merge: fix barge-in audio flow

ql-application/threadx/interface/custom_app/hal_key.c

@@ -187,6 +187,9 @@ static void key_thread_entry(void *parameter)
 					if(ql_rtos_get_systicks() - key_press_mic_time > KEY_LONG_PRESS_3S && power_down_status==0){
 						nfc_tx_click_count = 0;
 						power_down_status = 1;
+						/* Shutdown audio must own the codec.  Stop VAD and any
+						 * interrupted TTS before queueing the power-off prompt. */
+						stop_player_func();
 						app_play_tips_file(10);
 						power_down_status = 2;
 						//ql_rtos_task_sleep_s(3);

ql-application/threadx/interface/custom_app/http_downfile.c

@@ -34,6 +34,7 @@ const char *mp3_file[] ={
 	"volume_up_zh.mp3",//16  
 	"volume_down_zh.mp3",//17  
 	"ai_active_zh.mp3",//18
+	"barge_in_cute.mp3",//19  对话打断提示音
 };
 
 //http://banban.api.talkingq.com:8080/assets/roles/banban/tts_error.mp3
@@ -183,11 +184,13 @@ void app_file_check_download(void)
 	char filename_buff[64]={0x00};
 	int ret = 0, cnt = 0;
 	
-	for(index=1;index<19;index++){	
+	for(index=1;index<20;index++){
-		
+
-		if(index==4||index==5||index==6||index==7||index==8){
+		/* barge_in_cute.mp3 is shipped in customer_fs and is not hosted by the
-			continue;
+		 * legacy prompt download service. */
-		}
+		if(index==4||index==5||index==6||index==7||index==8||index==19){
+			continue;
+		}
 		
 		if(strlen(mp3_file[index])>0){
 			sprintf(filename_buff, "U:/%s", mp3_file[index]);
@@ -203,14 +206,16 @@ down_loop:
 	}
 }
 
-void app_file_check_del(void)
+void app_file_check_del(void)
 {
 	int index = 1;
 	char filename_buff[64]={0x00};
 	
-	for(index=1;index<19;index++){	
+	for(index=1;index<20;index++){
 		
-		if(index==4||index==5||index==6||index==7||index==8){
+		/* Keep the bundled barge-in prompt across app-only FOTA.  Unlike the
+		 * server-managed prompts, it cannot be restored by the downloader. */
+		if(index==4||index==5||index==6||index==7||index==8||index==19){
 			continue;
 		}
 		

ql-application/threadx/interface/custom_app/ws_talkingq.c

@@ -49,11 +49,14 @@ static uint8_t sender_buf[PCM_FRAME_LENGTH]={0};
 static int16 mp3_buff[2*PCM_FRAME_LENGTH];
 static int16 *ppbuff=mp3_buff;
 
-static int onmessage_status = -1;
+static int onmessage_status = -1;
-volatile int tts_playing = 0;
+volatile int tts_playing = 0;
-unsigned char wss_connect_start = 0;
+unsigned char wss_connect_start = 0;
 static volatile int player_mq_couter = 0;
-static unsigned char audio_in_task_cancel = 0;
+/* Set while a barge-in prompt bridges the old and new sessions.  The old
+ * server response can still deliver URLs after its player queue was cleared. */
+static volatile int barge_in_transition = 0;
+static unsigned char audio_in_task_cancel = 0;
 
 static unsigned char xiaoice_wss_connect_status = 0; 
 static unsigned char wss_connect_counter = 0;
@@ -299,6 +302,14 @@ static void onmessage_text(rws_socket socket, const char *text, const unsigned i
 			return;
 		}
 
+		/* The new session id is generated only when recording starts.  Until
+		 * then, every TTS message on this socket belongs to the interrupted
+		 * session and must not be allowed to refill the cleared player queue. */
+		if (barge_in_transition) {
+			printf("drop stale websocket message during barge-in\r\n");
+			return;
+		}
+
 	//	printf("message(txt), %d(byte): %s ", len, text);
 		
 		char url_buff[128]={0x00};
@@ -743,9 +754,12 @@ void generateSessionId(void)
 
 
 #if 1
-void sendStartPacket(void) {
+void sendStartPacket(void) {
-    generateSessionId();
+    generateSessionId();
-	generateDeviceId();
+	/* sessionId now identifies the new conversation.  Late URLs from the old
+	 * conversation will fail the existing session-id check below. */
+	barge_in_transition = 0;
+	generateDeviceId();
     gSequenceNumber = 0;
 
     AudioPacket startPacket = {0};
@@ -862,17 +876,23 @@ void sendRecordingAudioPacket(AudioPacket *packet) {
 
 extern const char *mp3_file[];
 
