chunk.stride_mut() = (2 * crate::audio::PLAYBACK_CHANNELS) as _;
*chunk.size_mut() = total_size as _;
}
}
@@ -422,7 +427,7 @@ fn run_playback(
let mut audio_info = spa::param::audio::AudioInfoRaw::new();
audio_info.set_format(spa::param::audio::AudioFormat::S16LE);
audio_info.set_rate(48000);
audio_info.set_channels(1); // Mono
audio_info.set_channels(crate::audio::PLAYBACK_CHANNELS as u32); // Stereo playback
let obj = pw::spa::pod::Object {
type_: pw::spa::utils::SpaTypes::ObjectParamFormat.as_raw(),
@@ -450,7 +455,7 @@ fn run_playback(
// `frames_to_produce`). `requested()`, not the buffer size, now governs
// per-cycle output, so this is a generous max rather than a hard pin.
const MAX_QUANTUM_FRAMES: i32 = 8192;
const STRIDE: i32 = 2; // S16LE mono = 2 bytes/frame
const STRIDE: i32 = 2 * crate::audio::PLAYBACK_CHANNELS as i32; // S16LE stereo
let buffers_obj = pw::spa::pod::Object {
type_: pw::spa::utils::SpaTypes::ObjectParamBuffers.as_raw(),
id: pw::spa::param::ParamType::Buffers.as_raw(),
@@ -555,7 +560,7 @@ fn run_playback(
if verbose || du > 0 || dd > 0 {
crate::log_msg(&format!(
"playout-health: fill={fill} samples (~{}ms) | underrun +{du} samples/s (total {u}) | dropped +{dd} frames/s (total {d}) | quantum={q} frames, {dc} callbacks/s",
fill / 48,
fill / (48 * crate::audio::PLAYBACK_CHANNELS),
));
}
}
+55 -8
View File
@@ -22,6 +22,8 @@ use std::path::{Path, PathBuf};
const SAMPLE_RATE: u32 = 48_000;
const BITS_PER_SAMPLE: u16 = 16;
const CHANNELS: u16 = 1;
const RIFF_DATA_OVERHEAD: u64 = 36;
const MAX_RIFF_DATA_BYTES: u64 = u32::MAX as u64 - RIFF_DATA_OVERHEAD;
/// Cap on buffered mic samples (~200ms). Bounds how far recording lag can drift
/// if the capture clock runs persistently faster than playout — past this we drop
@@ -34,7 +36,7 @@ const MAX_MIC_FIFO: usize = SAMPLE_RATE as usize / 5;
pub struct WavWriter {
file: File,
/// Bytes of PCM data written so far (for the size fields).
data_bytes: u32,
data_bytes: u64,
}
impl WavWriter {
@@ -42,7 +44,10 @@ impl WavWriter {
pub fn new(path: &Path) -> io::Result<Self> {
let mut file = File::create(path)?;
file.write_all(&Self::header(0))?;
Ok(Self { file, data_bytes: 0 })
Ok(Self {
file,
data_bytes: 0,
})
}
/// The 44-byte canonical WAV/PCM header for the given data length in bytes.
@@ -68,21 +73,40 @@ impl WavWriter {
/// Append PCM samples to the data chunk.
pub fn write_samples(&mut self, samples: &[i16]) -> io::Result<()> {
let added_bytes = u64::try_from(samples.len())
.ok()
.and_then(|len| len.checked_mul(2))
.ok_or_else(|| io::Error::other("WAV sample buffer too large"))?;
let new_data_bytes = self
.data_bytes
.checked_add(added_bytes)
.ok_or_else(|| io::Error::other("WAV data size overflow"))?;
if new_data_bytes > MAX_RIFF_DATA_BYTES {
return Err(io::Error::other("WAV too large for RIFF"));
}
let mut buf = Vec::with_capacity(samples.len() * 2);
for &s in samples {
buf.extend_from_slice(&s.to_le_bytes());
}
self.file.write_all(&buf)?;
self.data_bytes += (samples.len() * 2) as u32;
self.data_bytes = new_data_bytes;
Ok(())
}
/// Patch the RIFF + data size fields and flush. Consumes the writer.
