From 76004429c4594a2dca2633413ee197ea91b70429 Mon Sep 17 00:00:00 2001
From: Miriam Zimmerman <mutexlox@signal.org>
Date: Tue, 22 Oct 2024 11:15:33 -0400
Subject: [PATCH] Fix stereo -> mono downmixing.

Rather than summing the two channels, average them. This prevents
integer overflow and clipping or needlessly increasing volume.
---
 Cargo.toml                    |  1 +
 src/backend/buffer_manager.rs | 31 ++++++++++++++++++++++++++++---
 src/lib.rs                    |  2 ++
 3 files changed, 31 insertions(+), 3 deletions(-)

diff --git a/Cargo.toml b/Cargo.toml
index 95ebb14f..7c7ca582 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -15,6 +15,7 @@ cubeb-backend = "0.13"
 float-cmp = "0.6"
 libc = "0.2"
 mach = "0.3"
+num = "0.4.3"
 audio-mixer = "0.2"
 ringbuf = "0.2.6"
 triple_buffer = "5.0.5"
diff --git a/src/backend/buffer_manager.rs b/src/backend/buffer_manager.rs
index 6f1c299b..4e76d4bd 100644
--- a/src/backend/buffer_manager.rs
+++ b/src/backend/buffer_manager.rs
@@ -4,6 +4,7 @@ use std::os::raw::c_void;
 use std::slice;
 
 use cubeb_backend::SampleFormat;
+use num::cast::AsPrimitive;
 
 use super::ringbuf::RingBuffer;
 
@@ -36,12 +37,25 @@ fn drop_first_n_channels_in_place<T: Copy>(
     }
 }
 
+trait DataType: AsPrimitive<f32> {
+    fn from_f32(v: f32) -> Self;
+}
+
+impl<T: AsPrimitive<f32>> DataType for T
+where
+    f32: AsPrimitive<T>,
+{
+    fn from_f32(v: f32) -> T {
+        v.as_()
+    }
+}
+
 // It can be that the a stereo microphone is in use, but the user asked for mono input. In this
 // particular case, downmix the stereo pair into a mono channel. In all other cases, simply drop
 // the remaining channels before appending to the ringbuffer, becauses there is no right or wrong
 // way to do this, unlike with the output side, where proper channel matrixing can be done.
 // Return the number of valid samples in the buffer.
-fn remix_or_drop_channels<T: Copy + std::ops::Add<Output = T>>(
+fn remix_or_drop_channels<T: DataType>(
     input_channels: usize,
     output_channels: usize,
     data: &mut [T],
@@ -56,7 +70,8 @@ fn remix_or_drop_channels<T: Copy + std::ops::Add<Output = T>>(
     if input_channels == 2 && output_channels == 1 {
         let mut read_idx = 0;
         for (write_idx, _) in (0..frame_count).enumerate() {
-            data[write_idx] = data[read_idx] + data[read_idx + 1];
+            let avg = (data[read_idx].as_() + data[read_idx + 1].as_()) / 2.0;
+            data[write_idx] = DataType::from_f32(avg);
             read_idx += 2;
         }
         return output_channels * frame_count;
@@ -76,7 +91,7 @@ fn remix_or_drop_channels<T: Copy + std::ops::Add<Output = T>>(
     output_channels * frame_count
 }
 
-fn process_data<T: Copy + std::ops::Add<Output = T>>(
+fn process_data<T: DataType>(
     data: *mut c_void,
     frame_count: usize,
     input_channel_count: usize,
@@ -353,3 +368,13 @@ impl fmt::Debug for BufferManager {
         Ok(())
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    #[test]
+    fn remix_stereo_ints() {
+        let mut data = [i16::MAX / 2 + 1, i16::MAX / 2 + 1];
+        assert_eq!(remix_or_drop_channels(2, 1, &mut data, 1), 1);
+    }
+}
diff --git a/src/lib.rs b/src/lib.rs
index bf1ae96f..8eda3dd2 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -12,6 +12,8 @@ extern crate cubeb_backend;
 extern crate float_cmp;
 extern crate mach;
 
+extern crate num;
+
 mod backend;
 mod capi;