You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

1460 lines
47 KiB

/*
* muxing functions for use within FFmpeg
* Copyright (c) 2000, 2001, 2002 Fabrice Bellard
*
* This file is part of FFmpeg.
*
* FFmpeg 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.
*
* FFmpeg 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 FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "avformat.h"
#include "internal.h"
#include "mux.h"
#include "version.h"
#include "libavcodec/bsf.h"
#include "libavcodec/internal.h"
#include "libavcodec/packet_internal.h"
#include "libavutil/opt.h"
#include "libavutil/dict.h"
#include "libavutil/timestamp.h"
#include "libavutil/avassert.h"
#include "libavutil/internal.h"
#include "libavutil/mathematics.h"
/**
* @file
* muxing functions for use within libavformat
*/
/* fraction handling */
/**
* f = val + (num / den) + 0.5.
*
* 'num' is normalized so that it is such as 0 <= num < den.
*
* @param f fractional number
* @param val integer value
* @param num must be >= 0
* @param den must be >= 1
*/
static void frac_init(FFFrac *f, int64_t val, int64_t num, int64_t den)
{
num += (den >> 1);
if (num >= den) {
val += num / den;
num = num % den;
}
f->val = val;
f->num = num;
f->den = den;
}
/**
* Fractional addition to f: f = f + (incr / f->den).
*
* @param f fractional number
* @param incr increment, can be positive or negative
*/
static void frac_add(FFFrac *f, int64_t incr)
{
int64_t num, den;
num = f->num + incr;
den = f->den;
if (num < 0) {
f->val += num / den;
num = num % den;
if (num < 0) {
num += den;
f->val--;
}
} else if (num >= den) {
f->val += num / den;
num = num % den;
}
f->num = num;
}
int avformat_alloc_output_context2(AVFormatContext **avctx, const AVOutputFormat *oformat,
const char *format, const char *filename)
{
AVFormatContext *s = avformat_alloc_context();
int ret = 0;
*avctx = NULL;
if (!s)
goto nomem;
if (!oformat) {
if (format) {
oformat = av_guess_format(format, NULL, NULL);
if (!oformat) {
av_log(s, AV_LOG_ERROR, "Requested output format '%s' is not a suitable output format\n", format);
ret = AVERROR(EINVAL);
goto error;
}
} else {
oformat = av_guess_format(NULL, filename, NULL);
if (!oformat) {
ret = AVERROR(EINVAL);
av_log(s, AV_LOG_ERROR, "Unable to find a suitable output format for '%s'\n",
filename);
goto error;
}
}
}
s->oformat = oformat;
if (s->oformat->priv_data_size > 0) {
s->priv_data = av_mallocz(s->oformat->priv_data_size);
if (!s->priv_data)
goto nomem;
if (s->oformat->priv_class) {
*(const AVClass**)s->priv_data= s->oformat->priv_class;
av_opt_set_defaults(s->priv_data);
}
} else
s->priv_data = NULL;
if (filename) {
if (!(s->url = av_strdup(filename)))
goto nomem;
}
*avctx = s;
return 0;
nomem:
av_log(s, AV_LOG_ERROR, "Out of memory\n");
ret = AVERROR(ENOMEM);
error:
avformat_free_context(s);
return ret;
}
static int validate_codec_tag(const AVFormatContext *s, const AVStream *st)
{
const AVCodecTag *avctag;
enum AVCodecID id = AV_CODEC_ID_NONE;
unsigned uppercase_tag = ff_toupper4(st->codecpar->codec_tag);
int64_t tag = -1;
/**
* Check that tag + id is in the table
* If neither is in the table -> OK
* If tag is in the table with another id -> FAIL
* If id is in the table with another tag -> FAIL unless strict < normal
*/
for (int n = 0; s->oformat->codec_tag[n]; n++) {
avctag = s->oformat->codec_tag[n];
while (avctag->id != AV_CODEC_ID_NONE) {
if (ff_toupper4(avctag->tag) == uppercase_tag) {
id = avctag->id;
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (id == st->codecpar->codec_id)
return 1;
}
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (avctag->id == st->codecpar->codec_id)
tag = avctag->tag;
avctag++;
}
}
if (id != AV_CODEC_ID_NONE)
return 0;
if (tag >= 0 && (s->strict_std_compliance >= FF_COMPLIANCE_NORMAL))
return 0;
return 1;
}
static int init_muxer(AVFormatContext *s, AVDictionary **options)
{
FFFormatContext *const si = ffformatcontext(s);
AVDictionary *tmp = NULL;
const AVOutputFormat *of = s->oformat;
AVDictionaryEntry *e;
int ret = 0;
if (options)
av_dict_copy(&tmp, *options, 0);
if ((ret = av_opt_set_dict(s, &tmp)) < 0)
goto fail;
if (s->priv_data && s->oformat->priv_class && *(const AVClass**)s->priv_data==s->oformat->priv_class &&
(ret = av_opt_set_dict2(s->priv_data, &tmp, AV_OPT_SEARCH_CHILDREN)) < 0)
goto fail;
if (!s->url && !(s->url = av_strdup(""))) {
ret = AVERROR(ENOMEM);
goto fail;
}
// some sanity checks
if (s->nb_streams == 0 && !(of->flags & AVFMT_NOSTREAMS)) {
av_log(s, AV_LOG_ERROR, "No streams to mux were specified\n");
ret = AVERROR(EINVAL);
goto fail;
}
for (unsigned i = 0; i < s->nb_streams; i++) {
AVStream *const st = s->streams[i];
FFStream *const sti = ffstream(st);
AVCodecParameters *const par = st->codecpar;
const AVCodecDescriptor *desc;
if (!st->time_base.num) {
/* fall back on the default timebase values */
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (par->codec_type == AVMEDIA_TYPE_AUDIO && par->sample_rate)
avpriv_set_pts_info(st, 64, 1, par->sample_rate);
else
avpriv_set_pts_info(st, 33, 1, 90000);
}
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
switch (par->codec_type) {
case AVMEDIA_TYPE_AUDIO:
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (par->sample_rate <= 0) {
av_log(s, AV_LOG_ERROR, "sample rate not set\n");
ret = AVERROR(EINVAL);
goto fail;
}
#if FF_API_OLD_CHANNEL_LAYOUT
FF_DISABLE_DEPRECATION_WARNINGS
/* if the caller is using the deprecated channel layout API,
* convert it to the new style */
if (!par->ch_layout.nb_channels &&
par->channels) {
if (par->channel_layout) {
av_channel_layout_from_mask(&par->ch_layout, par->channel_layout);
} else {
par->ch_layout.order = AV_CHANNEL_ORDER_UNSPEC;
par->ch_layout.nb_channels = par->channels;
}
}
FF_ENABLE_DEPRECATION_WARNINGS
#endif
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (!par->block_align)
par->block_align = par->ch_layout.nb_channels *
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
av_get_bits_per_sample(par->codec_id) >> 3;
break;
case AVMEDIA_TYPE_VIDEO:
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if ((par->width <= 0 || par->height <= 0) &&
!(of->flags & AVFMT_NODIMENSIONS)) {
av_log(s, AV_LOG_ERROR, "dimensions not set\n");
ret = AVERROR(EINVAL);
goto fail;
}
if (av_cmp_q(st->sample_aspect_ratio, par->sample_aspect_ratio)
&& fabs(av_q2d(st->sample_aspect_ratio) - av_q2d(par->sample_aspect_ratio)) > 0.004*av_q2d(st->sample_aspect_ratio)
) {
if (st->sample_aspect_ratio.num != 0 &&
st->sample_aspect_ratio.den != 0 &&
par->sample_aspect_ratio.num != 0 &&
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
par->sample_aspect_ratio.den != 0) {
av_log(s, AV_LOG_ERROR, "Aspect ratio mismatch between muxer "
"(%d/%d) and encoder layer (%d/%d)\n",
st->sample_aspect_ratio.num, st->sample_aspect_ratio.den,
par->sample_aspect_ratio.num,
par->sample_aspect_ratio.den);
ret = AVERROR(EINVAL);
goto fail;
}
}
break;
}
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
desc = avcodec_descriptor_get(par->codec_id);
if (desc && desc->props & AV_CODEC_PROP_REORDER)
sti->reorder = 1;
sti->is_intra_only = ff_is_intra_only(par->codec_id);
avformat/mux: Set AV_PKT_FLAG_KEY for is_intra_only packet The patch will make audio and subtitle packets be marked as AV_PKT_FLAG_KEY. For audio, it'll caused the audio sample to be sync sample. To verify ref/fate/movenc results: 1. Get the movenc test data [lmwang@vpn ffmpeg]$ libavformat/tests/movenc -w && mkdir -p audio_old && mv *.mp4 audio_old_ After applied the patch: [lmwang@vpn ffmpeg]$ make fate-movenc SAMPLES=../fate-suite [lmwang@vpn ffmpeg]$ libavformat/tests/movenc -w && mkdir -p audio_key && mv *.mp4 audio_key 2. Get l-smash and build boxdumper https://github.com/l-smash/l-smash.git 3. dump the box of crc change mp4 and diff -u [lmwang@vpn ffmpeg]$ ../l-smash/cli/boxdumper --box audio_key/non-empty-moov-no-elst.mp4 > audio_key/non-empty-moov-no-elst.log [lmwang@vpn ffmpeg]$ ../l-smash/cli/boxdumper --box audio_old/non-empty-moov-no-elst.mp4 > audio_old/non-empty-moov-no-elst.log [lmwang@vpn ffmpeg]$ diff -u audio_key/non-empty-moov-no-elst.log audio_old/non-empty-moov-no-elst.log - default_sample_flags = 0x02000000 - independent - sync sample + default_sample_flags = 0x01010000 + dependent + non-sync sample 4. have checked the change of crc are caused by default_sample_flags non-empty-moov.mp4, non-empty-moov-elst.mp4, non-empty-moov-no-elst.mp4, empty-moov.mp4, delay-moov-content.mp4, empty-moov-second-frag.mp4, empty-moov-second-frag-discont.mp4, delay-moov-second-frag-discont.mp4, delay-moov-elst-second-frag.mp4 etc 5 For subtitle, it'll effect for tests/ref/fate/binsub-movtextenc and tests/ref/fate/sub2video, that's expecting result for the subtitle is marked as keyframe. Below is the checking result of binsub-movtextenc: [lmwang@vpn ffmpeg]$ ./ffmpeg -i ../fate-suite/sub/MovText_capability_tester.mp4 -map 0 -scodec mov_text -f mp4 -flags +bitexact -fflags +bitexact -movflags frag_keyframe+empty_moov audio_key/binsub-movtextenc.mp4 [lmwang@vpn ffmpeg]$ ./ffmpeg -i ../fate-suite/sub/MovText_capability_tester.mp4 -map 0 -scodec mov_text -f mp4 -flags +bitexact -fflags +bitexact -movflags frag_keyframe+empty_moov audio_old/binsub-movtextenc.mp4 [lmwang@vpn ffmpeg]$../l-smash/cli/boxdumper audio_key/binsub-movtextenc.mp4 > audio_key/binsub-movtextenc.log [lmwang@vpn ffmpeg]$../l-smash/cli/boxdumper audio_old/binsub-movtextenc.mp4 > audio_old/binsub-movtextenc.log [lmwang@vpn ffmpeg]$ diff -u audio_key/binsub-movtextenc.log audio_old/binsub-movtextenc.log .... // the key difference is the flag for sync sample - flags = 0x000701 + flags = 0x000301 data-offset-present sample-duration-present sample-size-present - sample-flags-present sample_count = 6 - data_offset = 188 + data_offset = 164 sample[0] sample_duration = 1570000 sample_size = 21 - sample_flags = 0x02000000 - independent - sync sample - degradation_priority = 0 sample[1] sample_duration = 510000 sample_size = 2 - sample_flags = 0x01010000 - dependent - non-sync sample - degradation_priority = 0 sample[2] sample_duration = 1690000 sample_size = 9 - sample_flags = 0x02000000 - independent - sync sample - degradation_priority = 0 Suggested-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Suggested-by: Nicolas George <george@nsup.org> Signed-off-by: Limin Wang <lance.lmwang@gmail.com>
5 years ago
if (of->codec_tag) {
if ( par->codec_tag
&& par->codec_id == AV_CODEC_ID_RAWVIDEO
&& ( av_codec_get_tag(of->codec_tag, par->codec_id) == 0
|| av_codec_get_tag(of->codec_tag, par->codec_id) == MKTAG('r', 'a', 'w', ' '))
&& !validate_codec_tag(s, st)) {
// the current rawvideo encoding system ends up setting
// the wrong codec_tag for avi/mov, we override it here
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
par->codec_tag = 0;
}
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (par->codec_tag) {
if (!validate_codec_tag(s, st)) {
const uint32_t otag = av_codec_get_tag(s->oformat->codec_tag, par->codec_id);
av_log(s, AV_LOG_ERROR,
"Tag %s incompatible with output codec id '%d' (%s)\n",
av_fourcc2str(par->codec_tag), par->codec_id, av_fourcc2str(otag));
ret = AVERROR_INVALIDDATA;
goto fail;
}
} else
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
par->codec_tag = av_codec_get_tag(of->codec_tag, par->codec_id);
}
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (par->codec_type != AVMEDIA_TYPE_ATTACHMENT)
si->nb_interleaved_streams++;
}
si->interleave_packet = of->interleave_packet;
if (!si->interleave_packet)
si->interleave_packet = si->nb_interleaved_streams > 1 ?
