3 * Copyright (c) 2001 Fabrice Bellard
4 * Copyright (c) 2009 Baptiste Coudurier <baptiste dot coudurier at gmail dot com>
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
26 //#define DEBUG_METADATA
27 //#define MOV_EXPORT_ALL_METADATA
29 #include "libavutil/intreadwrite.h"
30 #include "libavutil/avstring.h"
34 #include "libavcodec/mpeg4audio.h"
35 #include "libavcodec/mpegaudiodata.h"
42 * First version by Francois Revol revol@free.fr
43 * Seek function by Gael Chardon gael.dev@4now.net
45 * Features and limitations:
46 * - reads most of the QT files I have (at least the structure),
47 * Sample QuickTime files with mp3 audio can be found at: http://www.3ivx.com/showcase.html
48 * - the code is quite ugly... maybe I won't do it recursive next time :-)
50 * Funny I didn't know about http://sourceforge.net/projects/qt-ffmpeg/
51 * when coding this :) (it's a writer anyway)
53 * Reference documents:
54 * http://www.geocities.com/xhelmboyx/quicktime/formats/qtm-layout.txt
56 * http://developer.apple.com/documentation/QuickTime/QTFF/
57 * http://developer.apple.com/documentation/QuickTime/QTFF/qtff.pdf
58 * QuickTime is a trademark of Apple (AFAIK :))
61 #include "qtpalette.h"
67 /* XXX: it's the first time I make a recursive parser I think... sorry if it's ugly :P */
69 /* those functions parse an atom */
71 0: continue to parse next atom
72 <0: error occurred, exit
74 /* links atom IDs to parse functions */
75 typedef struct MOVParseTableEntry {
77 int (*parse)(MOVContext *ctx, ByteIOContext *pb, MOVAtom atom);
80 static const MOVParseTableEntry mov_default_parse_table[];
82 static int mov_metadata_trkn(MOVContext *c, ByteIOContext *pb, unsigned len)
86 get_be16(pb); // unknown
87 snprintf(buf, sizeof(buf), "%d", get_be16(pb));
88 av_metadata_set(&c->fc->metadata, "track", buf);
90 get_be16(pb); // total tracks
95 static int mov_read_udta_string(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
97 #ifdef MOV_EXPORT_ALL_METADATA
100 char str[1024], key2[16], language[4] = {0};
101 const char *key = NULL;
103 int (*parse)(MOVContext*, ByteIOContext*, unsigned) = NULL;
106 case MKTAG(0xa9,'n','a','m'): key = "title"; break;
107 case MKTAG(0xa9,'a','u','t'):
108 case MKTAG(0xa9,'A','R','T'):
109 case MKTAG(0xa9,'w','r','t'): key = "author"; break;
110 case MKTAG(0xa9,'c','p','y'): key = "copyright"; break;
111 case MKTAG(0xa9,'c','m','t'):
112 case MKTAG(0xa9,'i','n','f'): key = "comment"; break;
113 case MKTAG(0xa9,'a','l','b'): key = "album"; break;
114 case MKTAG(0xa9,'d','a','y'): key = "year"; break;
115 case MKTAG(0xa9,'g','e','n'): key = "genre"; break;
116 case MKTAG(0xa9,'t','o','o'):
117 case MKTAG(0xa9,'e','n','c'): key = "muxer"; break;
118 case MKTAG( 't','r','k','n'): key = "track";
119 parse = mov_metadata_trkn; break;
122 if (c->itunes_metadata && atom.size > 8) {
123 int data_size = get_be32(pb);
124 int tag = get_le32(pb);
125 if (tag == MKTAG('d','a','t','a')) {
126 get_be32(pb); // type
127 get_be32(pb); // unknown
128 str_size = data_size - 16;
131 } else if (atom.size > 4 && key && !c->itunes_metadata) {
132 str_size = get_be16(pb); // string length
133 ff_mov_lang_to_iso639(get_be16(pb), language);
136 str_size = atom.size;
138 #ifdef MOV_EXPORT_ALL_METADATA
140 snprintf(tmp_key, 5, "%.4s", (char*)&atom.type);
150 str_size = FFMIN3(sizeof(str)-1, str_size, atom.size);
153 parse(c, pb, str_size);
155 get_buffer(pb, str, str_size);
157 av_metadata_set(&c->fc->metadata, key, str);
158 if (*language && strcmp(language, "und")) {
159 snprintf(key2, sizeof(key2), "%s-%s", key, language);
160 av_metadata_set(&c->fc->metadata, key2, str);
163 #ifdef DEBUG_METADATA
164 av_log(c->fc, AV_LOG_DEBUG, "lang \"%3s\" ", language);
165 av_log(c->fc, AV_LOG_DEBUG, "tag \"%s\" value \"%s\" atom \"%.4s\" %d %lld\n",
166 key, str, (char*)&atom.type, str_size, atom.size);
172 static int mov_read_default(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
174 int64_t total_size = 0;
179 a.offset = atom.offset;
182 atom.size = INT64_MAX;
183 while(((total_size + 8) < atom.size) && !url_feof(pb) && !err) {
184 int (*parse)(MOVContext*, ByteIOContext*, MOVAtom) = NULL;
188 a.size = get_be32(pb);
189 a.type = get_le32(pb);
193 dprintf(c->fc, "type: %08x %.4s sz: %"PRIx64" %"PRIx64" %"PRIx64"\n",
194 a.type, (char*)&a.type, a.size, atom.size, total_size);
195 if (a.size == 1) { /* 64 bit extended size */
196 a.size = get_be64(pb) - 8;
201 a.size = atom.size - total_size;
208 a.size = FFMIN(a.size, atom.size - total_size);
210 for (i = 0; mov_default_parse_table[i].type; i++)
211 if (mov_default_parse_table[i].type == a.type) {
212 parse = mov_default_parse_table[i].parse;
216 // container is user data
217 if (!parse && (atom.type == MKTAG('u','d','t','a') ||
218 atom.type == MKTAG('i','l','s','t')))
219 parse = mov_read_udta_string;
221 if (!parse) { /* skip leaf atoms data */
222 url_fskip(pb, a.size);
224 int64_t start_pos = url_ftell(pb);
226 err = parse(c, pb, a);
227 if (url_is_streamed(pb) && c->found_moov && c->found_mdat)
229 left = a.size - url_ftell(pb) + start_pos;
230 if (left > 0) /* skip garbage at atom end */
235 total_size += a.size;
238 if (!err && total_size < atom.size && atom.size < 0x7ffff)
239 url_fskip(pb, atom.size - total_size);
244 static int mov_read_dref(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
246 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
247 MOVStreamContext *sc = st->priv_data;
250 get_be32(pb); // version + flags
251 entries = get_be32(pb);
252 if (entries >= UINT_MAX / sizeof(*sc->drefs))
254 sc->drefs = av_mallocz(entries * sizeof(*sc->drefs));
256 return AVERROR(ENOMEM);
257 sc->drefs_count = entries;
259 for (i = 0; i < sc->drefs_count; i++) {
260 MOVDref *dref = &sc->drefs[i];
261 uint32_t size = get_be32(pb);
262 int64_t next = url_ftell(pb) + size - 4;
264 dref->type = get_le32(pb);
265 get_be32(pb); // version + flags
266 dprintf(c->fc, "type %.4s size %d\n", (char*)&dref->type, size);
268 if (dref->type == MKTAG('a','l','i','s') && size > 150) {
269 /* macintosh alias record */
270 uint16_t volume_len, len;
276 volume_len = get_byte(pb);
277 volume_len = FFMIN(volume_len, 27);
278 get_buffer(pb, volume, 27);
279 volume[volume_len] = 0;
280 av_log(c->fc, AV_LOG_DEBUG, "volume %s, len %d\n", volume, volume_len);
284 for (type = 0; type != -1 && url_ftell(pb) < next; ) {
287 av_log(c->fc, AV_LOG_DEBUG, "type %d, len %d\n", type, len);
290 if (type == 2) { // absolute path
292 dref->path = av_mallocz(len+1);
294 return AVERROR(ENOMEM);
295 get_buffer(pb, dref->path, len);
296 if (len > volume_len && !strncmp(dref->path, volume, volume_len)) {
298 memmove(dref->path, dref->path+volume_len, len);
301 for (j = 0; j < len; j++)
302 if (dref->path[j] == ':')
304 av_log(c->fc, AV_LOG_DEBUG, "path %s\n", dref->path);
309 url_fseek(pb, next, SEEK_SET);
314 static int mov_read_hdlr(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
320 if (c->fc->nb_streams < 1) // meta before first trak
323 st = c->fc->streams[c->fc->nb_streams-1];
325 get_byte(pb); /* version */
326 get_be24(pb); /* flags */
329 ctype = get_le32(pb);
330 type = get_le32(pb); /* component subtype */
332 dprintf(c->fc, "ctype= %c%c%c%c (0x%08x)\n", *((char *)&ctype), ((char *)&ctype)[1],
333 ((char *)&ctype)[2], ((char *)&ctype)[3], (int) ctype);
334 dprintf(c->fc, "stype= %c%c%c%c\n",
335 *((char *)&type), ((char *)&type)[1], ((char *)&type)[2], ((char *)&type)[3]);
337 if (type == MKTAG('v','i','d','e'))
338 st->codec->codec_type = CODEC_TYPE_VIDEO;
339 else if(type == MKTAG('s','o','u','n'))
340 st->codec->codec_type = CODEC_TYPE_AUDIO;
341 else if(type == MKTAG('m','1','a',' '))
342 st->codec->codec_id = CODEC_ID_MP2;
343 else if(type == MKTAG('s','u','b','p')) {
344 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
346 get_be32(pb); /* component manufacture */
347 get_be32(pb); /* component flags */
348 get_be32(pb); /* component flags mask */
351 return 0; /* nothing left to read */
353 url_fskip(pb, atom.size - (url_ftell(pb) - atom.offset));
357 static int mp4_read_descr_len(ByteIOContext *pb)
362 int c = get_byte(pb);
363 len = (len << 7) | (c & 0x7f);
370 static int mp4_read_descr(MOVContext *c, ByteIOContext *pb, int *tag)
374 len = mp4_read_descr_len(pb);
375 dprintf(c->fc, "MPEG4 description: tag=0x%02x len=%d\n", *tag, len);
