use tauri events instead of polling to fix race condition in recording stop

daw-backend/src/audio/engine.rs

@@ -762,62 +762,59 @@ impl Engine {
 
     /// Handle stopping a recording
     fn handle_stop_recording(&mut self) {
+        eprintln!("[STOP_RECORDING] handle_stop_recording called");
         if let Some(recording) = self.recording_state.take() {
             let clip_id = recording.clip_id;
             let track_id = recording.track_id;
+            let sample_rate = recording.sample_rate;
+            let channels = recording.channels;
 
-            // Finalize the recording and get temp file path
+            eprintln!("[STOP_RECORDING] Stopping recording for clip_id={}, track_id={}", clip_id, track_id);
+
+            // Finalize the recording (flush buffers, close file, get waveform and audio data)
             let frames_recorded = recording.frames_written;
+            eprintln!("[STOP_RECORDING] Calling finalize() - frames_recorded={}", frames_recorded);
             match recording.finalize() {
-                Ok(temp_file_path) => {
+                Ok((temp_file_path, waveform, audio_data)) => {
-                    eprintln!("Recording finalized: {} frames written to {:?}", frames_recorded, temp_file_path);
+                    eprintln!("[STOP_RECORDING] Finalize succeeded: {} frames written to {:?}, {} waveform peaks generated, {} samples in memory",
+                              frames_recorded, temp_file_path, waveform.len(), audio_data.len());
 
-                    // Load the recorded audio file
+                    // Add to pool using the in-memory audio data (no file loading needed!)
-                    match crate::io::AudioFile::load(&temp_file_path) {
-                        Ok(audio_file) => {
-                            // Generate waveform for UI
-                            let duration = audio_file.duration();
-                            let target_peaks = ((duration * 300.0) as usize).clamp(1000, 20000);
-                            let waveform = audio_file.generate_waveform_overview(target_peaks);
-
-                            // Add to pool
                     let pool_file = crate::audio::pool::AudioFile::new(
                         temp_file_path.clone(),
-                                audio_file.data,
+                        audio_data,
-                                audio_file.channels,
+                        channels,
-                                audio_file.sample_rate,
+                        sample_rate,
                     );
                     let pool_index = self.audio_pool.add_file(pool_file);
+                    eprintln!("[STOP_RECORDING] Added to pool at index {}", pool_index);
 
                     // Update the clip to reference the pool
                     if let Some(crate::audio::track::TrackNode::Audio(track)) = self.project.get_track_mut(track_id) {
                         if let Some(clip) = track.clips.iter_mut().find(|c| c.id == clip_id) {
                             clip.audio_pool_index = pool_index;
-                                    // Duration should already be set during recording progress updates
+                            eprintln!("[STOP_RECORDING] Updated clip {} with pool_index {}", clip_id, pool_index);
                         }
                     }
 
                     // Delete temp file
                     let _ = std::fs::remove_file(&temp_file_path);
 
-                            // Notify UI that recording has stopped (with waveform)
+                    // Send event with the incrementally-generated waveform
+                    eprintln!("[STOP_RECORDING] Pushing RecordingStopped event for clip_id={}, pool_index={}, waveform_peaks={}",
+                              clip_id, pool_index, waveform.len());
                     let _ = self.event_tx.push(AudioEvent::RecordingStopped(clip_id, pool_index, waveform));
+                    eprintln!("[STOP_RECORDING] RecordingStopped event pushed successfully");
                 }
                 Err(e) => {
-                            // Send error event
+                    eprintln!("[STOP_RECORDING] Finalize failed: {}", e);
-                            let _ = self.event_tx.push(AudioEvent::RecordingError(
-                                format!("Failed to load recorded audio: {}", e)
-                            ));
-                        }
-                    }
-                }
-                Err(e) => {
-                    // Send error event
                     let _ = self.event_tx.push(AudioEvent::RecordingError(
                         format!("Failed to finalize recording: {}", e)
                     ));
                 }
             }
+        } else {
+            eprintln!("[STOP_RECORDING] No active recording to stop");
         }
     }
 

daw-backend/src/audio/recording.rs

@@ -1,6 +1,6 @@
 /// Audio recording system for capturing microphone input
 use crate::audio::{ClipId, TrackId};
-use crate::io::WavWriter;
+use crate::io::{WavWriter, WaveformPeak};
 use std::path::PathBuf;
 