ff_interleave_packet_per_dts :
ff_interleave_packet_passthrough;
if (!s->priv_data && of->priv_data_size > 0) {
s->priv_data = av_mallocz(of->priv_data_size);
if (!s->priv_data) {
ret = AVERROR(ENOMEM);
goto fail;
}
if (of->priv_class) {
*(const AVClass **)s->priv_data = of->priv_class;
av_opt_set_defaults(s->priv_data);
if ((ret = av_opt_set_dict2(s->priv_data, &tmp, AV_OPT_SEARCH_CHILDREN)) < 0)
goto fail;
}
}
/* set muxer identification string */
if (!(s->flags & AVFMT_FLAG_BITEXACT)) {
av_dict_set(&s->metadata, "encoder", LIBAVFORMAT_IDENT, 0);
} else {
av_dict_set(&s->metadata, "encoder", NULL, 0);
}
for (e = NULL; e = av_dict_get(s->metadata, "encoder-", e, AV_DICT_IGNORE_SUFFIX); ) {
av_dict_set(&s->metadata, e->key, NULL, 0);
}
if (options) {
av_dict_free(options);
*options = tmp;
}
if (s->oformat->init) {
if ((ret = s->oformat->init(s)) < 0) {
if (s->oformat->deinit)
s->oformat->deinit(s);
return ret;
}
return ret == 0;
}
return 0;
fail:
av_dict_free(&tmp);
return ret;
}
static int init_pts(AVFormatContext *s)
{
avformat/mux: Preserve sync even if later packet has negative ts write_packet() has code to shift the packets timestamps to make them nonnegative or even make them start at ts zero; this code inspects every packet that is written and if a packet with negative timestamp (whether this is dts or pts depends upon another flag; basically: Matroska uses pts, everyone else dts) is encountered, this is offset to make the timestamp zero. All further packets will be offset accordingly (with the offset converted according to the streams' timebases). This is based around an assumption, namely that the timestamps are indeed non-decreasing, so that the first packet with negative timestamps is the first packet with timestamps. This assumption is often fulfilled given that the default interleavement function by default interleaves per dts; yet there are scenarios in which it may not be fulfilled: a) av_write_frame() instead of av_interleaved_write_frame() is used. b) The audio_preload option is used. c) When the timestamps that are made nonnegative/zero are pts (i.e. with Matroska), because the packet with the smallest dts is not necessarily the packet with the smallest pts. d) Possibly with custom interleavement functions. In these cases the relative sync of the first few packet(s) is offset relative to the later packets. This contradicts the documentation ("When shifting is enabled, all output timestamps are shifted by the same amount"). Therefore this commit changes this: As soon as the first packet with valid timestamps is output, it is checked and recorded whether the timestamps need to be shifted. Further packets are no longer checked for needing to be offset; instead they are simply offset. In the cases above this leads to packets with negative timestamps (and the appropriate warnings) instead of desync. This will mostly be fixed in the next commit. This commit also factors handling the avoid_negative_ts stuff out of write_packet() in order to be able to return immediately. Tickets #4536 and #5784 as well as the matroska-avoid-negative-ts-test are examples of c); as has been said, some timestamps are now negative, yet the ref file update does not show it because ffmpeg.c sanitizes the timestamps (-copyts disables it; ffprobe and mkvinfo also show the original timestamps). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
FFFormatContext *const si = ffformatcontext(s);
/* init PTS generation */
for (unsigned i = 0; i < s->nb_streams; i++) {
AVStream *const st = s->streams[i];
FFStream *const sti = ffstream(st);
int64_t den = AV_NOPTS_VALUE;
switch (st->codecpar->codec_type) {
case AVMEDIA_TYPE_AUDIO:
den = (int64_t)st->time_base.num * st->codecpar->sample_rate;
break;
case AVMEDIA_TYPE_VIDEO:
den = (int64_t)st->time_base.num * st->time_base.den;
break;
default:
break;
}
if (!sti->priv_pts)
sti->priv_pts = av_mallocz(sizeof(*sti->priv_pts));
if (!sti->priv_pts)
return AVERROR(ENOMEM);
if (den != AV_NOPTS_VALUE) {
if (den <= 0)
return AVERROR_INVALIDDATA;
frac_init(sti->priv_pts, 0, 0, den);
}
}
avformat/mux: Preserve sync even if later packet has negative ts write_packet() has code to shift the packets timestamps to make them nonnegative or even make them start at ts zero; this code inspects every packet that is written and if a packet with negative timestamp (whether this is dts or pts depends upon another flag; basically: Matroska uses pts, everyone else dts) is encountered, this is offset to make the timestamp zero. All further packets will be offset accordingly (with the offset converted according to the streams' timebases). This is based around an assumption, namely that the timestamps are indeed non-decreasing, so that the first packet with negative timestamps is the first packet with timestamps. This assumption is often fulfilled given that the default interleavement function by default interleaves per dts; yet there are scenarios in which it may not be fulfilled: a) av_write_frame() instead of av_interleaved_write_frame() is used. b) The audio_preload option is used. c) When the timestamps that are made nonnegative/zero are pts (i.e. with Matroska), because the packet with the smallest dts is not necessarily the packet with the smallest pts. d) Possibly with custom interleavement functions. In these cases the relative sync of the first few packet(s) is offset relative to the later packets. This contradicts the documentation ("When shifting is enabled, all output timestamps are shifted by the same amount"). Therefore this commit changes this: As soon as the first packet with valid timestamps is output, it is checked and recorded whether the timestamps need to be shifted. Further packets are no longer checked for needing to be offset; instead they are simply offset. In the cases above this leads to packets with negative timestamps (and the appropriate warnings) instead of desync. This will mostly be fixed in the next commit. This commit also factors handling the avoid_negative_ts stuff out of write_packet() in order to be able to return immediately. Tickets #4536 and #5784 as well as the matroska-avoid-negative-ts-test are examples of c); as has been said, some timestamps are now negative, yet the ref file update does not show it because ffmpeg.c sanitizes the timestamps (-copyts disables it; ffprobe and mkvinfo also show the original timestamps). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
si->avoid_negative_ts_status = AVOID_NEGATIVE_TS_UNKNOWN;
if (s->avoid_negative_ts < 0) {
av_assert2(s->avoid_negative_ts == AVFMT_AVOID_NEG_TS_AUTO);
if (s->oformat->flags & (AVFMT_TS_NEGATIVE | AVFMT_NOTIMESTAMPS)) {
s->avoid_negative_ts = AVFMT_AVOID_NEG_TS_DISABLED;
avformat/mux: Preserve sync even if later packet has negative ts write_packet() has code to shift the packets timestamps to make them nonnegative or even make them start at ts zero; this code inspects every packet that is written and if a packet with negative timestamp (whether this is dts or pts depends upon another flag; basically: Matroska uses pts, everyone else dts) is encountered, this is offset to make the timestamp zero. All further packets will be offset accordingly (with the offset converted according to the streams' timebases). This is based around an assumption, namely that the timestamps are indeed non-decreasing, so that the first packet with negative timestamps is the first packet with timestamps. This assumption is often fulfilled given that the default interleavement function by default interleaves per dts; yet there are scenarios in which it may not be fulfilled: a) av_write_frame() instead of av_interleaved_write_frame() is used. b) The audio_preload option is used. c) When the timestamps that are made nonnegative/zero are pts (i.e. with Matroska), because the packet with the smallest dts is not necessarily the packet with the smallest pts. d) Possibly with custom interleavement functions. In these cases the relative sync of the first few packet(s) is offset relative to the later packets. This contradicts the documentation ("When shifting is enabled, all output timestamps are shifted by the same amount"). Therefore this commit changes this: As soon as the first packet with valid timestamps is output, it is checked and recorded whether the timestamps need to be shifted. Further packets are no longer checked for needing to be offset; instead they are simply offset. In the cases above this leads to packets with negative timestamps (and the appropriate warnings) instead of desync. This will mostly be fixed in the next commit. This commit also factors handling the avoid_negative_ts stuff out of write_packet() in order to be able to return immediately. Tickets #4536 and #5784 as well as the matroska-avoid-negative-ts-test are examples of c); as has been said, some timestamps are now negative, yet the ref file update does not show it because ffmpeg.c sanitizes the timestamps (-copyts disables it; ffprobe and mkvinfo also show the original timestamps). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
si->avoid_negative_ts_status = AVOID_NEGATIVE_TS_DISABLED;
} else
s->avoid_negative_ts = AVFMT_AVOID_NEG_TS_MAKE_NON_NEGATIVE;
avformat/mux: Preserve sync even if later packet has negative ts write_packet() has code to shift the packets timestamps to make them nonnegative or even make them start at ts zero; this code inspects every packet that is written and if a packet with negative timestamp (whether this is dts or pts depends upon another flag; basically: Matroska uses pts, everyone else dts) is encountered, this is offset to make the timestamp zero. All further packets will be offset accordingly (with the offset converted according to the streams' timebases). This is based around an assumption, namely that the timestamps are indeed non-decreasing, so that the first packet with negative timestamps is the first packet with timestamps. This assumption is often fulfilled given that the default interleavement function by default interleaves per dts; yet there are scenarios in which it may not be fulfilled: a) av_write_frame() instead of av_interleaved_write_frame() is used. b) The audio_preload option is used. c) When the timestamps that are made nonnegative/zero are pts (i.e. with Matroska), because the packet with the smallest dts is not necessarily the packet with the smallest pts. d) Possibly with custom interleavement functions. In these cases the relative sync of the first few packet(s) is offset relative to the later packets. This contradicts the documentation ("When shifting is enabled, all output timestamps are shifted by the same amount"). Therefore this commit changes this: As soon as the first packet with valid timestamps is output, it is checked and recorded whether the timestamps need to be shifted. Further packets are no longer checked for needing to be offset; instead they are simply offset. In the cases above this leads to packets with negative timestamps (and the appropriate warnings) instead of desync. This will mostly be fixed in the next commit. This commit also factors handling the avoid_negative_ts stuff out of write_packet() in order to be able to return immediately. Tickets #4536 and #5784 as well as the matroska-avoid-negative-ts-test are examples of c); as has been said, some timestamps are now negative, yet the ref file update does not show it because ffmpeg.c sanitizes the timestamps (-copyts disables it; ffprobe and mkvinfo also show the original timestamps). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
} else if (s->avoid_negative_ts == AVFMT_AVOID_NEG_TS_DISABLED)
si->avoid_negative_ts_status = AVOID_NEGATIVE_TS_DISABLED;
return 0;
}
static void flush_if_needed(AVFormatContext *s)
{
if (s->pb && s->pb->error >= 0) {
if (s->flush_packets == 1 || s->flags & AVFMT_FLAG_FLUSH_PACKETS)
avio_flush(s->pb);
else if (s->flush_packets && !(s->oformat->flags & AVFMT_NOFILE))
avio_write_marker(s->pb, AV_NOPTS_VALUE, AVIO_DATA_MARKER_FLUSH_POINT);
}
}
static void deinit_muxer(AVFormatContext *s)
{
FFFormatContext *const si = ffformatcontext(s);
if (s->oformat && s->oformat->deinit && si->initialized)
s->oformat->deinit(s);
si->initialized =
si->streams_initialized = 0;
}
int avformat_init_output(AVFormatContext *s, AVDictionary **options)
{
FFFormatContext *const si = ffformatcontext(s);
int ret = 0;
if ((ret = init_muxer(s, options)) < 0)
return ret;
si->initialized = 1;
si->streams_initialized = ret;
if (s->oformat->init && ret) {
if ((ret = init_pts(s)) < 0)
return ret;
return AVSTREAM_INIT_IN_INIT_OUTPUT;
}
return AVSTREAM_INIT_IN_WRITE_HEADER;
}
int avformat_write_header(AVFormatContext *s, AVDictionary **options)
{
FFFormatContext *const si = ffformatcontext(s);
int already_initialized = si->initialized;
int streams_already_initialized = si->streams_initialized;