pub fn finalize(mut self) -> io::Result<()> {
let data_bytes = u32::try_from(self.data_bytes)
.map_err(|_| io::Error::other("WAV too large for RIFF"))?;
let riff_size = self
.data_bytes
.checked_add(RIFF_DATA_OVERHEAD)
.and_then(|size| u32::try_from(size).ok())
.ok_or_else(|| io::Error::other("WAV too large for RIFF"))?;
self.file.seek(SeekFrom::Start(4))?;
self.file.write_all(&(36 + self.data_bytes).to_le_bytes())?;
self.file.write_all(&riff_size.to_le_bytes())?;
self.file.seek(SeekFrom::Start(40))?;
self.file.write_all(&self.data_bytes.to_le_bytes())?;
self.file.write_all(&data_bytes.to_le_bytes())?;
self.file.flush()?;
Ok(())
}
@@ -209,11 +233,29 @@ mod tests {
let _ = std::fs::remove_file(&path);
}
#[test]
fn wav_writer_rejects_data_that_would_overflow_riff_header() {
let dir = std::env::temp_dir();
let path = dir.join(format!("peerspeak-overflow-{}.wav", std::process::id()));
let mut w = WavWriter::new(&path).unwrap();
w.data_bytes = MAX_RIFF_DATA_BYTES - 1;
let before_len = std::fs::metadata(&path).unwrap().len();
let err = w.write_samples(&[0]).unwrap_err();
assert_eq!(err.kind(), io::ErrorKind::Other);
assert_eq!(w.data_bytes, MAX_RIFF_DATA_BYTES - 1);
assert_eq!(std::fs::metadata(&path).unwrap().len(), before_len);
drop(w);
let _ = std::fs::remove_file(&path);
}
#[test]
fn mic_is_summed_with_mix_when_present() {
let dir = std::env::temp_dir();
let mut r = Recorder {
writer: WavWriter::new(&dir.join(format!("ps-sum-{}.wav", std::process::id()))).unwrap(),
writer: WavWriter::new(&dir.join(format!("ps-sum-{}.wav", std::process::id())))
.unwrap(),
mic_fifo: VecDeque::new(),
path: PathBuf::new(),
};
@@ -223,7 +265,11 @@ mod tests {
r.write_frame(&[10, 20]).unwrap();
assert_eq!(r.mic_fifo.len(), 1, "two samples consumed, one mic left");
r.write_frame(&[0, 0]).unwrap();
assert_eq!(r.mic_fifo.len(), 0, "remaining mic sample consumed; rest is silence");
assert_eq!(
r.mic_fifo.len(),
0,
"remaining mic sample consumed; rest is silence"
);
let _ = r.finalize();
}
@@ -231,7 +277,8 @@ mod tests {
fn mic_fifo_is_capped() {
let dir = std::env::temp_dir();
let mut r = Recorder {
writer: WavWriter::new(&dir.join(format!("ps-cap-{}.wav", std::process::id()))).unwrap(),
writer: WavWriter::new(&dir.join(format!("ps-cap-{}.wav", std::process::id())))
.unwrap(),
mic_fifo: VecDeque::new(),
path: PathBuf::new(),
};
+4 -2
View File
@@ -26,7 +26,7 @@ use std::time::Duration;
use peerspeak::audio::AudioBackend;
use peerspeak::audio::pipewire_impl::PipeWireBackend;
use peerspeak::core::jitter::FRAME_SAMPLES; // 960 samples = 20ms @ 48kHz mono
use peerspeak::core::jitter::FRAME_SAMPLES; // 960 mono frames = 20ms @ 48kHz
const SAMPLE_RATE: f32 = 48_000.0;
@@ -69,11 +69,13 @@ async fn main() {
tokio::time::sleep(Duration::from_millis(2)).await;
continue;
}
let mut frame = Vec::with_capacity(FRAME_SAMPLES);
let mut frame = Vec::with_capacity(FRAME_SAMPLES * peerspeak::audio::PLAYBACK_CHANNELS);
for _ in 0..FRAME_SAMPLES {
let t = n as f32 / SAMPLE_RATE;
// 0.25 amplitude: clearly audible but not harsh.
let sample = (0.25 * i16::MAX as f32 * (2.0 * std::f32::consts::PI * freq * t).sin()) as i16;
// Stereo playback bus: duplicate the probe tone to L/R.
frame.push(sample);
frame.push(sample);
n += 1;
}
+27 -1
View File
@@ -1,6 +1,7 @@
use crate::notify::Sound;
use crate::theme::AppTheme;
use serde::{Deserialize, Serialize};
use std::collections::HashMap;
use std::fs;
use std::path::PathBuf;
@@ -232,6 +233,17 @@ pub struct AppConfig {
/// capped (see `recents`). Defaulted empty so older configs upgrade cleanly.
#[serde(default)]
pub recents: Vec<crate::recents::Recent>,
/// Per-peer listener-side EQ settings, keyed by peer node id string. Local
/// preference only; never sent to peers.
#[serde(default)]
pub peer_eq: HashMap<String, crate::audio::eq::EqSettings>,
/// Per-peer listener-side pan (`-1.0` left, `0.0` center, `1.0` right),
/// keyed by peer node id string. Local preference only.
#[serde(default)]
pub peer_pan: HashMap<String, f32>,
/// Focused app-local keyboard shortcuts.
#[serde(default)]
pub hotkeys: crate::hotkeys::HotkeyMap,
/// Last window size (px), restored as the initial size on next launch.
/// Saved on close.
#[serde(default = "default_window_width")]
@@ -287,6 +299,9 @@ impl Default for AppConfig {
sound_reconnect_failed_enabled: true,
pixelpass_path: None,
recents: Vec::new(),
peer_eq: HashMap::new(),
peer_pan: HashMap::new(),
hotkeys: crate::hotkeys::HotkeyMap::default(),
window_width: default_window_width(),
window_height: default_window_height(),
window_x: None,
@@ -411,6 +426,18 @@ mod tests {
assert_eq!(deserialized.window_height, 760.0);
// Configs predating the recents list load an empty list.
assert!(deserialized.recents.is_empty());
// Configs predating per-peer listener shaping load flat/center/default
// shortcut settings.
assert!(deserialized.peer_eq.is_empty());
assert!(deserialized.peer_pan.is_empty());
assert_eq!(
crate::hotkeys::format_binding(
deserialized
.hotkeys
.binding(crate::hotkeys::HotkeyAction::PushToTalk)
),
"Space"
);
}
#[test]
@@ -596,4 +623,3 @@ mod tests {
assert_eq!(config.noise_gate_threshold, 0.01);
}
}
+62 -3
View File
@@ -59,6 +59,10 @@ const PRIME_TIMEOUT_TICKS: usize = 25;
/// badly behind, so we drop the oldest and resync rather than grow unbounded.
const MAX_BUFFERED_FRAMES: usize = 32;
/// Sequence discontinuities larger than this (~10s at 20ms/frame) are treated
/// as a restarted/new stream, not ordinary packet loss or reordering.
const MAX_REASONABLE_SEQ_GAP: u32 = 500;
pub struct JitterBuffer {
decoder: OpusDecoder,
/// Reorder window: sequence number -> encoded Opus payload.