 /// State of an active recording session
@@ -29,6 +29,14 @@ pub struct RecordingState {
     pub paused: bool,
     /// Number of samples remaining to skip (to discard stale buffer data)
     pub samples_to_skip: usize,
+    /// Waveform peaks generated incrementally during recording
+    pub waveform: Vec<WaveformPeak>,
+    /// Temporary buffer for collecting samples for next waveform peak
+    pub waveform_buffer: Vec<f32>,
+    /// Number of frames per waveform peak
+    pub frames_per_peak: usize,
+    /// All recorded audio data accumulated in memory (for fast finalization)
+    pub audio_data: Vec<f32>,
 }
 
 impl RecordingState {
@@ -45,6 +53,11 @@ impl RecordingState {
     ) -> Self {
         let flush_interval_frames = (sample_rate as f64 * flush_interval_seconds) as usize;
 
+        // Calculate frames per waveform peak
+        // Target ~300 peaks per second with minimum 1000 samples per peak
+        let target_peaks_per_second = 300;
+        let frames_per_peak = (sample_rate / target_peaks_per_second).max(1000) as usize;
+
         Self {
             track_id,
             clip_id,
@@ -58,6 +71,10 @@ impl RecordingState {
             flush_interval_frames,
             paused: false,
             samples_to_skip: 0, // Will be set by engine when it knows buffer size
+            waveform: Vec::new(),
+            waveform_buffer: Vec::new(),
+            frames_per_peak,
+            audio_data: Vec::new(),
         }
     }
 
@@ -68,8 +85,8 @@ impl RecordingState {
             return Ok(false);
         }
 
-        // Skip stale samples from the buffer
+        // Determine which samples to process
-        if self.samples_to_skip > 0 {
+        let samples_to_process = if self.samples_to_skip > 0 {
             let to_skip = self.samples_to_skip.min(samples.len());
             self.samples_to_skip -= to_skip;
 
@@ -79,12 +96,22 @@ impl RecordingState {
             }
 
             // Skip partial batch and process the rest
-            self.buffer.extend_from_slice(&samples[to_skip..]);
+            &samples[to_skip..]
         } else {
-            self.buffer.extend_from_slice(samples);
+            samples
-        }
+        };
 
-        // Check if we should flush
+        // Add to disk buffer
+        self.buffer.extend_from_slice(samples_to_process);
+
+        // Add to audio data (accumulate in memory for fast finalization)
+        self.audio_data.extend_from_slice(samples_to_process);
+
+        // Add to waveform buffer and generate peaks incrementally
+        self.waveform_buffer.extend_from_slice(samples_to_process);
+        self.generate_waveform_peaks();
+
+        // Check if we should flush to disk
         let frames_in_buffer = self.buffer.len() / self.channels as usize;
         if frames_in_buffer >= self.flush_interval_frames {
             self.flush()?;
@@ -94,6 +121,28 @@ impl RecordingState {
         Ok(false)
     }
 
+    /// Generate waveform peaks from accumulated samples
+    /// This is called incrementally as samples arrive
+    fn generate_waveform_peaks(&mut self) {
+        let samples_per_peak = self.frames_per_peak * self.channels as usize;
+
+        while self.waveform_buffer.len() >= samples_per_peak {
+            let mut min = 0.0f32;
+            let mut max = 0.0f32;
+
+            // Scan all samples for this peak
+            for sample in &self.waveform_buffer[..samples_per_peak] {
+                min = min.min(*sample);
+                max = max.max(*sample);
+            }
+
+            self.waveform.push(WaveformPeak { min, max });
+
+            // Remove processed samples from waveform buffer
+            self.waveform_buffer.drain(..samples_per_peak);
+        }
+    }
+
     /// Flush accumulated samples to disk
     pub fn flush(&mut self) -> Result<(), std::io::Error> {
         if self.buffer.is_empty() {
@@ -121,15 +170,28 @@ impl RecordingState {
         total_frames as f64 / self.sample_rate as f64
     }
 
-    /// Finalize the recording and return the temp file path
+    /// Finalize the recording and return the temp file path, waveform, and audio data
-    pub fn finalize(mut self) -> Result<PathBuf, std::io::Error> {
+    pub fn finalize(mut self) -> Result<(PathBuf, Vec<WaveformPeak>, Vec<f32>), std::io::Error> {
-        // Flush any remaining samples
+        // Flush any remaining samples to disk
         self.flush()?;
 