int ret = 0;
if (!already_initialized)
if ((ret = avformat_init_output(s, options)) < 0)
return ret;
if (!(s->oformat->flags & AVFMT_NOFILE) && s->pb)
avio_write_marker(s->pb, AV_NOPTS_VALUE, AVIO_DATA_MARKER_HEADER);
if (s->oformat->write_header) {
ret = s->oformat->write_header(s);
if (ret >= 0 && s->pb && s->pb->error < 0)
ret = s->pb->error;
if (ret < 0)
goto fail;
flush_if_needed(s);
}
if (!(s->oformat->flags & AVFMT_NOFILE) && s->pb)
avio_write_marker(s->pb, AV_NOPTS_VALUE, AVIO_DATA_MARKER_UNKNOWN);
if (!si->streams_initialized) {
if ((ret = init_pts(s)) < 0)
goto fail;
}
return streams_already_initialized;
fail:
deinit_muxer(s);
return ret;
}
#define AV_PKT_FLAG_UNCODED_FRAME 0x2000
#if FF_API_COMPUTE_PKT_FIELDS2
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
FF_DISABLE_DEPRECATION_WARNINGS
//FIXME merge with compute_pkt_fields
static int compute_muxer_pkt_fields(AVFormatContext *s, AVStream *st, AVPacket *pkt)
{
FFFormatContext *const si = ffformatcontext(s);
FFStream *const sti = ffstream(st);
int delay = st->codecpar->video_delay;
int frame_size;
if (!si->missing_ts_warning &&
!(s->oformat->flags & AVFMT_NOTIMESTAMPS) &&
(!(st->disposition & AV_DISPOSITION_ATTACHED_PIC) || (st->disposition & AV_DISPOSITION_TIMED_THUMBNAILS)) &&
(pkt->pts == AV_NOPTS_VALUE || pkt->dts == AV_NOPTS_VALUE)) {
av_log(s, AV_LOG_WARNING,
"Timestamps are unset in a packet for stream %d. "
"This is deprecated and will stop working in the future. "
"Fix your code to set the timestamps properly\n", st->index);
si->missing_ts_warning = 1;
}
if (s->debug & FF_FDEBUG_TS)
av_log(s, AV_LOG_DEBUG, "compute_muxer_pkt_fields: pts:%s dts:%s cur_dts:%s b:%d size:%d st:%d\n",
av_ts2str(pkt->pts), av_ts2str(pkt->dts), av_ts2str(sti->cur_dts), delay, pkt->size, pkt->stream_index);
if (pkt->pts == AV_NOPTS_VALUE && pkt->dts != AV_NOPTS_VALUE && delay == 0)
pkt->pts = pkt->dts;
//XXX/FIXME this is a temporary hack until all encoders output pts
if ((pkt->pts == 0 || pkt->pts == AV_NOPTS_VALUE) && pkt->dts == AV_NOPTS_VALUE && !delay) {
static int warned;
if (!warned) {
av_log(s, AV_LOG_WARNING, "Encoder did not produce proper pts, making some up.\n");
warned = 1;
}
pkt->dts =
// pkt->pts= st->cur_dts;
pkt->pts = sti->priv_pts->val;
}
//calculate dts from pts
if (pkt->pts != AV_NOPTS_VALUE && pkt->dts == AV_NOPTS_VALUE && delay <= MAX_REORDER_DELAY) {
sti->pts_buffer[0] = pkt->pts;
for (int i = 1; i < delay + 1 && sti->pts_buffer[i] == AV_NOPTS_VALUE; i++)
sti->pts_buffer[i] = pkt->pts + (i - delay - 1) * pkt->duration;
for (int i = 0; i<delay && sti->pts_buffer[i] > sti->pts_buffer[i + 1]; i++)
FFSWAP(int64_t, sti->pts_buffer[i], sti->pts_buffer[i + 1]);
pkt->dts = sti->pts_buffer[0];
}
if (sti->cur_dts && sti->cur_dts != AV_NOPTS_VALUE &&
((!(s->oformat->flags & AVFMT_TS_NONSTRICT) &&
st->codecpar->codec_type != AVMEDIA_TYPE_SUBTITLE &&
st->codecpar->codec_type != AVMEDIA_TYPE_DATA &&
sti->cur_dts >= pkt->dts) || sti->cur_dts > pkt->dts)) {
av_log(s, AV_LOG_ERROR,
"Application provided invalid, non monotonically increasing dts to muxer in stream %d: %s >= %s\n",
st->index, av_ts2str(sti->cur_dts), av_ts2str(pkt->dts));
return AVERROR(EINVAL);
}
if (pkt->dts != AV_NOPTS_VALUE && pkt->pts != AV_NOPTS_VALUE && pkt->pts < pkt->dts) {
av_log(s, AV_LOG_ERROR,
"pts (%s) < dts (%s) in stream %d\n",
av_ts2str(pkt->pts), av_ts2str(pkt->dts),
st->index);
return AVERROR(EINVAL);
}
if (s->debug & FF_FDEBUG_TS)
av_log(s, AV_LOG_DEBUG, "av_write_frame: pts2:%s dts2:%s\n",
av_ts2str(pkt->pts), av_ts2str(pkt->dts));
sti->cur_dts = pkt->dts;
sti->priv_pts->val = pkt->dts;
/* update pts */
switch (st->codecpar->codec_type) {
case AVMEDIA_TYPE_AUDIO:
frame_size = (pkt->flags & AV_PKT_FLAG_UNCODED_FRAME) ?
avformat/mux: Make uncoded frames av_packet_unref() compatible Currently uncoded frames (i.e. packets whose data actually points to an AVFrame) are not refcounted. As a consequence, calling av_packet_unref() on them will not free them, but may simply make sure that they leak by losing the pointer to the frame. This commit changes this by actually making uncoded frames refcounted. In order not to rely on sizeof(AVFrame) (which is not part of the public API and so must not be used here in libavformat) the packet's data is changed to a (padded) buffer containing just a pointer to an AVFrame. Said buffer is owned by an AVBuffer with a custom free function that frees the frame as well as the buffer. Thereby the pointer/the AVBuffer owns the AVFrame. Said ownership can actually be transferred by copying and resetting the pointer, as might happen when actually writing the uncoded frames in AVOutputFormat.write_uncoded_frame() (although currently no muxer makes use of this possibility). This makes packets containing uncoded frames compatible with av_packet_unref(). This already has three advantages in interleaved mode: 1. If an error happens at the preparatory steps (before the packet is put into the interleavement queue), the frame is properly freed. 2. If the trailer is never written, the frames still in the interleavement queue will now be properly freed by ff_packet_list_free(). 3. The custom code for moving the packet to the packet list in ff_interleave_add_packet() can be removed. It will also simplify fixing further memleaks in future commits. Suggested-by: Marton Balint <cus@passwd.hu> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
5 years ago
(*(AVFrame **)pkt->data)->nb_samples :
av_get_audio_frame_duration2(st->codecpar, pkt->size);
/* HACK/FIXME, we skip the initial 0 size packets as they are most
* likely equal to the encoder delay, but it would be better if we
* had the real timestamps from the encoder */
if (frame_size >= 0 && (pkt->size || sti->priv_pts->num != sti->priv_pts->den >> 1 || sti->priv_pts->val)) {
frac_add(sti->priv_pts, (int64_t)st->time_base.den * frame_size);
}
break;
case AVMEDIA_TYPE_VIDEO:
frac_add(sti->priv_pts, (int64_t)st->time_base.den * st->time_base.num);
break;
}
return 0;
}
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
FF_ENABLE_DEPRECATION_WARNINGS
#endif
static void guess_pkt_duration(AVFormatContext *s, AVStream *st, AVPacket *pkt)
{
if (pkt->duration < 0 && st->codecpar->codec_type != AVMEDIA_TYPE_SUBTITLE) {
av_log(s, AV_LOG_WARNING, "Packet with invalid duration %"PRId64" in stream %d\n",
pkt->duration, pkt->stream_index);
pkt->duration = 0;
}
if (pkt->duration)
return;
switch (st->codecpar->codec_type) {
case AVMEDIA_TYPE_VIDEO:
if (st->avg_frame_rate.num > 0 && st->avg_frame_rate.den > 0) {
pkt->duration = av_rescale_q(1, av_inv_q(st->avg_frame_rate),
st->time_base);
} else if (st->time_base.num * 1000LL > st->time_base.den)
pkt->duration = 1;
break;
case AVMEDIA_TYPE_AUDIO: {
int frame_size = av_get_audio_frame_duration2(st->codecpar, pkt->size);
if (frame_size && st->codecpar->sample_rate) {
pkt->duration = av_rescale_q(frame_size,
(AVRational){1, st->codecpar->sample_rate},
st->time_base);
}
break;
}
}
}
avformat/mux: Preserve sync even if later packet has negative ts write_packet() has code to shift the packets timestamps to make them nonnegative or even make them start at ts zero; this code inspects every packet that is written and if a packet with negative timestamp (whether this is dts or pts depends upon another flag; basically: Matroska uses pts, everyone else dts) is encountered, this is offset to make the timestamp zero. All further packets will be offset accordingly (with the offset converted according to the streams' timebases). This is based around an assumption, namely that the timestamps are indeed non-decreasing, so that the first packet with negative timestamps is the first packet with timestamps. This assumption is often fulfilled given that the default interleavement function by default interleaves per dts; yet there are scenarios in which it may not be fulfilled: a) av_write_frame() instead of av_interleaved_write_frame() is used. b) The audio_preload option is used. c) When the timestamps that are made nonnegative/zero are pts (i.e. with Matroska), because the packet with the smallest dts is not necessarily the packet with the smallest pts. d) Possibly with custom interleavement functions. In these cases the relative sync of the first few packet(s) is offset relative to the later packets. This contradicts the documentation ("When shifting is enabled, all output timestamps are shifted by the same amount"). Therefore this commit changes this: As soon as the first packet with valid timestamps is output, it is checked and recorded whether the timestamps need to be shifted. Further packets are no longer checked for needing to be offset; instead they are simply offset. In the cases above this leads to packets with negative timestamps (and the appropriate warnings) instead of desync. This will mostly be fixed in the next commit. This commit also factors handling the avoid_negative_ts stuff out of write_packet() in order to be able to return immediately. Tickets #4536 and #5784 as well as the matroska-avoid-negative-ts-test are examples of c); as has been said, some timestamps are now negative, yet the ref file update does not show it because ffmpeg.c sanitizes the timestamps (-copyts disables it; ffprobe and mkvinfo also show the original timestamps). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
static void handle_avoid_negative_ts(FFFormatContext *si, FFStream *sti,
AVPacket *pkt)
{
AVFormatContext *const s = &si->pub;
int64_t offset;
if (!AVOID_NEGATIVE_TS_ENABLED(si->avoid_negative_ts_status))
return;
if (si->avoid_negative_ts_status == AVOID_NEGATIVE_TS_UNKNOWN) {
int use_pts = si->avoid_negative_ts_use_pts;
int64_t ts = use_pts ? pkt->pts : pkt->dts;
AVRational tb = sti->pub.time_base;
avformat/mux: Preserve sync even if later packet has negative ts write_packet() has code to shift the packets timestamps to make them nonnegative or even make them start at ts zero; this code inspects every packet that is written and if a packet with negative timestamp (whether this is dts or pts depends upon another flag; basically: Matroska uses pts, everyone else dts) is encountered, this is offset to make the timestamp zero. All further packets will be offset accordingly (with the offset converted according to the streams' timebases). This is based around an assumption, namely that the timestamps are indeed non-decreasing, so that the first packet with negative timestamps is the first packet with timestamps. This assumption is often fulfilled given that the default interleavement function by default interleaves per dts; yet there are scenarios in which it may not be fulfilled: a) av_write_frame() instead of av_interleaved_write_frame() is used. b) The audio_preload option is used. c) When the timestamps that are made nonnegative/zero are pts (i.e. with Matroska), because the packet with the smallest dts is not necessarily the packet with the smallest pts. d) Possibly with custom interleavement functions. In these cases the relative sync of the first few packet(s) is offset relative to the later packets. This contradicts the documentation ("When shifting is enabled, all output timestamps are shifted by the same amount"). Therefore this commit changes this: As soon as the first packet with valid timestamps is output, it is checked and recorded whether the timestamps need to be shifted. Further packets are no longer checked for needing to be offset; instead they are simply offset. In the cases above this leads to packets with negative timestamps (and the appropriate warnings) instead of desync. This will mostly be fixed in the next commit. This commit also factors handling the avoid_negative_ts stuff out of write_packet() in order to be able to return immediately. Tickets #4536 and #5784 as well as the matroska-avoid-negative-ts-test are examples of c); as has been said, some timestamps are now negative, yet the ref file update does not show it because ffmpeg.c sanitizes the timestamps (-copyts disables it; ffprobe and mkvinfo also show the original timestamps). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
if (ts == AV_NOPTS_VALUE)
return;
ts -= sti->lowest_ts_allowed;
/* Peek into the muxing queue to improve our estimate
* of the lowest timestamp if av_interleaved_write_frame() is used. */
for (const PacketListEntry *pktl = si->packet_buffer.head;
pktl; pktl = pktl->next) {
AVRational cmp_tb = s->streams[pktl->pkt.stream_index]->time_base;
int64_t cmp_ts = use_pts ? pktl->pkt.pts : pktl->pkt.dts;
if (cmp_ts == AV_NOPTS_VALUE)
continue;
cmp_ts -= ffstream(s->streams[pktl->pkt.stream_index])->lowest_ts_allowed;
if (s->output_ts_offset)
cmp_ts += av_rescale_q(s->output_ts_offset, AV_TIME_BASE_Q, cmp_tb);
if (av_compare_ts(cmp_ts, cmp_tb, ts, tb) < 0) {
ts = cmp_ts;
tb = cmp_tb;
}
}
