mirror of
https://git.ffmpeg.org/ffmpeg.git
synced 2024-12-14 03:04:53 +00:00
2912e87a6c
Signed-off-by: Mans Rullgard <mans@mansr.com>
571 lines
18 KiB
C
571 lines
18 KiB
C
/*
|
|
* Realmedia RTSP protocol (RDT) support.
|
|
* Copyright (c) 2007 Ronald S. Bultje
|
|
*
|
|
* This file is part of Libav.
|
|
*
|
|
* Libav is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU Lesser General Public
|
|
* License as published by the Free Software Foundation; either
|
|
* version 2.1 of the License, or (at your option) any later version.
|
|
*
|
|
* Libav is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
* Lesser General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU Lesser General Public
|
|
* License along with Libav; if not, write to the Free Software
|
|
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|
*/
|
|
|
|
/**
|
|
* @file
|
|
* @brief Realmedia RTSP protocol (RDT) support
|
|
* @author Ronald S. Bultje <rbultje@ronald.bitfreak.net>
|
|
*/
|
|
|
|
#include "avformat.h"
|
|
#include "libavutil/avstring.h"
|
|
#include "rtpdec.h"
|
|
#include "rdt.h"
|
|
#include "libavutil/base64.h"
|
|
#include "libavutil/md5.h"
|
|
#include "rm.h"
|
|
#include "internal.h"
|
|
#include "avio_internal.h"
|
|
#include "libavcodec/get_bits.h"
|
|
|
|
struct RDTDemuxContext {
|
|
AVFormatContext *ic; /**< the containing (RTSP) demux context */
|
|
/** Each RDT stream-set (represented by one RTSPStream) can contain
|
|
* multiple streams (of the same content, but with possibly different
|
|
* codecs/bitrates). Each such stream is represented by one AVStream
|
|
* in the AVFormatContext, and this variable points to the offset in
|
|
* that array such that the first is the first stream of this set. */
|
|
AVStream **streams;
|
|
int n_streams; /**< streams with identifical content in this set */
|
|
void *dynamic_protocol_context;
|
|
DynamicPayloadPacketHandlerProc parse_packet;
|
|
uint32_t prev_timestamp;
|
|
int prev_set_id, prev_stream_id;
|
|
};
|
|
|
|
RDTDemuxContext *
|
|
ff_rdt_parse_open(AVFormatContext *ic, int first_stream_of_set_idx,
|
|
void *priv_data, RTPDynamicProtocolHandler *handler)
|
|
{
|
|
RDTDemuxContext *s = av_mallocz(sizeof(RDTDemuxContext));
|
|
if (!s)
|
|
return NULL;
|
|
|
|
s->ic = ic;
|
|
s->streams = &ic->streams[first_stream_of_set_idx];
|
|
do {
|
|
s->n_streams++;
|
|
} while (first_stream_of_set_idx + s->n_streams < ic->nb_streams &&
|
|
s->streams[s->n_streams]->id == s->streams[0]->id);
|
|
s->prev_set_id = -1;
|
|
s->prev_stream_id = -1;
|
|
s->prev_timestamp = -1;
|
|
s->parse_packet = handler ? handler->parse_packet : NULL;
|
|
s->dynamic_protocol_context = priv_data;
|
|
|
|
return s;
|
|
}
|
|
|
|
void
|
|
ff_rdt_parse_close(RDTDemuxContext *s)
|
|
{
|
|
av_free(s);
|
|
}
|
|
|
|
struct PayloadContext {
|
|
AVFormatContext *rmctx;
|
|
int nb_rmst;
|
|
RMStream **rmst;
|
|