+        // Generate final waveform peak from any remaining samples
+        if !self.waveform_buffer.is_empty() {
+            let mut min = 0.0f32;
+            let mut max = 0.0f32;
+
+            for sample in &self.waveform_buffer {
+                min = min.min(*sample);
+                max = max.max(*sample);
+            }
+
+            self.waveform.push(WaveformPeak { min, max });
+        }
+
         // Finalize the WAV file
         self.writer.finalize()?;
 
-        Ok(self.temp_file_path)
+        Ok((self.temp_file_path, self.waveform, self.audio_data))
     }
 
     /// Pause recording

daw-backend/src/lib.rs

@@ -22,18 +22,27 @@ pub use io::{load_midi_file, AudioFile, WaveformPeak, WavWriter};
 
 use cpal::traits::{DeviceTrait, HostTrait, StreamTrait};
 
+/// Trait for emitting audio events to external systems (UI, logging, etc.)
+/// This allows the DAW backend to remain framework-agnostic
+pub trait EventEmitter: Send + Sync {
+    /// Emit an audio event
+    fn emit(&self, event: AudioEvent);
+}
+
 /// Simple audio system that handles cpal initialization internally
 pub struct AudioSystem {
     pub controller: EngineController,
     pub stream: cpal::Stream,
-    pub event_rx: rtrb::Consumer<AudioEvent>,
     pub sample_rate: u32,
     pub channels: u32,
 }
 
 impl AudioSystem {
     /// Initialize the audio system with default input and output devices
-    pub fn new() -> Result<Self, String> {
+    ///
+    /// # Arguments
+    /// * `event_emitter` - Optional event emitter for pushing events to external systems
+    pub fn new(event_emitter: Option<std::sync::Arc<dyn EventEmitter>>) -> Result<Self, String> {
         let host = cpal::default_host();
 
         // Get output device
@@ -84,10 +93,15 @@ impl AudioSystem {
                 eprintln!("Warning: No input device available, recording will be disabled");
                 // Start output stream and return without input
                 output_stream.play().map_err(|e| e.to_string())?;
+
+                // Spawn emitter thread if provided
+                if let Some(emitter) = event_emitter {
+                    Self::spawn_emitter_thread(event_rx, emitter);
+                }
+
                 return Ok(Self {
                     controller,
                     stream: output_stream,
-                    event_rx,
                     sample_rate,
                     channels,
                 });
@@ -106,10 +120,15 @@ impl AudioSystem {
             Err(e) => {
                 eprintln!("Warning: Could not get input config: {}, recording will be disabled", e);
                 output_stream.play().map_err(|e| e.to_string())?;
+
+                // Spawn emitter thread if provided
+                if let Some(emitter) = event_emitter {
+                    Self::spawn_emitter_thread(event_rx, emitter);
+                }
+
                 return Ok(Self {
                     controller,
                     stream: output_stream,
-                    event_rx,
                     sample_rate,
                     channels,
                 });
@@ -138,12 +157,31 @@ impl AudioSystem {
         // Leak the input stream to keep it alive
         Box::leak(Box::new(input_stream));
 
+        // Spawn emitter thread if provided
+        if let Some(emitter) = event_emitter {
+            Self::spawn_emitter_thread(event_rx, emitter);
+        }
+
         Ok(Self {
             controller,
             stream: output_stream,
-            event_rx,
             sample_rate,
             channels,
         })
     }
+
+    /// Spawn a background thread to emit events from the ringbuffer
+    fn spawn_emitter_thread(mut event_rx: rtrb::Consumer<AudioEvent>, emitter: std::sync::Arc<dyn EventEmitter>) {
+        std::thread::spawn(move || {
+            loop {
+                // Wait for events and emit them
+                if let Ok(event) = event_rx.pop() {
+                    emitter.emit(event);
+                } else {
+                    // No events available, sleep briefly to avoid busy-waiting
+                    std::thread::sleep(std::time::Duration::from_millis(1));
+                }
+            }
+        });
+    }
 }

src-tauri/src/audio.rs

@@ -1,5 +1,6 @@
-use daw_backend::{AudioEvent, AudioSystem, EngineController, WaveformPeak};
+use daw_backend::{AudioEvent, AudioSystem, EngineController, EventEmitter, WaveformPeak};
 use std::sync::{Arc, Mutex};
+use tauri::{Emitter, Manager};
 