379 #define MP4ESDescrTag 0x03
380 #define MP4DecConfigDescrTag 0x04
381 #define MP4DecSpecificDescrTag 0x05
383 static const AVCodecTag mp4_audio_types[] = {
384 { CODEC_ID_MP3ON4, 29 }, /* old mp3on4 draft */
385 { CODEC_ID_MP3ON4, 32 }, /* layer 1 */
386 { CODEC_ID_MP3ON4, 33 }, /* layer 2 */
387 { CODEC_ID_MP3ON4, 34 }, /* layer 3 */
388 { CODEC_ID_NONE, 0 },
391 static int mov_read_esds(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
393 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
396 get_be32(pb); /* version + flags */
397 len = mp4_read_descr(c, pb, &tag);
398 if (tag == MP4ESDescrTag) {
399 get_be16(pb); /* ID */
400 get_byte(pb); /* priority */
402 get_be16(pb); /* ID */
404 len = mp4_read_descr(c, pb, &tag);
405 if (tag == MP4DecConfigDescrTag) {
406 int object_type_id = get_byte(pb);
407 get_byte(pb); /* stream type */
408 get_be24(pb); /* buffer size db */
409 get_be32(pb); /* max bitrate */
410 get_be32(pb); /* avg bitrate */
412 st->codec->codec_id= codec_get_id(ff_mp4_obj_type, object_type_id);
413 dprintf(c->fc, "esds object type id %d\n", object_type_id);
414 len = mp4_read_descr(c, pb, &tag);
415 if (tag == MP4DecSpecificDescrTag) {
416 dprintf(c->fc, "Specific MPEG4 header len=%d\n", len);
417 if((uint64_t)len > (1<<30))
419 st->codec->extradata = av_mallocz(len + FF_INPUT_BUFFER_PADDING_SIZE);
420 if (!st->codec->extradata)
421 return AVERROR(ENOMEM);
422 get_buffer(pb, st->codec->extradata, len);
423 st->codec->extradata_size = len;
424 if (st->codec->codec_id == CODEC_ID_AAC) {
425 MPEG4AudioConfig cfg;
426 ff_mpeg4audio_get_config(&cfg, st->codec->extradata,
427 st->codec->extradata_size);
428 if (cfg.chan_config > 7)
430 st->codec->channels = ff_mpeg4audio_channels[cfg.chan_config];
431 if (cfg.object_type == 29 && cfg.sampling_index < 3) // old mp3on4
432 st->codec->sample_rate = ff_mpa_freq_tab[cfg.sampling_index];
434 st->codec->sample_rate = cfg.sample_rate; // ext sample rate ?
435 dprintf(c->fc, "mp4a config channels %d obj %d ext obj %d "
436 "sample rate %d ext sample rate %d\n", st->codec->channels,
437 cfg.object_type, cfg.ext_object_type,
438 cfg.sample_rate, cfg.ext_sample_rate);
439 if (!(st->codec->codec_id = codec_get_id(mp4_audio_types,
441 st->codec->codec_id = CODEC_ID_AAC;
448 static int mov_read_pasp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
450 const int num = get_be32(pb);
451 const int den = get_be32(pb);
452 AVStream * const st = c->fc->streams[c->fc->nb_streams-1];
454 if ((st->sample_aspect_ratio.den != 1 || st->sample_aspect_ratio.num) && // default
455 (den != st->sample_aspect_ratio.den || num != st->sample_aspect_ratio.num))
456 av_log(c->fc, AV_LOG_WARNING,
457 "sample aspect ratio already set to %d:%d, overriding by 'pasp' atom\n",
458 st->sample_aspect_ratio.num, st->sample_aspect_ratio.den);
459 st->sample_aspect_ratio.num = num;
460 st->sample_aspect_ratio.den = den;
465 /* this atom contains actual media data */
466 static int mov_read_mdat(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
468 if(atom.size == 0) /* wrong one (MP4) */
471 return 0; /* now go for moov */
474 static int mov_read_ftyp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
476 uint32_t type = get_le32(pb);
478 if (type != MKTAG('q','t',' ',' '))
480 av_log(c->fc, AV_LOG_DEBUG, "ISO: File Type Major Brand: %.4s\n",(char *)&type);
481 get_be32(pb); /* minor version */
482 url_fskip(pb, atom.size - 8);
486 /* this atom should contain all header atoms */
487 static int mov_read_moov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
489 if (mov_read_default(c, pb, atom) < 0)
491 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
492 /* so we don't parse the whole file if over a network */
494 return 0; /* now go for mdat */
497 static int mov_read_moof(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
499 c->fragment.moof_offset = url_ftell(pb) - 8;
500 dprintf(c->fc, "moof offset %llx\n", c->fragment.moof_offset);
501 return mov_read_default(c, pb, atom);
504 static int mov_read_mdhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
506 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
507 MOVStreamContext *sc = st->priv_data;
508 int version = get_byte(pb);
509 char language[4] = {0};
513 return -1; /* unsupported */
515 get_be24(pb); /* flags */
520 get_be32(pb); /* creation time */
521 get_be32(pb); /* modification time */
524 sc->time_scale = get_be32(pb);
525 st->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
527 lang = get_be16(pb); /* language */
528 if (ff_mov_lang_to_iso639(lang, language))
529 av_metadata_set(&st->metadata, "language", language);
530 get_be16(pb); /* quality */
535 static int mov_read_mvhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
537 int version = get_byte(pb); /* version */
538 get_be24(pb); /* flags */
544 get_be32(pb); /* creation time */
545 get_be32(pb); /* modification time */
547 c->time_scale = get_be32(pb); /* time scale */
549 dprintf(c->fc, "time scale = %i\n", c->time_scale);
551 c->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
552 get_be32(pb); /* preferred scale */
554 get_be16(pb); /* preferred volume */
556 url_fskip(pb, 10); /* reserved */
558 url_fskip(pb, 36); /* display matrix */
560 get_be32(pb); /* preview time */
561 get_be32(pb); /* preview duration */
562 get_be32(pb); /* poster time */
563 get_be32(pb); /* selection time */
564 get_be32(pb); /* selection duration */
565 get_be32(pb); /* current time */
566 get_be32(pb); /* next track ID */
571 static int mov_read_smi(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
573 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
575 if((uint64_t)atom.size > (1<<30))
578 // currently SVQ3 decoder expect full STSD header - so let's fake it
579 // this should be fixed and just SMI header should be passed
580 av_free(st->codec->extradata);
581 st->codec->extradata = av_mallocz(atom.size + 0x5a + FF_INPUT_BUFFER_PADDING_SIZE);
582 if (!st->codec->extradata)
583 return AVERROR(ENOMEM);
584 st->codec->extradata_size = 0x5a + atom.size;
585 memcpy(st->codec->extradata, "SVQ3", 4); // fake
586 get_buffer(pb, st->codec->extradata + 0x5a, atom.size);
587 dprintf(c->fc, "Reading SMI %"PRId64" %s\n", atom.size, st->codec->extradata + 0x5a);
591 static int mov_read_enda(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
593 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
594 int little_endian = get_be16(pb);
596 dprintf(c->fc, "enda %d\n", little_endian);
597 if (little_endian == 1) {
598 switch (st->codec->codec_id) {
599 case CODEC_ID_PCM_S24BE:
600 st->codec->codec_id = CODEC_ID_PCM_S24LE;
602 case CODEC_ID_PCM_S32BE:
603 st->codec->codec_id = CODEC_ID_PCM_S32LE;
605 case CODEC_ID_PCM_F32BE:
606 st->codec->codec_id = CODEC_ID_PCM_F32LE;
608 case CODEC_ID_PCM_F64BE:
609 st->codec->codec_id = CODEC_ID_PCM_F64LE;
618 /* FIXME modify qdm2/svq3/h264 decoders to take full atom as extradata */
619 static int mov_read_extradata(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
625 if (c->fc->nb_streams < 1) // will happen with jp2 files
627 st= c->fc->streams[c->fc->nb_streams-1];
628 size= (uint64_t)st->codec->extradata_size + atom.size + 8 + FF_INPUT_BUFFER_PADDING_SIZE;
629 if(size > INT_MAX || (uint64_t)atom.size > INT_MAX)
631 buf= av_realloc(st->codec->extradata, size);
634 st->codec->extradata= buf;
635 buf+= st->codec->extradata_size;
636 st->codec->extradata_size= size - FF_INPUT_BUFFER_PADDING_SIZE;
637 AV_WB32( buf , atom.size + 8);
638 AV_WL32( buf + 4, atom.type);
639 get_buffer(pb, buf + 8, atom.size);
643 static int mov_read_wave(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
645 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
647 if((uint64_t)atom.size > (1<<30))
650 if (st->codec->codec_id == CODEC_ID_QDM2) {
651 // pass all frma atom to codec, needed at least for QDM2
652 av_free(st->codec->extradata);
653 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
654 if (!st->codec->extradata)
655 return AVERROR(ENOMEM);
656 st->codec->extradata_size = atom.size;
657 get_buffer(pb, st->codec->extradata, atom.size);
658 } else if (atom.size > 8) { /* to read frma, esds atoms */
659 if (mov_read_default(c, pb, atom) < 0)
662 url_fskip(pb, atom.size);
667 * This function reads atom content and puts data in extradata without tag
668 * nor size unlike mov_read_extradata.
