2 * muxing functions for use within Libav
3 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard
5 * This file is part of Libav.
7 * Libav is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * Libav is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 #include "avio_internal.h"
25 #include "libavcodec/internal.h"
26 #include "libavcodec/bytestream.h"
27 #include "libavutil/opt.h"
28 #include "libavutil/dict.h"
29 #include "libavutil/pixdesc.h"
32 #include "libavutil/avassert.h"
33 #include "libavutil/avstring.h"
34 #include "libavutil/internal.h"
35 #include "libavutil/mathematics.h"
36 #include "libavutil/parseutils.h"
37 #include "libavutil/time.h"
39 #include "audiointerleave.h"
51 * muxing functions for use within Libav
54 /* fraction handling */
57 * f = val + (num / den) + 0.5.
59 * 'num' is normalized so that it is such as 0 <= num < den.
61 * @param f fractional number
62 * @param val integer value
63 * @param num must be >= 0
64 * @param den must be >= 1
66 static void frac_init(AVFrac *f, int64_t val, int64_t num, int64_t den)
79 * Fractional addition to f: f = f + (incr / f->den).
81 * @param f fractional number
82 * @param incr increment, can be positive or negative
84 static void frac_add(AVFrac *f, int64_t incr)
97 } else if (num >= den) {
104 static int validate_codec_tag(AVFormatContext *s, AVStream *st)
106 const AVCodecTag *avctag;
108 enum AVCodecID id = AV_CODEC_ID_NONE;
109 unsigned int tag = 0;
112 * Check that tag + id is in the table
113 * If neither is in the table -> OK
114 * If tag is in the table with another id -> FAIL
115 * If id is in the table with another tag -> FAIL unless strict < normal
117 for (n = 0; s->oformat->codec_tag[n]; n++) {
118 avctag = s->oformat->codec_tag[n];
119 while (avctag->id != AV_CODEC_ID_NONE) {
120 if (avpriv_toupper4(avctag->tag) == avpriv_toupper4(st->codec->codec_tag)) {
122 if (id == st->codec->codec_id)
125 if (avctag->id == st->codec->codec_id)
130 if (id != AV_CODEC_ID_NONE)
132 if (tag && (st->codec->strict_std_compliance >= FF_COMPLIANCE_NORMAL))
138 static int init_muxer(AVFormatContext *s, AVDictionary **options)
142 AVDictionary *tmp = NULL;
143 AVCodecContext *codec = NULL;
144 AVOutputFormat *of = s->oformat;
147 av_dict_copy(&tmp, *options, 0);
149 if ((ret = av_opt_set_dict(s, &tmp)) < 0)
152 // some sanity checks
153 if (s->nb_streams == 0 && !(of->flags & AVFMT_NOSTREAMS)) {
154 av_log(s, AV_LOG_ERROR, "no streams\n");
155 ret = AVERROR(EINVAL);
159 for (i = 0; i < s->nb_streams; i++) {
163 switch (codec->codec_type) {
164 case AVMEDIA_TYPE_AUDIO:
165 if (codec->sample_rate <= 0) {
166 av_log(s, AV_LOG_ERROR, "sample rate not set\n");
167 ret = AVERROR(EINVAL);
170 if (!codec->block_align)
171 codec->block_align = codec->channels *
172 av_get_bits_per_sample(codec->codec_id) >> 3;
174 case AVMEDIA_TYPE_VIDEO:
175 if (codec->time_base.num <= 0 ||
176 codec->time_base.den <= 0) { //FIXME audio too?