uint8_t *mlti_data;
|
|
unsigned int mlti_data_size;
|
|
char buffer[RTP_MAX_PACKET_LENGTH + FF_INPUT_BUFFER_PADDING_SIZE];
|
|
int audio_pkt_cnt; /**< remaining audio packets in rmdec */
|
|
};
|
|
|
|
void
|
|
ff_rdt_calc_response_and_checksum(char response[41], char chksum[9],
|
|
const char *challenge)
|
|
{
|
|
int ch_len = strlen (challenge), i;
|
|
unsigned char zres[16],
|
|
buf[64] = { 0xa1, 0xe9, 0x14, 0x9d, 0x0e, 0x6b, 0x3b, 0x59 };
|
|
#define XOR_TABLE_SIZE 37
|
|
const unsigned char xor_table[XOR_TABLE_SIZE] = {
|
|
0x05, 0x18, 0x74, 0xd0, 0x0d, 0x09, 0x02, 0x53,
|
|
0xc0, 0x01, 0x05, 0x05, 0x67, 0x03, 0x19, 0x70,
|
|
0x08, 0x27, 0x66, 0x10, 0x10, 0x72, 0x08, 0x09,
|
|
0x63, 0x11, 0x03, 0x71, 0x08, 0x08, 0x70, 0x02,
|
|
0x10, 0x57, 0x05, 0x18, 0x54 };
|
|
|
|
/* some (length) checks */
|
|
if (ch_len == 40) /* what a hack... */
|
|
ch_len = 32;
|
|
else if (ch_len > 56)
|
|
ch_len = 56;
|
|
memcpy(buf + 8, challenge, ch_len);
|
|
|
|
/* xor challenge bytewise with xor_table */
|
|
for (i = 0; i < XOR_TABLE_SIZE; i++)
|
|
buf[8 + i] ^= xor_table[i];
|
|
|
|
av_md5_sum(zres, buf, 64);
|
|
ff_data_to_hex(response, zres, 16, 1);
|
|
|
|
/* add tail */
|
|
strcpy (response + 32, "01d0a8e3");
|
|
|
|
/* calculate checksum */
|
|
for (i = 0; i < 8; i++)
|
|
chksum[i] = response[i * 4];
|
|
chksum[8] = 0;
|
|
}
|
|
|
|
static int
|
|
rdt_load_mdpr (PayloadContext *rdt, AVStream *st, int rule_nr)
|
|
{
|
|
AVIOContext pb;
|
|
int size;
|
|
uint32_t tag;
|
|
|
|
/**
|
|
* Layout of the MLTI chunk:
|
|
* 4: MLTI
|
|
* 2: number of streams
|
|
* Then for each stream ([number_of_streams] times):
|
|
* 2: mdpr index
|
|
* 2: number of mdpr chunks
|
|
* Then for each mdpr chunk ([number_of_mdpr_chunks] times):
|
|
* 4: size
|
|
* [size]: data
|
|
* we skip MDPR chunks until we reach the one of the stream
|
|
* we're interested in, and forward that ([size]+[data]) to
|
|
* the RM demuxer to parse the stream-specific header data.
|
|
*/
|
|
if (!rdt->mlti_data)
|
|
return -1;
|
|
ffio_init_context(&pb, rdt->mlti_data, rdt->mlti_data_size, 0,
|
|
NULL, NULL, NULL, NULL);
|
|
tag = avio_rl32(&pb);
|
|
if (tag == MKTAG('M', 'L', 'T', 'I')) {
|
|
int num, chunk_nr;
|
|
|
|
/* read index of MDPR chunk numbers */
|
|
num = avio_rb16(&pb);
|
|
if (rule_nr < 0 || rule_nr >= num)
|
|
return -1;
|
|
avio_skip(&pb, rule_nr * 2);
|
|
chunk_nr = avio_rb16(&pb);
|
|
avio_skip(&pb, (num - 1 - rule_nr) * 2);
|
|
|
|
/* read MDPR chunks */
|
|
num = avio_rb16(&pb);
|
|
if (chunk_nr >= num)
|
|
return -1;
|
|
while (chunk_nr--)
|
|
avio_skip(&pb, avio_rb32(&pb));
|
|
size = avio_rb32(&pb);
|
|
} else {
|
|
size = rdt->mlti_data_size;
|
|
avio_seek(&pb, 0, SEEK_SET);
|
|
}
|
|
if (ff_rm_read_mdpr_codecdata(rdt->rmctx, &pb, st, rdt->rmst[st->index], size) < 0)
|
|
return -1;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/**
|
|
* Actual data handling.
