3 * Copyright (c) 2001 Fabrice Bellard.
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
30 #include "libavcodec/mpeg4audio.h"
31 #include "libavcodec/mpegaudiodata.h"
38 * First version by Francois Revol revol@free.fr
39 * Seek function by Gael Chardon gael.dev@4now.net
41 * Features and limitations:
42 * - reads most of the QT files I have (at least the structure),
43 * Sample QuickTime files with mp3 audio can be found at: http://www.3ivx.com/showcase.html
44 * - the code is quite ugly... maybe I won't do it recursive next time :-)
46 * Funny I didn't know about http://sourceforge.net/projects/qt-ffmpeg/
47 * when coding this :) (it's a writer anyway)
49 * Reference documents:
50 * http://www.geocities.com/xhelmboyx/quicktime/formats/qtm-layout.txt
52 * http://developer.apple.com/documentation/QuickTime/QTFF/
53 * http://developer.apple.com/documentation/QuickTime/QTFF/qtff.pdf
54 * QuickTime is a trademark of Apple (AFAIK :))
57 #include "qtpalette.h"
63 /* the QuickTime file format is quite convoluted...
64 * it has lots of index tables, each indexing something in another one...
65 * Here we just use what is needed to read the chunks
82 int64_t size; /* total size (excluding the size and type fields) */
85 struct MOVParseTableEntry;
89 uint64_t base_data_offset;
105 typedef struct MOVStreamContext {
107 int ffindex; /* the ffmpeg stream id */
109 unsigned int chunk_count;
110 int64_t *chunk_offsets;
111 unsigned int stts_count;
112 MOV_stts_t *stts_data;
113 unsigned int ctts_count;
114 MOV_stts_t *ctts_data;
115 unsigned int edit_count; /* number of 'edit' (elst atom) */
116 unsigned int sample_to_chunk_sz;
117 MOV_stsc_t *sample_to_chunk;
118 int sample_to_ctime_index;
119 int sample_to_ctime_sample;
120 unsigned int sample_size;
121 unsigned int sample_count;
123 unsigned int keyframe_count;
128 unsigned int bytes_per_frame;
129 unsigned int samples_per_frame;
130 int dv_audio_container;
131 int pseudo_stream_id; ///< -1 means demux all ids
132 int16_t audio_cid; ///< stsd audio compression id
133 unsigned drefs_count;
136 int wrong_dts; ///< dts are wrong due to negative ctts
139 typedef struct MOVContext {
142 int64_t duration; /* duration of the longest track */
143 int found_moov; /* when both 'moov' and 'mdat' sections has been found */
144 int found_mdat; /* we suppose we have enough data to read the file */
145 AVPaletteControl palette_control;
146 DVDemuxContext *dv_demux;
147 AVFormatContext *dv_fctx;
148 int isom; /* 1 if file is ISO Media (mp4/3gp) */
149 MOVFragment fragment; ///< current fragment in moof atom
150 MOVTrackExt *trex_data;
155 /* XXX: it's the first time I make a recursive parser I think... sorry if it's ugly :P */
157 /* those functions parse an atom */
159 0: continue to parse next atom
160 <0: error occurred, exit
162 /* links atom IDs to parse functions */
163 typedef struct MOVParseTableEntry {
165 int (*parse)(MOVContext *ctx, ByteIOContext *pb, MOV_atom_t atom);
166 } MOVParseTableEntry;
168 static const MOVParseTableEntry mov_default_parse_table[];
170 static int mov_read_default(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
172 int64_t total_size = 0;
177 a.offset = atom.offset;
180 atom.size = INT64_MAX;
181 while(((total_size + 8) < atom.size) && !url_feof(pb) && !err) {
185 a.size = get_be32(pb);
186 a.type = get_le32(pb);
190 dprintf(c->fc, "type: %08x %.4s sz: %"PRIx64" %"PRIx64" %"PRIx64"\n",
191 a.type, (char*)&a.type, a.size, atom.size, total_size);
192 if (a.size == 1) { /* 64 bit extended size */
193 a.size = get_be64(pb) - 8;
198 a.size = atom.size - total_size;
205 a.size = FFMIN(a.size, atom.size - total_size);
207 for (i = 0; mov_default_parse_table[i].type != 0
208 && mov_default_parse_table[i].type != a.type; i++)
211 if (mov_default_parse_table[i].type == 0) { /* skip leaf atoms data */
212 url_fskip(pb, a.size);
214 int64_t start_pos = url_ftell(pb);
216 err = mov_default_parse_table[i].parse(c, pb, a);
217 if (url_is_streamed(pb) && c->found_moov && c->found_mdat)
219 left = a.size - url_ftell(pb) + start_pos;
220 if (left > 0) /* skip garbage at atom end */
225 total_size += a.size;
228 if (!err && total_size < atom.size && atom.size < 0x7ffff)
229 url_fskip(pb, atom.size - total_size);
234 static int mov_read_dref(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
236 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
237 MOVStreamContext *sc = st->priv_data;
240 get_be32(pb); // version + flags
241 entries = get_be32(pb);
242 if (entries >= UINT_MAX / sizeof(*sc->drefs))
244 sc->drefs_count = entries;
245 sc->drefs = av_mallocz(entries * sizeof(*sc->drefs));
247 for (i = 0; i < sc->drefs_count; i++) {
248 MOV_dref_t *dref = &sc->drefs[i];
249 uint32_t size = get_be32(pb);
250 int64_t next = url_ftell(pb) + size - 4;
252 dref->type = get_le32(pb);
253 get_be32(pb); // version + flags
254 dprintf(c->fc, "type %.4s size %d\n", (char*)&dref->type, size);
256 if (dref->type == MKTAG('a','l','i','s') && size > 150) {
257 /* macintosh alias record */
258 uint16_t volume_len, len;
264 volume_len = get_byte(pb);
265 volume_len = FFMIN(volume_len, 27);
266 get_buffer(pb, volume, 27);
267 volume[volume_len] = 0;
268 av_log(c->fc, AV_LOG_DEBUG, "volume %s, len %d\n", volume, volume_len);
272 for (type = 0; type != -1 && url_ftell(pb) < next; ) {
275 av_log(c->fc, AV_LOG_DEBUG, "type %d, len %d\n", type, len);
278 if (type == 2) { // absolute path
280 dref->path = av_mallocz(len+1);
282 return AVERROR(ENOMEM);
283 get_buffer(pb, dref->path, len);
284 if (len > volume_len && !strncmp(dref->path, volume, volume_len)) {
286 memmove(dref->path, dref->path+volume_len, len);
289 for (j = 0; j < len; j++)
290 if (dref->path[j] == ':')
292 av_log(c->fc, AV_LOG_DEBUG, "path %s\n", dref->path);
297 url_fseek(pb, next, SEEK_SET);
302 static int mov_read_hdlr(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
304 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
308 get_byte(pb); /* version */
309 get_be24(pb); /* flags */
312 ctype = get_le32(pb);
313 type = get_le32(pb); /* component subtype */
315 dprintf(c->fc, "ctype= %c%c%c%c (0x%08x)\n", *((char *)&ctype), ((char *)&ctype)[1],
316 ((char *)&ctype)[2], ((char *)&ctype)[3], (int) ctype);
317 dprintf(c->fc, "stype= %c%c%c%c\n",
318 *((char *)&type), ((char *)&type)[1], ((char *)&type)[2], ((char *)&type)[3]);
321 if (type == MKTAG('v','i','d','e'))
322 st->codec->codec_type = CODEC_TYPE_VIDEO;
323 else if(type == MKTAG('s','o','u','n'))
324 st->codec->codec_type = CODEC_TYPE_AUDIO;
325 else if(type == MKTAG('m','1','a',' '))
326 st->codec->codec_id = CODEC_ID_MP2;
327 else if(type == MKTAG('s','u','b','p')) {
328 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
330 get_be32(pb); /* component manufacture */
331 get_be32(pb); /* component flags */
332 get_be32(pb); /* component flags mask */
335 return 0; /* nothing left to read */
337 url_fskip(pb, atom.size - (url_ftell(pb) - atom.offset));
341 static int mp4_read_descr_len(ByteIOContext *pb)
346 int c = get_byte(pb);
347 len = (len << 7) | (c & 0x7f);
354 static int mp4_read_descr(MOVContext *c, ByteIOContext *pb, int *tag)
358 len = mp4_read_descr_len(pb);
359 dprintf(c->fc, "MPEG4 description: tag=0x%02x len=%d\n", *tag, len);
363 #define MP4ESDescrTag 0x03
364 #define MP4DecConfigDescrTag 0x04
365 #define MP4DecSpecificDescrTag 0x05
367 static const AVCodecTag mp4_audio_types[] = {