 #[derive(serde::Serialize)]
 pub struct AudioFileMetadata {
@@ -12,7 +13,6 @@ pub struct AudioFileMetadata {
 
 pub struct AudioState {
     controller: Option<EngineController>,
-    event_rx: Option<rtrb::Consumer<AudioEvent>>,
     sample_rate: u32,
     channels: u32,
     next_track_id: u32,
@@ -23,7 +23,6 @@ impl Default for AudioState {
     fn default() -> Self {
         Self {
             controller: None,
-            event_rx: None,
             sample_rate: 0,
             channels: 0,
             next_track_id: 0,
@@ -32,8 +31,42 @@ impl Default for AudioState {
     }
 }
 
+/// Implementation of EventEmitter that uses Tauri's event system
+struct TauriEventEmitter {
+    app_handle: tauri::AppHandle,
+}
+
+impl EventEmitter for TauriEventEmitter {
+    fn emit(&self, event: AudioEvent) {
+        // Serialize the event to the format expected by the frontend
+        let serialized_event = match event {
+            AudioEvent::RecordingStarted(track_id, clip_id) => {
+                SerializedAudioEvent::RecordingStarted { track_id, clip_id }
+            }
+            AudioEvent::RecordingProgress(clip_id, duration) => {
+                SerializedAudioEvent::RecordingProgress { clip_id, duration }
+            }
+            AudioEvent::RecordingStopped(clip_id, pool_index, waveform) => {
+                SerializedAudioEvent::RecordingStopped { clip_id, pool_index, waveform }
+            }
+            AudioEvent::RecordingError(message) => {
+                SerializedAudioEvent::RecordingError { message }
+            }
+            _ => return, // Ignore other event types for now
+        };
+
+        // Emit the event via Tauri
+        if let Err(e) = self.app_handle.emit("audio-event", serialized_event) {
+            eprintln!("Failed to emit audio event: {}", e);
+        }
+    }
+}
+
 #[tauri::command]
-pub async fn audio_init(state: tauri::State<'_, Arc<Mutex<AudioState>>>) -> Result<String, String> {
+pub async fn audio_init(
+    state: tauri::State<'_, Arc<Mutex<AudioState>>>,
+    app_handle: tauri::AppHandle,
+) -> Result<String, String> {
     let mut audio_state = state.lock().unwrap();
 
     // Check if already initialized - if so, reset DAW state (for hot-reload)
@@ -47,8 +80,11 @@ pub async fn audio_init(state: tauri::State<'_, Arc<Mutex<AudioState>>>) -> Resu
         ));
     }
 
+    // Create TauriEventEmitter
+    let emitter = Arc::new(TauriEventEmitter { app_handle });
+
     // AudioSystem handles all cpal initialization internally
-    let system = AudioSystem::new()?;
+    let system = AudioSystem::new(Some(emitter))?;
 
     let info = format!(
         "Audio initialized: {} Hz, {} ch",
@@ -60,7 +96,6 @@ pub async fn audio_init(state: tauri::State<'_, Arc<Mutex<AudioState>>>) -> Resu
     Box::leak(Box::new(system.stream));
 
     audio_state.controller = Some(system.controller);
-    audio_state.event_rx = Some(system.event_rx);
     audio_state.sample_rate = system.sample_rate;
     audio_state.channels = system.channels;
     audio_state.next_track_id = 0;
@@ -310,7 +345,7 @@ pub async fn audio_resume_recording(
     }
 }
 
-#[derive(serde::Serialize)]
+#[derive(serde::Serialize, Clone)]
 #[serde(tag = "type")]
 pub enum SerializedAudioEvent {
     RecordingStarted { track_id: u32, clip_id: u32 },
@@ -319,34 +354,4 @@ pub enum SerializedAudioEvent {
     RecordingError { message: String },
 }
 