-int app_play_tips_file(int index)
+int app_play_tips_file(int index)
-{
+{
-	int file_valid_check = 0;
+	int file_valid_check = 0;
-	
+	int access_ret = 0;
-	printf("xiaoice_play_index:%d\n", index);
+
-	
+	printf("xiaoice_play_index:%d\n", index);
-		//正在录音过程中不允许播放
+
-//	if(xiaoice_get_tts_play_status() || get_start_mic_flag()==1) return -1;
+		//正在录音过程中不允许播放
-    if(power_down_status!=1){
+//	if(xiaoice_get_tts_play_status() || get_start_mic_flag()==1) return -1;
-		if(vad_audio_is_recording() || tts_playing!=0 || fota_start==1 || power_down_status==2 || get_sleep_mode()==1) return -1;
+    if(power_down_status!=1){
-    }
+		if(vad_audio_is_recording() || tts_playing!=0 || fota_start==1 || power_down_status==2 || get_sleep_mode()==1) {
+			printf("tips blocked:index=%d vad=%d tts=%d fota=%d power=%d sleep=%d\r\n",
+			       index, vad_audio_is_recording(), tts_playing, fota_start,
+			       power_down_status, get_sleep_mode());
+			return -1;
+		}
+    }
 	
 	char filename_buff[64]={0x00};
 	
@@ -881,15 +901,18 @@ int app_play_tips_file(int index)
 	if(index==0) {
 		file_valid_check = 0;
 	}else{ 
-		if(ql_access(filename_buff, 0)==0){	
+		access_ret = ql_access(filename_buff, 0);
-			file_valid_check = 1;
+		if(access_ret==0){
-		}else{
+			file_valid_check = 1;
-			file_valid_check = 2;
+		}else{
-		}
+			printf("tips file missing:index=%d path=%s access=%d\r\n",
+			       index, filename_buff, access_ret);
+			file_valid_check = 2;
+		}
 	}
 
 	if(file_valid_check!=2){
-		if(index==18){
+		if(index==18 || index==19){
 			add_file_to_play_player_queue(filename_buff, 2);
 		}else{
 			add_file_to_play_player_queue(filename_buff, 0);
@@ -1001,26 +1024,46 @@ void app_player_thread(void *arvg)
 
 int tts_play_interrupt_func(void)
 {
+	int playback_interrupted = 0;
+
 	//如果在vad状态，则直接返回
 	if(vad_audio_is_recording()||g_mic_working) return -1;
 
 	g_mic_working = 1;//等待转为vad状态就清0 或者拾音超时 或者ws没有正常
-	
+
-	xiaoice_close_all_audio();
+	/* Remember whether this key press is interrupting TTS.  tts_playing == 2
+	 * is only a cancellation signal for the old player; it must not leak into
+	 * the next conversation, otherwise prompts and error audio are discarded. */
+	playback_interrupted = (tts_playing != 0 || player_mq_couter > 0);
+	barge_in_transition = playback_interrupted;
+	xiaoice_close_all_audio();
 
 	talkingq_ws_connect();
 
-	if(get_xiaoice_connected_status()!=2) {
+	if(get_xiaoice_connected_status()!=2) {
-		g_mic_working = 0;
+		/* Let the caller play its local network-error prompt. */
-		return -2;
+		tts_play_status = 0;
-	}
+		tts_playing = 0;
-	
+		barge_in_transition = 0;
-	if(tts_playing==2){
+		g_mic_working = 0;
-		printf("vad_start_func0\n");
+		return -2;
-		vad_start_func();//如果是打断就不播报
+	}
+
+	if(playback_interrupted){
+		printf("barge_in_prompt\n");
+		/* The old player has finished consuming the cancellation state. */
+		tts_play_status = 0;
+		tts_playing = 0;
+		/* Use the same local-prompt flow as ai_active_zh.mp3. Queue type 2
+		 * starts VAD after playback; fall back to VAD if the file is absent. */
+		if(app_play_tips_file(19) != 0){
+			vad_start_func();
+		}
 	}else{
 		app_play_tips_file(18);
 	}
+
+	return 0;
 }
 
 void stop_player_func(void)
@@ -1028,6 +1071,7 @@ void stop_player_func(void)
 	vad_audio_stop_record();
 	xiaoice_close_all_audio();
 
+	tts_play_status = 0;
 	tts_playing = 0;
 }
 
@@ -1194,9 +1238,16 @@ void xiaoice_close_all_audio(void)
 
 		timeout = hal_os_up_time_ms();
 
-		while(tts_play_status&&tts_playing!=2){
+		/* The queue counter also includes the item currently owned by the
+		 * player task.  Wait for that item to observe tts_playing == 2 and
+		 * finish cleanup before the cancellation state is reused/reset. */
+		while(player_mq_couter>0){
 			ql_rtos_task_sleep_ms(20);
-			if(hal_os_up_time_ms()-timeout>10000) break;
+			if(hal_os_up_time_ms()-timeout>OTA_AUDIO_STOP_TIMEOUT_MS){
+				printf("wait player stop timeout, counter:%d\r\n",
+				       player_mq_couter);
+				break;
+			}
 		}
 	}
 #endif