368 { CODEC_ID_MP3ON4, 29 }, /* old mp3on4 draft */
369 { CODEC_ID_MP3ON4, 32 }, /* layer 1 */
370 { CODEC_ID_MP3ON4, 33 }, /* layer 2 */
371 { CODEC_ID_MP3ON4, 34 }, /* layer 3 */
372 { CODEC_ID_NONE, 0 },
375 static int mov_read_esds(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
377 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
380 get_be32(pb); /* version + flags */
381 len = mp4_read_descr(c, pb, &tag);
382 if (tag == MP4ESDescrTag) {
383 get_be16(pb); /* ID */
384 get_byte(pb); /* priority */
386 get_be16(pb); /* ID */
388 len = mp4_read_descr(c, pb, &tag);
389 if (tag == MP4DecConfigDescrTag) {
390 int object_type_id = get_byte(pb);
391 get_byte(pb); /* stream type */
392 get_be24(pb); /* buffer size db */
393 get_be32(pb); /* max bitrate */
394 get_be32(pb); /* avg bitrate */
396 st->codec->codec_id= codec_get_id(ff_mp4_obj_type, object_type_id);
397 dprintf(c->fc, "esds object type id %d\n", object_type_id);
398 len = mp4_read_descr(c, pb, &tag);
399 if (tag == MP4DecSpecificDescrTag) {
400 dprintf(c->fc, "Specific MPEG4 header len=%d\n", len);
401 if((uint64_t)len > (1<<30))
403 st->codec->extradata = av_mallocz(len + FF_INPUT_BUFFER_PADDING_SIZE);
404 if (!st->codec->extradata)
405 return AVERROR(ENOMEM);
406 get_buffer(pb, st->codec->extradata, len);
407 st->codec->extradata_size = len;
408 if (st->codec->codec_id == CODEC_ID_AAC) {
409 MPEG4AudioConfig cfg;
410 ff_mpeg4audio_get_config(&cfg, st->codec->extradata,
411 st->codec->extradata_size);
412 if (cfg.chan_config > 7)
414 st->codec->channels = ff_mpeg4audio_channels[cfg.chan_config];
415 if (cfg.object_type == 29 && cfg.sampling_index < 3) // old mp3on4
416 st->codec->sample_rate = ff_mpa_freq_tab[cfg.sampling_index];
418 st->codec->sample_rate = cfg.sample_rate; // ext sample rate ?
419 dprintf(c->fc, "mp4a config channels %d obj %d ext obj %d "
420 "sample rate %d ext sample rate %d\n", st->codec->channels,
421 cfg.object_type, cfg.ext_object_type,
422 cfg.sample_rate, cfg.ext_sample_rate);
423 if (!(st->codec->codec_id = codec_get_id(mp4_audio_types,
425 st->codec->codec_id = CODEC_ID_AAC;
432 /* this atom contains actual media data */
433 static int mov_read_mdat(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
435 if(atom.size == 0) /* wrong one (MP4) */
438 return 0; /* now go for moov */
441 static int mov_read_ftyp(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
443 uint32_t type = get_le32(pb);
445 if (type != MKTAG('q','t',' ',' '))
447 av_log(c->fc, AV_LOG_DEBUG, "ISO: File Type Major Brand: %.4s\n",(char *)&type);
448 get_be32(pb); /* minor version */
449 url_fskip(pb, atom.size - 8);
453 /* this atom should contain all header atoms */
454 static int mov_read_moov(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
456 if (mov_read_default(c, pb, atom) < 0)
458 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
459 /* so we don't parse the whole file if over a network */
461 return 0; /* now go for mdat */
464 static int mov_read_moof(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
466 c->fragment.moof_offset = url_ftell(pb) - 8;
467 dprintf(c->fc, "moof offset %llx\n", c->fragment.moof_offset);
468 return mov_read_default(c, pb, atom);
471 static int mov_read_mdhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
473 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
474 MOVStreamContext *sc = st->priv_data;
475 int version = get_byte(pb);
479 return -1; /* unsupported */
481 get_be24(pb); /* flags */
486 get_be32(pb); /* creation time */
487 get_be32(pb); /* modification time */
490 sc->time_scale = get_be32(pb);
491 st->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
493 lang = get_be16(pb); /* language */
494 ff_mov_lang_to_iso639(lang, st->language);
495 get_be16(pb); /* quality */
500 static int mov_read_mvhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
502 int version = get_byte(pb); /* version */
503 get_be24(pb); /* flags */
509 get_be32(pb); /* creation time */
510 get_be32(pb); /* modification time */
512 c->time_scale = get_be32(pb); /* time scale */
514 dprintf(c->fc, "time scale = %i\n", c->time_scale);
516 c->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
517 get_be32(pb); /* preferred scale */
519 get_be16(pb); /* preferred volume */
521 url_fskip(pb, 10); /* reserved */
523 url_fskip(pb, 36); /* display matrix */
525 get_be32(pb); /* preview time */
526 get_be32(pb); /* preview duration */
527 get_be32(pb); /* poster time */
528 get_be32(pb); /* selection time */
529 get_be32(pb); /* selection duration */
530 get_be32(pb); /* current time */
531 get_be32(pb); /* next track ID */
536 static int mov_read_smi(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
538 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
540 if((uint64_t)atom.size > (1<<30))
543 // currently SVQ3 decoder expect full STSD header - so let's fake it
544 // this should be fixed and just SMI header should be passed
545 av_free(st->codec->extradata);
546 st->codec->extradata = av_mallocz(atom.size + 0x5a + FF_INPUT_BUFFER_PADDING_SIZE);
547 if (!st->codec->extradata)
548 return AVERROR(ENOMEM);
549 st->codec->extradata_size = 0x5a + atom.size;
550 memcpy(st->codec->extradata, "SVQ3", 4); // fake
551 get_buffer(pb, st->codec->extradata + 0x5a, atom.size);
552 dprintf(c->fc, "Reading SMI %"PRId64" %s\n", atom.size, st->codec->extradata + 0x5a);
556 static int mov_read_enda(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
558 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
559 int little_endian = get_be16(pb);
562 switch (st->codec->codec_id) {
563 case CODEC_ID_PCM_S24BE:
564 st->codec->codec_id = CODEC_ID_PCM_S24LE;
566 case CODEC_ID_PCM_S32BE:
567 st->codec->codec_id = CODEC_ID_PCM_S32LE;
569 case CODEC_ID_PCM_F32BE:
570 st->codec->codec_id = CODEC_ID_PCM_F32LE;
572 case CODEC_ID_PCM_F64BE:
573 st->codec->codec_id = CODEC_ID_PCM_F64LE;
582 /* FIXME modify qdm2/svq3/h264 decoders to take full atom as extradata */
583 static int mov_read_extradata(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
585 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
586 uint64_t size= (uint64_t)st->codec->extradata_size + atom.size + 8 + FF_INPUT_BUFFER_PADDING_SIZE;
588 if(size > INT_MAX || (uint64_t)atom.size > INT_MAX)
590 buf= av_realloc(st->codec->extradata, size);
593 st->codec->extradata= buf;
594 buf+= st->codec->extradata_size;
595 st->codec->extradata_size= size - FF_INPUT_BUFFER_PADDING_SIZE;
596 AV_WB32( buf , atom.size + 8);
597 AV_WL32( buf + 4, atom.type);
598 get_buffer(pb, buf + 8, atom.size);
602 static int mov_read_wave(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
604 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
606 if((uint64_t)atom.size > (1<<30))
609 if (st->codec->codec_id == CODEC_ID_QDM2) {
610 // pass all frma atom to codec, needed at least for QDM2
611 av_free(st->codec->extradata);
612 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
613 if (!st->codec->extradata)
614 return AVERROR(ENOMEM);
615 st->codec->extradata_size = atom.size;
616 get_buffer(pb, st->codec->extradata, atom.size);
617 } else if (atom.size > 8) { /* to read frma, esds atoms */
618 if (mov_read_default(c, pb, atom) < 0)
621 url_fskip(pb, atom.size);
626 * This function reads atom content and puts data in extradata without tag
627 * nor size unlike mov_read_extradata.
