Use the number of channels of the stream, not the device, for draining computation.

This is right after the callback, that has the number of channels the
user requested, and can be different from the number of channels of the
device.

1 files changed, 3 insertions, 2 deletions

diff --git a/src/cubeb_audiounit.cpp b/src/cubeb_audiounit.cpp
index bdd1db0..e0c8fc6 100644
--- a/src/cubeb_audiounit.cpp
+++ b/src/cubeb_audiounit.cpp
@@ -693,10 +693,11 @@ audiounit_output_callback(void * user_ptr,
   /* Post process output samples. */
   if (stm->draining) {
     /* Clear missing frames (silence) */
-    size_t missing_samples = (output_frames - outframes) * stm->output_desc.mChannelsPerFrame;
+    size_t channels = stm->output_stream_params.channels;
+    size_t missing_samples = (output_frames - outframes) * channels;
     size_t size_sample = cubeb_sample_size(stm->output_stream_params.format);
     /* number of bytes that have been filled with valid audio by the callback. */
-    size_t audio_byte_count = outframes * stm->output_desc.mChannelsPerFrame * size_sample;
+    size_t audio_byte_count = outframes * channels * size_sample;
     PodZero((uint8_t*)output_buffer + audio_byte_count,
             missing_samples * size_sample);
   }