FS-7500: initial idea to decode video in core

src/include/private/switch_core_pvt.h

@@ -134,6 +134,8 @@ struct switch_core_session {
 	switch_mutex_t *resample_mutex;
 	switch_mutex_t *codec_read_mutex;
 	switch_mutex_t *codec_write_mutex;
+	switch_mutex_t *video_codec_read_mutex;
+	switch_mutex_t *video_codec_write_mutex;
 	switch_thread_cond_t *cond;
 	switch_mutex_t *frame_read_mutex;
 
@@ -166,6 +168,16 @@ struct switch_core_session {
 	switch_frame_t enc_read_frame;
 	uint8_t raw_read_buf[SWITCH_RECOMMENDED_BUFFER_SIZE];
 	uint8_t enc_read_buf[SWITCH_RECOMMENDED_BUFFER_SIZE];
+
+	/* video frame.data being trated differently than audio, allocate a dynamic data buffer if necessary*/
+	switch_buffer_t *video_raw_write_buffer;
+	switch_frame_t video_raw_write_frame;
+	// switch_frame_t video_enc_write_frame;
+
+	switch_buffer_t *video_raw_read_buffer;
+	switch_frame_t video_raw_read_frame;
+	// switch_frame_t video_enc_read_frame;
+
 	switch_codec_t bug_codec;
 	uint32_t read_frame_count;
 	uint32_t track_duration;

src/include/switch_module_interfaces.h

@@ -611,6 +611,13 @@ struct switch_codec_fmtp {
 
 };
 
+struct switch_picture {
+	uint32_t width;      /* the picture width */
+	uint32_t height;     /* the picture height */
+	uint8_t *planes[4];  /* pointer to the top left pixel for each plane */
+	uint32_t stride[4];  /* stride between rows for each plane */
+};
+
 /*! an abstract handle to a codec module */
 struct switch_codec {
 	/*! the codec interface table this handle uses */
@@ -632,6 +639,10 @@ struct switch_codec {
 	struct switch_codec *next;
 	switch_core_session_t *session;
 	switch_frame_t *cur_frame;
+	/*! raw picture for encode */
+	switch_picture_t enc_picture;
+	/*! decoded picture */
+	switch_picture_t dec_picture;
 };
 
 /*! \brief A table of settings and callbacks that define a paticular implementation of a codec */

src/include/switch_types.h

@@ -541,6 +541,7 @@ SWITCH_DECLARE_DATA extern switch_filenames SWITCH_GLOBAL_filenames;
 #define SWITCH_MAX_SAMPLE_LEN 48
 #define SWITCH_BYTES_PER_SAMPLE 2	/* slin is 2 bytes per sample */
 #define SWITCH_RECOMMENDED_BUFFER_SIZE 8192
+#define SWITCH_RECOMMENDED_VIDEO_BUFFER_SIZE 4096 * 1024 /* Fixme: Just Make sure it's big enough for now */
 #define SWITCH_MAX_CODECS 50
 #define SWITCH_MAX_STATE_HANDLERS 30
 #define SWITCH_CORE_QUEUE_LEN 100000
@@ -1472,10 +1473,12 @@ typedef enum {
 <pre>
 SFF_CNG        = (1 <<  0) - Frame represents comfort noise
 SFF_RAW_RTP    = (1 <<  1) - Frame has raw rtp accessible
-SFF_RTP_HEADER = (1 << 2)  - Get the rtp header from the frame header
-SFF_PLC        = (1 << 3)  - Frame has generated PLC data
-SFF_RFC2833    = (1 << 4)  - Frame has rfc2833 dtmf data
-SFF_DYNAMIC    = (1 << 5)  - Frame is dynamic and should be freed
+SFF_RTP_HEADER = (1 <<  2) - Get the rtp header from the frame header
+SFF_PLC        = (1 <<  3) - Frame has generated PLC data
+SFF_RFC2833    = (1 <<  4) - Frame has rfc2833 dtmf data
+SFF_DYNAMIC    = (1 <<  5) - Frame is dynamic and should be freed
+SFF_MARKER     = (1 << 11) - Frame flag has Marker set, only set by encoder
+SFF_WAIT_KEY_FRAME = (1 << 12) - Need a key from before could decode
 </pre>
  */
 typedef enum {
@@ -1490,7 +1493,9 @@ typedef enum {
 	SFF_ZRTP = (1 << 7),
 	SFF_UDPTL_PACKET = (1 << 8),
 	SFF_NOT_AUDIO = (1 << 9),
-	SFF_RTCP = (1 << 10)
+	SFF_RTCP = (1 << 10),
+	SFF_MARKER = (1 << 11),
+	SFF_WAIT_KEY_FRAME = (1 << 12)
 } switch_frame_flag_enum_t;
 typedef uint32_t switch_frame_flag_t;
 