avformat/mux: Preserve sync even if later packet has negative ts write_packet() has code to shift the packets timestamps to make them nonnegative or even make them start at ts zero; this code inspects every packet that is written and if a packet with negative timestamp (whether this is dts or pts depends upon another flag; basically: Matroska uses pts, everyone else dts) is encountered, this is offset to make the timestamp zero. All further packets will be offset accordingly (with the offset converted according to the streams' timebases). This is based around an assumption, namely that the timestamps are indeed non-decreasing, so that the first packet with negative timestamps is the first packet with timestamps. This assumption is often fulfilled given that the default interleavement function by default interleaves per dts; yet there are scenarios in which it may not be fulfilled: a) av_write_frame() instead of av_interleaved_write_frame() is used. b) The audio_preload option is used. c) When the timestamps that are made nonnegative/zero are pts (i.e. with Matroska), because the packet with the smallest dts is not necessarily the packet with the smallest pts. d) Possibly with custom interleavement functions. In these cases the relative sync of the first few packet(s) is offset relative to the later packets. This contradicts the documentation ("When shifting is enabled, all output timestamps are shifted by the same amount"). Therefore this commit changes this: As soon as the first packet with valid timestamps is output, it is checked and recorded whether the timestamps need to be shifted. Further packets are no longer checked for needing to be offset; instead they are simply offset. In the cases above this leads to packets with negative timestamps (and the appropriate warnings) instead of desync. This will mostly be fixed in the next commit. This commit also factors handling the avoid_negative_ts stuff out of write_packet() in order to be able to return immediately. Tickets #4536 and #5784 as well as the matroska-avoid-negative-ts-test are examples of c); as has been said, some timestamps are now negative, yet the ref file update does not show it because ffmpeg.c sanitizes the timestamps (-copyts disables it; ffprobe and mkvinfo also show the original timestamps). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
if (ts < 0 ||
ts > 0 && s->avoid_negative_ts == AVFMT_AVOID_NEG_TS_MAKE_ZERO) {
for (unsigned i = 0; i < s->nb_streams; i++) {
AVStream *const st2 = s->streams[i];
FFStream *const sti2 = ffstream(st2);
sti2->mux_ts_offset = av_rescale_q_rnd(-ts, tb,
avformat/mux: Preserve sync even if later packet has negative ts write_packet() has code to shift the packets timestamps to make them nonnegative or even make them start at ts zero; this code inspects every packet that is written and if a packet with negative timestamp (whether this is dts or pts depends upon another flag; basically: Matroska uses pts, everyone else dts) is encountered, this is offset to make the timestamp zero. All further packets will be offset accordingly (with the offset converted according to the streams' timebases). This is based around an assumption, namely that the timestamps are indeed non-decreasing, so that the first packet with negative timestamps is the first packet with timestamps. This assumption is often fulfilled given that the default interleavement function by default interleaves per dts; yet there are scenarios in which it may not be fulfilled: a) av_write_frame() instead of av_interleaved_write_frame() is used. b) The audio_preload option is used. c) When the timestamps that are made nonnegative/zero are pts (i.e. with Matroska), because the packet with the smallest dts is not necessarily the packet with the smallest pts. d) Possibly with custom interleavement functions. In these cases the relative sync of the first few packet(s) is offset relative to the later packets. This contradicts the documentation ("When shifting is enabled, all output timestamps are shifted by the same amount"). Therefore this commit changes this: As soon as the first packet with valid timestamps is output, it is checked and recorded whether the timestamps need to be shifted. Further packets are no longer checked for needing to be offset; instead they are simply offset. In the cases above this leads to packets with negative timestamps (and the appropriate warnings) instead of desync. This will mostly be fixed in the next commit. This commit also factors handling the avoid_negative_ts stuff out of write_packet() in order to be able to return immediately. Tickets #4536 and #5784 as well as the matroska-avoid-negative-ts-test are examples of c); as has been said, some timestamps are now negative, yet the ref file update does not show it because ffmpeg.c sanitizes the timestamps (-copyts disables it; ffprobe and mkvinfo also show the original timestamps). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
st2->time_base,
AV_ROUND_UP);
}
}
si->avoid_negative_ts_status = AVOID_NEGATIVE_TS_KNOWN;
}
offset = sti->mux_ts_offset;
if (pkt->dts != AV_NOPTS_VALUE)
pkt->dts += offset;
if (pkt->pts != AV_NOPTS_VALUE)
pkt->pts += offset;
if (si->avoid_negative_ts_use_pts) {
if (pkt->pts != AV_NOPTS_VALUE && pkt->pts < sti->lowest_ts_allowed) {
avformat/mux: Preserve sync even if later packet has negative ts write_packet() has code to shift the packets timestamps to make them nonnegative or even make them start at ts zero; this code inspects every packet that is written and if a packet with negative timestamp (whether this is dts or pts depends upon another flag; basically: Matroska uses pts, everyone else dts) is encountered, this is offset to make the timestamp zero. All further packets will be offset accordingly (with the offset converted according to the streams' timebases). This is based around an assumption, namely that the timestamps are indeed non-decreasing, so that the first packet with negative timestamps is the first packet with timestamps. This assumption is often fulfilled given that the default interleavement function by default interleaves per dts; yet there are scenarios in which it may not be fulfilled: a) av_write_frame() instead of av_interleaved_write_frame() is used. b) The audio_preload option is used. c) When the timestamps that are made nonnegative/zero are pts (i.e. with Matroska), because the packet with the smallest dts is not necessarily the packet with the smallest pts. d) Possibly with custom interleavement functions. In these cases the relative sync of the first few packet(s) is offset relative to the later packets. This contradicts the documentation ("When shifting is enabled, all output timestamps are shifted by the same amount"). Therefore this commit changes this: As soon as the first packet with valid timestamps is output, it is checked and recorded whether the timestamps need to be shifted. Further packets are no longer checked for needing to be offset; instead they are simply offset. In the cases above this leads to packets with negative timestamps (and the appropriate warnings) instead of desync. This will mostly be fixed in the next commit. This commit also factors handling the avoid_negative_ts stuff out of write_packet() in order to be able to return immediately. Tickets #4536 and #5784 as well as the matroska-avoid-negative-ts-test are examples of c); as has been said, some timestamps are now negative, yet the ref file update does not show it because ffmpeg.c sanitizes the timestamps (-copyts disables it; ffprobe and mkvinfo also show the original timestamps). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
av_log(s, AV_LOG_WARNING, "failed to avoid negative "
"pts %s in stream %d.\n"
"Try -avoid_negative_ts 1 as a possible workaround.\n",
av_ts2str(pkt->pts),
pkt->stream_index
);
}
} else {
if (pkt->dts != AV_NOPTS_VALUE && pkt->dts < sti->lowest_ts_allowed) {
avformat/mux: Preserve sync even if later packet has negative ts write_packet() has code to shift the packets timestamps to make them nonnegative or even make them start at ts zero; this code inspects every packet that is written and if a packet with negative timestamp (whether this is dts or pts depends upon another flag; basically: Matroska uses pts, everyone else dts) is encountered, this is offset to make the timestamp zero. All further packets will be offset accordingly (with the offset converted according to the streams' timebases). This is based around an assumption, namely that the timestamps are indeed non-decreasing, so that the first packet with negative timestamps is the first packet with timestamps. This assumption is often fulfilled given that the default interleavement function by default interleaves per dts; yet there are scenarios in which it may not be fulfilled: a) av_write_frame() instead of av_interleaved_write_frame() is used. b) The audio_preload option is used. c) When the timestamps that are made nonnegative/zero are pts (i.e. with Matroska), because the packet with the smallest dts is not necessarily the packet with the smallest pts. d) Possibly with custom interleavement functions. In these cases the relative sync of the first few packet(s) is offset relative to the later packets. This contradicts the documentation ("When shifting is enabled, all output timestamps are shifted by the same amount"). Therefore this commit changes this: As soon as the first packet with valid timestamps is output, it is checked and recorded whether the timestamps need to be shifted. Further packets are no longer checked for needing to be offset; instead they are simply offset. In the cases above this leads to packets with negative timestamps (and the appropriate warnings) instead of desync. This will mostly be fixed in the next commit. This commit also factors handling the avoid_negative_ts stuff out of write_packet() in order to be able to return immediately. Tickets #4536 and #5784 as well as the matroska-avoid-negative-ts-test are examples of c); as has been said, some timestamps are now negative, yet the ref file update does not show it because ffmpeg.c sanitizes the timestamps (-copyts disables it; ffprobe and mkvinfo also show the original timestamps). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
av_log(s, AV_LOG_WARNING,
"Packets poorly interleaved, failed to avoid negative "
"timestamp %s in stream %d.\n"
"Try -max_interleave_delta 0 as a possible workaround.\n",
av_ts2str(pkt->dts),
pkt->stream_index
);
}
}
}
/**
* Shift timestamps and call muxer; the original pts/dts are not kept.
*
* FIXME: this function should NEVER get undefined pts/dts beside when the
* AVFMT_NOTIMESTAMPS is set.
* Those additional safety checks should be dropped once the correct checks
* are set in the callers.
*/
static int write_packet(AVFormatContext *s, AVPacket *pkt)
{
FFFormatContext *const si = ffformatcontext(s);
avformat/mux: Fix double-free when using AVPacket.opaque_ref Up until now, ff_write_chained() copied the packet (manually, not with av_packet_move_ref()) from a packet given to it to a stack packet whose timing and stream_index is then modified before being sent to another muxer via av_(interleaved_)write_frame(). Afterwards it is intended to sync the fields of the packet relevant to freeing again; yet this only encompasses buf, side_data and side_data_elems and not the newly added opaque_ref. The other fields are not synced so that the returned packet can have a size > 0 and data != NULL despite its buf being NULL (this always happens in the interleaved codepath; before commit fe251f77c80b0512ab8907902e1dbed3f4fe1aad it could also happen in the noninterleaved one). This leads to double-frees if the interleaved codepath is used and opaque_ref is set. This commit therefore changes this by directly reusing the packet instead of a spare packet. Given that av_write_frame() does not change the packet given to it, one only needs to restore the timing information to return it as it was; for the interleaved codepath it is not possible to do likewise*, because av_interleaved_write_frame() takes ownership of the packets given to it and returns blank packets. But precisely because of this users of the interleaved codepath have no legitimate expectation that their packet will be returned unchanged. In line with av_interleaved_write_frame() ff_write_chained() therefore returns blank packets when using the interleaved codepath. Making the only user of said codepath compatible with this was trivial. *: Unless one wanted to create a full new reference. Reviewed-by: Lynne <dev@lynne.ee> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
AVStream *const st = s->streams[pkt->stream_index];
FFStream *const sti = ffstream(st);
int ret;
// If the timestamp offsetting below is adjusted, adjust
// ff_interleaved_peek similarly.