177 av_log(s, AV_LOG_ERROR, "time base not set\n");
178 ret = AVERROR(EINVAL);
182 if ((codec->width <= 0 || codec->height <= 0) &&
183 !(of->flags & AVFMT_NODIMENSIONS)) {
184 av_log(s, AV_LOG_ERROR, "dimensions not set\n");
185 ret = AVERROR(EINVAL);
189 if (av_cmp_q(st->sample_aspect_ratio,
190 codec->sample_aspect_ratio)) {
191 if (st->sample_aspect_ratio.num != 0 &&
192 st->sample_aspect_ratio.den != 0 &&
193 codec->sample_aspect_ratio.den != 0 &&
194 codec->sample_aspect_ratio.den != 0) {
195 av_log(s, AV_LOG_ERROR, "Aspect ratio mismatch between muxer "
196 "(%d/%d) and encoder layer (%d/%d)\n",
197 st->sample_aspect_ratio.num, st->sample_aspect_ratio.den,
198 codec->sample_aspect_ratio.num,
199 codec->sample_aspect_ratio.den);
200 ret = AVERROR(EINVAL);
208 if (codec->codec_tag &&
209 codec->codec_id == AV_CODEC_ID_RAWVIDEO &&
210 !av_codec_get_tag(of->codec_tag, codec->codec_id) &&
211 !validate_codec_tag(s, st)) {
212 // the current rawvideo encoding system ends up setting
213 // the wrong codec_tag for avi, we override it here
214 codec->codec_tag = 0;
216 if (codec->codec_tag) {
217 if (!validate_codec_tag(s, st)) {
219 av_get_codec_tag_string(tagbuf, sizeof(tagbuf), codec->codec_tag);
220 av_log(s, AV_LOG_ERROR,
221 "Tag %s/0x%08x incompatible with output codec id '%d'\n",
222 tagbuf, codec->codec_tag, codec->codec_id);
223 ret = AVERROR_INVALIDDATA;
227 codec->codec_tag = av_codec_get_tag(of->codec_tag, codec->codec_id);
230 if (of->flags & AVFMT_GLOBALHEADER &&
231 !(codec->flags & CODEC_FLAG_GLOBAL_HEADER))
232 av_log(s, AV_LOG_WARNING,
233 "Codec for stream %d does not use global headers "
234 "but container format requires global headers\n", i);
236 if (codec->codec_type != AVMEDIA_TYPE_ATTACHMENT)
237 s->internal->nb_interleaved_streams++;
240 if (!s->priv_data && of->priv_data_size > 0) {
241 s->priv_data = av_mallocz(of->priv_data_size);
243 ret = AVERROR(ENOMEM);
246 if (of->priv_class) {
247 *(const AVClass **)s->priv_data = of->priv_class;
248 av_opt_set_defaults(s->priv_data);
249 if ((ret = av_opt_set_dict(s->priv_data, &tmp)) < 0)
254 /* set muxer identification string */
255 if (s->nb_streams && !(s->streams[0]->codec->flags & CODEC_FLAG_BITEXACT)) {
256 av_dict_set(&s->metadata, "encoder", LIBAVFORMAT_IDENT, 0);
260 av_dict_free(options);
271 static int init_pts(AVFormatContext *s)
276 /* init PTS generation */
277 for (i = 0; i < s->nb_streams; i++) {
278 int64_t den = AV_NOPTS_VALUE;
281 switch (st->codec->codec_type) {
282 case AVMEDIA_TYPE_AUDIO:
283 den = (int64_t)st->time_base.num * st->codec->sample_rate;
285 case AVMEDIA_TYPE_VIDEO:
286 den = (int64_t)st->time_base.num * st->codec->time_base.den;
291 if (den != AV_NOPTS_VALUE) {
293 return AVERROR_INVALIDDATA;
295 frac_init(&st->pts, 0, 0, den);
302 int avformat_write_header(AVFormatContext *s, AVDictionary **options)
306 if (ret = init_muxer(s, options))
309 if (s->oformat->write_header) {
310 ret = s->oformat->write_header(s);
315 if ((ret = init_pts(s)) < 0)