@@ -116,6 +120,14 @@ impl JitterBuffer {
}
}
fn reset_to_stream(&mut self, seq: u32, payload: Vec<u8>) {
self.packets.clear();
self.packets.insert(seq, payload);
self.next_seq = None;
self.clean_run = 0;
self.buffering_ticks = 0;
}
/// Store a received packet, dropping ones we've already played past and
/// bounding total depth.
pub fn insert(&mut self, seq: u32, payload: Vec<u8>) {
@@ -124,9 +136,19 @@ impl JitterBuffer {
if let Some(next) = self.next_seq
&& seq_before(seq, next)
{
if next.wrapping_sub(seq) > MAX_REASONABLE_SEQ_GAP {
self.reset_to_stream(seq, payload);
return;
}
self.note_disruption();
return;
}
if let Some(next) = self.next_seq
&& seq.wrapping_sub(next) > MAX_REASONABLE_SEQ_GAP
{
self.reset_to_stream(seq, payload);
return;
}
self.packets.insert(seq, payload);
while self.packets.len() > MAX_BUFFERED_FRAMES {
@@ -283,6 +305,44 @@ mod tests {
assert_eq!(jb.packets.len(), 1); // only seq 7 remains buffered
}
#[test]
fn far_behind_sequence_resets_as_restarted_stream() {
let mut jb = JitterBuffer::new().unwrap();
jb.next_seq = Some(5_000);
jb.packets.insert(5_000, vec![9]);
jb.clean_run = 12;
jb.buffering_ticks = 4;
jb.insert(0, vec![1]);
assert_eq!(jb.next_seq, None);
assert_eq!(jb.packets.len(), 1);
assert_eq!(jb.packets.get(&0).map(Vec::as_slice), Some(&[1][..]));
assert_eq!(jb.clean_run, 0);
assert_eq!(jb.buffering_ticks, 0);
}
#[test]
fn far_ahead_sequence_resets_to_bound_plc_run() {
let mut jb = JitterBuffer::new().unwrap();
jb.next_seq = Some(10);
jb.packets.insert(10, vec![9]);
jb.clean_run = 12;
jb.buffering_ticks = 4;
let jumped_seq = 10 + MAX_REASONABLE_SEQ_GAP + 1;
jb.insert(jumped_seq, vec![2]);
assert_eq!(jb.next_seq, None);
assert_eq!(jb.packets.len(), 1);
assert_eq!(
jb.packets.get(&jumped_seq).map(Vec::as_slice),
Some(&[2][..])
);
assert_eq!(jb.clean_run, 0);
assert_eq!(jb.buffering_ticks, 0);
}
#[test]
fn test_seq_before_ordering() {
// Basic ordering
@@ -358,7 +418,7 @@ mod tests {
fn is_idle_reflects_buffer_state() {
let mut enc = OpusEncoder::new(48000, Channels::Mono, Application::Voip).unwrap();
let mut jb = JitterBuffer::new().unwrap();
// Fresh buffer
assert!(jb.is_idle());
@@ -369,7 +429,7 @@ mod tests {
// Prime (3 frames)
jb.insert(1, frame(&mut enc, 1000));
jb.insert(2, frame(&mut enc, 1000));
// Drain past the end so it underruns
assert!(jb.pop_frame().is_some());
assert!(jb.pop_frame().is_some());
@@ -635,4 +695,3 @@ mod tests {
assert_eq!(jb.clean_run, 0);
}
}
+10
View File
@@ -11,6 +11,10 @@ pub enum CoreCommand {
/// for a NEW room; it's ignored when joining (the label rides in the ticket).
Join { name: String, ticket: String, room_name: String, input_device: Option<String>, output_device: Option<String>, echo_cancellation: bool, avatar: crate::avatar::Avatar },
Leave,
/// Orderly app shutdown: finalize recordings, leave any active room, stop local
/// audio/screen-share work, close the persistent network stack, then ack with
/// [`UiEvent::ShutdownComplete`].
Shutdown,
ToggleMute,
/// Change our avatar (W4) and re-announce it to the room over presence.
SetAvatar(crate::avatar::Avatar),
@@ -18,6 +22,10 @@ pub enum CoreCommand {
SetPttMode(bool),
SetPttActive(bool),
SetPeerVolume(EndpointId, f32),
/// Listener-side per-peer EQ. Local only; never leaves this app instance.
SetPeerEq(EndpointId, crate::audio::eq::EqSettings),
/// Listener-side per-peer pan. Local only; never leaves this app instance.
SetPeerPan(EndpointId, f32),
/// Locally mute/unmute a peer: when muted, their audio is decoded (so levels
/// still show) but not mixed into our output.
SetPeerMuted(EndpointId, bool),
@@ -122,5 +130,7 @@ pub enum UiEvent {
/// Discoverable. Distinct from a user-driven change so the GUI knows to update
/// without having issued the command itself.
PresenceModeReverted { mode: PresenceMode },
/// Core finished orderly app shutdown and the GUI can exit.