if (s->output_ts_offset) {
int64_t offset = av_rescale_q(s->output_ts_offset, AV_TIME_BASE_Q, st->time_base);
if (pkt->dts != AV_NOPTS_VALUE)
pkt->dts += offset;
if (pkt->pts != AV_NOPTS_VALUE)
pkt->pts += offset;
}
avformat/mux: Preserve sync even if later packet has negative ts write_packet() has code to shift the packets timestamps to make them nonnegative or even make them start at ts zero; this code inspects every packet that is written and if a packet with negative timestamp (whether this is dts or pts depends upon another flag; basically: Matroska uses pts, everyone else dts) is encountered, this is offset to make the timestamp zero. All further packets will be offset accordingly (with the offset converted according to the streams' timebases). This is based around an assumption, namely that the timestamps are indeed non-decreasing, so that the first packet with negative timestamps is the first packet with timestamps. This assumption is often fulfilled given that the default interleavement function by default interleaves per dts; yet there are scenarios in which it may not be fulfilled: a) av_write_frame() instead of av_interleaved_write_frame() is used. b) The audio_preload option is used. c) When the timestamps that are made nonnegative/zero are pts (i.e. with Matroska), because the packet with the smallest dts is not necessarily the packet with the smallest pts. d) Possibly with custom interleavement functions. In these cases the relative sync of the first few packet(s) is offset relative to the later packets. This contradicts the documentation ("When shifting is enabled, all output timestamps are shifted by the same amount"). Therefore this commit changes this: As soon as the first packet with valid timestamps is output, it is checked and recorded whether the timestamps need to be shifted. Further packets are no longer checked for needing to be offset; instead they are simply offset. In the cases above this leads to packets with negative timestamps (and the appropriate warnings) instead of desync. This will mostly be fixed in the next commit. This commit also factors handling the avoid_negative_ts stuff out of write_packet() in order to be able to return immediately. Tickets #4536 and #5784 as well as the matroska-avoid-negative-ts-test are examples of c); as has been said, some timestamps are now negative, yet the ref file update does not show it because ffmpeg.c sanitizes the timestamps (-copyts disables it; ffprobe and mkvinfo also show the original timestamps). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
handle_avoid_negative_ts(si, sti, pkt);
if ((pkt->flags & AV_PKT_FLAG_UNCODED_FRAME)) {
avformat/mux: Make uncoded frames av_packet_unref() compatible Currently uncoded frames (i.e. packets whose data actually points to an AVFrame) are not refcounted. As a consequence, calling av_packet_unref() on them will not free them, but may simply make sure that they leak by losing the pointer to the frame. This commit changes this by actually making uncoded frames refcounted. In order not to rely on sizeof(AVFrame) (which is not part of the public API and so must not be used here in libavformat) the packet's data is changed to a (padded) buffer containing just a pointer to an AVFrame. Said buffer is owned by an AVBuffer with a custom free function that frees the frame as well as the buffer. Thereby the pointer/the AVBuffer owns the AVFrame. Said ownership can actually be transferred by copying and resetting the pointer, as might happen when actually writing the uncoded frames in AVOutputFormat.write_uncoded_frame() (although currently no muxer makes use of this possibility). This makes packets containing uncoded frames compatible with av_packet_unref(). This already has three advantages in interleaved mode: 1. If an error happens at the preparatory steps (before the packet is put into the interleavement queue), the frame is properly freed. 2. If the trailer is never written, the frames still in the interleavement queue will now be properly freed by ff_packet_list_free(). 3. The custom code for moving the packet to the packet list in ff_interleave_add_packet() can be removed. It will also simplify fixing further memleaks in future commits. Suggested-by: Marton Balint <cus@passwd.hu> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
5 years ago
AVFrame **frame = (AVFrame **)pkt->data;
av_assert0(pkt->size == sizeof(*frame));
ret = s->oformat->write_uncoded_frame(s, pkt->stream_index, frame, 0);
} else {
ret = s->oformat->write_packet(s, pkt);
}
if (s->pb && ret >= 0) {
flush_if_needed(s);
if (s->pb->error < 0)
ret = s->pb->error;
}
if (ret >= 0)
avformat/mux: Fix double-free when using AVPacket.opaque_ref Up until now, ff_write_chained() copied the packet (manually, not with av_packet_move_ref()) from a packet given to it to a stack packet whose timing and stream_index is then modified before being sent to another muxer via av_(interleaved_)write_frame(). Afterwards it is intended to sync the fields of the packet relevant to freeing again; yet this only encompasses buf, side_data and side_data_elems and not the newly added opaque_ref. The other fields are not synced so that the returned packet can have a size > 0 and data != NULL despite its buf being NULL (this always happens in the interleaved codepath; before commit fe251f77c80b0512ab8907902e1dbed3f4fe1aad it could also happen in the noninterleaved one). This leads to double-frees if the interleaved codepath is used and opaque_ref is set. This commit therefore changes this by directly reusing the packet instead of a spare packet. Given that av_write_frame() does not change the packet given to it, one only needs to restore the timing information to return it as it was; for the interleaved codepath it is not possible to do likewise*, because av_interleaved_write_frame() takes ownership of the packets given to it and returns blank packets. But precisely because of this users of the interleaved codepath have no legitimate expectation that their packet will be returned unchanged. In line with av_interleaved_write_frame() ff_write_chained() therefore returns blank packets when using the interleaved codepath. Making the only user of said codepath compatible with this was trivial. *: Unless one wanted to create a full new reference. Reviewed-by: Lynne <dev@lynne.ee> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
st->nb_frames++;
return ret;
}
static int check_packet(AVFormatContext *s, AVPacket *pkt)
{
if (pkt->stream_index < 0 || pkt->stream_index >= s->nb_streams) {
av_log(s, AV_LOG_ERROR, "Invalid packet stream index: %d\n",
pkt->stream_index);
return AVERROR(EINVAL);
}
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (s->streams[pkt->stream_index]->codecpar->codec_type == AVMEDIA_TYPE_ATTACHMENT) {
av_log(s, AV_LOG_ERROR, "Received a packet for an attachment stream.\n");
return AVERROR(EINVAL);
}
return 0;
}
static int prepare_input_packet(AVFormatContext *s, AVStream *st, AVPacket *pkt)
{
FFStream *const sti = ffstream(st);
#if !FF_API_COMPUTE_PKT_FIELDS2
/* sanitize the timestamps */
if (!(s->oformat->flags & AVFMT_NOTIMESTAMPS)) {
/* when there is no reordering (so dts is equal to pts), but
* only one of them is set, set the other as well */
if (!sti->reorder) {
if (pkt->pts == AV_NOPTS_VALUE && pkt->dts != AV_NOPTS_VALUE)
pkt->pts = pkt->dts;
if (pkt->dts == AV_NOPTS_VALUE && pkt->pts != AV_NOPTS_VALUE)
pkt->dts = pkt->pts;
}
/* check that the timestamps are set */
if (pkt->pts == AV_NOPTS_VALUE || pkt->dts == AV_NOPTS_VALUE) {
av_log(s, AV_LOG_ERROR,
"Timestamps are unset in a packet for stream %d\n", st->index);
return AVERROR(EINVAL);
}
/* check that the dts are increasing (or at least non-decreasing,
* if the format allows it */
if (sti->cur_dts != AV_NOPTS_VALUE &&
((!(s->oformat->flags & AVFMT_TS_NONSTRICT) && sti->cur_dts >= pkt->dts) ||
sti->cur_dts > pkt->dts)) {
av_log(s, AV_LOG_ERROR,
"Application provided invalid, non monotonically increasing "
"dts to muxer in stream %d: %" PRId64 " >= %" PRId64 "\n",
st->index, sti->cur_dts, pkt->dts);
return AVERROR(EINVAL);
}
if (pkt->pts < pkt->dts) {
av_log(s, AV_LOG_ERROR, "pts %" PRId64 " < dts %" PRId64 " in stream %d\n",
pkt->pts, pkt->dts, st->index);
return AVERROR(EINVAL);
}
}
#endif
avformat/mux: Set AV_PKT_FLAG_KEY for is_intra_only packet The patch will make audio and subtitle packets be marked as AV_PKT_FLAG_KEY. For audio, it'll caused the audio sample to be sync sample. To verify ref/fate/movenc results: 1. Get the movenc test data [lmwang@vpn ffmpeg]$ libavformat/tests/movenc -w && mkdir -p audio_old && mv *.mp4 audio_old_ After applied the patch: [lmwang@vpn ffmpeg]$ make fate-movenc SAMPLES=../fate-suite [lmwang@vpn ffmpeg]$ libavformat/tests/movenc -w && mkdir -p audio_key && mv *.mp4 audio_key 2. Get l-smash and build boxdumper https://github.com/l-smash/l-smash.git 3. dump the box of crc change mp4 and diff -u [lmwang@vpn ffmpeg]$ ../l-smash/cli/boxdumper --box audio_key/non-empty-moov-no-elst.mp4 > audio_key/non-empty-moov-no-elst.log [lmwang@vpn ffmpeg]$ ../l-smash/cli/boxdumper --box audio_old/non-empty-moov-no-elst.mp4 > audio_old/non-empty-moov-no-elst.log [lmwang@vpn ffmpeg]$ diff -u audio_key/non-empty-moov-no-elst.log audio_old/non-empty-moov-no-elst.log - default_sample_flags = 0x02000000 - independent - sync sample + default_sample_flags = 0x01010000 + dependent + non-sync sample 4. have checked the change of crc are caused by default_sample_flags non-empty-moov.mp4, non-empty-moov-elst.mp4, non-empty-moov-no-elst.mp4, empty-moov.mp4, delay-moov-content.mp4, empty-moov-second-frag.mp4, empty-moov-second-frag-discont.mp4, delay-moov-second-frag-discont.mp4, delay-moov-elst-second-frag.mp4 etc 5 For subtitle, it'll effect for tests/ref/fate/binsub-movtextenc and tests/ref/fate/sub2video, that's expecting result for the subtitle is marked as keyframe. Below is the checking result of binsub-movtextenc: [lmwang@vpn ffmpeg]$ ./ffmpeg -i ../fate-suite/sub/MovText_capability_tester.mp4 -map 0 -scodec mov_text -f mp4 -flags +bitexact -fflags +bitexact -movflags frag_keyframe+empty_moov audio_key/binsub-movtextenc.mp4 [lmwang@vpn ffmpeg]$ ./ffmpeg -i ../fate-suite/sub/MovText_capability_tester.mp4 -map 0 -scodec mov_text -f mp4 -flags +bitexact -fflags +bitexact -movflags frag_keyframe+empty_moov audio_old/binsub-movtextenc.mp4 [lmwang@vpn ffmpeg]$../l-smash/cli/boxdumper audio_key/binsub-movtextenc.mp4 > audio_key/binsub-movtextenc.log [lmwang@vpn ffmpeg]$../l-smash/cli/boxdumper audio_old/binsub-movtextenc.mp4 > audio_old/binsub-movtextenc.log [lmwang@vpn ffmpeg]$ diff -u audio_key/binsub-movtextenc.log audio_old/binsub-movtextenc.log .... // the key difference is the flag for sync sample - flags = 0x000701 + flags = 0x000301 data-offset-present sample-duration-present sample-size-present - sample-flags-present sample_count = 6 - data_offset = 188 + data_offset = 164 sample[0] sample_duration = 1570000 sample_size = 21 - sample_flags = 0x02000000 - independent - sync sample - degradation_priority = 0 sample[1] sample_duration = 510000 sample_size = 2 - sample_flags = 0x01010000 - dependent - non-sync sample - degradation_priority = 0 sample[2] sample_duration = 1690000 sample_size = 9 - sample_flags = 0x02000000 - independent - sync sample - degradation_priority = 0 Suggested-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Suggested-by: Nicolas George <george@nsup.org> Signed-off-by: Limin Wang <lance.lmwang@gmail.com>
5 years ago
/* update flags */
if (sti->is_intra_only)
avformat/mux: Set AV_PKT_FLAG_KEY for is_intra_only packet The patch will make audio and subtitle packets be marked as AV_PKT_FLAG_KEY. For audio, it'll caused the audio sample to be sync sample. To verify ref/fate/movenc results: 1. Get the movenc test data [lmwang@vpn ffmpeg]$ libavformat/tests/movenc -w && mkdir -p audio_old && mv *.mp4 audio_old_ After applied the patch: [lmwang@vpn ffmpeg]$ make fate-movenc SAMPLES=../fate-suite [lmwang@vpn ffmpeg]$ libavformat/tests/movenc -w && mkdir -p audio_key && mv *.mp4 audio_key 2. Get l-smash and build boxdumper https://github.com/l-smash/l-smash.git 3. dump the box of crc change mp4 and diff -u [lmwang@vpn ffmpeg]$ ../l-smash/cli/boxdumper --box audio_key/non-empty-moov-no-elst.mp4 > audio_key/non-empty-moov-no-elst.log [lmwang@vpn ffmpeg]$ ../l-smash/cli/boxdumper --box audio_old/non-empty-moov-no-elst.mp4 > audio_old/non-empty-moov-no-elst.log [lmwang@vpn ffmpeg]$ diff -u audio_key/non-empty-moov-no-elst.log audio_old/non-empty-moov-no-elst.log - default_sample_flags = 0x02000000 - independent - sync sample + default_sample_flags = 0x01010000 + dependent + non-sync sample 4. have checked the change of crc are caused by default_sample_flags non-empty-moov.mp4, non-empty-moov-elst.mp4, non-empty-moov-no-elst.mp4, empty-moov.mp4, delay-moov-content.mp4, empty-moov-second-frag.mp4, empty-moov-second-frag-discont.mp4, delay-moov-second-frag-discont.mp4, delay-moov-elst-second-frag.mp4 etc 5 For subtitle, it'll effect for tests/ref/fate/binsub-movtextenc and tests/ref/fate/sub2video, that's expecting result for the subtitle is marked as keyframe. Below is the checking result of binsub-movtextenc: [lmwang@vpn ffmpeg]$ ./ffmpeg -i ../fate-suite/sub/MovText_capability_tester.mp4 -map 0 -scodec mov_text -f mp4 -flags +bitexact -fflags +bitexact -movflags frag_keyframe+empty_moov audio_key/binsub-movtextenc.mp4 [lmwang@vpn ffmpeg]$ ./ffmpeg -i ../fate-suite/sub/MovText_capability_tester.mp4 -map 0 -scodec mov_text -f mp4 -flags +bitexact -fflags +bitexact -movflags frag_keyframe+empty_moov audio_old/binsub-movtextenc.mp4 [lmwang@vpn ffmpeg]$../l-smash/cli/boxdumper audio_key/binsub-movtextenc.mp4 > audio_key/binsub-movtextenc.log [lmwang@vpn ffmpeg]$../l-smash/cli/boxdumper audio_old/binsub-movtextenc.mp4 > audio_old/binsub-movtextenc.log [lmwang@vpn ffmpeg]$ diff -u audio_key/binsub-movtextenc.log audio_old/binsub-movtextenc.log .... // the key difference is the flag for sync sample - flags = 0x000701 + flags = 0x000301 data-offset-present sample-duration-present sample-size-present - sample-flags-present sample_count = 6 - data_offset = 188 + data_offset = 164 sample[0] sample_duration = 1570000 sample_size = 21 - sample_flags = 0x02000000 - independent - sync sample - degradation_priority = 0 sample[1] sample_duration = 510000 sample_size = 2 - sample_flags = 0x01010000 - dependent - non-sync sample - degradation_priority = 0 sample[2] sample_duration = 1690000 sample_size = 9 - sample_flags = 0x02000000 - independent - sync sample - degradation_priority = 0 Suggested-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Suggested-by: Nicolas George <george@nsup.org> Signed-off-by: Limin Wang <lance.lmwang@gmail.com>