-#[tauri::command]
+// audio_get_events command removed - events are now pushed via Tauri event system
-pub async fn audio_get_events(
-    state: tauri::State<'_, Arc<Mutex<AudioState>>>,
-) -> Result<Vec<SerializedAudioEvent>, String> {
-    let mut audio_state = state.lock().unwrap();
-    let mut events = Vec::new();
-
-    if let Some(event_rx) = &mut audio_state.event_rx {
-        // Poll all available events
-        while let Ok(event) = event_rx.pop() {
-            match event {
-                AudioEvent::RecordingStarted(track_id, clip_id) => {
-                    events.push(SerializedAudioEvent::RecordingStarted { track_id, clip_id });
-                }
-                AudioEvent::RecordingProgress(clip_id, duration) => {
-                    events.push(SerializedAudioEvent::RecordingProgress { clip_id, duration });
-                }
-                AudioEvent::RecordingStopped(clip_id, pool_index, waveform) => {
-                    events.push(SerializedAudioEvent::RecordingStopped { clip_id, pool_index, waveform });
-                }
-                AudioEvent::RecordingError(message) => {
-                    events.push(SerializedAudioEvent::RecordingError { message });
-                }
-                // Ignore other event types for now
-                _ => {}
-            }
-        }
-    }
-
-    Ok(events)
-}

src-tauri/src/lib.rs

@@ -207,7 +207,6 @@ pub fn run() {
         audio::audio_stop_recording,
         audio::audio_pause_recording,
         audio::audio_resume_recording,
-        audio::audio_get_events,
       ])
       // .manage(window_counter)
       .build(tauri::generate_context!())

src/main.js

@@ -1,4 +1,5 @@
 const { invoke } = window.__TAURI__.core;
+const { listen } = window.__TAURI__.event;
 import * as fitCurve from "/fit-curve.js";
 import { Bezier } from "/bezier.js";
 import { Quadtree } from "./quadtree.js";
@@ -923,12 +924,13 @@ async function playPause() {
   } else {
     // Stop recording if active
     if (context.isRecording) {
+      console.log('playPause - stopping recording for clip:', context.recordingClipId);
       try {
         await invoke('audio_stop_recording');
         context.isRecording = false;
         context.recordingTrackId = null;
         context.recordingClipId = null;
-        console.log('Recording stopped by play/pause');
+        console.log('Recording stopped by play/pause button');
 
         // Update record button appearance if it exists
         if (context.recordButton) {
@@ -969,9 +971,6 @@ function advanceFrame() {
     context.timelineWidget.timelineState.currentTime = context.activeObject.currentTime;
   }
 
-  // Poll for audio events (recording progress, etc.)
-  pollAudioEvents();
-
   // Redraw stage and timeline
   updateUI();
   if (context.timelineWidget?.requestRedraw) {
@@ -981,6 +980,11 @@ function advanceFrame() {
   if (playing) {
     const duration = context.activeObject.duration;
 
+    // Debug logging for recording
+    if (context.isRecording) {
+      console.log('advanceFrame - recording active, currentTime:', context.activeObject.currentTime, 'duration:', duration, 'isRecording:', context.isRecording);
+    }
+
     // Check if we've reached the end (but allow infinite playback when recording)
     if (context.isRecording || (duration > 0 && context.activeObject.currentTime < duration)) {
       // Continue playing
@@ -988,6 +992,18 @@ function advanceFrame() {
     } else {
       // Animation finished
       playing = false;
+
+      // Stop DAW backend audio playback
+      invoke('audio_stop').catch(error => {
+        console.error('Failed to stop audio playback:', error);
+      });
+
+      // Update play/pause button appearance
+      if (context.playPauseButton) {
+        context.playPauseButton.className = "playback-btn playback-btn-play";
+        context.playPauseButton.title = "Play";
+      }
+
       for (let audioTrack of context.activeObject.audioTracks) {
         for (let i in audioTrack.sounds) {
           let sound = audioTrack.sounds[i];
@@ -998,22 +1014,18 @@ function advanceFrame() {
   }
 }
 
-async function pollAudioEvents() {
+// Handle audio events pushed from Rust via Tauri event system
-  const { invoke } = window.__TAURI__.core;
+async function handleAudioEvent(event) {
-
-  try {
-    const events = await invoke('audio_get_events');
-
-    for (const event of events) {
   switch (event.type) {
     case 'RecordingStarted':
-          console.log('Recording started - track:', event.track_id, 'clip:', event.clip_id);
+      console.log('[FRONTEND] RecordingStarted - track:', event.track_id, 'clip:', event.clip_id);
       context.recordingClipId = event.clip_id;
 