321 //FIXME merge with compute_pkt_fields
322 static int compute_pkt_fields2(AVFormatContext *s, AVStream *st, AVPacket *pkt)
324 int delay = FFMAX(st->codec->has_b_frames, !!st->codec->max_b_frames);
325 int num, den, frame_size, i;
327 av_dlog(s, "compute_pkt_fields2: pts:%" PRId64 " dts:%" PRId64 " cur_dts:%" PRId64 " b:%d size:%d st:%d\n",
328 pkt->pts, pkt->dts, st->cur_dts, delay, pkt->size, pkt->stream_index);
330 /* if(pkt->pts == AV_NOPTS_VALUE && pkt->dts == AV_NOPTS_VALUE)
331 * return AVERROR(EINVAL);*/
334 if (pkt->duration == 0) {
335 ff_compute_frame_duration(&num, &den, st, NULL, pkt);
337 pkt->duration = av_rescale(1, num * (int64_t)st->time_base.den * st->codec->ticks_per_frame, den * (int64_t)st->time_base.num);
341 if (pkt->pts == AV_NOPTS_VALUE && pkt->dts != AV_NOPTS_VALUE && delay == 0)
344 //XXX/FIXME this is a temporary hack until all encoders output pts
345 if ((pkt->pts == 0 || pkt->pts == AV_NOPTS_VALUE) && pkt->dts == AV_NOPTS_VALUE && !delay) {
347 // pkt->pts= st->cur_dts;
348 pkt->pts = st->pts.val;
351 //calculate dts from pts
352 if (pkt->pts != AV_NOPTS_VALUE && pkt->dts == AV_NOPTS_VALUE && delay <= MAX_REORDER_DELAY) {
353 st->pts_buffer[0] = pkt->pts;
354 for (i = 1; i < delay + 1 && st->pts_buffer[i] == AV_NOPTS_VALUE; i++)
355 st->pts_buffer[i] = pkt->pts + (i - delay - 1) * pkt->duration;
356 for (i = 0; i<delay && st->pts_buffer[i] > st->pts_buffer[i + 1]; i++)
357 FFSWAP(int64_t, st->pts_buffer[i], st->pts_buffer[i + 1]);
359 pkt->dts = st->pts_buffer[0];
362 if (st->cur_dts && st->cur_dts != AV_NOPTS_VALUE &&
363 ((!(s->oformat->flags & AVFMT_TS_NONSTRICT) &&
364 st->cur_dts >= pkt->dts) || st->cur_dts > pkt->dts)) {
365 av_log(s, AV_LOG_ERROR,
366 "Application provided invalid, non monotonically increasing dts to muxer in stream %d: %" PRId64 " >= %" PRId64 "\n",
367 st->index, st->cur_dts, pkt->dts);
368 return AVERROR(EINVAL);
370 if (pkt->dts != AV_NOPTS_VALUE && pkt->pts != AV_NOPTS_VALUE && pkt->pts < pkt->dts) {
371 av_log(s, AV_LOG_ERROR, "pts < dts in stream %d\n", st->index);
372 return AVERROR(EINVAL);
375 av_dlog(s, "av_write_frame: pts2:%"PRId64" dts2:%"PRId64"\n",
377 st->cur_dts = pkt->dts;
378 st->pts.val = pkt->dts;
381 switch (st->codec->codec_type) {
382 case AVMEDIA_TYPE_AUDIO:
383 frame_size = ff_get_audio_frame_size(st->codec, pkt->size, 1);
385 /* HACK/FIXME, we skip the initial 0 size packets as they are most
386 * likely equal to the encoder delay, but it would be better if we
387 * had the real timestamps from the encoder */
388 if (frame_size >= 0 && (pkt->size || st->pts.num != st->pts.den >> 1 || st->pts.val)) {
389 frac_add(&st->pts, (int64_t)st->time_base.den * frame_size);
392 case AVMEDIA_TYPE_VIDEO:
393 frac_add(&st->pts, (int64_t)st->time_base.den * st->codec->time_base.num);
402 * FIXME: this function should NEVER get undefined pts/dts beside when the
403 * AVFMT_NOTIMESTAMPS is set.
404 * Those additional safety checks should be dropped once the correct checks
405 * are set in the callers.