5 years ago
pkt->flags |= AV_PKT_FLAG_KEY;
if (!pkt->data && !pkt->side_data_elems) {
/* Such empty packets signal EOS for the BSF API; so sanitize
* the packet by allocating data of size 0 (+ padding). */
av_buffer_unref(&pkt->buf);
return av_packet_make_refcounted(pkt);
}
return 0;
}
#define CHUNK_START 0x1000
int ff_interleave_add_packet(AVFormatContext *s, AVPacket *pkt,
int (*compare)(AVFormatContext *, const AVPacket *, const AVPacket *))
{
int ret;
FFFormatContext *const si = ffformatcontext(s);
PacketListEntry **next_point, *this_pktl;
AVStream *st = s->streams[pkt->stream_index];
FFStream *const sti = ffstream(st);
int chunked = s->max_chunk_size || s->max_chunk_duration;
this_pktl = av_malloc(sizeof(*this_pktl));
if (!this_pktl) {
av_packet_unref(pkt);
return AVERROR(ENOMEM);
}
avformat/mux: Make uncoded frames av_packet_unref() compatible Currently uncoded frames (i.e. packets whose data actually points to an AVFrame) are not refcounted. As a consequence, calling av_packet_unref() on them will not free them, but may simply make sure that they leak by losing the pointer to the frame. This commit changes this by actually making uncoded frames refcounted. In order not to rely on sizeof(AVFrame) (which is not part of the public API and so must not be used here in libavformat) the packet's data is changed to a (padded) buffer containing just a pointer to an AVFrame. Said buffer is owned by an AVBuffer with a custom free function that frees the frame as well as the buffer. Thereby the pointer/the AVBuffer owns the AVFrame. Said ownership can actually be transferred by copying and resetting the pointer, as might happen when actually writing the uncoded frames in AVOutputFormat.write_uncoded_frame() (although currently no muxer makes use of this possibility). This makes packets containing uncoded frames compatible with av_packet_unref(). This already has three advantages in interleaved mode: 1. If an error happens at the preparatory steps (before the packet is put into the interleavement queue), the frame is properly freed. 2. If the trailer is never written, the frames still in the interleavement queue will now be properly freed by ff_packet_list_free(). 3. The custom code for moving the packet to the packet list in ff_interleave_add_packet() can be removed. It will also simplify fixing further memleaks in future commits. Suggested-by: Marton Balint <cus@passwd.hu> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
5 years ago
if ((ret = av_packet_make_refcounted(pkt)) < 0) {
av_free(this_pktl);
av_packet_unref(pkt);
avformat/mux: Make uncoded frames av_packet_unref() compatible Currently uncoded frames (i.e. packets whose data actually points to an AVFrame) are not refcounted. As a consequence, calling av_packet_unref() on them will not free them, but may simply make sure that they leak by losing the pointer to the frame. This commit changes this by actually making uncoded frames refcounted. In order not to rely on sizeof(AVFrame) (which is not part of the public API and so must not be used here in libavformat) the packet's data is changed to a (padded) buffer containing just a pointer to an AVFrame. Said buffer is owned by an AVBuffer with a custom free function that frees the frame as well as the buffer. Thereby the pointer/the AVBuffer owns the AVFrame. Said ownership can actually be transferred by copying and resetting the pointer, as might happen when actually writing the uncoded frames in AVOutputFormat.write_uncoded_frame() (although currently no muxer makes use of this possibility). This makes packets containing uncoded frames compatible with av_packet_unref(). This already has three advantages in interleaved mode: 1. If an error happens at the preparatory steps (before the packet is put into the interleavement queue), the frame is properly freed. 2. If the trailer is never written, the frames still in the interleavement queue will now be properly freed by ff_packet_list_free(). 3. The custom code for moving the packet to the packet list in ff_interleave_add_packet() can be removed. It will also simplify fixing further memleaks in future commits. Suggested-by: Marton Balint <cus@passwd.hu> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
5 years ago
return ret;
}
av_packet_move_ref(&this_pktl->pkt, pkt);
pkt = &this_pktl->pkt;
if (sti->last_in_packet_buffer) {
next_point = &(sti->last_in_packet_buffer->next);
} else {
next_point = &si->packet_buffer.head;
}
if (chunked) {
uint64_t max= av_rescale_q_rnd(s->max_chunk_duration, AV_TIME_BASE_Q, st->time_base, AV_ROUND_UP);
sti->interleaver_chunk_size += pkt->size;
sti->interleaver_chunk_duration += pkt->duration;
if ( (s->max_chunk_size && sti->interleaver_chunk_size > s->max_chunk_size)
|| (max && sti->interleaver_chunk_duration > max)) {
sti->interleaver_chunk_size = 0;
pkt->flags |= CHUNK_START;
if (max && sti->interleaver_chunk_duration > max) {
int64_t syncoffset = (st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO)*max/2;
int64_t syncto = av_rescale(pkt->dts + syncoffset, 1, max)*max - syncoffset;
sti->interleaver_chunk_duration += (pkt->dts - syncto)/8 - max;
} else
sti->interleaver_chunk_duration = 0;
}
}
if (*next_point) {
if (chunked && !(pkt->flags & CHUNK_START))
goto next_non_null;
if (compare(s, &si->packet_buffer.tail->pkt, pkt)) {
while ( *next_point
&& ((chunked && !((*next_point)->pkt.flags&CHUNK_START))
|| !compare(s, &(*next_point)->pkt, pkt)))
next_point = &(*next_point)->next;
if (*next_point)
goto next_non_null;
} else {
next_point = &(si->packet_buffer.tail->next);
}
}
av_assert1(!*next_point);
si->packet_buffer.tail = this_pktl;
next_non_null:
this_pktl->next = *next_point;
sti->last_in_packet_buffer = *next_point = this_pktl;
return 0;
}
static int interleave_compare_dts(AVFormatContext *s, const AVPacket *next,
const AVPacket *pkt)
{
AVStream *st = s->streams[pkt->stream_index];
AVStream *st2 = s->streams[next->stream_index];
int comp = av_compare_ts(next->dts, st2->time_base, pkt->dts,
st->time_base);
if (s->audio_preload) {
int preload = st ->codecpar->codec_type == AVMEDIA_TYPE_AUDIO;
int preload2 = st2->codecpar->codec_type == AVMEDIA_TYPE_AUDIO;
if (preload != preload2) {
int64_t ts, ts2;
preload *= s->audio_preload;
preload2 *= s->audio_preload;
ts = av_rescale_q(pkt ->dts, st ->time_base, AV_TIME_BASE_Q) - preload;
ts2= av_rescale_q(next->dts, st2->time_base, AV_TIME_BASE_Q) - preload2;
if (ts == ts2) {
ts = ((uint64_t)pkt ->dts*st ->time_base.num*AV_TIME_BASE - (uint64_t)preload *st ->time_base.den)*st2->time_base.den
- ((uint64_t)next->dts*st2->time_base.num*AV_TIME_BASE - (uint64_t)preload2*st2->time_base.den)*st ->time_base.den;
ts2 = 0;
}
comp = (ts2 > ts) - (ts2 < ts);
}
}
if (comp == 0)
return pkt->stream_index < next->stream_index;
return comp > 0;
}
int ff_interleave_packet_per_dts(AVFormatContext *s, AVPacket *pkt,
int flush, int has_packet)
{
FFFormatContext *const si = ffformatcontext(s);
int stream_count = 0;
int noninterleaved_count = 0;
int ret;
int eof = flush;
if (has_packet) {
if ((ret = ff_interleave_add_packet(s, pkt, interleave_compare_dts)) < 0)
return ret;
}
for (unsigned i = 0; i < s->nb_streams; i++) {
const AVStream *const st = s->streams[i];
const FFStream *const sti = cffstream(st);
const AVCodecParameters *const par = st->codecpar;
if (sti->last_in_packet_buffer) {
++stream_count;
} else if (par->codec_type != AVMEDIA_TYPE_ATTACHMENT &&
par->codec_id != AV_CODEC_ID_VP8 &&
par->codec_id != AV_CODEC_ID_VP9) {
++noninterleaved_count;
}
}
if (si->nb_interleaved_streams == stream_count)
flush = 1;
if (s->max_interleave_delta > 0 &&
si->packet_buffer.head &&
!flush &&
si->nb_interleaved_streams == stream_count+noninterleaved_count
) {
AVPacket *const top_pkt = &si->packet_buffer.head->pkt;
int64_t delta_dts = INT64_MIN;
int64_t top_dts = av_rescale_q(top_pkt->dts,
s->streams[top_pkt->stream_index]->time_base,
AV_TIME_BASE_Q);
for (unsigned i = 0; i < s->nb_streams; i++) {
const AVStream *const st = s->streams[i];
const FFStream *const sti = cffstream(st);
const PacketListEntry *const last = sti->last_in_packet_buffer;
int64_t last_dts;
if (!last)
continue;
last_dts = av_rescale_q(last->pkt.dts,
st->time_base,
AV_TIME_BASE_Q);
delta_dts = FFMAX(delta_dts, last_dts - top_dts);
}
if (delta_dts > s->max_interleave_delta) {
av_log(s, AV_LOG_DEBUG,
"Delay between the first packet and last packet in the "
"muxing queue is %"PRId64" > %"PRId64": forcing output\n",
delta_dts, s->max_interleave_delta);
flush = 1;
}
}
if (si->packet_buffer.head &&
eof &&
(s->flags & AVFMT_FLAG_SHORTEST) &&
si->shortest_end == AV_NOPTS_VALUE) {
AVPacket *const top_pkt = &si->packet_buffer.head->pkt;
si->shortest_end = av_rescale_q(top_pkt->dts,
s->streams[top_pkt->stream_index]->time_base,
AV_TIME_BASE_Q);
}
if (si->shortest_end != AV_NOPTS_VALUE) {
while (si->packet_buffer.head) {
PacketListEntry *pktl = si->packet_buffer.head;
AVPacket *const top_pkt = &pktl->pkt;
AVStream *const st = s->streams[top_pkt->stream_index];
FFStream *const sti = ffstream(st);
int64_t top_dts = av_rescale_q(top_pkt->dts, st->time_base,
AV_TIME_BASE_Q);
if (si->shortest_end + 1 >= top_dts)
break;
si->packet_buffer.head = pktl->next;
if (!si->packet_buffer.head)
si->packet_buffer.tail = NULL;
if (sti->last_in_packet_buffer == pktl)
sti->last_in_packet_buffer = NULL;
av_packet_unref(&pktl->pkt);
av_freep(&pktl);
flush = 0;
}
}
if (stream_count && flush) {
PacketListEntry *pktl = si->packet_buffer.head;
AVStream *const st = s->streams[pktl->pkt.stream_index];
FFStream *const sti = ffstream(st);
if (sti->last_in_packet_buffer == pktl)
sti->last_in_packet_buffer = NULL;
avpriv_packet_list_get(&si->packet_buffer, pkt);
return 1;
} else {
return 0;
}
}
int ff_interleave_packet_passthrough(AVFormatContext *s, AVPacket *pkt,
int flush, int has_packet)
{
return has_packet;
}
int ff_get_muxer_ts_offset(AVFormatContext *s, int stream_index, int64_t *offset)
{
AVStream *st;
if (stream_index < 0 || stream_index >= s->nb_streams)
return AVERROR(EINVAL);
st = s->streams[stream_index];
*offset = ffstream(st)->mux_ts_offset;
if (s->output_ts_offset)
*offset += av_rescale_q(s->output_ts_offset, AV_TIME_BASE_Q, st->time_base);
return 0;
}
const AVPacket *ff_interleaved_peek(AVFormatContext *s, int stream)
{
FFFormatContext *const si = ffformatcontext(s);
PacketListEntry *pktl = si->packet_buffer.head;
while (pktl) {
if (pktl->pkt.stream_index == stream) {
return &pktl->pkt;
}
pktl = pktl->next;
}
return NULL;
}
static int check_bitstream(AVFormatContext *s, FFStream *sti, AVPacket *pkt)
{
int ret;
if (!(s->flags & AVFMT_FLAG_AUTO_BSF))
return 1;
if (s->oformat->check_bitstream) {
if (!sti->bitstream_checked) {
if ((ret = s->oformat->check_bitstream(s, &sti->pub, pkt)) < 0)
return ret;
else if (ret == 1)
sti->bitstream_checked = 1;
}
}
return 1;
}
static int interleaved_write_packet(AVFormatContext *s, AVPacket *pkt,
int flush, int has_packet)
{
FFFormatContext *const si = ffformatcontext(s);
for (;; ) {
int ret = si->interleave_packet(s, pkt, flush, has_packet);
if (ret <= 0)
return ret;
has_packet = 0;
ret = write_packet(s, pkt);
av_packet_unref(pkt);
if (ret < 0)
return ret;
}
}
static int write_packet_common(AVFormatContext *s, AVStream *st, AVPacket *pkt, int interleaved)
{
int ret;
if (s->debug & FF_FDEBUG_TS)
av_log(s, AV_LOG_DEBUG, "%s size:%d dts:%s pts:%s\n", __FUNCTION__,
pkt->size, av_ts2str(pkt->dts), av_ts2str(pkt->pts));
guess_pkt_duration(s, st, pkt);
#if FF_API_COMPUTE_PKT_FIELDS2
if ((ret = compute_muxer_pkt_fields(s, st, pkt)) < 0 && !(s->oformat->flags & AVFMT_NOTIMESTAMPS))
return ret;
#endif
if (interleaved) {
if (pkt->dts == AV_NOPTS_VALUE && !(s->oformat->flags & AVFMT_NOTIMESTAMPS))
return AVERROR(EINVAL);
return interleaved_write_packet(s, pkt, 0, 1);
} else {
return write_packet(s, pkt);
}
}
static int write_packets_from_bsfs(AVFormatContext *s, AVStream *st, AVPacket *pkt, int interleaved)
{
FFStream *const sti = ffstream(st);
AVBSFContext *const bsfc = sti->bsfc;
int ret;
if ((ret = av_bsf_send_packet(bsfc, pkt)) < 0) {
av_log(s, AV_LOG_ERROR,
"Failed to send packet to filter %s for stream %d\n",
bsfc->filter->name, st->index);
return ret;
}
do {
ret = av_bsf_receive_packet(bsfc, pkt);
if (ret < 0) {
if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF)
return 0;
av_log(s, AV_LOG_ERROR, "Error applying bitstream filters to an output "
"packet for stream #%d: %s\n", st->index, av_err2str(ret));
if (!(s->error_recognition & AV_EF_EXPLODE) && ret != AVERROR(ENOMEM))
continue;
return ret;
}
av_packet_rescale_ts(pkt, bsfc->time_base_out, st->time_base);
ret = write_packet_common(s, st, pkt, interleaved);
if (ret >= 0 && !interleaved) // a successful write_packet_common already unrefed pkt for interleaved
av_packet_unref(pkt);
} while (ret >= 0);
return ret;
}
static int write_packets_common(AVFormatContext *s, AVPacket *pkt, int interleaved)
{
AVStream *st;
FFStream *sti;
int ret = check_packet(s, pkt);
if (ret < 0)
return ret;
st = s->streams[pkt->stream_index];
sti = ffstream(st);
ret = prepare_input_packet(s, st, pkt);
if (ret < 0)
return ret;
ret = check_bitstream(s, sti, pkt);
if (ret < 0)
return ret;
if (sti->bsfc) {
return write_packets_from_bsfs(s, st, pkt, interleaved);
} else {
return write_packet_common(s, st, pkt, interleaved);
}
}
int av_write_frame(AVFormatContext *s, AVPacket *in)
{
FFFormatContext *const si = ffformatcontext(s);
AVPacket *pkt = si->parse_pkt;
int ret;
if (!in) {
if (s->oformat->flags & AVFMT_ALLOW_FLUSH) {
ret = s->oformat->write_packet(s, NULL);
flush_if_needed(s);
if (ret >= 0 && s->pb && s->pb->error < 0)
ret = s->pb->error;
return ret;
}
return 1;
}
if (in->flags & AV_PKT_FLAG_UNCODED_FRAME) {
pkt = in;
} else {
/* We don't own in, so we have to make sure not to modify it.