|
|
*/
|
|
|
|
int
|
|
ff_rdt_parse_header(const uint8_t *buf, int len,
|
|
int *pset_id, int *pseq_no, int *pstream_id,
|
|
int *pis_keyframe, uint32_t *ptimestamp)
|
|
{
|
|
GetBitContext gb;
|
|
int consumed = 0, set_id, seq_no, stream_id, is_keyframe,
|
|
len_included, need_reliable;
|
|
uint32_t timestamp;
|
|
|
|
/* skip status packets */
|
|
while (len >= 5 && buf[1] == 0xFF /* status packet */) {
|
|
int pkt_len;
|
|
|
|
if (!(buf[0] & 0x80))
|
|
return -1; /* not followed by a data packet */
|
|
|
|
pkt_len = AV_RB16(buf+3);
|
|
buf += pkt_len;
|
|
len -= pkt_len;
|
|
consumed += pkt_len;
|
|
}
|
|
if (len < 16)
|
|
return -1;
|
|
/**
|
|
* Layout of the header (in bits):
|
|
* 1: len_included
|
|
* Flag indicating whether this header includes a length field;
|
|
* this can be used to concatenate multiple RDT packets in a
|
|
* single UDP/TCP data frame and is used to precede RDT data
|
|
* by stream status packets
|
|
* 1: need_reliable
|
|
* Flag indicating whether this header includes a "reliable
|
|
* sequence number"; these are apparently sequence numbers of
|
|
* data packets alone. For data packets, this flag is always
|
|
* set, according to the Real documentation [1]
|
|
* 5: set_id
|
|
* ID of a set of streams of identical content, possibly with
|
|
* different codecs or bitrates
|
|
* 1: is_reliable
|
|
* Flag set for certain streams deemed less tolerable for packet
|
|
* loss
|
|
* 16: seq_no
|
|
* Packet sequence number; if >=0xFF00, this is a non-data packet
|
|
* containing stream status info, the second byte indicates the
|
|
* type of status packet (see wireshark docs / source code [2])
|
|
* if (len_included) {
|
|
* 16: packet_len
|
|
* } else {
|
|
* packet_len = remainder of UDP/TCP frame
|
|
* }
|
|
* 1: is_back_to_back
|
|
* Back-to-Back flag; used for timing, set for one in every 10
|
|
* packets, according to the Real documentation [1]
|
|
* 1: is_slow_data
|
|
* Slow-data flag; currently unused, according to Real docs [1]
|
|
* 5: stream_id
|
|
* ID of the stream within this particular set of streams
|
|
* 1: is_no_keyframe
|
|
* Non-keyframe flag (unset if packet belongs to a keyframe)
|
|
* 32: timestamp (PTS)
|
|
* if (set_id == 0x1F) {
|
|
* 16: set_id (extended set-of-streams ID; see set_id)
|
|
* }
|
|
* if (need_reliable) {
|
|
* 16: reliable_seq_no
|
|
* Reliable sequence number (see need_reliable)
|
|
* }
|
|
* if (stream_id == 0x3F) {
|
|
* 16: stream_id (extended stream ID; see stream_id)
|
|
* }
|
|
* [1] https://protocol.helixcommunity.org/files/2005/devdocs/RDT_Feature_Level_20.txt
|
|
* [2] http://www.wireshark.org/docs/dfref/r/rdt.html and
|
|