ShutdownComplete,
Error(String),
}
+152 -8
View File
@@ -2,6 +2,7 @@ pub mod messages;
pub mod jitter;
use crate::audio::{AudioBackend, pipewire_impl::PipeWireBackend};
use crate::audio::eq::{Eq, EqSettings};
use crate::codec::{AudioEncoder, opus_impl::OpusEncoder};
use crate::core::jitter::{JitterBuffer, FRAME_SAMPLES};
use crate::network::{
@@ -49,6 +50,12 @@ impl CoreController {
pub fn send(&self, cmd: CoreCommand) -> bool {
self.cmd_tx.try_send(cmd).is_ok()
}
/// Clone the command sender for asynchronous one-shot sends that should wait
/// for channel capacity instead of failing immediately on a full queue.
pub fn command_sender(&self) -> mpsc::Sender<CoreCommand> {
self.cmd_tx.clone()
}
}
/// How long a peer may stay "reconnecting" after a transient drop before we give
@@ -214,6 +221,7 @@ fn stop_mic_monitor(backend: &PipeWireBackend, monitor: Option<MicMonitor>) {
/// limiter (see [`crate::audio::limiter`]) can ride it down to the ceiling instead
/// of the old hard clip shattering loud moments. Peers shorter than `frame_len`
/// contribute 0 past their end; an empty peer set yields a silent bus.
#[cfg(test)]
fn mix_frames(peer_frames: &[Vec<i16>], frame_len: usize) -> Vec<i32> {
let mut mixed = vec![0i32; frame_len];
for frame in peer_frames {
@@ -224,6 +232,44 @@ fn mix_frames(peer_frames: &[Vec<i16>], frame_len: usize) -> Vec<i32> {
mixed
}
/// Sum per-peer mono frames into one interleaved stereo `i32` bus. Center pan is
/// a special exact dual-mono path so the default listener mix is bit-for-bit the
/// old mono sum duplicated to both ears.
fn mix_stereo_frames(peer_frames: &[(Vec<i16>, f32)], frame_len: usize) -> Vec<i32> {
let mut mixed = vec![0i32; frame_len * crate::audio::PLAYBACK_CHANNELS];
for (frame, pan) in peer_frames {
if pan.abs() <= f32::EPSILON {
for (i, &sample) in frame.iter().take(frame_len).enumerate() {
let idx = i * crate::audio::PLAYBACK_CHANNELS;
let s = sample as i32;
mixed[idx] += s;
mixed[idx + 1] += s;
}
continue;
}
let (left_gain, right_gain) = crate::audio::pan::playback_pan_gains(*pan);
for (i, &sample) in frame.iter().take(frame_len).enumerate() {
let idx = i * crate::audio::PLAYBACK_CHANNELS;
let x = sample as f32;
mixed[idx] += (x * left_gain).round() as i32;
mixed[idx + 1] += (x * right_gain).round() as i32;
}
}
mixed
}
/// Fold an interleaved stereo frame to mono for the existing mixed WAV writers.
/// Center/default pan folds back to the exact old mono mix.
fn stereo_to_mono(stereo: &[i16]) -> Vec<i16> {
let mut mono = Vec::with_capacity(stereo.len() / crate::audio::PLAYBACK_CHANNELS);
for pair in stereo.chunks_exact(crate::audio::PLAYBACK_CHANNELS) {
let sum = pair[0] as i32 + pair[1] as i32;
mono.push((sum / 2).clamp(i16::MIN as i32, i16::MAX as i32) as i16);
}
mono
}
/// Handles the transport's per-peer link-state stream (`ConnEvent`): arms/cancels
/// reconnect grace timers, tracks which peers we've linked with, and forwards
/// link state to the UI. Pulled out of the conn-event task as a unit so the
@@ -665,6 +711,8 @@ async fn run_core_loop(
let is_multitrack = Arc::new(AtomicBool::new(false));
let mut recording_mode = RecordingMode::default();
let peer_volumes = Arc::new(Mutex::new(HashMap::<EndpointId, f32>::new()));
let peer_eq = Arc::new(Mutex::new(HashMap::<EndpointId, EqSettings>::new()));
let peer_pan = Arc::new(Mutex::new(HashMap::<EndpointId, f32>::new()));
// Peers locally muted by us: decoded for level metering but not mixed.
let locally_muted = Arc::new(Mutex::new(HashSet::<EndpointId>::new()));
let mut current_name = "Anonymous".to_string();
@@ -814,6 +862,23 @@ async fn run_core_loop(
}
};
match cmd {
CoreCommand::Shutdown => {
crate::log_msg("Core shutdown requested");
// Finalize recordings while capture/mixer feeders are still alive.
stop_recording(&recorder, &is_recording, &multitrack, &is_multitrack, &ui_tx).await;
stop_mic_monitor(&audio_backend, mic_monitor.take());
if let Some(session) = active_session.take() {
session.shutdown(audio_backend.clone()).await;
net.audio_router.clear();
}
*current_room.lock().unwrap() = None;
net.shutdown().await;
let _ = ui_tx.send(UiEvent::ShutdownComplete).await;
break;
}
CoreCommand::Join { name, ticket, room_name, input_device, output_device, echo_cancellation, avatar } => {
current_name = name.clone();
current_avatar = avatar;
@@ -1104,6 +1169,8 @@ async fn run_core_loop(
let jitter_mixer = jitter.clone();
let is_deafened_clone = is_deafened.clone();
let peer_volumes_mixer = peer_volumes.clone();
let peer_eq_mixer = peer_eq.clone();
let peer_pan_mixer = peer_pan.clone();
let locally_muted_mixer = locally_muted.clone();
let output_gain_mixer = output_gain.clone();
let ui_tx_mixer = ui_tx.clone();
@@ -1117,6 +1184,9 @@ async fn run_core_loop(
// the ceiling instead of hard-clipping. State carries across
// frames (see audio::limiter).