avformat/mux: Fix double-free when using AVPacket.opaque_ref Up until now, ff_write_chained() copied the packet (manually, not with av_packet_move_ref()) from a packet given to it to a stack packet whose timing and stream_index is then modified before being sent to another muxer via av_(interleaved_)write_frame(). Afterwards it is intended to sync the fields of the packet relevant to freeing again; yet this only encompasses buf, side_data and side_data_elems and not the newly added opaque_ref. The other fields are not synced so that the returned packet can have a size > 0 and data != NULL despite its buf being NULL (this always happens in the interleaved codepath; before commit fe251f77c80b0512ab8907902e1dbed3f4fe1aad it could also happen in the noninterleaved one). This leads to double-frees if the interleaved codepath is used and opaque_ref is set. This commit therefore changes this by directly reusing the packet instead of a spare packet. Given that av_write_frame() does not change the packet given to it, one only needs to restore the timing information to return it as it was; for the interleaved codepath it is not possible to do likewise*, because av_interleaved_write_frame() takes ownership of the packets given to it and returns blank packets. But precisely because of this users of the interleaved codepath have no legitimate expectation that their packet will be returned unchanged. In line with av_interleaved_write_frame() ff_write_chained() therefore returns blank packets when using the interleaved codepath. Making the only user of said codepath compatible with this was trivial. *: Unless one wanted to create a full new reference. Reviewed-by: Lynne <dev@lynne.ee> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
* (ff_write_chained() relies on this fact.)
* The following avoids copying in's data unnecessarily.
* Copying side data is unavoidable as a bitstream filter
* may change it, e.g. free it on errors. */
pkt->data = in->data;
pkt->size = in->size;
ret = av_packet_copy_props(pkt, in);
if (ret < 0)
return ret;
if (in->buf) {
pkt->buf = av_buffer_ref(in->buf);
if (!pkt->buf) {
ret = AVERROR(ENOMEM);
goto fail;
}
}
}
ret = write_packets_common(s, pkt, 0/*non-interleaved*/);
fail:
// Uncoded frames using the noninterleaved codepath are also freed here
av_packet_unref(pkt);
return ret;
}
int av_interleaved_write_frame(AVFormatContext *s, AVPacket *pkt)
{
int ret;
if (pkt) {
ret = write_packets_common(s, pkt, 1/*interleaved*/);
if (ret < 0)
av_packet_unref(pkt);
return ret;
} else {
av_log(s, AV_LOG_TRACE, "av_interleaved_write_frame FLUSH\n");
return interleaved_write_packet(s, ffformatcontext(s)->parse_pkt, 1/*flush*/, 0);
}
}
int av_write_trailer(AVFormatContext *s)
{
FFFormatContext *const si = ffformatcontext(s);
AVPacket *const pkt = si->parse_pkt;
int ret1, ret = 0;
for (unsigned i = 0; i < s->nb_streams; i++) {
AVStream *const st = s->streams[i];
FFStream *const sti = ffstream(st);
if (sti->bsfc) {
ret1 = write_packets_from_bsfs(s, st, pkt, 1/*interleaved*/);
if (ret1 < 0)
av_packet_unref(pkt);
if (ret >= 0)
ret = ret1;
}
}
ret1 = interleaved_write_packet(s, pkt, 1, 0);
if (ret >= 0)
ret = ret1;
if (s->oformat->write_trailer) {
if (!(s->oformat->flags & AVFMT_NOFILE) && s->pb)
avio_write_marker(s->pb, AV_NOPTS_VALUE, AVIO_DATA_MARKER_TRAILER);
if (ret >= 0) {
ret = s->oformat->write_trailer(s);
} else {
s->oformat->write_trailer(s);
}
}
deinit_muxer(s);
if (s->pb)
avio_flush(s->pb);
if (ret == 0)
ret = s->pb ? s->pb->error : 0;
for (unsigned i = 0; i < s->nb_streams; i++) {
av_freep(&s->streams[i]->priv_data);
av_freep(&ffstream(s->streams[i])->index_entries);
}
if (s->oformat->priv_class)
av_opt_free(s->priv_data);
av_freep(&s->priv_data);
av_packet_unref(si->pkt);
return ret;
}
int av_get_output_timestamp(struct AVFormatContext *s, int stream,
int64_t *dts, int64_t *wall)
{
if (!s->oformat || !s->oformat->get_output_timestamp)
return AVERROR(ENOSYS);
s->oformat->get_output_timestamp(s, stream, dts, wall);
return 0;
}
int ff_stream_add_bitstream_filter(AVStream *st, const char *name, const char *args)
{
int ret;
const AVBitStreamFilter *bsf;
FFStream *const sti = ffstream(st);
AVBSFContext *bsfc;
av_assert0(!sti->bsfc);
if (!(bsf = av_bsf_get_by_name(name))) {
av_log(NULL, AV_LOG_ERROR, "Unknown bitstream filter '%s'\n", name);
return AVERROR_BSF_NOT_FOUND;
}
if ((ret = av_bsf_alloc(bsf, &bsfc)) < 0)
return ret;
bsfc->time_base_in = st->time_base;
if ((ret = avcodec_parameters_copy(bsfc->par_in, st->codecpar)) < 0) {
av_bsf_free(&bsfc);
return ret;
}
if (args && bsfc->filter->priv_class) {
if ((ret = av_set_options_string(bsfc->priv_data, args, "=", ":")) < 0) {
av_bsf_free(&bsfc);
return ret;
}
}
if ((ret = av_bsf_init(bsfc)) < 0) {
av_bsf_free(&bsfc);
return ret;
}
sti->bsfc = bsfc;
av_log(NULL, AV_LOG_VERBOSE,
"Automatically inserted bitstream filter '%s'; args='%s'\n",
name, args ? args : "");
return 1;
}
int ff_write_chained(AVFormatContext *dst, int dst_stream, AVPacket *pkt,
AVFormatContext *src, int interleave)
{
avformat/mux: Fix double-free when using AVPacket.opaque_ref Up until now, ff_write_chained() copied the packet (manually, not with av_packet_move_ref()) from a packet given to it to a stack packet whose timing and stream_index is then modified before being sent to another muxer via av_(interleaved_)write_frame(). Afterwards it is intended to sync the fields of the packet relevant to freeing again; yet this only encompasses buf, side_data and side_data_elems and not the newly added opaque_ref. The other fields are not synced so that the returned packet can have a size > 0 and data != NULL despite its buf being NULL (this always happens in the interleaved codepath; before commit fe251f77c80b0512ab8907902e1dbed3f4fe1aad it could also happen in the noninterleaved one). This leads to double-frees if the interleaved codepath is used and opaque_ref is set. This commit therefore changes this by directly reusing the packet instead of a spare packet. Given that av_write_frame() does not change the packet given to it, one only needs to restore the timing information to return it as it was; for the interleaved codepath it is not possible to do likewise*, because av_interleaved_write_frame() takes ownership of the packets given to it and returns blank packets. But precisely because of this users of the interleaved codepath have no legitimate expectation that their packet will be returned unchanged. In line with av_interleaved_write_frame() ff_write_chained() therefore returns blank packets when using the interleaved codepath. Making the only user of said codepath compatible with this was trivial. *: Unless one wanted to create a full new reference. Reviewed-by: Lynne <dev@lynne.ee> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
int64_t pts = pkt->pts, dts = pkt->dts, duration = pkt->duration;
int stream_index = pkt->stream_index;
AVRational time_base = pkt->time_base;
int ret;
avformat/mux: Fix double-free when using AVPacket.opaque_ref Up until now, ff_write_chained() copied the packet (manually, not with av_packet_move_ref()) from a packet given to it to a stack packet whose timing and stream_index is then modified before being sent to another muxer via av_(interleaved_)write_frame(). Afterwards it is intended to sync the fields of the packet relevant to freeing again; yet this only encompasses buf, side_data and side_data_elems and not the newly added opaque_ref. The other fields are not synced so that the returned packet can have a size > 0 and data != NULL despite its buf being NULL (this always happens in the interleaved codepath; before commit fe251f77c80b0512ab8907902e1dbed3f4fe1aad it could also happen in the noninterleaved one). This leads to double-frees if the interleaved codepath is used and opaque_ref is set. This commit therefore changes this by directly reusing the packet instead of a spare packet. Given that av_write_frame() does not change the packet given to it, one only needs to restore the timing information to return it as it was; for the interleaved codepath it is not possible to do likewise*, because av_interleaved_write_frame() takes ownership of the packets given to it and returns blank packets. But precisely because of this users of the interleaved codepath have no legitimate expectation that their packet will be returned unchanged. In line with av_interleaved_write_frame() ff_write_chained() therefore returns blank packets when using the interleaved codepath. Making the only user of said codepath compatible with this was trivial. *: Unless one wanted to create a full new reference. Reviewed-by: Lynne <dev@lynne.ee> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
pkt->stream_index = dst_stream;
avformat/mux: Fix double-free when using AVPacket.opaque_ref Up until now, ff_write_chained() copied the packet (manually, not with av_packet_move_ref()) from a packet given to it to a stack packet whose timing and stream_index is then modified before being sent to another muxer via av_(interleaved_)write_frame(). Afterwards it is intended to sync the fields of the packet relevant to freeing again; yet this only encompasses buf, side_data and side_data_elems and not the newly added opaque_ref. The other fields are not synced so that the returned packet can have a size > 0 and data != NULL despite its buf being NULL (this always happens in the interleaved codepath; before commit fe251f77c80b0512ab8907902e1dbed3f4fe1aad it could also happen in the noninterleaved one). This leads to double-frees if the interleaved codepath is used and opaque_ref is set. This commit therefore changes this by directly reusing the packet instead of a spare packet. Given that av_write_frame() does not change the packet given to it, one only needs to restore the timing information to return it as it was; for the interleaved codepath it is not possible to do likewise*, because av_interleaved_write_frame() takes ownership of the packets given to it and returns blank packets. But precisely because of this users of the interleaved codepath have no legitimate expectation that their packet will be returned unchanged. In line with av_interleaved_write_frame() ff_write_chained() therefore returns blank packets when using the interleaved codepath. Making the only user of said codepath compatible with this was trivial. *: Unless one wanted to create a full new reference. Reviewed-by: Lynne <dev@lynne.ee> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
av_packet_rescale_ts(pkt,
src->streams[stream_index]->time_base,
dst->streams[dst_stream]->time_base);
avformat/mux: Fix double-free when using AVPacket.opaque_ref Up until now, ff_write_chained() copied the packet (manually, not with av_packet_move_ref()) from a packet given to it to a stack packet whose timing and stream_index is then modified before being sent to another muxer via av_(interleaved_)write_frame(). Afterwards it is intended to sync the fields of the packet relevant to freeing again; yet this only encompasses buf, side_data and side_data_elems and not the newly added opaque_ref. The other fields are not synced so that the returned packet can have a size > 0 and data != NULL despite its buf being NULL (this always happens in the interleaved codepath; before commit fe251f77c80b0512ab8907902e1dbed3f4fe1aad it could also happen in the noninterleaved one). This leads to double-frees if the interleaved codepath is used and opaque_ref is set. This commit therefore changes this by directly reusing the packet instead of a spare packet. Given that av_write_frame() does not change the packet given to it, one only needs to restore the timing information to return it as it was; for the interleaved codepath it is not possible to do likewise*, because av_interleaved_write_frame() takes ownership of the packets given to it and returns blank packets. But precisely because of this users of the interleaved codepath have no legitimate expectation that their packet will be returned unchanged. In line with av_interleaved_write_frame() ff_write_chained() therefore returns blank packets when using the interleaved codepath. Making the only user of said codepath compatible with this was trivial. *: Unless one wanted to create a full new reference. Reviewed-by: Lynne <dev@lynne.ee> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