670 static int mov_read_glbl(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
672 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
674 if((uint64_t)atom.size > (1<<30))
677 av_free(st->codec->extradata);
678 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
679 if (!st->codec->extradata)
680 return AVERROR(ENOMEM);
681 st->codec->extradata_size = atom.size;
682 get_buffer(pb, st->codec->extradata, atom.size);
686 static int mov_read_stco(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
688 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
689 MOVStreamContext *sc = st->priv_data;
690 unsigned int i, entries;
692 get_byte(pb); /* version */
693 get_be24(pb); /* flags */
695 entries = get_be32(pb);
697 if(entries >= UINT_MAX/sizeof(int64_t))
700 sc->chunk_offsets = av_malloc(entries * sizeof(int64_t));
701 if (!sc->chunk_offsets)
702 return AVERROR(ENOMEM);
703 sc->chunk_count = entries;
705 if (atom.type == MKTAG('s','t','c','o'))
706 for(i=0; i<entries; i++)
707 sc->chunk_offsets[i] = get_be32(pb);
708 else if (atom.type == MKTAG('c','o','6','4'))
709 for(i=0; i<entries; i++)
710 sc->chunk_offsets[i] = get_be64(pb);
718 * Compute codec id for 'lpcm' tag.
719 * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
721 static enum CodecID mov_get_lpcm_codec_id(int bps, int flags)
723 if (flags & 1) { // floating point
724 if (flags & 2) { // big endian
725 if (bps == 32) return CODEC_ID_PCM_F32BE;
726 else if (bps == 64) return CODEC_ID_PCM_F64BE;
728 if (bps == 32) return CODEC_ID_PCM_F32LE;
729 else if (bps == 64) return CODEC_ID_PCM_F64LE;
735 if (flags & 4) return CODEC_ID_PCM_S8;
736 else return CODEC_ID_PCM_U8;
737 else if (bps == 16) return CODEC_ID_PCM_S16BE;
738 else if (bps == 24) return CODEC_ID_PCM_S24BE;
739 else if (bps == 32) return CODEC_ID_PCM_S32BE;
742 if (flags & 4) return CODEC_ID_PCM_S8;
743 else return CODEC_ID_PCM_U8;
744 else if (bps == 16) return CODEC_ID_PCM_S16LE;
745 else if (bps == 24) return CODEC_ID_PCM_S24LE;
746 else if (bps == 32) return CODEC_ID_PCM_S32LE;
749 return CODEC_ID_NONE;
752 static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
754 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
755 MOVStreamContext *sc = st->priv_data;
756 int j, entries, pseudo_stream_id;
758 get_byte(pb); /* version */
759 get_be24(pb); /* flags */
761 entries = get_be32(pb);
763 for(pseudo_stream_id=0; pseudo_stream_id<entries; pseudo_stream_id++) {
764 //Parsing Sample description table
767 MOVAtom a = { 0, 0, 0 };
768 int64_t start_pos = url_ftell(pb);
769 int size = get_be32(pb); /* size */
770 uint32_t format = get_le32(pb); /* data format */
772 get_be32(pb); /* reserved */
773 get_be16(pb); /* reserved */
774 dref_id = get_be16(pb);
776 if (st->codec->codec_tag &&
777 st->codec->codec_tag != format &&
778 (c->fc->video_codec_id ? codec_get_id(codec_movvideo_tags, format) != c->fc->video_codec_id
779 : st->codec->codec_tag != MKTAG('j','p','e','g'))
781 /* Multiple fourcc, we skip JPEG. This is not correct, we should
782 * export it as a separate AVStream but this needs a few changes
783 * in the MOV demuxer, patch welcome. */
784 av_log(c->fc, AV_LOG_WARNING, "multiple fourcc not supported\n");
785 url_fskip(pb, size - (url_ftell(pb) - start_pos));
788 sc->pseudo_stream_id = st->codec->codec_tag ? -1 : pseudo_stream_id;
789 sc->dref_id= dref_id;
791 st->codec->codec_tag = format;
792 id = codec_get_id(codec_movaudio_tags, format);
793 if (id<=0 && (format&0xFFFF) == 'm'+('s'<<8))
794 id = codec_get_id(codec_wav_tags, bswap_32(format)&0xFFFF);
796 if (st->codec->codec_type != CODEC_TYPE_VIDEO && id > 0) {
797 st->codec->codec_type = CODEC_TYPE_AUDIO;
798 } else if (st->codec->codec_type != CODEC_TYPE_AUDIO && /* do not overwrite codec type */
799 format && format != MKTAG('m','p','4','s')) { /* skip old asf mpeg4 tag */
800 id = codec_get_id(codec_movvideo_tags, format);
802 id = codec_get_id(codec_bmp_tags, format);
804 st->codec->codec_type = CODEC_TYPE_VIDEO;
805 else if(st->codec->codec_type == CODEC_TYPE_DATA){
806 id = codec_get_id(ff_codec_movsubtitle_tags, format);
808 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
812 dprintf(c->fc, "size=%d 4CC= %c%c%c%c codec_type=%d\n", size,
813 (format >> 0) & 0xff, (format >> 8) & 0xff, (format >> 16) & 0xff,
814 (format >> 24) & 0xff, st->codec->codec_type);
816 if(st->codec->codec_type==CODEC_TYPE_VIDEO) {
817 uint8_t codec_name[32];
818 unsigned int color_depth;
821 st->codec->codec_id = id;
822 get_be16(pb); /* version */
823 get_be16(pb); /* revision level */
824 get_be32(pb); /* vendor */
825 get_be32(pb); /* temporal quality */
826 get_be32(pb); /* spatial quality */
828 st->codec->width = get_be16(pb); /* width */
829 st->codec->height = get_be16(pb); /* height */
831 get_be32(pb); /* horiz resolution */
832 get_be32(pb); /* vert resolution */
833 get_be32(pb); /* data size, always 0 */
834 get_be16(pb); /* frames per samples */
836 get_buffer(pb, codec_name, 32); /* codec name, pascal string */
837 if (codec_name[0] <= 31) {
838 memcpy(st->codec->codec_name, &codec_name[1],codec_name[0]);
839 st->codec->codec_name[codec_name[0]] = 0;
842 st->codec->bits_per_coded_sample = get_be16(pb); /* depth */
843 st->codec->color_table_id = get_be16(pb); /* colortable id */
844 dprintf(c->fc, "depth %d, ctab id %d\n",
845 st->codec->bits_per_coded_sample, st->codec->color_table_id);
846 /* figure out the palette situation */
847 color_depth = st->codec->bits_per_coded_sample & 0x1F;
848 color_greyscale = st->codec->bits_per_coded_sample & 0x20;
850 /* if the depth is 2, 4, or 8 bpp, file is palettized */
851 if ((color_depth == 2) || (color_depth == 4) ||
852 (color_depth == 8)) {
853 /* for palette traversal */
854 unsigned int color_start, color_count, color_end;
855 unsigned char r, g, b;
857 st->codec->palctrl = av_malloc(sizeof(*st->codec->palctrl));
858 if (color_greyscale) {
859 int color_index, color_dec;
860 /* compute the greyscale palette */
861 st->codec->bits_per_coded_sample = color_depth;
862 color_count = 1 << color_depth;
864 color_dec = 256 / (color_count - 1);
865 for (j = 0; j < color_count; j++) {
866 r = g = b = color_index;
867 st->codec->palctrl->palette[j] =
868 (r << 16) | (g << 8) | (b);
869 color_index -= color_dec;
873 } else if (st->codec->color_table_id) {
874 const uint8_t *color_table;
875 /* if flag bit 3 is set, use the default palette */
876 color_count = 1 << color_depth;
877 if (color_depth == 2)
878 color_table = ff_qt_default_palette_4;
879 else if (color_depth == 4)
880 color_table = ff_qt_default_palette_16;
882 color_table = ff_qt_default_palette_256;
884 for (j = 0; j < color_count; j++) {
885 r = color_table[j * 4 + 0];
886 g = color_table[j * 4 + 1];
887 b = color_table[j * 4 + 2];