629 static int mov_read_glbl(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
631 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
633 if((uint64_t)atom.size > (1<<30))
636 av_free(st->codec->extradata);
637 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
638 if (!st->codec->extradata)
639 return AVERROR(ENOMEM);
640 st->codec->extradata_size = atom.size;
641 get_buffer(pb, st->codec->extradata, atom.size);
645 static int mov_read_stco(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
647 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
648 MOVStreamContext *sc = st->priv_data;
649 unsigned int i, entries;
651 get_byte(pb); /* version */
652 get_be24(pb); /* flags */
654 entries = get_be32(pb);
656 if(entries >= UINT_MAX/sizeof(int64_t))
659 sc->chunk_count = entries;
660 sc->chunk_offsets = av_malloc(entries * sizeof(int64_t));
661 if (!sc->chunk_offsets)
663 if (atom.type == MKTAG('s','t','c','o'))
664 for(i=0; i<entries; i++)
665 sc->chunk_offsets[i] = get_be32(pb);
666 else if (atom.type == MKTAG('c','o','6','4'))
667 for(i=0; i<entries; i++)
668 sc->chunk_offsets[i] = get_be64(pb);
676 * Compute codec id for 'lpcm' tag.
677 * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
679 static enum CodecID mov_get_lpcm_codec_id(int bps, int flags)
681 if (flags & 1) { // floating point
682 if (flags & 2) { // big endian
683 if (bps == 32) return CODEC_ID_PCM_F32BE;
684 else if (bps == 64) return CODEC_ID_PCM_F64BE;
686 if (bps == 32) return CODEC_ID_PCM_F32LE;
687 else if (bps == 64) return CODEC_ID_PCM_F64LE;
693 if (flags & 4) return CODEC_ID_PCM_S8;
694 else return CODEC_ID_PCM_U8;
695 else if (bps == 16) return CODEC_ID_PCM_S16BE;
696 else if (bps == 24) return CODEC_ID_PCM_S24BE;
697 else if (bps == 32) return CODEC_ID_PCM_S32BE;
700 if (flags & 4) return CODEC_ID_PCM_S8;
701 else return CODEC_ID_PCM_U8;
702 else if (bps == 16) return CODEC_ID_PCM_S16LE;
703 else if (bps == 24) return CODEC_ID_PCM_S24LE;
704 else if (bps == 32) return CODEC_ID_PCM_S32LE;
707 return CODEC_ID_NONE;
710 static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
712 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
713 MOVStreamContext *sc = st->priv_data;
714 int j, entries, pseudo_stream_id;
716 get_byte(pb); /* version */
717 get_be24(pb); /* flags */
719 entries = get_be32(pb);
721 for(pseudo_stream_id=0; pseudo_stream_id<entries; pseudo_stream_id++) {
722 //Parsing Sample description table
725 MOV_atom_t a = { 0, 0, 0 };
726 int64_t start_pos = url_ftell(pb);
727 int size = get_be32(pb); /* size */
728 uint32_t format = get_le32(pb); /* data format */
730 get_be32(pb); /* reserved */
731 get_be16(pb); /* reserved */
732 dref_id = get_be16(pb);
734 if (st->codec->codec_tag &&
735 st->codec->codec_tag != format &&
736 (c->fc->video_codec_id ? codec_get_id(codec_movvideo_tags, format) != c->fc->video_codec_id
737 : st->codec->codec_tag != MKTAG('j','p','e','g'))
739 /* Multiple fourcc, we skip JPEG. This is not correct, we should
740 * export it as a separate AVStream but this needs a few changes
741 * in the MOV demuxer, patch welcome. */
742 av_log(c->fc, AV_LOG_WARNING, "multiple fourcc not supported\n");
743 url_fskip(pb, size - (url_ftell(pb) - start_pos));
746 sc->pseudo_stream_id = st->codec->codec_tag ? -1 : pseudo_stream_id;
747 sc->dref_id= dref_id;
749 st->codec->codec_tag = format;
750 id = codec_get_id(codec_movaudio_tags, format);
751 if (id<=0 && (format&0xFFFF) == 'm'+('s'<<8))
752 id = codec_get_id(codec_wav_tags, bswap_32(format)&0xFFFF);
754 if (st->codec->codec_type != CODEC_TYPE_VIDEO && id > 0) {
755 st->codec->codec_type = CODEC_TYPE_AUDIO;
756 } else if (st->codec->codec_type != CODEC_TYPE_AUDIO && /* do not overwrite codec type */
757 format && format != MKTAG('m','p','4','s')) { /* skip old asf mpeg4 tag */
758 id = codec_get_id(codec_movvideo_tags, format);
760 id = codec_get_id(codec_bmp_tags, format);
762 st->codec->codec_type = CODEC_TYPE_VIDEO;
763 else if(st->codec->codec_type == CODEC_TYPE_DATA){
764 id = codec_get_id(ff_codec_movsubtitle_tags, format);
766 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
770 dprintf(c->fc, "size=%d 4CC= %c%c%c%c codec_type=%d\n", size,
771 (format >> 0) & 0xff, (format >> 8) & 0xff, (format >> 16) & 0xff,
772 (format >> 24) & 0xff, st->codec->codec_type);
774 if(st->codec->codec_type==CODEC_TYPE_VIDEO) {
775 uint8_t codec_name[32];
776 unsigned int color_depth;
779 st->codec->codec_id = id;
780 get_be16(pb); /* version */
781 get_be16(pb); /* revision level */
782 get_be32(pb); /* vendor */
783 get_be32(pb); /* temporal quality */