3 years ago
if (!interleave) {
ret = av_write_frame(dst, pkt);
/* We only have to backup and restore the fields that
* we changed ourselves, because av_write_frame() does not
* modify the packet given to it. */
pkt->pts = pts;
pkt->dts = dts;
pkt->duration = duration;
pkt->stream_index = stream_index;
pkt->time_base = time_base;
} else
ret = av_interleaved_write_frame(dst, pkt);
return ret;
}
avformat/mux: Make uncoded frames av_packet_unref() compatible Currently uncoded frames (i.e. packets whose data actually points to an AVFrame) are not refcounted. As a consequence, calling av_packet_unref() on them will not free them, but may simply make sure that they leak by losing the pointer to the frame. This commit changes this by actually making uncoded frames refcounted. In order not to rely on sizeof(AVFrame) (which is not part of the public API and so must not be used here in libavformat) the packet's data is changed to a (padded) buffer containing just a pointer to an AVFrame. Said buffer is owned by an AVBuffer with a custom free function that frees the frame as well as the buffer. Thereby the pointer/the AVBuffer owns the AVFrame. Said ownership can actually be transferred by copying and resetting the pointer, as might happen when actually writing the uncoded frames in AVOutputFormat.write_uncoded_frame() (although currently no muxer makes use of this possibility). This makes packets containing uncoded frames compatible with av_packet_unref(). This already has three advantages in interleaved mode: 1. If an error happens at the preparatory steps (before the packet is put into the interleavement queue), the frame is properly freed. 2. If the trailer is never written, the frames still in the interleavement queue will now be properly freed by ff_packet_list_free(). 3. The custom code for moving the packet to the packet list in ff_interleave_add_packet() can be removed. It will also simplify fixing further memleaks in future commits. Suggested-by: Marton Balint <cus@passwd.hu> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
5 years ago
static void uncoded_frame_free(void *unused, uint8_t *data)
{
av_frame_free((AVFrame **)data);
av_free(data);
}
static int write_uncoded_frame_internal(AVFormatContext *s, int stream_index,
AVFrame *frame, int interleaved)
{
FFFormatContext *const si = ffformatcontext(s);
AVPacket *pkt = si->parse_pkt;
av_assert0(s->oformat);
avformat/mux: Make uncoded frames av_packet_unref() compatible Currently uncoded frames (i.e. packets whose data actually points to an AVFrame) are not refcounted. As a consequence, calling av_packet_unref() on them will not free them, but may simply make sure that they leak by losing the pointer to the frame. This commit changes this by actually making uncoded frames refcounted. In order not to rely on sizeof(AVFrame) (which is not part of the public API and so must not be used here in libavformat) the packet's data is changed to a (padded) buffer containing just a pointer to an AVFrame. Said buffer is owned by an AVBuffer with a custom free function that frees the frame as well as the buffer. Thereby the pointer/the AVBuffer owns the AVFrame. Said ownership can actually be transferred by copying and resetting the pointer, as might happen when actually writing the uncoded frames in AVOutputFormat.write_uncoded_frame() (although currently no muxer makes use of this possibility). This makes packets containing uncoded frames compatible with av_packet_unref(). This already has three advantages in interleaved mode: 1. If an error happens at the preparatory steps (before the packet is put into the interleavement queue), the frame is properly freed. 2. If the trailer is never written, the frames still in the interleavement queue will now be properly freed by ff_packet_list_free(). 3. The custom code for moving the packet to the packet list in ff_interleave_add_packet() can be removed. It will also simplify fixing further memleaks in future commits. Suggested-by: Marton Balint <cus@passwd.hu> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
5 years ago
if (!s->oformat->write_uncoded_frame) {
av_frame_free(&frame);
return AVERROR(ENOSYS);
avformat/mux: Make uncoded frames av_packet_unref() compatible Currently uncoded frames (i.e. packets whose data actually points to an AVFrame) are not refcounted. As a consequence, calling av_packet_unref() on them will not free them, but may simply make sure that they leak by losing the pointer to the frame. This commit changes this by actually making uncoded frames refcounted. In order not to rely on sizeof(AVFrame) (which is not part of the public API and so must not be used here in libavformat) the packet's data is changed to a (padded) buffer containing just a pointer to an AVFrame. Said buffer is owned by an AVBuffer with a custom free function that frees the frame as well as the buffer. Thereby the pointer/the AVBuffer owns the AVFrame. Said ownership can actually be transferred by copying and resetting the pointer, as might happen when actually writing the uncoded frames in AVOutputFormat.write_uncoded_frame() (although currently no muxer makes use of this possibility). This makes packets containing uncoded frames compatible with av_packet_unref(). This already has three advantages in interleaved mode: 1. If an error happens at the preparatory steps (before the packet is put into the interleavement queue), the frame is properly freed. 2. If the trailer is never written, the frames still in the interleavement queue will now be properly freed by ff_packet_list_free(). 3. The custom code for moving the packet to the packet list in ff_interleave_add_packet() can be removed. It will also simplify fixing further memleaks in future commits. Suggested-by: Marton Balint <cus@passwd.hu> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
5 years ago
}
if (!frame) {
pkt = NULL;
} else {
avformat/mux: Make uncoded frames av_packet_unref() compatible Currently uncoded frames (i.e. packets whose data actually points to an AVFrame) are not refcounted. As a consequence, calling av_packet_unref() on them will not free them, but may simply make sure that they leak by losing the pointer to the frame. This commit changes this by actually making uncoded frames refcounted. In order not to rely on sizeof(AVFrame) (which is not part of the public API and so must not be used here in libavformat) the packet's data is changed to a (padded) buffer containing just a pointer to an AVFrame. Said buffer is owned by an AVBuffer with a custom free function that frees the frame as well as the buffer. Thereby the pointer/the AVBuffer owns the AVFrame. Said ownership can actually be transferred by copying and resetting the pointer, as might happen when actually writing the uncoded frames in AVOutputFormat.write_uncoded_frame() (although currently no muxer makes use of this possibility). This makes packets containing uncoded frames compatible with av_packet_unref(). This already has three advantages in interleaved mode: 1. If an error happens at the preparatory steps (before the packet is put into the interleavement queue), the frame is properly freed. 2. If the trailer is never written, the frames still in the interleavement queue will now be properly freed by ff_packet_list_free(). 3. The custom code for moving the packet to the packet list in ff_interleave_add_packet() can be removed. It will also simplify fixing further memleaks in future commits. Suggested-by: Marton Balint <cus@passwd.hu> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
5 years ago
size_t bufsize = sizeof(frame) + AV_INPUT_BUFFER_PADDING_SIZE;
AVFrame **framep = av_mallocz(bufsize);
if (!framep)
goto fail;
pkt->buf = av_buffer_create((void *)framep, bufsize,
avformat/mux: Make uncoded frames av_packet_unref() compatible Currently uncoded frames (i.e. packets whose data actually points to an AVFrame) are not refcounted. As a consequence, calling av_packet_unref() on them will not free them, but may simply make sure that they leak by losing the pointer to the frame. This commit changes this by actually making uncoded frames refcounted. In order not to rely on sizeof(AVFrame) (which is not part of the public API and so must not be used here in libavformat) the packet's data is changed to a (padded) buffer containing just a pointer to an AVFrame. Said buffer is owned by an AVBuffer with a custom free function that frees the frame as well as the buffer. Thereby the pointer/the AVBuffer owns the AVFrame. Said ownership can actually be transferred by copying and resetting the pointer, as might happen when actually writing the uncoded frames in AVOutputFormat.write_uncoded_frame() (although currently no muxer makes use of this possibility). This makes packets containing uncoded frames compatible with av_packet_unref(). This already has three advantages in interleaved mode: 1. If an error happens at the preparatory steps (before the packet is put into the interleavement queue), the frame is properly freed. 2. If the trailer is never written, the frames still in the interleavement queue will now be properly freed by ff_packet_list_free(). 3. The custom code for moving the packet to the packet list in ff_interleave_add_packet() can be removed. It will also simplify fixing further memleaks in future commits. Suggested-by: Marton Balint <cus@passwd.hu> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
5 years ago
uncoded_frame_free, NULL, 0);
if (!pkt->buf) {
avformat/mux: Make uncoded frames av_packet_unref() compatible Currently uncoded frames (i.e. packets whose data actually points to an AVFrame) are not refcounted. As a consequence, calling av_packet_unref() on them will not free them, but may simply make sure that they leak by losing the pointer to the frame. This commit changes this by actually making uncoded frames refcounted. In order not to rely on sizeof(AVFrame) (which is not part of the public API and so must not be used here in libavformat) the packet's data is changed to a (padded) buffer containing just a pointer to an AVFrame. Said buffer is owned by an AVBuffer with a custom free function that frees the frame as well as the buffer. Thereby the pointer/the AVBuffer owns the AVFrame. Said ownership can actually be transferred by copying and resetting the pointer, as might happen when actually writing the uncoded frames in AVOutputFormat.write_uncoded_frame() (although currently no muxer makes use of this possibility). This makes packets containing uncoded frames compatible with av_packet_unref(). This already has three advantages in interleaved mode: 1. If an error happens at the preparatory steps (before the packet is put into the interleavement queue), the frame is properly freed. 2. If the trailer is never written, the frames still in the interleavement queue will now be properly freed by ff_packet_list_free(). 3. The custom code for moving the packet to the packet list in ff_interleave_add_packet() can be removed. It will also simplify fixing further memleaks in future commits. Suggested-by: Marton Balint <cus@passwd.hu> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
5 years ago
av_free(framep);
fail:
av_frame_free(&frame);
return AVERROR(ENOMEM);
}
*framep = frame;
pkt->data = (void *)framep;
pkt->size = sizeof(frame);
pkt->pts =
pkt->dts = frame->pts;
#if FF_API_PKT_DURATION
FF_DISABLE_DEPRECATION_WARNINGS
if (frame->pkt_duration)
pkt->duration = frame->pkt_duration;
else
FF_ENABLE_DEPRECATION_WARNINGS
#endif
pkt->duration = frame->duration;
pkt->stream_index = stream_index;
pkt->flags |= AV_PKT_FLAG_UNCODED_FRAME;
}
return interleaved ? av_interleaved_write_frame(s, pkt) :
av_write_frame(s, pkt);
}
int av_write_uncoded_frame(AVFormatContext *s, int stream_index,
AVFrame *frame)
{
return write_uncoded_frame_internal(s, stream_index, frame, 0);
}
int av_interleaved_write_uncoded_frame(AVFormatContext *s, int stream_index,
AVFrame *frame)
{
return write_uncoded_frame_internal(s, stream_index, frame, 1);
}
int av_write_uncoded_frame_query(AVFormatContext *s, int stream_index)
{
av_assert0(s->oformat);
if (!s->oformat->write_uncoded_frame)
return AVERROR(ENOSYS);
return s->oformat->write_uncoded_frame(s, stream_index, NULL,
AV_WRITE_UNCODED_FRAME_QUERY);
}