888 st->codec->palctrl->palette[j] =
889 (r << 16) | (g << 8) | (b);
892 /* load the palette from the file */
893 color_start = get_be32(pb);
894 color_count = get_be16(pb);
895 color_end = get_be16(pb);
896 if ((color_start <= 255) &&
897 (color_end <= 255)) {
898 for (j = color_start; j <= color_end; j++) {
899 /* each R, G, or B component is 16 bits;
900 * only use the top 8 bits; skip alpha bytes
910 st->codec->palctrl->palette[j] =
911 (r << 16) | (g << 8) | (b);
915 st->codec->palctrl->palette_changed = 1;
917 } else if(st->codec->codec_type==CODEC_TYPE_AUDIO) {
918 int bits_per_sample, flags;
919 uint16_t version = get_be16(pb);
921 st->codec->codec_id = id;
922 get_be16(pb); /* revision level */
923 get_be32(pb); /* vendor */
925 st->codec->channels = get_be16(pb); /* channel count */
926 dprintf(c->fc, "audio channels %d\n", st->codec->channels);
927 st->codec->bits_per_coded_sample = get_be16(pb); /* sample size */
929 sc->audio_cid = get_be16(pb);
930 get_be16(pb); /* packet size = 0 */
932 st->codec->sample_rate = ((get_be32(pb) >> 16));
934 //Read QT version 1 fields. In version 0 these do not exist.
935 dprintf(c->fc, "version =%d, isom =%d\n",version,c->isom);
938 sc->samples_per_frame = get_be32(pb);
939 get_be32(pb); /* bytes per packet */
940 sc->bytes_per_frame = get_be32(pb);
941 get_be32(pb); /* bytes per sample */
942 } else if(version==2) {
943 get_be32(pb); /* sizeof struct only */
944 st->codec->sample_rate = av_int2dbl(get_be64(pb)); /* float 64 */
945 st->codec->channels = get_be32(pb);
946 get_be32(pb); /* always 0x7F000000 */
947 st->codec->bits_per_coded_sample = get_be32(pb); /* bits per channel if sound is uncompressed */
948 flags = get_be32(pb); /* lcpm format specific flag */
949 sc->bytes_per_frame = get_be32(pb); /* bytes per audio packet if constant */
950 sc->samples_per_frame = get_be32(pb); /* lpcm frames per audio packet if constant */
951 if (format == MKTAG('l','p','c','m'))
952 st->codec->codec_id = mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, flags);
956 switch (st->codec->codec_id) {
957 case CODEC_ID_PCM_S8:
958 case CODEC_ID_PCM_U8:
959 if (st->codec->bits_per_coded_sample == 16)
960 st->codec->codec_id = CODEC_ID_PCM_S16BE;
962 case CODEC_ID_PCM_S16LE:
963 case CODEC_ID_PCM_S16BE:
964 if (st->codec->bits_per_coded_sample == 8)
965 st->codec->codec_id = CODEC_ID_PCM_S8;
966 else if (st->codec->bits_per_coded_sample == 24)
967 st->codec->codec_id =
968 st->codec->codec_id == CODEC_ID_PCM_S16BE ?
969 CODEC_ID_PCM_S24BE : CODEC_ID_PCM_S24LE;
971 /* set values for old format before stsd version 1 appeared */
973 sc->samples_per_frame = 6;
974 sc->bytes_per_frame = 2*st->codec->channels;
977 sc->samples_per_frame = 6;
978 sc->bytes_per_frame = 1*st->codec->channels;
980 case CODEC_ID_ADPCM_IMA_QT:
981 sc->samples_per_frame = 64;
982 sc->bytes_per_frame = 34*st->codec->channels;
985 sc->samples_per_frame = 160;
986 sc->bytes_per_frame = 33;
992 bits_per_sample = av_get_bits_per_sample(st->codec->codec_id);
993 if (bits_per_sample) {
994 st->codec->bits_per_coded_sample = bits_per_sample;
995 sc->sample_size = (bits_per_sample >> 3) * st->codec->channels;
997 } else if(st->codec->codec_type==CODEC_TYPE_SUBTITLE){
998 // ttxt stsd contains display flags, justification, background
999 // color, fonts, and default styles, so fake an atom to read it
1000 MOVAtom fake_atom = { .size = size - (url_ftell(pb) - start_pos) };
1001 mov_read_glbl(c, pb, fake_atom);
1002 st->codec->codec_id= id;
1003 st->codec->width = sc->width;
1004 st->codec->height = sc->height;
1006 /* other codec type, just skip (rtp, mp4s, tmcd ...) */
1007 url_fskip(pb, size - (url_ftell(pb) - start_pos));
1009 /* this will read extra atoms at the end (wave, alac, damr, avcC, SMI ...) */
1010 a.size = size - (url_ftell(pb) - start_pos);
1012 if (mov_read_default(c, pb, a) < 0)
1014 } else if (a.size > 0)
1015 url_fskip(pb, a.size);
1018 if(st->codec->codec_type==CODEC_TYPE_AUDIO && st->codec->sample_rate==0 && sc->time_scale>1)
1019 st->codec->sample_rate= sc->time_scale;
1021 /* special codec parameters handling */
1022 switch (st->codec->codec_id) {
1023 #if CONFIG_DV_DEMUXER
1024 case CODEC_ID_DVAUDIO:
1025 c->dv_fctx = avformat_alloc_context();
1026 c->dv_demux = dv_init_demux(c->dv_fctx);
1028 av_log(c->fc, AV_LOG_ERROR, "dv demux context init error\n");
1031 sc->dv_audio_container = 1;
1032 st->codec->codec_id = CODEC_ID_PCM_S16LE;
1035 /* no ifdef since parameters are always those */
1036 case CODEC_ID_QCELP:
1037 st->codec->frame_size= 160;
1038 st->codec->channels= 1; /* really needed */
1040 case CODEC_ID_AMR_NB:
1041 case CODEC_ID_AMR_WB:
1042 st->codec->frame_size= sc->samples_per_frame;
1043 st->codec->channels= 1; /* really needed */
1044 /* force sample rate for amr, stsd in 3gp does not store sample rate */
1045 if (st->codec->codec_id == CODEC_ID_AMR_NB)
1046 st->codec->sample_rate = 8000;
1047 else if (st->codec->codec_id == CODEC_ID_AMR_WB)
1048 st->codec->sample_rate = 16000;
1052 st->codec->codec_type = CODEC_TYPE_AUDIO; /* force type after stsd for m1a hdlr */
1053 st->need_parsing = AVSTREAM_PARSE_FULL;
1056 case CODEC_ID_ADPCM_MS:
1057 case CODEC_ID_ADPCM_IMA_WAV:
1058 st->codec->block_align = sc->bytes_per_frame;
1061 if (st->codec->extradata_size == 36) {
1062 st->codec->frame_size = AV_RB32(st->codec->extradata+12);
1063 st->codec->channels = AV_RB8 (st->codec->extradata+21);
1073 static int mov_read_stsc(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1075 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1076 MOVStreamContext *sc = st->priv_data;
1077 unsigned int i, entries;
1079 get_byte(pb); /* version */
1080 get_be24(pb); /* flags */
1082 entries = get_be32(pb);
1084 dprintf(c->fc, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
1086 if(entries >= UINT_MAX / sizeof(*sc->stsc_data))
1088 sc->stsc_data = av_malloc(entries * sizeof(*sc->stsc_data));
1090 return AVERROR(ENOMEM);
1091 sc->stsc_count = entries;
1093 for(i=0; i<entries; i++) {
1094 sc->stsc_data[i].first = get_be32(pb);
1095 sc->stsc_data[i].count = get_be32(pb);
1096 sc->stsc_data[i].id = get_be32(pb);
1101 static int mov_read_stss(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1103 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1104 MOVStreamContext *sc = st->priv_data;
1105 unsigned int i, entries;
1107 get_byte(pb); /* version */
1108 get_be24(pb); /* flags */
1110 entries = get_be32(pb);
1112 dprintf(c->fc, "keyframe_count = %d\n", entries);
1114 if(entries >= UINT_MAX / sizeof(int))
1116 sc->keyframes = av_malloc(entries * sizeof(int));
1118 return AVERROR(ENOMEM);
1119 sc->keyframe_count = entries;
1121 for(i=0; i<entries; i++) {
1122 sc->keyframes[i] = get_be32(pb);