408 static int write_packet(AVFormatContext *s, AVPacket *pkt)
411 if (!(s->oformat->flags & (AVFMT_TS_NEGATIVE | AVFMT_NOTIMESTAMPS))) {
412 AVRational time_base = s->streams[pkt->stream_index]->time_base;
415 if (!s->offset && pkt->dts != AV_NOPTS_VALUE && pkt->dts < 0) {
416 s->offset = -pkt->dts;
417 s->offset_timebase = time_base;
420 offset = av_rescale_q(s->offset, s->offset_timebase, time_base);
422 if (pkt->dts != AV_NOPTS_VALUE)
424 if (pkt->pts != AV_NOPTS_VALUE)
427 ret = s->oformat->write_packet(s, pkt);
429 if (s->pb && ret >= 0 && s->flags & AVFMT_FLAG_FLUSH_PACKETS)
435 static int check_packet(AVFormatContext *s, AVPacket *pkt)
440 if (pkt->stream_index < 0 || pkt->stream_index >= s->nb_streams) {
441 av_log(s, AV_LOG_ERROR, "Invalid packet stream index: %d\n",
443 return AVERROR(EINVAL);
446 if (s->streams[pkt->stream_index]->codec->codec_type == AVMEDIA_TYPE_ATTACHMENT) {
447 av_log(s, AV_LOG_ERROR, "Received a packet for an attachment stream.\n");
448 return AVERROR(EINVAL);
454 int av_write_frame(AVFormatContext *s, AVPacket *pkt)
458 ret = check_packet(s, pkt);
463 if (s->oformat->flags & AVFMT_ALLOW_FLUSH)
464 return s->oformat->write_packet(s, pkt);
468 ret = compute_pkt_fields2(s, s->streams[pkt->stream_index], pkt);
470 if (ret < 0 && !(s->oformat->flags & AVFMT_NOTIMESTAMPS))
473 ret = write_packet(s, pkt);
476 s->streams[pkt->stream_index]->nb_frames++;
480 void ff_interleave_add_packet(AVFormatContext *s, AVPacket *pkt,
481 int (*compare)(AVFormatContext *, AVPacket *, AVPacket *))
483 AVPacketList **next_point, *this_pktl;
485 this_pktl = av_mallocz(sizeof(AVPacketList));
486 this_pktl->pkt = *pkt;
487 #if FF_API_DESTRUCT_PACKET
488 FF_DISABLE_DEPRECATION_WARNINGS
489 pkt->destruct = NULL; // do not free original but only the copy
490 FF_ENABLE_DEPRECATION_WARNINGS
493 av_dup_packet(&this_pktl->pkt); // duplicate the packet if it uses non-alloced memory
495 if (s->streams[pkt->stream_index]->last_in_packet_buffer) {
496 next_point = &(s->streams[pkt->stream_index]->last_in_packet_buffer->next);
498 next_point = &s->packet_buffer;
501 if (compare(s, &s->packet_buffer_end->pkt, pkt)) {
502 while (!compare(s, &(*next_point)->pkt, pkt))
503 next_point = &(*next_point)->next;
506 next_point = &(s->packet_buffer_end->next);
509 assert(!*next_point);
511 s->packet_buffer_end = this_pktl;
514 this_pktl->next = *next_point;
516 s->streams[pkt->stream_index]->last_in_packet_buffer =
517 *next_point = this_pktl;
520 static int interleave_compare_dts(AVFormatContext *s, AVPacket *next,
523 AVStream *st = s->streams[pkt->stream_index];
524 AVStream *st2 = s->streams[next->stream_index];
525 int comp = av_compare_ts(next->dts, st2->time_base, pkt->dts,
529 return pkt->stream_index < next->stream_index;
533 int ff_interleave_packet_per_dts(AVFormatContext *s, AVPacket *out,
534 AVPacket *pkt, int flush)
537 int stream_count = 0;
541 ff_interleave_add_packet(s, pkt, interleave_compare_dts);
544 if (s->max_interleave_delta > 0 && s->packet_buffer && !flush) {
545 AVPacket *top_pkt = &s->packet_buffer->pkt;
546 int64_t delta_dts = INT64_MIN;
547 int64_t top_dts = av_rescale_q(top_pkt->dts,
548 s->streams[top_pkt->stream_index]->time_base,
551 for (i = 0; i < s->nb_streams; i++) {
553 const AVPacketList *last = s->streams[i]->last_in_packet_buffer;
558 last_dts = av_rescale_q(last->pkt.dts,
559 s->streams[i]->time_base,
561 delta_dts = FFMAX(delta_dts, last_dts - top_dts);
565 if (delta_dts > s->max_interleave_delta) {
566 av_log(s, AV_LOG_DEBUG,
567 "Delay between the first packet and last packet in the "
568 "muxing queue is %"PRId64" > %"PRId64": forcing output\n",
569 delta_dts, s->max_interleave_delta);
573 for (i = 0; i < s->nb_streams; i++)
574 stream_count += !!s->streams[i]->last_in_packet_buffer;
578 if (stream_count && (s->internal->nb_interleaved_streams == stream_count || flush)) {
579 pktl = s->packet_buffer;
582 s->packet_buffer = pktl->next;
583 if (!s->packet_buffer)
584 s->packet_buffer_end = NULL;
586 if (s->streams[out->stream_index]->last_in_packet_buffer == pktl)
587 s->streams[out->stream_index]->last_in_packet_buffer = NULL;
597 * Interleave an AVPacket correctly so it can be muxed.