let mut limiter = crate::audio::limiter::SoftLimiter::new(48_000);
// Per-peer EQ filter state. Settings are live-cloned each
// cycle; state is rebuilt only when a peer's EQ changes.
let mut peer_eqs: HashMap<EndpointId, Eq> = HashMap::new();
// When the ring is at/above target we have nothing to do; nap
// briefly and re-check. Short enough (relative to the ~60ms
// target and ~21ms device quantum) that we always refill well
@@ -1140,8 +1210,11 @@ async fn run_core_loop(
}
let current_volumes = peer_volumes_mixer.lock().await.clone();
let current_eq = peer_eq_mixer.lock().await.clone();
let current_pans = peer_pan_mixer.lock().await.clone();
let muted_peers = locally_muted_mixer.lock().await.clone();
let mut peer_frames = Vec::new();
let mut peer_frames: Vec<(Vec<i16>, f32)> = Vec::new();
let mut peers_seen = HashSet::new();
// Multitrack stem capture: tap each peer's RAW decoded frame
// (pre-volume, pre-mute, pre-limiter) so the stems are clean
@@ -1167,10 +1240,31 @@ async fn run_core_loop(
let vol = current_volumes.get(&peer_id).copied().unwrap_or(1.0);
apply_volume(&mut frame, vol);
let eq_settings = current_eq
.get(&peer_id)
.copied()
.unwrap_or_default()
.clamped();
if eq_settings.is_flat() {
peer_eqs.remove(&peer_id);
} else {
let needs_rebuild = peer_eqs
.get(&peer_id)
.map(|eq| eq.settings() != eq_settings)
.unwrap_or(true);
if needs_rebuild {
peer_eqs.insert(peer_id, Eq::new(eq_settings));
}
if let Some(eq) = peer_eqs.get_mut(&peer_id) {
eq.process_frame(&mut frame);
}
}
// Level is recorded even for locally-muted peers so
// the UI still shows that they're speaking.
let peak = level_peaks.entry(peer_id).or_insert(0.0);
*peak = peak.max(frame_level(&frame));
peers_seen.insert(peer_id);
// Locally muted: decoded above (jitter buffer advances,
// level shown) but not mixed into our output.
@@ -1178,16 +1272,23 @@ async fn run_core_loop(
continue;
}
peer_frames.push(frame);
let pan = current_pans
.get(&peer_id)
.copied()
.unwrap_or(0.0)
.clamp(-1.0, 1.0);
peer_frames.push((frame, pan));
}
}
peer_eqs.retain(|id, _| peers_seen.contains(id) || current_eq.contains_key(id));
// Lossless i32 sum, then the limiter applies the master
// output gain (in f32, so a boost past the ceiling is
// limited too) and rides peaks down to the ceiling.
let mixed_sum = mix_frames(&peer_frames, FRAME_SAMPLES);
let mixed_sum = mix_stereo_frames(&peer_frames, FRAME_SAMPLES);
let out_gain = f32::from_bits(output_gain_mixer.load(Ordering::Relaxed));
let mixed = limiter.process(&mixed_sum, out_gain);
let record_mix = stereo_to_mono(&mixed);
// Record the true call audio, independent of local deafen —
// deafen only silences our own monitor, not what the call
@@ -1200,7 +1301,7 @@ async fn run_core_loop(
for (id, f) in &stems {
mt.write_peer(*id, f)?;
}
mt.write_mix(&mixed)?;
mt.write_mix(&record_mix)?;
mt.end_cycle()
})();
if let Err(e) = res {
@@ -1209,13 +1310,13 @@ async fn run_core_loop(
}
} else if is_recording_mixer.load(Ordering::Relaxed)
&& let Some(rec) = recorder_mixer.lock().unwrap().as_mut()
&& let Err(e) = rec.write_frame(&mixed)
&& let Err(e) = rec.write_frame(&record_mix)
{
crate::log_msg(&format!("Recording write failed: {e}"));
}
let frame_to_send = if is_deafened_clone.load(Ordering::Relaxed) {
vec![0i16; FRAME_SAMPLES]
vec![0i16; mixed.len()]
} else {
mixed
};
@@ -1522,6 +1623,26 @@ async fn run_core_loop(
guard.insert(peer_id, vol);
}
CoreCommand::SetPeerEq(peer_id, settings) => {
let settings = settings.clamped();
let mut guard = peer_eq.lock().await;
if settings.is_flat() {
guard.remove(&peer_id);
} else {
guard.insert(peer_id, settings);
}
}
CoreCommand::SetPeerPan(peer_id, pan) => {
let pan = pan.clamp(-1.0, 1.0);
let mut guard = peer_pan.lock().await;
if pan.abs() <= 0.001 {
guard.remove(&peer_id);
} else {
guard.insert(peer_id, pan);
}
}
CoreCommand::SetPeerMuted(peer_id, muted) => {
let mut guard = locally_muted.lock().await;
if muted {
@@ -1865,7 +1986,10 @@ async fn run_core_loop(
#[cfg(test)]
mod tests {
use super::{apply_volume, frame_level, mix_frames, MicLevelMeter, MIC_LEVEL_REPORT_SAMPLES};
use super::{
apply_volume, frame_level, mix_frames, mix_stereo_frames, stereo_to_mono, MicLevelMeter,
MIC_LEVEL_REPORT_SAMPLES,
};
/// A frame of constant amplitude with the given sample count.