1123 //dprintf(c->fc, "keyframes[]=%d\n", sc->keyframes[i]);
1128 static int mov_read_stsz(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1130 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1131 MOVStreamContext *sc = st->priv_data;
1132 unsigned int i, entries, sample_size;
1134 get_byte(pb); /* version */
1135 get_be24(pb); /* flags */
1137 sample_size = get_be32(pb);
1138 if (!sc->sample_size) /* do not overwrite value computed in stsd */
1139 sc->sample_size = sample_size;
1140 entries = get_be32(pb);
1142 dprintf(c->fc, "sample_size = %d sample_count = %d\n", sc->sample_size, entries);
1144 sc->sample_count = entries;
1148 if(entries >= UINT_MAX / sizeof(int))
1150 sc->sample_sizes = av_malloc(entries * sizeof(int));
1151 if (!sc->sample_sizes)
1152 return AVERROR(ENOMEM);
1154 for(i=0; i<entries; i++)
1155 sc->sample_sizes[i] = get_be32(pb);
1159 static int mov_read_stts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1161 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1162 MOVStreamContext *sc = st->priv_data;
1163 unsigned int i, entries;
1165 int64_t total_sample_count=0;
1167 get_byte(pb); /* version */
1168 get_be24(pb); /* flags */
1169 entries = get_be32(pb);
1171 dprintf(c->fc, "track[%i].stts.entries = %i\n", c->fc->nb_streams-1, entries);
1173 if(entries >= UINT_MAX / sizeof(*sc->stts_data))
1175 sc->stts_data = av_malloc(entries * sizeof(*sc->stts_data));
1177 return AVERROR(ENOMEM);
1178 sc->stts_count = entries;
1180 for(i=0; i<entries; i++) {
1181 int sample_duration;
1184 sample_count=get_be32(pb);
1185 sample_duration = get_be32(pb);
1186 sc->stts_data[i].count= sample_count;
1187 sc->stts_data[i].duration= sample_duration;
1189 sc->time_rate= av_gcd(sc->time_rate, sample_duration);
1191 dprintf(c->fc, "sample_count=%d, sample_duration=%d\n",sample_count,sample_duration);
1193 duration+=(int64_t)sample_duration*sample_count;
1194 total_sample_count+=sample_count;
1197 st->nb_frames= total_sample_count;
1199 st->duration= duration;
1203 static int mov_read_ctts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1205 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1206 MOVStreamContext *sc = st->priv_data;
1207 unsigned int i, entries;
1209 get_byte(pb); /* version */
1210 get_be24(pb); /* flags */
1211 entries = get_be32(pb);
1213 dprintf(c->fc, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
1215 if(entries >= UINT_MAX / sizeof(*sc->ctts_data))
1217 sc->ctts_data = av_malloc(entries * sizeof(*sc->ctts_data));
1219 return AVERROR(ENOMEM);
1220 sc->ctts_count = entries;
1222 for(i=0; i<entries; i++) {
1223 int count =get_be32(pb);
1224 int duration =get_be32(pb);
1228 st->codec->has_b_frames = 1;
1230 sc->ctts_data[i].count = count;
1231 sc->ctts_data[i].duration= duration;
1233 sc->time_rate= av_gcd(sc->time_rate, FFABS(duration));
1238 static void mov_build_index(MOVContext *mov, AVStream *st)
1240 MOVStreamContext *sc = st->priv_data;
1241 int64_t current_offset;
1242 int64_t current_dts = 0;
1243 unsigned int stts_index = 0;
1244 unsigned int stsc_index = 0;
1245 unsigned int stss_index = 0;
1248 /* adjust first dts according to edit list */
1249 if (sc->time_offset) {
1250 assert(sc->time_offset % sc->time_rate == 0);
1251 current_dts = - (sc->time_offset / sc->time_rate);
1254 /* only use old uncompressed audio chunk demuxing when stts specifies it */
1255 if (!(st->codec->codec_type == CODEC_TYPE_AUDIO &&
1256 sc->stts_count == 1 && sc->stts_data[0].duration == 1)) {
1257 unsigned int current_sample = 0;
1258 unsigned int stts_sample = 0;
1259 unsigned int keyframe, sample_size;
1260 unsigned int distance = 0;
1261 int key_off = sc->keyframes && sc->keyframes[0] == 1;
1263 st->nb_frames = sc->sample_count;
1264 for (i = 0; i < sc->chunk_count; i++) {
1265 current_offset = sc->chunk_offsets[i];
1266 if (stsc_index + 1 < sc->stsc_count &&
1267 i + 1 == sc->stsc_data[stsc_index + 1].first)
1269 for (j = 0; j < sc->stsc_data[stsc_index].count; j++) {
1270 if (current_sample >= sc->sample_count) {
1271 av_log(mov->fc, AV_LOG_ERROR, "wrong sample count\n");
1274 keyframe = !sc->keyframe_count || current_sample+key_off == sc->keyframes[stss_index];
1277 if (stss_index + 1 < sc->keyframe_count)
1280 sample_size = sc->sample_size > 0 ? sc->sample_size : sc->sample_sizes[current_sample];
1281 if(sc->pseudo_stream_id == -1 ||
1282 sc->stsc_data[stsc_index].id - 1 == sc->pseudo_stream_id) {
1283 av_add_index_entry(st, current_offset, current_dts, sample_size, distance,
1284 keyframe ? AVINDEX_KEYFRAME : 0);
1285 dprintf(mov->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1286 "size %d, distance %d, keyframe %d\n", st->index, current_sample,
1287 current_offset, current_dts, sample_size, distance, keyframe);
1290 current_offset += sample_size;
1291 assert(sc->stts_data[stts_index].duration % sc->time_rate == 0);
1292 current_dts += sc->stts_data[stts_index].duration / sc->time_rate;
1296 if (stts_index + 1 < sc->stts_count && stts_sample == sc->stts_data[stts_index].count) {
1302 } else { /* read whole chunk */
1303 unsigned int chunk_samples, chunk_size, chunk_duration;
1304 unsigned int frames = 1;
1305 for (i = 0; i < sc->chunk_count; i++) {
1306 current_offset = sc->chunk_offsets[i];
1307 if (stsc_index + 1 < sc->stsc_count &&
1308 i + 1 == sc->stsc_data[stsc_index + 1].first)
1310 chunk_samples = sc->stsc_data[stsc_index].count;
1311 /* get chunk size, beware of alaw/ulaw/mace */
1312 if (sc->samples_per_frame > 0 &&
1313 (chunk_samples * sc->bytes_per_frame % sc->samples_per_frame == 0)) {
1314 if (sc->samples_per_frame < 160)
1315 chunk_size = chunk_samples * sc->bytes_per_frame / sc->samples_per_frame;
1317 chunk_size = sc->bytes_per_frame;
1318 frames = chunk_samples / sc->samples_per_frame;
1319 chunk_samples = sc->samples_per_frame;
1322 chunk_size = chunk_samples * sc->sample_size;
1323 for (j = 0; j < frames; j++) {
1324 av_add_index_entry(st, current_offset, current_dts, chunk_size, 0, AVINDEX_KEYFRAME);
1325 /* get chunk duration */
1327 while (chunk_samples > 0) {
1328 if (chunk_samples < sc->stts_data[stts_index].count) {
1329 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1330 sc->stts_data[stts_index].count -= chunk_samples;
1333 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1334 chunk_samples -= sc->stts_data[stts_index].count;
1335 if (stts_index + 1 < sc->stts_count)
1339 current_offset += sc->bytes_per_frame;
1340 dprintf(mov->fc, "AVIndex stream %d, chunk %d, offset %"PRIx64", dts %"PRId64", "
1341 "size %d, duration %d\n", st->index, i, current_offset, current_dts,
1342 chunk_size, chunk_duration);
1343 assert(chunk_duration % sc->time_rate == 0);
1344 current_dts += chunk_duration / sc->time_rate;