784 get_be32(pb); /* spatial quality */
786 st->codec->width = get_be16(pb); /* width */
787 st->codec->height = get_be16(pb); /* height */
789 get_be32(pb); /* horiz resolution */
790 get_be32(pb); /* vert resolution */
791 get_be32(pb); /* data size, always 0 */
792 get_be16(pb); /* frames per samples */
794 get_buffer(pb, codec_name, 32); /* codec name, pascal string */
795 if (codec_name[0] <= 31) {
796 memcpy(st->codec->codec_name, &codec_name[1],codec_name[0]);
797 st->codec->codec_name[codec_name[0]] = 0;
800 st->codec->bits_per_coded_sample = get_be16(pb); /* depth */
801 st->codec->color_table_id = get_be16(pb); /* colortable id */
802 dprintf(c->fc, "depth %d, ctab id %d\n",
803 st->codec->bits_per_coded_sample, st->codec->color_table_id);
804 /* figure out the palette situation */
805 color_depth = st->codec->bits_per_coded_sample & 0x1F;
806 color_greyscale = st->codec->bits_per_coded_sample & 0x20;
808 /* if the depth is 2, 4, or 8 bpp, file is palettized */
809 if ((color_depth == 2) || (color_depth == 4) ||
810 (color_depth == 8)) {
811 /* for palette traversal */
812 unsigned int color_start, color_count, color_end;
813 unsigned char r, g, b;
815 if (color_greyscale) {
816 int color_index, color_dec;
817 /* compute the greyscale palette */
818 st->codec->bits_per_coded_sample = color_depth;
819 color_count = 1 << color_depth;
821 color_dec = 256 / (color_count - 1);
822 for (j = 0; j < color_count; j++) {
823 r = g = b = color_index;
824 c->palette_control.palette[j] =
825 (r << 16) | (g << 8) | (b);
826 color_index -= color_dec;
830 } else if (st->codec->color_table_id) {
831 const uint8_t *color_table;
832 /* if flag bit 3 is set, use the default palette */
833 color_count = 1 << color_depth;
834 if (color_depth == 2)
835 color_table = ff_qt_default_palette_4;
836 else if (color_depth == 4)
837 color_table = ff_qt_default_palette_16;
839 color_table = ff_qt_default_palette_256;
841 for (j = 0; j < color_count; j++) {
842 r = color_table[j * 4 + 0];
843 g = color_table[j * 4 + 1];
844 b = color_table[j * 4 + 2];
845 c->palette_control.palette[j] =
846 (r << 16) | (g << 8) | (b);
849 /* load the palette from the file */
850 color_start = get_be32(pb);
851 color_count = get_be16(pb);
852 color_end = get_be16(pb);
853 if ((color_start <= 255) &&
854 (color_end <= 255)) {
855 for (j = color_start; j <= color_end; j++) {
856 /* each R, G, or B component is 16 bits;
857 * only use the top 8 bits; skip alpha bytes
867 c->palette_control.palette[j] =
868 (r << 16) | (g << 8) | (b);
872 st->codec->palctrl = &c->palette_control;
873 st->codec->palctrl->palette_changed = 1;
875 st->codec->palctrl = NULL;
876 } else if(st->codec->codec_type==CODEC_TYPE_AUDIO) {
877 int bits_per_sample, flags;
878 uint16_t version = get_be16(pb);
880 st->codec->codec_id = id;
881 get_be16(pb); /* revision level */
882 get_be32(pb); /* vendor */
884 st->codec->channels = get_be16(pb); /* channel count */
885 dprintf(c->fc, "audio channels %d\n", st->codec->channels);
886 st->codec->bits_per_coded_sample = get_be16(pb); /* sample size */
888 sc->audio_cid = get_be16(pb);
889 get_be16(pb); /* packet size = 0 */
891 st->codec->sample_rate = ((get_be32(pb) >> 16));
893 //Read QT version 1 fields. In version 0 these do not exist.
894 dprintf(c->fc, "version =%d, isom =%d\n",version,c->isom);
897 sc->samples_per_frame = get_be32(pb);
898 get_be32(pb); /* bytes per packet */
899 sc->bytes_per_frame = get_be32(pb);
900 get_be32(pb); /* bytes per sample */
901 } else if(version==2) {
902 get_be32(pb); /* sizeof struct only */
903 st->codec->sample_rate = av_int2dbl(get_be64(pb)); /* float 64 */
904 st->codec->channels = get_be32(pb);
905 get_be32(pb); /* always 0x7F000000 */
906 st->codec->bits_per_coded_sample = get_be32(pb); /* bits per channel if sound is uncompressed */
907 flags = get_be32(pb); /* lcpm format specific flag */
908 sc->bytes_per_frame = get_be32(pb); /* bytes per audio packet if constant */
909 sc->samples_per_frame = get_be32(pb); /* lpcm frames per audio packet if constant */
910 if (format == MKTAG('l','p','c','m'))
911 st->codec->codec_id = mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, flags);
915 switch (st->codec->codec_id) {
916 case CODEC_ID_PCM_S8:
917 case CODEC_ID_PCM_U8:
918 if (st->codec->bits_per_coded_sample == 16)
919 st->codec->codec_id = CODEC_ID_PCM_S16BE;
921 case CODEC_ID_PCM_S16LE:
922 case CODEC_ID_PCM_S16BE:
923 if (st->codec->bits_per_coded_sample == 8)
924 st->codec->codec_id = CODEC_ID_PCM_S8;
925 else if (st->codec->bits_per_coded_sample == 24)
926 st->codec->codec_id =
927 st->codec->codec_id == CODEC_ID_PCM_S16BE ?