fn frame(amp: i16, len: usize) -> Vec<i16> {
@@ -1924,6 +2048,27 @@ mod tests {
assert_eq!(mixed, vec![100i32, -200, 300, -400]);
}
#[test]
fn centered_stereo_mix_is_exact_dual_mono() {
let a = vec![100, -200, 300, -400];
let b = vec![50, 200, -100, 400];
let mixed = mix_stereo_frames(&[(a, 0.0), (b, 0.0)], 4);
assert_eq!(mixed, vec![150, 150, 0, 0, 200, 200, 0, 0]);
}
#[test]
fn hard_left_pan_only_contributes_left_channel() {
let frame = vec![100, 200];
let mixed = mix_stereo_frames(&[(frame, -1.0)], 2);
assert_eq!(mixed, vec![141, 0, 283, 0]);
}
#[test]
fn stereo_fold_down_averages_pairs() {
let mono = stereo_to_mono(&[100, 100, 200, 0, i16::MAX, i16::MAX]);
assert_eq!(mono, vec![100, 100, i16::MAX]);
}
#[test]
fn two_peers_sum_sample_by_sample() {
let a = vec![100, -200, 300, -400];
@@ -2038,4 +2183,3 @@ mod tests {
assert!((level - 0.5).abs() < 1e-3, "mid-range level was {level}");
}
}
+284
View File
@@ -0,0 +1,284 @@
//! Focused, app-local keyboard shortcuts.
//!
//! These helpers are intentionally pure: key serialization, formatting, lookup,
//! and conflict detection live here, while iced event handling stays at the app
//! edge. There are no OS-global shortcuts.
use iced::keyboard;
use serde::{Deserialize, Serialize};
/// A serializable key identity. Modifiers are deliberately out of scope for this
/// first pass; iced delivers the focused app key and we compare that exact key.
#[derive(Debug, Clone, PartialEq, Eq, Hash, Serialize, Deserialize)]
pub enum KeyBinding {
Named(String),
Character(String),
}
impl KeyBinding {
pub fn from_key(key: &keyboard::Key) -> Option<Self> {
match key {
keyboard::Key::Named(named) => Some(Self::Named(format!("{named:?}"))),
keyboard::Key::Character(ch) => {
let s = ch.to_string();
if s.is_empty() {
None
} else {
Some(Self::Character(s.to_lowercase()))
}
}
keyboard::Key::Unidentified => None,
}
}
pub fn label(&self) -> String {
match self {
KeyBinding::Named(name) => name.clone(),
KeyBinding::Character(ch) => ch.to_uppercase(),
}
}
}
/// Parse a hand-editable binding string from config/docs/tests. Empty and
/// `"unset"` are unbound.
pub fn parse_binding(input: &str) -> Option<KeyBinding> {
let trimmed = input.trim();
if trimmed.is_empty() || trimmed.eq_ignore_ascii_case("unset") {
return None;
}
if trimmed.chars().count() == 1 {
Some(KeyBinding::Character(trimmed.to_lowercase()))
} else {
Some(KeyBinding::Named(trimmed.to_string()))
}
}
pub fn format_binding(binding: Option<&KeyBinding>) -> String {
binding
.map(KeyBinding::label)
.unwrap_or_else(|| "unset".to_string())
}
#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, Serialize, Deserialize)]
pub enum HotkeyAction {
ToggleMute,
ToggleDeafen,
OpenSettings,
PushToTalk,
LeaveRoom,
}
impl HotkeyAction {
pub const ALL: [HotkeyAction; 5] = [
HotkeyAction::ToggleMute,
HotkeyAction::ToggleDeafen,
HotkeyAction::OpenSettings,
HotkeyAction::PushToTalk,
HotkeyAction::LeaveRoom,
];
pub fn label(self) -> &'static str {
match self {
HotkeyAction::ToggleMute => "Toggle mute",
HotkeyAction::ToggleDeafen => "Toggle deafen",
HotkeyAction::OpenSettings => "Open Settings",
HotkeyAction::PushToTalk => "Push-to-talk",
HotkeyAction::LeaveRoom => "Leave room",
}
}
pub fn tier(self) -> HotkeyTier {
match self {
HotkeyAction::ToggleMute
| HotkeyAction::ToggleDeafen
| HotkeyAction::OpenSettings => HotkeyTier::AppWide,
HotkeyAction::PushToTalk | HotkeyAction::LeaveRoom => HotkeyTier::RoomOnly,
}
}
}
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum HotkeyTier {
AppWide,
RoomOnly,
}
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub struct HotkeyContext {
pub in_call: bool,
}
impl HotkeyContext {
fn allows(self, action: HotkeyAction) -> bool {
matches!(action.tier(), HotkeyTier::AppWide) || self.in_call
}
}
/// Persisted shortcut map. Defaults preserve the old Space push-to-talk binding
/// and add a few function-key app shortcuts that do not collide with typing.