598 * @param out the interleaved packet will be output here
599 * @param in the input packet
600 * @param flush 1 if no further packets are available as input and all
601 * remaining packets should be output
602 * @return 1 if a packet was output, 0 if no packet could be output,
603 * < 0 if an error occurred
605 static int interleave_packet(AVFormatContext *s, AVPacket *out, AVPacket *in, int flush)
607 if (s->oformat->interleave_packet) {
608 int ret = s->oformat->interleave_packet(s, out, in, flush);
613 return ff_interleave_packet_per_dts(s, out, in, flush);
616 int av_interleaved_write_frame(AVFormatContext *s, AVPacket *pkt)
620 ret = check_packet(s, pkt);
625 AVStream *st = s->streams[pkt->stream_index];
627 //FIXME/XXX/HACK drop zero sized packets
628 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO && pkt->size == 0) {
633 av_dlog(s, "av_interleaved_write_frame size:%d dts:%" PRId64 " pts:%" PRId64 "\n",
634 pkt->size, pkt->dts, pkt->pts);
635 if ((ret = compute_pkt_fields2(s, st, pkt)) < 0 && !(s->oformat->flags & AVFMT_NOTIMESTAMPS))
638 if (pkt->dts == AV_NOPTS_VALUE && !(s->oformat->flags & AVFMT_NOTIMESTAMPS)) {
639 ret = AVERROR(EINVAL);
643 av_dlog(s, "av_interleaved_write_frame FLUSH\n");
649 int ret = interleave_packet(s, &opkt, pkt, flush);
651 memset(pkt, 0, sizeof(*pkt));
655 if (ret <= 0) //FIXME cleanup needed for ret<0 ?
658 ret = write_packet(s, &opkt);
660 s->streams[opkt.stream_index]->nb_frames++;
662 av_free_packet(&opkt);
668 av_packet_unref(pkt);
672 int av_write_trailer(AVFormatContext *s)
678 ret = interleave_packet(s, &pkt, NULL, 1);
679 if (ret < 0) //FIXME cleanup needed for ret<0 ?
684 ret = write_packet(s, &pkt);
686 s->streams[pkt.stream_index]->nb_frames++;
688 av_free_packet(&pkt);
694 if (s->oformat->write_trailer)
695 ret = s->oformat->write_trailer(s);
697 if (!(s->oformat->flags & AVFMT_NOFILE))
701 for (i = 0; i < s->nb_streams; i++) {
702 av_freep(&s->streams[i]->priv_data);
703 av_freep(&s->streams[i]->index_entries);
705 if (s->oformat->priv_class)
706 av_opt_free(s->priv_data);
707 av_freep(&s->priv_data);
711 int ff_write_chained(AVFormatContext *dst, int dst_stream, AVPacket *pkt,
712 AVFormatContext *src)
717 local_pkt.stream_index = dst_stream;
718 if (pkt->pts != AV_NOPTS_VALUE)
719 local_pkt.pts = av_rescale_q(pkt->pts,
720 src->streams[pkt->stream_index]->time_base,
721 dst->streams[dst_stream]->time_base);
722 if (pkt->dts != AV_NOPTS_VALUE)
723 local_pkt.dts = av_rescale_q(pkt->dts,
724 src->streams[pkt->stream_index]->time_base,
725 dst->streams[dst_stream]->time_base);
726 return av_write_frame(dst, &local_pkt);