       // Create the clip object in the audio track
       const recordingTrack = context.activeObject.audioTracks.find(t => t.audioTrackId === event.track_id);
       if (recordingTrack) {
         const startTime = context.activeObject.currentTime || 0;
+        console.log('[FRONTEND] Creating clip object for clip', event.clip_id, 'on track', event.track_id, 'at time', startTime);
         recordingTrack.clips.push({
           clipId: event.clip_id,
           poolIndex: null,  // Will be set when recording stops
@@ -1028,6 +1040,8 @@ async function pollAudioEvents() {
         if (context.timelineWidget?.requestRedraw) {
           context.timelineWidget.requestRedraw();
         }
+      } else {
+        console.error('[FRONTEND] Could not find audio track', event.track_id, 'for RecordingStarted event');
       }
       break;
 
@@ -1038,11 +1052,21 @@ async function pollAudioEvents() {
       break;
 
     case 'RecordingStopped':
-          console.log('Recording stopped - clip:', event.clip_id, 'pool_index:', event.pool_index, 'waveform peaks:', event.waveform?.length);
+      console.log('[FRONTEND] RecordingStopped event - clip:', event.clip_id, 'pool_index:', event.pool_index, 'waveform peaks:', event.waveform?.length);
+      console.log('[FRONTEND] Current recording state - isRecording:', context.isRecording, 'recordingClipId:', context.recordingClipId);
       await finalizeRecording(event.clip_id, event.pool_index, event.waveform);
+
+      // Always clear recording state when we receive RecordingStopped
+      console.log('[FRONTEND] Clearing recording state after RecordingStopped event');
       context.isRecording = false;
       context.recordingTrackId = null;
       context.recordingClipId = null;
+
+      // Update record button appearance
+      if (context.recordButton) {
+        context.recordButton.className = "playback-btn playback-btn-record";
+        context.recordButton.title = "Record";
+      }
       break;
 
     case 'RecordingError':
@@ -1053,12 +1077,13 @@ async function pollAudioEvents() {
       context.recordingClipId = null;
       break;
   }
-    }
-  } catch (error) {
-    // Silently ignore errors - polling happens frequently
-  }
 }
 
+// Set up Tauri event listener for audio events
+listen('audio-event', (tauriEvent) => {
+  handleAudioEvent(tauriEvent.payload);
+});
+
 function updateRecordingClipDuration(clipId, duration) {
   // Find the clip in the active object's audio tracks and update its duration
   for (const audioTrack of context.activeObject.audioTracks) {
@@ -1155,14 +1180,15 @@ async function toggleRecording() {
 
   if (context.isRecording) {
     // Stop recording
+    console.log('[FRONTEND] toggleRecording - stopping recording for clip:', context.recordingClipId);
     try {
       await invoke('audio_stop_recording');
       context.isRecording = false;
       context.recordingTrackId = null;
       context.recordingClipId = null;
-      console.log('Recording stopped');
+      console.log('[FRONTEND] Recording stopped via toggle button');
     } catch (error) {
-      console.error('Failed to stop recording:', error);
+      console.error('[FRONTEND] Failed to stop recording:', error);
     }
   } else {
     // Start recording - check if activeLayer is an audio track
@@ -1180,6 +1206,7 @@ async function toggleRecording() {
     // Start recording at current playhead position
     const startTime = context.activeObject.currentTime || 0;
 
+    console.log('[FRONTEND] Starting recording on track', audioTrack.audioTrackId, 'at time', startTime);
     try {
       await invoke('audio_start_recording', {
         trackId: audioTrack.audioTrackId,
@@ -1187,14 +1214,14 @@ async function toggleRecording() {
       });
       context.isRecording = true;
       context.recordingTrackId = audioTrack.audioTrackId;
-      console.log('Recording started on track', audioTrack.audioTrackId, 'at time', startTime);
+      console.log('[FRONTEND] Recording started successfully, waiting for RecordingStarted event');
 
       // Start playback so the timeline moves (if not already playing)
       if (!playing) {
         await playPause();
       }
     } catch (error) {
-      console.error('Failed to start recording:', error);
+      console.error('[FRONTEND] Failed to start recording:', error);
       alert('Failed to start recording: ' + error);
     }
   }