928 CODEC_ID_PCM_S24BE : CODEC_ID_PCM_S24LE;
930 /* set values for old format before stsd version 1 appeared */
932 sc->samples_per_frame = 6;
933 sc->bytes_per_frame = 2*st->codec->channels;
936 sc->samples_per_frame = 6;
937 sc->bytes_per_frame = 1*st->codec->channels;
939 case CODEC_ID_ADPCM_IMA_QT:
940 sc->samples_per_frame = 64;
941 sc->bytes_per_frame = 34*st->codec->channels;
944 sc->samples_per_frame = 160;
945 sc->bytes_per_frame = 33;
951 bits_per_sample = av_get_bits_per_sample(st->codec->codec_id);
952 if (bits_per_sample) {
953 st->codec->bits_per_coded_sample = bits_per_sample;
954 sc->sample_size = (bits_per_sample >> 3) * st->codec->channels;
956 } else if(st->codec->codec_type==CODEC_TYPE_SUBTITLE){
957 st->codec->codec_id= id;
959 /* other codec type, just skip (rtp, mp4s, tmcd ...) */
960 url_fskip(pb, size - (url_ftell(pb) - start_pos));
962 /* this will read extra atoms at the end (wave, alac, damr, avcC, SMI ...) */
963 a.size = size - (url_ftell(pb) - start_pos);
965 if (mov_read_default(c, pb, a) < 0)
967 } else if (a.size > 0)
968 url_fskip(pb, a.size);
971 if(st->codec->codec_type==CODEC_TYPE_AUDIO && st->codec->sample_rate==0 && sc->time_scale>1)
972 st->codec->sample_rate= sc->time_scale;
974 /* special codec parameters handling */
975 switch (st->codec->codec_id) {
976 #ifdef CONFIG_DV_DEMUXER
977 case CODEC_ID_DVAUDIO:
978 c->dv_fctx = av_alloc_format_context();
979 c->dv_demux = dv_init_demux(c->dv_fctx);
981 av_log(c->fc, AV_LOG_ERROR, "dv demux context init error\n");
984 sc->dv_audio_container = 1;
985 st->codec->codec_id = CODEC_ID_PCM_S16LE;
988 /* no ifdef since parameters are always those */
990 case CODEC_ID_AMR_NB:
991 case CODEC_ID_AMR_WB:
992 st->codec->frame_size= sc->samples_per_frame;
993 st->codec->channels= 1; /* really needed */
994 /* force sample rate for amr, stsd in 3gp does not store sample rate */
995 if (st->codec->codec_id != CODEC_ID_QCELP)
996 st->codec->sample_rate = 8000;
1000 st->codec->codec_type = CODEC_TYPE_AUDIO; /* force type after stsd for m1a hdlr */
1001 st->need_parsing = AVSTREAM_PARSE_FULL;
1004 case CODEC_ID_ADPCM_MS:
1005 case CODEC_ID_ADPCM_IMA_WAV:
1006 st->codec->block_align = sc->bytes_per_frame;
1009 if (st->codec->extradata_size == 36)
1010 st->codec->frame_size = AV_RB32((st->codec->extradata+12));
1019 static int mov_read_stsc(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1021 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1022 MOVStreamContext *sc = st->priv_data;
1023 unsigned int i, entries;
1025 get_byte(pb); /* version */
1026 get_be24(pb); /* flags */
1028 entries = get_be32(pb);
1030 if(entries >= UINT_MAX / sizeof(MOV_stsc_t))
1033 dprintf(c->fc, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
1035 sc->sample_to_chunk_sz = entries;
1036 sc->sample_to_chunk = av_malloc(entries * sizeof(MOV_stsc_t));
1037 if (!sc->sample_to_chunk)
1039 for(i=0; i<entries; i++) {
1040 sc->sample_to_chunk[i].first = get_be32(pb);
1041 sc->sample_to_chunk[i].count = get_be32(pb);
1042 sc->sample_to_chunk[i].id = get_be32(pb);
1047 static int mov_read_stss(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1049 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1050 MOVStreamContext *sc = st->priv_data;
1051 unsigned int i, entries;
1053 get_byte(pb); /* version */
1054 get_be24(pb); /* flags */
1056 entries = get_be32(pb);
1058 if(entries >= UINT_MAX / sizeof(int))
1061 sc->keyframe_count = entries;
1063 dprintf(c->fc, "keyframe_count = %d\n", sc->keyframe_count);
1065 sc->keyframes = av_malloc(entries * sizeof(int));
1068 for(i=0; i<entries; i++) {
1069 sc->keyframes[i] = get_be32(pb);
1070 //dprintf(c->fc, "keyframes[]=%d\n", sc->keyframes[i]);
1075 static int mov_read_stsz(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1077 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1078 MOVStreamContext *sc = st->priv_data;
1079 unsigned int i, entries, sample_size;
1081 get_byte(pb); /* version */
1082 get_be24(pb); /* flags */
1084 sample_size = get_be32(pb);
1085 if (!sc->sample_size) /* do not overwrite value computed in stsd */
1086 sc->sample_size = sample_size;
1087 entries = get_be32(pb);
1088 if(entries >= UINT_MAX / sizeof(int))
1091 sc->sample_count = entries;
1095 dprintf(c->fc, "sample_size = %d sample_count = %d\n", sc->sample_size, sc->sample_count);
1097 sc->sample_sizes = av_malloc(entries * sizeof(int));
1098 if (!sc->sample_sizes)
1100 for(i=0; i<entries; i++)
1101 sc->sample_sizes[i] = get_be32(pb);
1105 static int mov_read_stts(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1107 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1108 MOVStreamContext *sc = st->priv_data;
1109 unsigned int i, entries;
1111 int64_t total_sample_count=0;
1113 get_byte(pb); /* version */
1114 get_be24(pb); /* flags */
1115 entries = get_be32(pb);
1116 if(entries >= UINT_MAX / sizeof(MOV_stts_t))
1119 sc->stts_count = entries;
1120 sc->stts_data = av_malloc(entries * sizeof(MOV_stts_t));
1123 dprintf(c->fc, "track[%i].stts.entries = %i\n", c->fc->nb_streams-1, entries);
1127 for(i=0; i<entries; i++) {
1128 int sample_duration;
1131 sample_count=get_be32(pb);
1132 sample_duration = get_be32(pb);
1133 sc->stts_data[i].count= sample_count;
1134 sc->stts_data[i].duration= sample_duration;
1136 sc->time_rate= ff_gcd(sc->time_rate, sample_duration);
1138 dprintf(c->fc, "sample_count=%d, sample_duration=%d\n",sample_count,sample_duration);
1140 duration+=(int64_t)sample_duration*sample_count;
1141 total_sample_count+=sample_count;
1144 st->nb_frames= total_sample_count;
1146 st->duration= duration;
1150 static int mov_read_ctts(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1152 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1153 MOVStreamContext *sc = st->priv_data;
1154 unsigned int i, entries;
1156 get_byte(pb); /* version */
1157 get_be24(pb); /* flags */
1158 entries = get_be32(pb);
1159 if(entries >= UINT_MAX / sizeof(MOV_stts_t))
1162 sc->ctts_count = entries;
1163 sc->ctts_data = av_malloc(entries * sizeof(MOV_stts_t));
1166 dprintf(c->fc, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
1168 for(i=0; i<entries; i++) {
1169 int count =get_be32(pb);
1170 int duration =get_be32(pb);
1174 st->codec->has_b_frames = 1;
1176 sc->ctts_data[i].count = count;
1177 sc->ctts_data[i].duration= duration;
1179 sc->time_rate= ff_gcd(sc->time_rate, FFABS(duration));