* http://anonsvn.wireshark.org/viewvc/trunk/epan/dissectors/packet-rdt.c
|
|
*/
|
|
init_get_bits(&gb, buf, len << 3);
|
|
len_included = get_bits1(&gb);
|
|
need_reliable = get_bits1(&gb);
|
|
set_id = get_bits(&gb, 5);
|
|
skip_bits(&gb, 1);
|
|
seq_no = get_bits(&gb, 16);
|
|
if (len_included)
|
|
skip_bits(&gb, 16);
|
|
skip_bits(&gb, 2);
|
|
stream_id = get_bits(&gb, 5);
|
|
is_keyframe = !get_bits1(&gb);
|
|
timestamp = get_bits_long(&gb, 32);
|
|
if (set_id == 0x1f)
|
|
set_id = get_bits(&gb, 16);
|
|
if (need_reliable)
|
|
skip_bits(&gb, 16);
|
|
if (stream_id == 0x1f)
|
|
stream_id = get_bits(&gb, 16);
|
|
|
|
if (pset_id) *pset_id = set_id;
|
|
if (pseq_no) *pseq_no = seq_no;
|
|
if (pstream_id) *pstream_id = stream_id;
|
|
if (pis_keyframe) *pis_keyframe = is_keyframe;
|
|
if (ptimestamp) *ptimestamp = timestamp;
|
|
|
|
return consumed + (get_bits_count(&gb) >> 3);
|
|
}
|
|
|
|
/**< return 0 on packet, no more left, 1 on packet, 1 on partial packet... */
|
|
static int
|
|
rdt_parse_packet (AVFormatContext *ctx, PayloadContext *rdt, AVStream *st,
|
|
AVPacket *pkt, uint32_t *timestamp,
|
|
const uint8_t *buf, int len, int flags)
|
|
{
|
|
int seq = 1, res;
|
|
AVIOContext pb;
|
|
|
|
if (rdt->audio_pkt_cnt == 0) {
|
|
int pos;
|
|
|
|
ffio_init_context(&pb, buf, len, 0, NULL, NULL, NULL, NULL);
|
|
flags = (flags & RTP_FLAG_KEY) ? 2 : 0;
|
|
res = ff_rm_parse_packet (rdt->rmctx, &pb, st, rdt->rmst[st->index], len, pkt,
|
|
&seq, flags, *timestamp);
|
|
pos = avio_tell(&pb);
|
|
if (res < 0)
|
|
return res;
|
|
if (res > 0) {
|
|
if (st->codec->codec_id == CODEC_ID_AAC) {
|
|
memcpy (rdt->buffer, buf + pos, len - pos);
|
|
rdt->rmctx->pb = avio_alloc_context (rdt->buffer, len - pos, 0,
|
|
NULL, NULL, NULL, NULL);
|
|
}
|
|
goto get_cache;
|
|
}
|
|
} else {
|
|
get_cache:
|
|
rdt->audio_pkt_cnt =
|
|
ff_rm_retrieve_cache (rdt->rmctx, rdt->rmctx->pb,
|
|
st, rdt->rmst[st->index], pkt);
|
|
if (rdt->audio_pkt_cnt == 0 &&
|
|
st->codec->codec_id == CODEC_ID_AAC)
|
|
av_freep(&rdt->rmctx->pb);
|
|
}
|
|
pkt->stream_index = st->index;
|
|
pkt->pts = *timestamp;
|
|
|
|
return rdt->audio_pkt_cnt > 0;
|
|
}
|
|
|
|
int
|
|
ff_rdt_parse_packet(RDTDemuxContext *s, AVPacket *pkt,
|
|
uint8_t **bufptr, int len)
|
|
{
|
|
uint8_t *buf = bufptr ? *bufptr : NULL;
|
|
int seq_no, flags = 0, stream_id, set_id, is_keyframe;
|
|
uint32_t timestamp;
|
|
int rv= 0;
|
|
|
|
if (!s->parse_packet)
|
|
return -1;
|
|
|
|
if (!buf && s->prev_stream_id != -1) {
|
|
/* return the next packets, if any */
|
|
timestamp= 0; ///< Should not be used if buf is NULL, but should be set to the timestamp of the packet returned....