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
pub struct HotkeyMap {
#[serde(default = "default_mute")]
pub toggle_mute: Option<KeyBinding>,
#[serde(default = "default_deafen")]
pub toggle_deafen: Option<KeyBinding>,
#[serde(default = "default_settings")]
pub open_settings: Option<KeyBinding>,
#[serde(default = "default_ptt")]
pub push_to_talk: Option<KeyBinding>,
#[serde(default)]
pub leave_room: Option<KeyBinding>,
}
impl Default for HotkeyMap {
fn default() -> Self {
Self {
toggle_mute: default_mute(),
toggle_deafen: default_deafen(),
open_settings: default_settings(),
push_to_talk: default_ptt(),
leave_room: None,
}
}
}
fn named(name: &str) -> Option<KeyBinding> {
Some(KeyBinding::Named(name.to_string()))
}
fn default_mute() -> Option<KeyBinding> {
named("F9")
}
fn default_deafen() -> Option<KeyBinding> {
named("F10")
}
fn default_settings() -> Option<KeyBinding> {
named("F2")
}
fn default_ptt() -> Option<KeyBinding> {
named("Space")
}
impl HotkeyMap {
pub fn binding(&self, action: HotkeyAction) -> Option<&KeyBinding> {
match action {
HotkeyAction::ToggleMute => self.toggle_mute.as_ref(),
HotkeyAction::ToggleDeafen => self.toggle_deafen.as_ref(),
HotkeyAction::OpenSettings => self.open_settings.as_ref(),
HotkeyAction::PushToTalk => self.push_to_talk.as_ref(),
HotkeyAction::LeaveRoom => self.leave_room.as_ref(),
}
}
pub fn set_binding(&mut self, action: HotkeyAction, binding: Option<KeyBinding>) {
match action {
HotkeyAction::ToggleMute => self.toggle_mute = binding,
HotkeyAction::ToggleDeafen => self.toggle_deafen = binding,
HotkeyAction::OpenSettings => self.open_settings = binding,
HotkeyAction::PushToTalk => self.push_to_talk = binding,
HotkeyAction::LeaveRoom => self.leave_room = binding,
}
}
pub fn lookup_key(&self, key: &keyboard::Key, context: HotkeyContext) -> Option<HotkeyAction> {
let pressed = KeyBinding::from_key(key)?;
HotkeyAction::ALL
.into_iter()
.find(|&action| context.allows(action) && self.binding(action) == Some(&pressed))
}
pub fn lookup_binding(
&self,
binding: &KeyBinding,
context: HotkeyContext,
) -> Option<HotkeyAction> {
HotkeyAction::ALL
.into_iter()
.find(|&action| context.allows(action) && self.binding(action) == Some(binding))
}
pub fn conflicts(&self) -> Vec<HotkeyConflict> {
let mut conflicts = Vec::new();
let actions = HotkeyAction::ALL;
for i in 0..actions.len() {
for j in (i + 1)..actions.len() {
let a = actions[i];
let b = actions[j];
if let (Some(ab), Some(bb)) = (self.binding(a), self.binding(b))
&& ab == bb
{
conflicts.push(HotkeyConflict {
binding: ab.clone(),
first: a,
second: b,
});
}
}
}
conflicts
}
}
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct HotkeyConflict {
pub binding: KeyBinding,
pub first: HotkeyAction,
pub second: HotkeyAction,
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn unset_actions_format_as_unset() {
assert_eq!(format_binding(None), "unset");
assert_eq!(parse_binding("unset"), None);
assert_eq!(parse_binding(""), None);
}
#[test]
fn duplicate_binding_is_detected() {
let mut map = HotkeyMap::default();
map.set_binding(HotkeyAction::ToggleMute, parse_binding("M"));
map.set_binding(HotkeyAction::ToggleDeafen, parse_binding("m"));
let conflicts = map.conflicts();
assert_eq!(conflicts.len(), 1);
assert_eq!(conflicts[0].first, HotkeyAction::ToggleMute);
assert_eq!(conflicts[0].second, HotkeyAction::ToggleDeafen);
}
#[test]
fn lookup_respects_room_tier() {
let mut map = HotkeyMap::default();
map.set_binding(HotkeyAction::LeaveRoom, parse_binding("Escape"));
let binding = parse_binding("Escape").unwrap();
assert_eq!(
map.lookup_binding(&binding, HotkeyContext { in_call: false }),
None,
"room-only shortcuts should not fire outside a call"
);
assert_eq!(
map.lookup_binding(&binding, HotkeyContext { in_call: true }),
Some(HotkeyAction::LeaveRoom)
);
}
#[test]
fn default_ptt_is_space() {
let map = HotkeyMap::default();
assert_eq!(
format_binding(map.binding(HotkeyAction::PushToTalk)),
"Space"
);
}
#[test]
fn parse_single_character_case_folds() {
assert_eq!(parse_binding("M"), Some(KeyBinding::Character("m".to_string())));
assert_eq!(format_binding(parse_binding("m").as_ref()), "M");
}
}
+1 -1
View File
@@ -16,6 +16,7 @@ pub mod sanitize;
pub mod avatar;
pub mod recents;
pub mod discovery;
pub mod hotkeys;
use std::path::PathBuf;
use std::sync::OnceLock;
@@ -60,4 +61,3 @@ pub fn log_msg(msg: &str) {
let _ = file.write_all(line.as_bytes());
}
}
+81
View File
@@ -0,0 +1,81 @@
# Codex task report - 2026-06-16
## W2 - Per-peer EQ
- Added `src/audio/eq.rs`: a 3-band listener-side RBJ biquad EQ (low shelf, mid peaking, high shelf) with per-peer state and flat bypass.
- Added local config persistence in `AppConfig.peer_eq`, keyed by peer node id string.
- Added local `CoreCommand::SetPeerEq` and mixer-side per-peer `Eq` state. EQ is applied after local volume and before pan/mix; raw multitrack stems remain pre-volume/pre-EQ.
- Added participant-card controls for Low/Mid/High gain sliders (-12 dB to +12 dB). Changes apply live and persist on slider release.
- Tests added for flat identity, low/high boost energy, coefficient finiteness, clamping, and hot-signal processing.
Unverified: subjective voice quality and zipper/noise behavior on real devices.
## W1 - Per-listener pan / stereo playback
- Added `src/audio/pan.rs`: constant-power `pan_gains()` with tests, plus playback gains that preserve the legacy default dual-mono center.
- Converted playback mix to interleaved stereo in `src/core/mod.rs`.