1184 static void mov_build_index(MOVContext *mov, AVStream *st)
1186 MOVStreamContext *sc = st->priv_data;
1187 int64_t current_offset;
1188 int64_t current_dts = 0;
1189 unsigned int stts_index = 0;
1190 unsigned int stsc_index = 0;
1191 unsigned int stss_index = 0;
1194 /* only use old uncompressed audio chunk demuxing when stts specifies it */
1195 if (!(st->codec->codec_type == CODEC_TYPE_AUDIO &&
1196 sc->stts_count == 1 && sc->stts_data[0].duration == 1)) {
1197 unsigned int current_sample = 0;
1198 unsigned int stts_sample = 0;
1199 unsigned int keyframe, sample_size;
1200 unsigned int distance = 0;
1201 int key_off = sc->keyframes && sc->keyframes[0] == 1;
1203 st->nb_frames = sc->sample_count;
1204 for (i = 0; i < sc->chunk_count; i++) {
1205 current_offset = sc->chunk_offsets[i];
1206 if (stsc_index + 1 < sc->sample_to_chunk_sz &&
1207 i + 1 == sc->sample_to_chunk[stsc_index + 1].first)
1209 for (j = 0; j < sc->sample_to_chunk[stsc_index].count; j++) {
1210 if (current_sample >= sc->sample_count) {
1211 av_log(mov->fc, AV_LOG_ERROR, "wrong sample count\n");
1214 keyframe = !sc->keyframe_count || current_sample+key_off == sc->keyframes[stss_index];
1217 if (stss_index + 1 < sc->keyframe_count)
1220 sample_size = sc->sample_size > 0 ? sc->sample_size : sc->sample_sizes[current_sample];
1221 if(sc->pseudo_stream_id == -1 ||
1222 sc->sample_to_chunk[stsc_index].id - 1 == sc->pseudo_stream_id) {
1223 av_add_index_entry(st, current_offset, current_dts, sample_size, distance,
1224 keyframe ? AVINDEX_KEYFRAME : 0);
1225 dprintf(mov->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1226 "size %d, distance %d, keyframe %d\n", st->index, current_sample,
1227 current_offset, current_dts, sample_size, distance, keyframe);
1229 current_offset += sample_size;
1230 assert(sc->stts_data[stts_index].duration % sc->time_rate == 0);
1231 current_dts += sc->stts_data[stts_index].duration / sc->time_rate;
1235 if (stts_index + 1 < sc->stts_count && stts_sample == sc->stts_data[stts_index].count) {
1241 } else { /* read whole chunk */
1242 unsigned int chunk_samples, chunk_size, chunk_duration;
1243 unsigned int frames = 1;
1244 for (i = 0; i < sc->chunk_count; i++) {
1245 current_offset = sc->chunk_offsets[i];
1246 if (stsc_index + 1 < sc->sample_to_chunk_sz &&
1247 i + 1 == sc->sample_to_chunk[stsc_index + 1].first)
1249 chunk_samples = sc->sample_to_chunk[stsc_index].count;
1250 /* get chunk size, beware of alaw/ulaw/mace */
1251 if (sc->samples_per_frame > 0 &&
1252 (chunk_samples * sc->bytes_per_frame % sc->samples_per_frame == 0)) {
1253 if (sc->samples_per_frame < 160)
1254 chunk_size = chunk_samples * sc->bytes_per_frame / sc->samples_per_frame;
1256 chunk_size = sc->bytes_per_frame;
1257 frames = chunk_samples / sc->samples_per_frame;
1258 chunk_samples = sc->samples_per_frame;
1261 chunk_size = chunk_samples * sc->sample_size;
1262 for (j = 0; j < frames; j++) {
1263 av_add_index_entry(st, current_offset, current_dts, chunk_size, 0, AVINDEX_KEYFRAME);
1264 /* get chunk duration */
1266 while (chunk_samples > 0) {
1267 if (chunk_samples < sc->stts_data[stts_index].count) {
1268 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1269 sc->stts_data[stts_index].count -= chunk_samples;
1272 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1273 chunk_samples -= sc->stts_data[stts_index].count;
1274 if (stts_index + 1 < sc->stts_count)
1278 current_offset += sc->bytes_per_frame;
1279 dprintf(mov->fc, "AVIndex stream %d, chunk %d, offset %"PRIx64", dts %"PRId64", "
1280 "size %d, duration %d\n", st->index, i, current_offset, current_dts,
1281 chunk_size, chunk_duration);
1282 assert(chunk_duration % sc->time_rate == 0);
1283 current_dts += chunk_duration / sc->time_rate;
1288 /* adjust sample count to avindex entries */
1289 sc->sample_count = st->nb_index_entries;
1292 static int mov_read_trak(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1295 MOVStreamContext *sc;
1298 st = av_new_stream(c->fc, c->fc->nb_streams);
1299 if (!st) return AVERROR(ENOMEM);
1300 sc = av_mallocz(sizeof(MOVStreamContext));
1301 if (!sc) return AVERROR(ENOMEM);
1304 st->codec->codec_type = CODEC_TYPE_DATA;
1305 st->start_time = 0; /* XXX: check */
1307 if ((ret = mov_read_default(c, pb, atom)) < 0)
1311 if(sc->chunk_count && (!sc->stts_count || !sc->sample_to_chunk_sz ||
1312 (!sc->sample_size && !sc->sample_count))){
1313 av_log(c->fc, AV_LOG_ERROR, "stream %d, missing mandatory atoms, broken header\n",
1315 sc->sample_count = 0; //ignore track
1321 sc->time_scale= c->time_scale;
1322 av_set_pts_info(st, 64, sc->time_rate, sc->time_scale);
1324 if (st->codec->codec_type == CODEC_TYPE_AUDIO &&
1325 !st->codec->frame_size && sc->stts_count == 1)
1326 st->codec->frame_size = av_rescale(sc->time_rate, st->codec->sample_rate, sc->time_scale);
1328 if(st->duration != AV_NOPTS_VALUE){
1329 assert(st->duration % sc->time_rate == 0);
1330 st->duration /= sc->time_rate;
1332 sc->ffindex = st->index;
1333 mov_build_index(c, st);
1335 if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
1336 if (url_fopen(&sc->pb, sc->drefs[sc->dref_id-1].path, URL_RDONLY) < 0)
1337 av_log(c->fc, AV_LOG_ERROR, "stream %d, error opening file %s: %s\n",
1338 st->index, sc->drefs[sc->dref_id-1].path, strerror(errno));
1342 switch (st->codec->codec_id) {
1343 #ifdef CONFIG_H261_DECODER
1346 #ifdef CONFIG_H263_DECODER
1349 #ifdef CONFIG_MPEG4_DECODER
1350 case CODEC_ID_MPEG4:
1352 st->codec->width= 0; /* let decoder init width/height */
1353 st->codec->height= 0;
1357 /* Do not need those anymore. */
1358 av_freep(&sc->chunk_offsets);
1359 av_freep(&sc->sample_to_chunk);
1360 av_freep(&sc->sample_sizes);
1361 av_freep(&sc->keyframes);
1362 av_freep(&sc->stts_data);
1367 static void mov_parse_udta_string(ByteIOContext *pb, char *str, int size)
1369 uint16_t str_size = get_be16(pb); /* string length */;
1371 get_be16(pb); /* skip language */
1372 get_buffer(pb, str, FFMIN(size, str_size));
1375 static int mov_read_udta(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1377 uint64_t end = url_ftell(pb) + atom.size;
1379 while (url_ftell(pb) + 8 < end) {
1380 uint32_t tag_size = get_be32(pb);
1381 uint32_t tag = get_le32(pb);
1382 uint64_t next = url_ftell(pb) + tag_size - 8;
1384 if (tag_size < 8 || next > end) // stop if tag_size is wrong
1388 case MKTAG(0xa9,'n','a','m'):