|
|
rv= s->parse_packet(s->ic, s->dynamic_protocol_context,
|
|
s->streams[s->prev_stream_id],
|
|
pkt, ×tamp, NULL, 0, flags);
|
|
return rv;
|
|
}
|
|
|
|
if (len < 12)
|
|
return -1;
|
|
rv = ff_rdt_parse_header(buf, len, &set_id, &seq_no, &stream_id, &is_keyframe, ×tamp);
|
|
if (rv < 0)
|
|
return rv;
|
|
if (is_keyframe &&
|
|
(set_id != s->prev_set_id || timestamp != s->prev_timestamp ||
|
|
stream_id != s->prev_stream_id)) {
|
|
flags |= RTP_FLAG_KEY;
|
|
s->prev_set_id = set_id;
|
|
s->prev_timestamp = timestamp;
|
|
}
|
|
s->prev_stream_id = stream_id;
|
|
buf += rv;
|
|
len -= rv;
|
|
|
|
if (s->prev_stream_id >= s->n_streams) {
|
|
s->prev_stream_id = -1;
|
|
return -1;
|
|
}
|
|
|
|
rv = s->parse_packet(s->ic, s->dynamic_protocol_context,
|
|
s->streams[s->prev_stream_id],
|
|
pkt, ×tamp, buf, len, flags);
|
|
|
|
return rv;
|
|
}
|
|
|
|
void
|
|
ff_rdt_subscribe_rule (char *cmd, int size,
|
|
int stream_nr, int rule_nr)
|
|
{
|
|
av_strlcatf(cmd, size, "stream=%d;rule=%d,stream=%d;rule=%d",
|
|
stream_nr, rule_nr * 2, stream_nr, rule_nr * 2 + 1);
|
|
}
|
|
|
|
static unsigned char *
|
|
rdt_parse_b64buf (unsigned int *target_len, const char *p)
|
|
{
|
|
unsigned char *target;
|
|
int len = strlen(p);
|
|
if (*p == '\"') {
|
|
p++;
|
|
len -= 2; /* skip embracing " at start/end */
|
|
}
|
|
*target_len = len * 3 / 4;
|
|
target = av_mallocz(*target_len + FF_INPUT_BUFFER_PADDING_SIZE);
|
|
av_base64_decode(target, p, *target_len);
|
|
return target;
|
|
}
|
|
|
|
static int
|
|
rdt_parse_sdp_line (AVFormatContext *s, int st_index,
|
|
PayloadContext *rdt, const char *line)
|
|
{
|
|
AVStream *stream = s->streams[st_index];
|
|
const char *p = line;
|
|
|
|
if (av_strstart(p, "OpaqueData:buffer;", &p)) {
|
|
rdt->mlti_data = rdt_parse_b64buf(&rdt->mlti_data_size, p);
|
|
} else if (av_strstart(p, "StartTime:integer;", &p))
|
|
stream->first_dts = atoi(p);
|
|
else if (av_strstart(p, "ASMRuleBook:string;", &p)) {
|
|
int n, first = -1;
|
|
|
|
for (n = 0; n < s->nb_streams; n++)
|
|
if (s->streams[n]->id == stream->id) {
|
|
int count = s->streams[n]->index + 1;
|
|
if (first == -1) first = n;
|
|
if (rdt->nb_rmst < count) {
|
|
RMStream **rmst= av_realloc(rdt->rmst, count*sizeof(*rmst));
|
|
if (!rmst)
|
|
return AVERROR(ENOMEM);
|
|
memset(rmst + rdt->nb_rmst, 0,
|
|
(count - rdt->nb_rmst) * sizeof(*rmst));
|
|
rdt->rmst = rmst;
|
|
rdt->nb_rmst = count;
|
|
}
|
|
rdt->rmst[s->streams[n]->index] = ff_rm_alloc_rmstream();
|
|
rdt_load_mdpr(rdt, s->streams[n], (n - first) * 2);
|
|
|
|
if (s->streams[n]->codec->codec_id == CODEC_ID_AAC)
|
|
s->streams[n]->codec->frame_size = 1; // FIXME
|
|
}
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
static void
|
|
real_parse_asm_rule(AVStream *st, const char *p, const char *end)
|
|
{
|
|
do {
|
|
/* can be either averagebandwidth= or AverageBandwidth= */
|
|
if (sscanf(p, " %*1[Aa]verage%*1[Bb]andwidth=%d", &st->codec->bit_rate) == 1)
|
|
break;
|
|
if (!(p = strchr(p, ',')) || p > end)
|
|
p = end;
|
|
p++;
|
|
} while (p < end);
|
|
}
|
|
|
|
static AVStream *
|
|
add_dstream(AVFormatContext *s, AVStream *orig_st)
|
|
{
|
|
AVStream *st;
|
|
|
|
if (!(st = av_new_stream(s, orig_st->id)))
|
|
return NULL;
|
|
st->codec->codec_type = orig_st->codec->codec_type;
|
|
st->first_dts = orig_st->first_dts;
|
|
|
|
return st;
|
|
}
|
|
|
|
static void
|
|
real_parse_asm_rulebook(AVFormatContext *s, AVStream *orig_st,
|
|
const char *p)
|
|
{
|
|
const char *end;
|
|
int n_rules = 0, odd = 0;
|
|
AVStream *st;
|
|
|
|
/**
|
|
* The ASMRuleBook contains a list of comma-separated strings per rule,
|
|
* and each rule is separated by a ;. The last one also has a ; at the
|
|
* end so we can use it as delimiter.