- Switched PipeWire playback output to 2-channel S16LE and adjusted ring target/capacity/stride accounting in `src/audio/pipewire_impl.rs`.
- Kept capture, Opus encode/decode, jitter buffers, and network audio mono.
- Limiter now receives the interleaved stereo bus; shared limiter gain ducks both channels consistently.
- Mixed WAV and multitrack convenience mix fold the listener stereo mix back to mono before writing. Per-peer stems remain raw mono.
- Updated `audio_probe` to send dual-mono stereo frames.
- Added tests for exact center dual-mono behavior, hard-left pan contribution, and stereo fold-down.
Decision for senior sanity-check: pure pan law is constant-power, but playback scales it by sqrt(2) so pan=0 is exactly the old mono signal in both ears. This satisfies the "default behavior unchanged" guardrail at the cost of louder hard-panned extremes, which the existing limiter catches.
Unverified: real PipeWire stereo playback, underrun behavior on actual hardware, and recorded WAV listening checks.
## W5 - Focused hotkeys + info popup
- Added `src/hotkeys.rs`: serializable `KeyBinding`, `HotkeyAction`, `HotkeyMap`, parse/format/lookup, tier checks, and duplicate conflict detection.
- Added `AppConfig.hotkeys` with defaults: F9 mute, F10 deafen, F2 Settings, Space push-to-talk, Leave unset.
- Replaced the hard-coded PTT key capture with config-backed binding capture.
- Added Settings hotkey editor with Set/Clear per action and live conflict warnings.
- Added top-right hotkey info popup that lists every action and current binding, showing `unset` for unbound actions.
- Routed focused iced key events through the map. App-wide actions can fire from any screen while focused; room-only actions require an active call. PTT press/release still uses `SetPttActive`.
- Tests added for unset formatting, duplicate detection, room-tier lookup, defaults, and character parse/format.
Unverified: manual keyboard interaction in the GUI. No OS-global hooks were added.
## W3 - PipeWire pro-routing plan (not implemented)
I stopped at design for W3. The current backend already supports simple target-node routing through PipeWire stream property `node.target`, but true "pro routing" (explicit ports / manual graph links / no-autoconnect patching) would require backend changes that are not safely verifiable offline.
Proposed future scope:
- Expose two advanced route targets: capture source node and playback sink node, with optional future per-port routing.
- Enumerate available nodes with the existing `pw-cli list-objects Node` parser. For port-level routing, add a separate parser for `pw-cli list-objects Port` collecting `object.id`, `node.id`, `port.name`, direction, and channel position.
- For node-level routing, continue using PipeWire stream property `node.target` on stream creation. This is the low-risk path and matches current backend behavior.
- For explicit port routing, do not use `AUTOCONNECT`; instead capture the created PeerSpeak stream node/port ids from the PipeWire registry, then link with PipeWire-native APIs or `pw-link <source-port-id> <sink-port-id>`. Degrade by falling back to `node.target` autoconnect if any selected node/port is missing.
- Offline tests should cover pure routing-plan decisions: selected node exists/missing, selected port exists/missing, capture/playback direction mismatch, and fallback choice. Real-device tests still need a PipeWire graph.
Reason for not implementing: the current `run_playback` / `run_capture` code does not retain stream node or port ids, and changing `AUTOCONNECT` behavior plus adding manual `pw-link` calls could destabilize the working audio path. That matches the assignment's "bail if risky" instruction.
## Backlog A21/A22 - correctness fixes
- Fixed A21 in `src/core/jitter.rs`: implausibly large sequence discontinuities now reset the per-peer jitter stream instead of being treated as ordinary late packets or packet loss.
- The reset threshold is `500` frames, about 10 seconds at 20 ms/frame. That covers both same-identity sender restart back to sequence 0 and a faulty/malicious jump far ahead that would otherwise force a long PLC run.
- Added jitter regression tests for both far-behind restart and far-ahead jump cases.
- Fixed A22 in `src/audio/recorder.rs`: `WavWriter` now tracks data bytes as `u64`, checks additions before writing, and rejects data that cannot fit both the RIFF size field and the `data` chunk size field.
- Added a WAV overflow regression test that exercises the limit without creating a huge file.
Unverified: the same-identity peer restart has not been exercised in a live 2-machine call; the WAV fix is counter/size-field tested, not a real >12h recording.
## Backlog A14 - orderly window-close shutdown
- Added `CoreCommand::Shutdown` and `UiEvent::ShutdownComplete`.
- Window close now saves config, marks the GUI as closing, asynchronously queues `Shutdown`, and exits only after the core acknowledges completion or after a 5-second fallback timeout.
- Core shutdown finalizes active mixed/multitrack recordings before session teardown, stops the standalone mic monitor, runs `ActiveSession::shutdown()` for active calls, clears room presence/routing, closes the persistent network stack, sends `ShutdownComplete`, and ends the core loop.
- The shutdown command is queued with an awaited `mpsc::Sender::send` task instead of the best-effort `try_send`, so a full command queue does not immediately drop the close command.
Unverified: actual GUI window-close behavior during a live call/recording still needs a manual run; tests/builds only prove the path compiles and existing unit coverage still passes.
## Verification
- `cargo check` passed.
- `cargo test --lib` passed: 288 passed, 0 failed, 2 ignored.
- `cargo clippy --all-targets` passed.
- `cargo build --release` passed.
- Formatted the touched Rust files with `rustfmt --edition 2024`; I did not run repo-wide `cargo fmt` to avoid unrelated formatting churn.
No new dependencies were added. Runtime/manual/field verification is still pending for audio-device and 2-machine behavior.