1389 mov_parse_udta_string(pb, c->fc->title, sizeof(c->fc->title));
1391 case MKTAG(0xa9,'w','r','t'):
1392 mov_parse_udta_string(pb, c->fc->author, sizeof(c->fc->author));
1394 case MKTAG(0xa9,'c','p','y'):
1395 mov_parse_udta_string(pb, c->fc->copyright, sizeof(c->fc->copyright));
1397 case MKTAG(0xa9,'i','n','f'):
1398 mov_parse_udta_string(pb, c->fc->comment, sizeof(c->fc->comment));
1404 url_fseek(pb, next, SEEK_SET);
1410 static int mov_read_tkhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1415 int64_t disp_transform[2];
1416 int display_matrix[3][2];
1417 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1418 int version = get_byte(pb);
1420 get_be24(pb); /* flags */
1422 MOV_TRACK_ENABLED 0x0001
1423 MOV_TRACK_IN_MOVIE 0x0002
1424 MOV_TRACK_IN_PREVIEW 0x0004
1425 MOV_TRACK_IN_POSTER 0x0008
1432 get_be32(pb); /* creation time */
1433 get_be32(pb); /* modification time */
1435 st->id = (int)get_be32(pb); /* track id (NOT 0 !)*/
1436 get_be32(pb); /* reserved */
1437 st->start_time = 0; /* check */
1438 /* highlevel (considering edits) duration in movie timebase */
1439 (version == 1) ? get_be64(pb) : get_be32(pb);
1440 get_be32(pb); /* reserved */
1441 get_be32(pb); /* reserved */
1443 get_be16(pb); /* layer */
1444 get_be16(pb); /* alternate group */
1445 get_be16(pb); /* volume */
1446 get_be16(pb); /* reserved */
1448 //read in the display matrix (outlined in ISO 14496-12, Section 6.2.2)
1449 // they're kept in fixed point format through all calculations
1450 // ignore u,v,z b/c we don't need the scale factor to calc aspect ratio
1451 for (i = 0; i < 3; i++) {
1452 display_matrix[i][0] = get_be32(pb); // 16.16 fixed point
1453 display_matrix[i][1] = get_be32(pb); // 16.16 fixed point
1454 get_be32(pb); // 2.30 fixed point (not used)
1457 width = get_be32(pb); // 16.16 fixed point track width
1458 height = get_be32(pb); // 16.16 fixed point track height
1460 //transform the display width/height according to the matrix
1461 // skip this if the display matrix is the default identity matrix
1462 // to keep the same scale, use [width height 1<<16]
1463 if (width && height &&
1464 (display_matrix[0][0] != 65536 || display_matrix[0][1] ||
1465 display_matrix[1][0] || display_matrix[1][1] != 65536 ||
1466 display_matrix[2][0] || display_matrix[2][1])) {
1467 for (i = 0; i < 2; i++)
1469 (int64_t) width * display_matrix[0][i] +
1470 (int64_t) height * display_matrix[1][i] +
1471 ((int64_t) display_matrix[2][i] << 16);
1473 //sample aspect ratio is new width/height divided by old width/height
1474 st->sample_aspect_ratio = av_d2q(
1475 ((double) disp_transform[0] * height) /
1476 ((double) disp_transform[1] * width), INT_MAX);
1481 static int mov_read_tfhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1483 MOVFragment *frag = &c->fragment;
1484 MOVTrackExt *trex = NULL;
1485 int flags, track_id, i;
1487 get_byte(pb); /* version */
1488 flags = get_be24(pb);
1490 track_id = get_be32(pb);
1491 if (!track_id || track_id > c->fc->nb_streams)
1493 frag->track_id = track_id;
1494 for (i = 0; i < c->trex_count; i++)
1495 if (c->trex_data[i].track_id == frag->track_id) {
1496 trex = &c->trex_data[i];
1500 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding trex\n");
1504 if (flags & 0x01) frag->base_data_offset = get_be64(pb);
1505 else frag->base_data_offset = frag->moof_offset;
1506 if (flags & 0x02) frag->stsd_id = get_be32(pb);
1507 else frag->stsd_id = trex->stsd_id;
1509 frag->duration = flags & 0x08 ? get_be32(pb) : trex->duration;
1510 frag->size = flags & 0x10 ? get_be32(pb) : trex->size;
1511 frag->flags = flags & 0x20 ? get_be32(pb) : trex->flags;
1512 dprintf(c->fc, "frag flags 0x%x\n", frag->flags);
1516 static int mov_read_trex(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1520 if ((uint64_t)c->trex_count+1 >= UINT_MAX / sizeof(*c->trex_data))
1522 c->trex_data = av_realloc(c->trex_data, (c->trex_count+1)*sizeof(*c->trex_data));
1524 return AVERROR(ENOMEM);
1525 trex = &c->trex_data[c->trex_count++];
1526 get_byte(pb); /* version */
1527 get_be24(pb); /* flags */
1528 trex->track_id = get_be32(pb);
1529 trex->stsd_id = get_be32(pb);
1530 trex->duration = get_be32(pb);
1531 trex->size = get_be32(pb);
1532 trex->flags = get_be32(pb);
1536 static int mov_read_trun(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1538 MOVFragment *frag = &c->fragment;
1540 MOVStreamContext *sc;
1543 int data_offset = 0;
1544 unsigned entries, first_sample_flags = frag->flags;
1545 int flags, distance, i;
1547 if (!frag->track_id || frag->track_id > c->fc->nb_streams)
1549 st = c->fc->streams[frag->track_id-1];
1551 if (sc->pseudo_stream_id+1 != frag->stsd_id)
1553 get_byte(pb); /* version */
1554 flags = get_be24(pb);
1555 entries = get_be32(pb);
1556 dprintf(c->fc, "flags 0x%x entries %d\n", flags, entries);
1557 if (flags & 0x001) data_offset = get_be32(pb);
1558 if (flags & 0x004) first_sample_flags = get_be32(pb);
1559 if (flags & 0x800) {
1560 if ((uint64_t)entries+sc->ctts_count >= UINT_MAX/sizeof(*sc->ctts_data))
1562 sc->ctts_data = av_realloc(sc->ctts_data,
1563 (entries+sc->ctts_count)*sizeof(*sc->ctts_data));
1565 return AVERROR(ENOMEM);
1568 offset = frag->base_data_offset + data_offset;
1570 dprintf(c->fc, "first sample flags 0x%x\n", first_sample_flags);
1571 for (i = 0; i < entries; i++) {
1572 unsigned sample_size = frag->size;
1573 int sample_flags = i ? frag->flags : first_sample_flags;
1574 unsigned sample_duration = frag->duration;
1577 if (flags & 0x100) sample_duration = get_be32(pb);
1578 if (flags & 0x200) sample_size = get_be32(pb);
1579 if (flags & 0x400) sample_flags = get_be32(pb);
1580 if (flags & 0x800) {
1581 sc->ctts_data[sc->ctts_count].count = 1;
1582 sc->ctts_data[sc->ctts_count].duration = get_be32(pb);
1585 if ((keyframe = st->codec->codec_type == CODEC_TYPE_AUDIO ||
1586 (flags & 0x004 && !i && !sample_flags) || sample_flags & 0x2000000))
1588 av_add_index_entry(st, offset, dts, sample_size, distance,
1589 keyframe ? AVINDEX_KEYFRAME : 0);
1590 dprintf(c->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1591 "size %d, distance %d, keyframe %d\n", st->index, sc->sample_count+i,
1592 offset, dts, sample_size, distance, keyframe);
1594 assert(sample_duration % sc->time_rate == 0);
1595 dts += sample_duration / sc->time_rate;
1596 offset += sample_size;
1598 frag->moof_offset = offset;
1599 sc->sample_count = st->nb_index_entries;