@@ -2090,6 +2095,7 @@ typedef struct switch_caller_extension switch_caller_extension_t;
 typedef struct switch_caller_application switch_caller_application_t;
 typedef struct switch_state_handler_table switch_state_handler_table_t;
 typedef struct switch_timer switch_timer_t;
+typedef struct switch_picture switch_picture_t;
 typedef struct switch_codec switch_codec_t;
 typedef struct switch_core_thread_session switch_core_thread_session_t;
 typedef struct switch_codec_implementation switch_codec_implementation_t;

src/switch_core_io.c

@@ -104,8 +104,6 @@ SWITCH_DECLARE(switch_status_t) switch_core_session_read_video_frame(switch_core
 		goto done;
 	}
 
-	switch_assert(*frame != NULL);
-
 	if (switch_test_flag(*frame, SFF_CNG)) {
 		status = SWITCH_STATUS_SUCCESS;
 		goto done;

src/switch_core_media.c

@@ -4346,9 +4346,6 @@ SWITCH_DECLARE(int) switch_core_media_toggle_hold(switch_core_session_t *session
 
 #define BUF_SIZE (352 * 288 * 3 / 2 * 4) // big enough for 4CIF, looks like C doesn't like huge array
 #define FPS 15
-#define WIDTH 352
-#define HEIGHT 288
-#define SIZE WIDTH * HEIGHT
 
 static switch_status_t video_bridge_callback(switch_core_session_t *session, switch_bool_t video_transcoding, uint32_t *ts)
 {
@@ -4397,7 +4394,7 @@ static switch_status_t video_bridge_callback(switch_core_session_t *session, swi
 		uint32_t flag = 0;
 		uint32_t encoded_data_len = 1500;
 		uint32_t encoded_rate = 0;
-		switch_frame_t write_frame;
+		switch_frame_t write_frame = { 0 };
 
 #if 0
 		switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "%d/%s != %d/%s need transcoding!!!\n",
@@ -4413,13 +4410,13 @@ static switch_status_t video_bridge_callback(switch_core_session_t *session, swi
 
 		if (decoded_data_len < 3) return SWITCH_STATUS_SUCCESS;
 
-		decoded_data_len = 152064;
-
-		switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "decoded_data_len: %d %s\n", decoded_data_len, codec->implementation->iananame);
+		switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "%s decoded_data_len: %d size: %dx%d\n", codec->implementation->iananame, decoded_data_len, codec->dec_picture.width, codec->dec_picture.height);
 
 		write_frame.packet = rtp_buff;
 		write_frame.data = rtp_buff + 12;
 
+		other_codec->enc_picture.width = codec->dec_picture.width;
+		other_codec->enc_picture.height = codec->dec_picture.height;
 		encoded_data_len = 1500;
 		switch_core_codec_encode(other_codec, NULL, raw_buff, decoded_data_len, 0, rtp_buff+12, &encoded_data_len, &encoded_rate, &flag);
 
@@ -4428,7 +4425,7 @@ static switch_status_t video_bridge_callback(switch_core_session_t *session, swi
 
 			write_frame.datalen = encoded_data_len;
 			write_frame.packetlen = write_frame.datalen + 12;
-			write_frame.m = flag;
+			write_frame.m = flag & SFF_MARKER ? 1 : 0;
 			write_frame.timestamp = *ts;
 			if (write_frame.m) *ts += 90000 / FPS;
 

src/switch_core_session.c

@@ -1317,6 +1317,17 @@ SWITCH_DECLARE(void) switch_core_session_reset(switch_core_session_t *session, s
 	switch_buffer_destroy(&session->raw_read_buffer);
 	switch_mutex_unlock(session->codec_read_mutex);
 
+	switch_mutex_lock(session->video_codec_write_mutex);
+	switch_buffer_destroy(&session->video_raw_write_buffer);
+	switch_mutex_unlock(session->video_codec_write_mutex);
+
+	switch_mutex_lock(session->video_codec_read_mutex);
+	switch_buffer_destroy(&session->video_raw_read_buffer);
+	switch_mutex_unlock(session->video_codec_read_mutex);
+
+	//video_raw_read_frame.data is dynamically allocated if necessary, so wipe this also
+	switch_safe_free(session->video_raw_read_frame.data);
+
 	if (flush_dtmf) {
 		while ((has = switch_channel_has_dtmf(channel))) {
 			switch_channel_flush_dtmf(channel);
@@ -2419,6 +2430,8 @@ SWITCH_DECLARE(switch_core_session_t *) switch_core_session_request_uuid(switch_
 	switch_mutex_init(&session->resample_mutex, SWITCH_MUTEX_NESTED, session->pool);
 	switch_mutex_init(&session->codec_read_mutex, SWITCH_MUTEX_NESTED, session->pool);
 	switch_mutex_init(&session->codec_write_mutex, SWITCH_MUTEX_NESTED, session->pool);
+	switch_mutex_init(&session->video_codec_read_mutex, SWITCH_MUTEX_NESTED, session->pool);
+	switch_mutex_init(&session->video_codec_write_mutex, SWITCH_MUTEX_NESTED, session->pool);
 	switch_mutex_init(&session->frame_read_mutex, SWITCH_MUTEX_NESTED, session->pool);
 	switch_thread_rwlock_create(&session->bug_rwlock, session->pool);
 	switch_thread_cond_create(&session->cond, session->pool);