1350 static int mov_read_trak(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1353 MOVStreamContext *sc;
1356 st = av_new_stream(c->fc, c->fc->nb_streams);
1357 if (!st) return AVERROR(ENOMEM);
1358 sc = av_mallocz(sizeof(MOVStreamContext));
1359 if (!sc) return AVERROR(ENOMEM);
1362 st->codec->codec_type = CODEC_TYPE_DATA;
1363 sc->ffindex = st->index;
1365 if ((ret = mov_read_default(c, pb, atom)) < 0)
1369 if (sc->chunk_count && (!sc->stts_count || !sc->stsc_count ||
1370 (!sc->sample_size && !sc->sample_count))) {
1371 av_log(c->fc, AV_LOG_ERROR, "stream %d, missing mandatory atoms, broken header\n",
1378 if (!sc->time_scale)
1379 sc->time_scale = c->time_scale;
1381 av_set_pts_info(st, 64, sc->time_rate, sc->time_scale);
1383 if (st->codec->codec_type == CODEC_TYPE_AUDIO &&
1384 !st->codec->frame_size && sc->stts_count == 1) {
1385 st->codec->frame_size = av_rescale(sc->stts_data[0].duration,
1386 st->codec->sample_rate, sc->time_scale);
1387 dprintf(c->fc, "frame size %d\n", st->codec->frame_size);
1390 if (st->duration != AV_NOPTS_VALUE) {
1391 assert(st->duration % sc->time_rate == 0);
1392 st->duration /= sc->time_rate;
1395 mov_build_index(c, st);
1397 if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
1398 if (url_fopen(&sc->pb, sc->drefs[sc->dref_id-1].path, URL_RDONLY) < 0)
1399 av_log(c->fc, AV_LOG_ERROR, "stream %d, error opening file %s: %s\n",
1400 st->index, sc->drefs[sc->dref_id-1].path, strerror(errno));
1404 switch (st->codec->codec_id) {
1405 #if CONFIG_H261_DECODER
1408 #if CONFIG_H263_DECODER
1411 #if CONFIG_MPEG4_DECODER
1412 case CODEC_ID_MPEG4:
1414 st->codec->width = 0; /* let decoder init width/height */
1415 st->codec->height= 0;
1419 /* Do not need those anymore. */
1420 av_freep(&sc->chunk_offsets);
1421 av_freep(&sc->stsc_data);
1422 av_freep(&sc->sample_sizes);
1423 av_freep(&sc->keyframes);
1424 av_freep(&sc->stts_data);
1429 static int mov_read_ilst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1432 c->itunes_metadata = 1;
1433 ret = mov_read_default(c, pb, atom);
1434 c->itunes_metadata = 0;
1438 static int mov_read_meta(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1440 while (atom.size > 8) {
1441 uint32_t tag = get_le32(pb);
1443 if (tag == MKTAG('h','d','l','r')) {
1444 url_fseek(pb, -8, SEEK_CUR);
1446 return mov_read_default(c, pb, atom);
1452 static int mov_read_tkhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1457 int64_t disp_transform[2];
1458 int display_matrix[3][2];
1459 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1460 MOVStreamContext *sc = st->priv_data;
1461 int version = get_byte(pb);
1463 get_be24(pb); /* flags */
1465 MOV_TRACK_ENABLED 0x0001
1466 MOV_TRACK_IN_MOVIE 0x0002
1467 MOV_TRACK_IN_PREVIEW 0x0004
1468 MOV_TRACK_IN_POSTER 0x0008
1475 get_be32(pb); /* creation time */
1476 get_be32(pb); /* modification time */
1478 st->id = (int)get_be32(pb); /* track id (NOT 0 !)*/
1479 get_be32(pb); /* reserved */
1481 /* highlevel (considering edits) duration in movie timebase */
1482 (version == 1) ? get_be64(pb) : get_be32(pb);
1483 get_be32(pb); /* reserved */
1484 get_be32(pb); /* reserved */
1486 get_be16(pb); /* layer */
1487 get_be16(pb); /* alternate group */
1488 get_be16(pb); /* volume */
1489 get_be16(pb); /* reserved */
1491 //read in the display matrix (outlined in ISO 14496-12, Section 6.2.2)
1492 // they're kept in fixed point format through all calculations
1493 // ignore u,v,z b/c we don't need the scale factor to calc aspect ratio
1494 for (i = 0; i < 3; i++) {
1495 display_matrix[i][0] = get_be32(pb); // 16.16 fixed point
1496 display_matrix[i][1] = get_be32(pb); // 16.16 fixed point
1497 get_be32(pb); // 2.30 fixed point (not used)
1500 width = get_be32(pb); // 16.16 fixed point track width
1501 height = get_be32(pb); // 16.16 fixed point track height
1502 sc->width = width >> 16;
1503 sc->height = height >> 16;
1505 //transform the display width/height according to the matrix
1506 // skip this if the display matrix is the default identity matrix
1507 // to keep the same scale, use [width height 1<<16]
1508 if (width && height &&
1509 (display_matrix[0][0] != 65536 || display_matrix[0][1] ||
1510 display_matrix[1][0] || display_matrix[1][1] != 65536 ||
1511 display_matrix[2][0] || display_matrix[2][1])) {
1512 for (i = 0; i < 2; i++)
1514 (int64_t) width * display_matrix[0][i] +
1515 (int64_t) height * display_matrix[1][i] +
1516 ((int64_t) display_matrix[2][i] << 16);
1518 //sample aspect ratio is new width/height divided by old width/height
1519 st->sample_aspect_ratio = av_d2q(
1520 ((double) disp_transform[0] * height) /
1521 ((double) disp_transform[1] * width), INT_MAX);
1526 static int mov_read_tfhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1528 MOVFragment *frag = &c->fragment;
1529 MOVTrackExt *trex = NULL;
1530 int flags, track_id, i;
1532 get_byte(pb); /* version */
1533 flags = get_be24(pb);
1535 track_id = get_be32(pb);
1538 frag->track_id = track_id;
1539 for (i = 0; i < c->trex_count; i++)
1540 if (c->trex_data[i].track_id == frag->track_id) {
1541 trex = &c->trex_data[i];
1545 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding trex\n");
1549 if (flags & 0x01) frag->base_data_offset = get_be64(pb);
1550 else frag->base_data_offset = frag->moof_offset;
1551 if (flags & 0x02) frag->stsd_id = get_be32(pb);
1552 else frag->stsd_id = trex->stsd_id;
1554 frag->duration = flags & 0x08 ? get_be32(pb) : trex->duration;
1555 frag->size = flags & 0x10 ? get_be32(pb) : trex->size;
1556 frag->flags = flags & 0x20 ? get_be32(pb) : trex->flags;
1557 dprintf(c->fc, "frag flags 0x%x\n", frag->flags);
1561 static int mov_read_trex(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1565 if ((uint64_t)c->trex_count+1 >= UINT_MAX / sizeof(*c->trex_data))
1567 trex = av_realloc(c->trex_data, (c->trex_count+1)*sizeof(*c->trex_data));
1569 return AVERROR(ENOMEM);
1570 c->trex_data = trex;
1571 trex = &c->trex_data[c->trex_count++];
1572 get_byte(pb); /* version */
1573 get_be24(pb); /* flags */
1574 trex->track_id = get_be32(pb);
1575 trex->stsd_id = get_be32(pb);
1576 trex->duration = get_be32(pb);
1577 trex->size = get_be32(pb);
1578 trex->flags = get_be32(pb);
1582 static int mov_read_trun(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1584 MOVFragment *frag = &c->fragment;
1585 AVStream *st = NULL;
1586 MOVStreamContext *sc;
1589 int data_offset = 0;
1590 unsigned entries, first_sample_flags = frag->flags;
1591 int flags, distance, i;
1593 for (i = 0; i < c->fc->nb_streams; i++) {
1594 if (c->fc->streams[i]->id == frag->track_id) {
1595 st = c->fc->streams[i];
1600 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding track id %d\n", frag->track_id);
1604 if (sc->pseudo_stream_id+1 != frag->stsd_id)
1606 get_byte(pb); /* version */
1607 flags = get_be24(pb);
