No need to time sending the blocks, the response message will be delayed if the buffer on the device is full.

This commit is contained in:
Frans-Willem Hardijzer committed 2023-12-01 14:48:33 +01:00
1 parent ce4e4f2f5d
commit afabd05cef
1 file changed
+7 -8
+7 -8
View File
@@ -269,6 +269,7 @@ impl BcCamera {
let block_size = talk_config.audio_config.length_per_encoder / 2;
let sample_rate = talk_config.audio_config.sample_rate;
let block_duration = talk_config.audio_config.length_per_encoder as f32 / talk_config.audio_config.sample_rate as f32;
let msg = Bc {
meta: BcMeta {
@@ -334,7 +335,6 @@ impl BcCamera {
let target_chunks = full_block_size as usize * BLOCK_PER_PAYLOAD;
let mut end_of_stream = false;
let mut message_end = std::time::Instant::now();
while !end_of_stream {
let mut payload_bytes = vec![];
while payload_bytes.len() < target_chunks {
@@ -396,18 +396,17 @@ impl BcCamera {
}),
};
// Wait until 100ms before message end
let sleep_until = message_end - std::time::Duration::from_secs_f32(0.1);
std::thread::sleep(sleep_until - std::time::Instant::now());
sub.send(msg).await?;
// TODO: What to do against drift ?
message_end += std::time::Duration::from_secs_f32(play_length);
let _ = sub.recv().await?;
}
// Wait for the full message to play + 100ms to be sure, before issuing talk stop.
std::thread::sleep((message_end - std::time::Instant::now()) + std::time::Duration::from_secs_f32(0.1));
// By observation in the Reolink Doorbell POE the buffer seems to be roughly 10 chunks,
// Wait until those buffers are fully flushed out to the audio device before issuing a
// talk_stop
let buffer_duration = block_duration * (BLOCK_PER_PAYLOAD as f32) * 10.0f32;
std::thread::sleep(std::time::Duration::from_secs_f32(buffer_duration));
log::info!("Block: {}s, Buffer: {}s", block_duration, buffer_duration);
self.talk_stop().await?;