|
|
* Every rule occurs twice, once for when the RTSP packet header marker
|
|
* is set and once for if it isn't. We only read the first because we
|
|
* don't care much (that's what the "odd" variable is for).
|
|
* Each rule contains a set of one or more statements, optionally
|
|
* preceeded by a single condition. If there's a condition, the rule
|
|
* starts with a '#'. Multiple conditions are merged between brackets,
|
|
* so there are never multiple conditions spread out over separate
|
|
* statements. Generally, these conditions are bitrate limits (min/max)
|
|
* for multi-bitrate streams.
|
|
*/
|
|
if (*p == '\"') p++;
|
|
while (1) {
|
|
if (!(end = strchr(p, ';')))
|
|
break;
|
|
if (!odd && end != p) {
|
|
if (n_rules > 0)
|
|
st = add_dstream(s, orig_st);
|
|
else
|
|
st = orig_st;
|
|
if (!st)
|
|
break;
|
|
real_parse_asm_rule(st, p, end);
|
|
n_rules++;
|
|
}
|
|
p = end + 1;
|
|
odd ^= 1;
|
|
}
|
|
}
|
|
|
|
void
|
|
ff_real_parse_sdp_a_line (AVFormatContext *s, int stream_index,
|
|
const char *line)
|
|
{
|
|
const char *p = line;
|
|
|
|
if (av_strstart(p, "ASMRuleBook:string;", &p))
|
|
real_parse_asm_rulebook(s, s->streams[stream_index], p);
|
|
}
|
|
|
|
static PayloadContext *
|
|
rdt_new_context (void)
|
|
{
|
|
PayloadContext *rdt = av_mallocz(sizeof(PayloadContext));
|
|
|
|
av_open_input_stream(&rdt->rmctx, NULL, "", &ff_rdt_demuxer, NULL);
|
|
|
|
return rdt;
|
|
}
|
|
|
|
static void
|
|
rdt_free_context (PayloadContext *rdt)
|
|
{
|
|
int i;
|
|
|
|
for (i = 0; i < rdt->nb_rmst; i++)
|
|
if (rdt->rmst[i]) {
|
|
ff_rm_free_rmstream(rdt->rmst[i]);
|
|
av_freep(&rdt->rmst[i]);
|
|
}
|
|
if (rdt->rmctx)
|
|
av_close_input_stream(rdt->rmctx);
|
|
av_freep(&rdt->mlti_data);
|
|
av_freep(&rdt->rmst);
|
|
av_free(rdt);
|
|
}
|
|
|
|
#define RDT_HANDLER(n, s, t) \
|
|
static RTPDynamicProtocolHandler ff_rdt_ ## n ## _handler = { \
|
|
.enc_name = s, \
|
|
.codec_type = t, \
|
|
.codec_id = CODEC_ID_NONE, \
|
|
.parse_sdp_a_line = rdt_parse_sdp_line, \
|
|
.open = rdt_new_context, \
|
|
.close = rdt_free_context, \
|
|
.parse_packet = rdt_parse_packet \
|
|
}
|
|
|
|
RDT_HANDLER(live_video, "x-pn-multirate-realvideo-live", AVMEDIA_TYPE_VIDEO);
|
|
RDT_HANDLER(live_audio, "x-pn-multirate-realaudio-live", AVMEDIA_TYPE_AUDIO);
|
|
RDT_HANDLER(video, "x-pn-realvideo", AVMEDIA_TYPE_VIDEO);
|
|
RDT_HANDLER(audio, "x-pn-realaudio", AVMEDIA_TYPE_AUDIO);
|
|
|
|
void av_register_rdt_dynamic_payload_handlers(void)
|
|
{
|
|
ff_register_dynamic_payload_handler(&ff_rdt_video_handler);
|
|
ff_register_dynamic_payload_handler(&ff_rdt_audio_handler);
|
|
ff_register_dynamic_payload_handler(&ff_rdt_live_video_handler);
|
|
ff_register_dynamic_payload_handler(&ff_rdt_live_audio_handler);
|
|
}
|