1608 entries = get_be32(pb);
1609 dprintf(c->fc, "flags 0x%x entries %d\n", flags, entries);
1610 if (flags & 0x001) data_offset = get_be32(pb);
1611 if (flags & 0x004) first_sample_flags = get_be32(pb);
1612 if (flags & 0x800) {
1614 if ((uint64_t)entries+sc->ctts_count >= UINT_MAX/sizeof(*sc->ctts_data))
1616 ctts_data = av_realloc(sc->ctts_data,
1617 (entries+sc->ctts_count)*sizeof(*sc->ctts_data));
1619 return AVERROR(ENOMEM);
1620 sc->ctts_data = ctts_data;
1623 offset = frag->base_data_offset + data_offset;
1625 dprintf(c->fc, "first sample flags 0x%x\n", first_sample_flags);
1626 for (i = 0; i < entries; i++) {
1627 unsigned sample_size = frag->size;
1628 int sample_flags = i ? frag->flags : first_sample_flags;
1629 unsigned sample_duration = frag->duration;
1632 if (flags & 0x100) sample_duration = get_be32(pb);
1633 if (flags & 0x200) sample_size = get_be32(pb);
1634 if (flags & 0x400) sample_flags = get_be32(pb);
1635 if (flags & 0x800) {
1636 sc->ctts_data[sc->ctts_count].count = 1;
1637 sc->ctts_data[sc->ctts_count].duration = get_be32(pb);
1640 if ((keyframe = st->codec->codec_type == CODEC_TYPE_AUDIO ||
1641 (flags & 0x004 && !i && !sample_flags) || sample_flags & 0x2000000))
1643 av_add_index_entry(st, offset, dts, sample_size, distance,
1644 keyframe ? AVINDEX_KEYFRAME : 0);
1645 dprintf(c->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1646 "size %d, distance %d, keyframe %d\n", st->index, sc->sample_count+i,
1647 offset, dts, sample_size, distance, keyframe);
1649 assert(sample_duration % sc->time_rate == 0);
1650 dts += sample_duration / sc->time_rate;
1651 offset += sample_size;
1653 frag->moof_offset = offset;
1658 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
1659 /* like the files created with Adobe Premiere 5.0, for samples see */
1660 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
1661 static int mov_read_wide(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1666 return 0; /* continue */
1667 if (get_be32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
1668 url_fskip(pb, atom.size - 4);
1671 atom.type = get_le32(pb);
1674 if (atom.type != MKTAG('m','d','a','t')) {
1675 url_fskip(pb, atom.size);
1678 err = mov_read_mdat(c, pb, atom);
1682 static int mov_read_cmov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1687 uint8_t *moov_data; /* uncompressed data */
1688 long cmov_len, moov_len;
1691 get_be32(pb); /* dcom atom */
1692 if (get_le32(pb) != MKTAG('d','c','o','m'))
1694 if (get_le32(pb) != MKTAG('z','l','i','b')) {
1695 av_log(c->fc, AV_LOG_ERROR, "unknown compression for cmov atom !");
1698 get_be32(pb); /* cmvd atom */
1699 if (get_le32(pb) != MKTAG('c','m','v','d'))
1701 moov_len = get_be32(pb); /* uncompressed size */
1702 cmov_len = atom.size - 6 * 4;
1704 cmov_data = av_malloc(cmov_len);
1706 return AVERROR(ENOMEM);
1707 moov_data = av_malloc(moov_len);
1710 return AVERROR(ENOMEM);
1712 get_buffer(pb, cmov_data, cmov_len);
1713 if(uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
1714 goto free_and_return;
1715 if(init_put_byte(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL, NULL) != 0)
1716 goto free_and_return;
1717 atom.type = MKTAG('m','o','o','v');
1719 atom.size = moov_len;
1721 // { int fd = open("/tmp/uncompheader.mov", O_WRONLY | O_CREAT); write(fd, moov_data, moov_len); close(fd); }
1723 ret = mov_read_default(c, &ctx, atom);
1729 av_log(c->fc, AV_LOG_ERROR, "this file requires zlib support compiled in\n");
1734 /* edit list atom */
1735 static int mov_read_elst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1737 MOVStreamContext *sc = c->fc->streams[c->fc->nb_streams-1]->priv_data;
1740 get_byte(pb); /* version */
1741 get_be24(pb); /* flags */
1742 edit_count = get_be32(pb); /* entries */
1744 for(i=0; i<edit_count; i++){
1746 get_be32(pb); /* Track duration */
1747 time = get_be32(pb); /* Media time */
1748 get_be32(pb); /* Media rate */
1749 if (i == 0 && time != -1) {
1750 sc->time_offset = time;
1751 sc->time_rate = av_gcd(sc->time_rate, time);
1756 av_log(c->fc, AV_LOG_WARNING, "multiple edit list entries, "
1757 "a/v desync might occur, patch welcome\n");
1759 dprintf(c->fc, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, edit_count);
1763 static const MOVParseTableEntry mov_default_parse_table[] = {
1764 { MKTAG('a','v','s','s'), mov_read_extradata },
1765 { MKTAG('c','o','6','4'), mov_read_stco },
1766 { MKTAG('c','t','t','s'), mov_read_ctts }, /* composition time to sample */
1767 { MKTAG('d','i','n','f'), mov_read_default },
1768 { MKTAG('d','r','e','f'), mov_read_dref },
1769 { MKTAG('e','d','t','s'), mov_read_default },
1770 { MKTAG('e','l','s','t'), mov_read_elst },
1771 { MKTAG('e','n','d','a'), mov_read_enda },
1772 { MKTAG('f','i','e','l'), mov_read_extradata },
1773 { MKTAG('f','t','y','p'), mov_read_ftyp },
1774 { MKTAG('g','l','b','l'), mov_read_glbl },
1775 { MKTAG('h','d','l','r'), mov_read_hdlr },
1776 { MKTAG('i','l','s','t'), mov_read_ilst },
1777 { MKTAG('j','p','2','h'), mov_read_extradata },
1778 { MKTAG('m','d','a','t'), mov_read_mdat },
1779 { MKTAG('m','d','h','d'), mov_read_mdhd },
1780 { MKTAG('m','d','i','a'), mov_read_default },
1781 { MKTAG('m','e','t','a'), mov_read_meta },
1782 { MKTAG('m','i','n','f'), mov_read_default },
1783 { MKTAG('m','o','o','f'), mov_read_moof },
1784 { MKTAG('m','o','o','v'), mov_read_moov },
1785 { MKTAG('m','v','e','x'), mov_read_default },
1786 { MKTAG('m','v','h','d'), mov_read_mvhd },
1787 { MKTAG('S','M','I',' '), mov_read_smi }, /* Sorenson extension ??? */
1788 { MKTAG('a','l','a','c'), mov_read_extradata }, /* alac specific atom */
1789 { MKTAG('a','v','c','C'), mov_read_glbl },
1790 { MKTAG('p','a','s','p'), mov_read_pasp },
1791 { MKTAG('s','t','b','l'), mov_read_default },
1792 { MKTAG('s','t','c','o'), mov_read_stco },
1793 { MKTAG('s','t','s','c'), mov_read_stsc },
1794 { MKTAG('s','t','s','d'), mov_read_stsd }, /* sample description */
1795 { MKTAG('s','t','s','s'), mov_read_stss }, /* sync sample */
1796 { MKTAG('s','t','s','z'), mov_read_stsz }, /* sample size */
1797 { MKTAG('s','t','t','s'), mov_read_stts },
1798 { MKTAG('t','k','h','d'), mov_read_tkhd }, /* track header */
1799 { MKTAG('t','f','h','d'), mov_read_tfhd }, /* track fragment header */
1800 { MKTAG('t','r','a','k'), mov_read_trak },
1801 { MKTAG('t','r','a','f'), mov_read_default },
1802 { MKTAG('t','r','e','x'), mov_read_trex },
1803 { MKTAG('t','r','u','n'), mov_read_trun },
1804 { MKTAG('u','d','t','a'), mov_read_default },
1805 { MKTAG('w','a','v','e'), mov_read_wave },
1806 { MKTAG('e','s','d','s'), mov_read_esds },
1807 { MKTAG('w','i','d','e'), mov_read_wide }, /* place holder */
1808 { MKTAG('c','m','o','v'), mov_read_cmov },
1812 static int mov_probe(AVProbeData *p)
1814 unsigned int offset;
1818 /* check file header */
1821 /* ignore invalid offset */
1822 if ((offset + 8) > (unsigned int)p->buf_size)
1824 tag = AV_RL32(p->buf + offset + 4);
1826 /* check for obvious tags */
1827 case MKTAG('j','P',' ',' '): /* jpeg 2000 signature */
1828 case MKTAG('m','o','o','v'):
1829 case MKTAG('m','d','a','t'):
1830 case MKTAG('p','n','o','t'): /* detect movs with preview pics like ew.mov and april.mov */
1831 case MKTAG('u','d','t','a'): /* Packet Video PVAuthor adds this and a lot of more junk */
1832 case MKTAG('f','t','y','p'):
1833 return AVPROBE_SCORE_MAX;
1834 /* those are more common words, so rate then a bit less */
1835 case MKTAG('e','d','i','w'): /* xdcam files have reverted first tags */
1836 case MKTAG('w','i','d','e'):
1837 case MKTAG('f','r','e','e'):
1838 case MKTAG('j','u','n','k'):
1839 case MKTAG('p','i','c','t'):
1840 return AVPROBE_SCORE_MAX - 5;
1841 case MKTAG(0x82,0x82,0x7f,0x7d):
1842 case MKTAG('s','k','i','p'):
1843 case MKTAG('u','u','i','d'):
1844 case MKTAG('p','r','f','l'):
1845 offset = AV_RB32(p->buf+offset) + offset;
1846 /* if we only find those cause probedata is too small at least rate them */
1847 score = AVPROBE_SCORE_MAX - 50;
1850 /* unrecognized tag */
1857 static int mov_read_header(AVFormatContext *s, AVFormatParameters *ap)
1859 MOVContext *mov = s->priv_data;
1860 ByteIOContext *pb = s->pb;
1862 MOVAtom atom = { 0, 0, 0 };
1865 /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
1866 if(!url_is_streamed(pb))
1867 atom.size = url_fsize(pb);
1869 atom.size = INT64_MAX;
1871 /* check MOV header */
1872 if ((err = mov_read_default(mov, pb, atom)) < 0) {
1873 av_log(s, AV_LOG_ERROR, "error reading header: %d\n", err);
1876 if (!mov->found_moov) {
1877 av_log(s, AV_LOG_ERROR, "moov atom not found\n");
1880 dprintf(mov->fc, "on_parse_exit_offset=%lld\n", url_ftell(pb));
1885 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
1887 MOVContext *mov = s->priv_data;
1888 MOVStreamContext *sc = 0;
1889 AVIndexEntry *sample = 0;
1890 int64_t best_dts = INT64_MAX;
1893 for (i = 0; i < s->nb_streams; i++) {
1894 AVStream *st = s->streams[i];
1895 MOVStreamContext *msc = st->priv_data;
1896 if (st->discard != AVDISCARD_ALL && msc->pb && msc->current_sample < st->nb_index_entries) {
1897 AVIndexEntry *current_sample = &st->index_entries[msc->current_sample];
1898 int64_t dts = av_rescale(current_sample->timestamp * (int64_t)msc->time_rate,
1899 AV_TIME_BASE, msc->time_scale);
1900 dprintf(s, "stream %d, sample %d, dts %"PRId64"\n", i, msc->current_sample, dts);
1901 if (!sample || (url_is_streamed(s->pb) && current_sample->pos < sample->pos) ||
1902 (!url_is_streamed(s->pb) &&
1903 ((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
1904 ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
1905 (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
1906 sample = current_sample;
1913 mov->found_mdat = 0;
1914 if (!url_is_streamed(s->pb) ||
1915 mov_read_default(mov, s->pb, (MOVAtom){ 0, 0, INT64_MAX }) < 0 ||
1918 dprintf(s, "read fragments, offset 0x%llx\n", url_ftell(s->pb));
1921 /* must be done just before reading, to avoid infinite loop on sample */
1922 sc->current_sample++;
1923 if (url_fseek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
1924 av_log(mov->fc, AV_LOG_ERROR, "stream %d, offset 0x%"PRIx64": partial file\n",
1925 sc->ffindex, sample->pos);
1928 ret = av_get_packet(sc->pb, pkt, sample->size);
1931 #if CONFIG_DV_DEMUXER
1932 if (mov->dv_demux && sc->dv_audio_container) {
1933 dv_produce_packet(mov->dv_demux, pkt, pkt->data, pkt->size);
1936 if (dv_get_packet(mov->dv_demux, pkt) < 0)
1940 pkt->stream_index = sc->ffindex;
1941 pkt->dts = sample->timestamp;
1942 if (sc->ctts_data) {
1943 assert(sc->ctts_data[sc->ctts_index].duration % sc->time_rate == 0);
1944 pkt->pts = pkt->dts + sc->ctts_data[sc->ctts_index].duration / sc->time_rate;
1945 /* update ctts context */
1947 if (sc->ctts_index < sc->ctts_count &&
1948 sc->ctts_data[sc->ctts_index].count == sc->ctts_sample) {
1950 sc->ctts_sample = 0;
1953 pkt->dts = AV_NOPTS_VALUE;
1955 AVStream *st = s->streams[sc->ffindex];
1956 int64_t next_dts = (sc->current_sample < st->nb_index_entries) ?
1957 st->index_entries[sc->current_sample].timestamp : st->duration;
1958 pkt->duration = next_dts - pkt->dts;
1959 pkt->pts = pkt->dts;
1961 pkt->flags |= sample->flags & AVINDEX_KEYFRAME ? PKT_FLAG_KEY : 0;
1962 pkt->pos = sample->pos;
1963 dprintf(s, "stream %d, pts %"PRId64", dts %"PRId64", pos 0x%"PRIx64", duration %d\n",
1964 pkt->stream_index, pkt->pts, pkt->dts, pkt->pos, pkt->duration);
1968 static int mov_seek_stream(AVStream *st, int64_t timestamp, int flags)
1970 MOVStreamContext *sc = st->priv_data;
1971 int sample, time_sample;
1974 sample = av_index_search_timestamp(st, timestamp, flags);
1975 dprintf(st->codec, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
1976 if (sample < 0) /* not sure what to do */
1978 sc->current_sample = sample;
1979 dprintf(st->codec, "stream %d, found sample %d\n", st->index, sc->current_sample);
1980 /* adjust ctts index */
1981 if (sc->ctts_data) {
1983 for (i = 0; i < sc->ctts_count; i++) {
1984 int next = time_sample + sc->ctts_data[i].count;
1985 if (next > sc->current_sample) {
1987 sc->ctts_sample = sc->current_sample - time_sample;
1996 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
1999 int64_t seek_timestamp, timestamp;
2003 if (stream_index >= s->nb_streams)
2005 if (sample_time < 0)
2008 st = s->streams[stream_index];
2009 sample = mov_seek_stream(st, sample_time, flags);
2013 /* adjust seek timestamp to found sample timestamp */
2014 seek_timestamp = st->index_entries[sample].timestamp;
2016 for (i = 0; i < s->nb_streams; i++) {
2018 if (stream_index == i || st->discard == AVDISCARD_ALL)
2021 timestamp = av_rescale_q(seek_timestamp, s->streams[stream_index]->time_base, st->time_base);
2022 mov_seek_stream(st, timestamp, flags);
2027 static int mov_read_close(AVFormatContext *s)
2029 MOVContext *mov = s->priv_data;
2032 for (i = 0; i < s->nb_streams; i++) {
2033 AVStream *st = s->streams[i];
2034 MOVStreamContext *sc = st->priv_data;
2036 av_freep(&sc->ctts_data);
2037 for (j = 0; j < sc->drefs_count; j++)
2038 av_freep(&sc->drefs[j].path);
2039 av_freep(&sc->drefs);
2040 if (sc->pb && sc->pb != s->pb)
2043 av_freep(&st->codec->palctrl);
2046 if (mov->dv_demux) {
2047 for(i = 0; i < mov->dv_fctx->nb_streams; i++) {
2048 av_freep(&mov->dv_fctx->streams[i]->codec);
2049 av_freep(&mov->dv_fctx->streams[i]);
2051 av_freep(&mov->dv_fctx);
2052 av_freep(&mov->dv_demux);
2055 av_freep(&mov->trex_data);
2060 AVInputFormat mov_demuxer = {
2061 "mov,mp4,m4a,3gp,3g2,mj2",
2062 NULL_IF_CONFIG_SMALL("QuickTime/MPEG-4/Motion JPEG 2000 format"),