1604 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
1605 /* like the files created with Adobe Premiere 5.0, for samples see */
1606 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
1607 static int mov_read_wide(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1612 return 0; /* continue */
1613 if (get_be32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
1614 url_fskip(pb, atom.size - 4);
1617 atom.type = get_le32(pb);
1620 if (atom.type != MKTAG('m','d','a','t')) {
1621 url_fskip(pb, atom.size);
1624 err = mov_read_mdat(c, pb, atom);
1628 static int mov_read_cmov(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1633 uint8_t *moov_data; /* uncompressed data */
1634 long cmov_len, moov_len;
1637 get_be32(pb); /* dcom atom */
1638 if (get_le32(pb) != MKTAG('d','c','o','m'))
1640 if (get_le32(pb) != MKTAG('z','l','i','b')) {
1641 av_log(NULL, AV_LOG_ERROR, "unknown compression for cmov atom !");
1644 get_be32(pb); /* cmvd atom */
1645 if (get_le32(pb) != MKTAG('c','m','v','d'))
1647 moov_len = get_be32(pb); /* uncompressed size */
1648 cmov_len = atom.size - 6 * 4;
1650 cmov_data = av_malloc(cmov_len);
1653 moov_data = av_malloc(moov_len);
1658 get_buffer(pb, cmov_data, cmov_len);
1659 if(uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
1660 goto free_and_return;
1661 if(init_put_byte(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL, NULL) != 0)
1662 goto free_and_return;
1663 atom.type = MKTAG('m','o','o','v');
1665 atom.size = moov_len;
1667 // { int fd = open("/tmp/uncompheader.mov", O_WRONLY | O_CREAT); write(fd, moov_data, moov_len); close(fd); }
1669 ret = mov_read_default(c, &ctx, atom);
1675 av_log(c->fc, AV_LOG_ERROR, "this file requires zlib support compiled in\n");
1680 /* edit list atom */
1681 static int mov_read_elst(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1683 MOVStreamContext *sc = c->fc->streams[c->fc->nb_streams-1]->priv_data;
1686 get_byte(pb); /* version */
1687 get_be24(pb); /* flags */
1688 edit_count= sc->edit_count = get_be32(pb); /* entries */
1690 for(i=0; i<edit_count; i++){
1692 get_be32(pb); /* Track duration */
1693 time = get_be32(pb); /* Media time */
1694 get_be32(pb); /* Media rate */
1696 av_log(c->fc, AV_LOG_WARNING, "edit list not starting at 0, "
1697 "a/v desync might occur, patch welcome\n");
1699 dprintf(c->fc, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, sc->edit_count);
1703 static const MOVParseTableEntry mov_default_parse_table[] = {
1704 { MKTAG('a','v','s','s'), mov_read_extradata },
1705 { MKTAG('c','o','6','4'), mov_read_stco },
1706 { MKTAG('c','t','t','s'), mov_read_ctts }, /* composition time to sample */
1707 { MKTAG('d','i','n','f'), mov_read_default },
1708 { MKTAG('d','r','e','f'), mov_read_dref },
1709 { MKTAG('e','d','t','s'), mov_read_default },
1710 { MKTAG('e','l','s','t'), mov_read_elst },
1711 { MKTAG('e','n','d','a'), mov_read_enda },
1712 { MKTAG('f','i','e','l'), mov_read_extradata },
1713 { MKTAG('f','t','y','p'), mov_read_ftyp },
1714 { MKTAG('g','l','b','l'), mov_read_glbl },
1715 { MKTAG('h','d','l','r'), mov_read_hdlr },
1716 { MKTAG('j','p','2','h'), mov_read_extradata },
1717 { MKTAG('m','d','a','t'), mov_read_mdat },
1718 { MKTAG('m','d','h','d'), mov_read_mdhd },
1719 { MKTAG('m','d','i','a'), mov_read_default },
1720 { MKTAG('m','i','n','f'), mov_read_default },
1721 { MKTAG('m','o','o','f'), mov_read_moof },
1722 { MKTAG('m','o','o','v'), mov_read_moov },
1723 { MKTAG('m','v','e','x'), mov_read_default },
1724 { MKTAG('m','v','h','d'), mov_read_mvhd },
1725 { MKTAG('S','M','I',' '), mov_read_smi }, /* Sorenson extension ??? */
1726 { MKTAG('a','l','a','c'), mov_read_extradata }, /* alac specific atom */
1727 { MKTAG('a','v','c','C'), mov_read_glbl },
1728 { MKTAG('s','t','b','l'), mov_read_default },
1729 { MKTAG('s','t','c','o'), mov_read_stco },
1730 { MKTAG('s','t','s','c'), mov_read_stsc },
1731 { MKTAG('s','t','s','d'), mov_read_stsd }, /* sample description */
1732 { MKTAG('s','t','s','s'), mov_read_stss }, /* sync sample */
1733 { MKTAG('s','t','s','z'), mov_read_stsz }, /* sample size */
1734 { MKTAG('s','t','t','s'), mov_read_stts },
1735 { MKTAG('t','k','h','d'), mov_read_tkhd }, /* track header */
1736 { MKTAG('t','f','h','d'), mov_read_tfhd }, /* track fragment header */
1737 { MKTAG('t','r','a','k'), mov_read_trak },
1738 { MKTAG('t','r','a','f'), mov_read_default },
1739 { MKTAG('t','r','e','x'), mov_read_trex },
1740 { MKTAG('t','r','u','n'), mov_read_trun },
1741 { MKTAG('u','d','t','a'), mov_read_udta },
1742 { MKTAG('w','a','v','e'), mov_read_wave },
1743 { MKTAG('e','s','d','s'), mov_read_esds },
1744 { MKTAG('w','i','d','e'), mov_read_wide }, /* place holder */
1745 { MKTAG('c','m','o','v'), mov_read_cmov },
1749 static int mov_probe(AVProbeData *p)
1751 unsigned int offset;
1755 /* check file header */
1758 /* ignore invalid offset */
1759 if ((offset + 8) > (unsigned int)p->buf_size)
1761 tag = AV_RL32(p->buf + offset + 4);
1763 /* check for obvious tags */
1764 case MKTAG('j','P',' ',' '): /* jpeg 2000 signature */
1765 case MKTAG('m','o','o','v'):
1766 case MKTAG('m','d','a','t'):
1767 case MKTAG('p','n','o','t'): /* detect movs with preview pics like ew.mov and april.mov */
1768 case MKTAG('u','d','t','a'): /* Packet Video PVAuthor adds this and a lot of more junk */
1769 case MKTAG('f','t','y','p'):
1770 return AVPROBE_SCORE_MAX;
1771 /* those are more common words, so rate then a bit less */
1772 case MKTAG('e','d','i','w'): /* xdcam files have reverted first tags */
1773 case MKTAG('w','i','d','e'):
1774 case MKTAG('f','r','e','e'):
1775 case MKTAG('j','u','n','k'):
1776 case MKTAG('p','i','c','t'):
1777 return AVPROBE_SCORE_MAX - 5;
1778 case MKTAG(0x82,0x82,0x7f,0x7d):
1779 case MKTAG('s','k','i','p'):
1780 case MKTAG('u','u','i','d'):
1781 case MKTAG('p','r','f','l'):
1782 offset = AV_RB32(p->buf+offset) + offset;
1783 /* if we only find those cause probedata is too small at least rate them */
1784 score = AVPROBE_SCORE_MAX - 50;
1787 /* unrecognized tag */
1794 static int mov_read_header(AVFormatContext *s, AVFormatParameters *ap)
1796 MOVContext *mov = s->priv_data;
1797 ByteIOContext *pb = s->pb;
1799 MOV_atom_t atom = { 0, 0, 0 };
1802 /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
1803 if(!url_is_streamed(pb))
1804 atom.size = url_fsize(pb);
1806 atom.size = INT64_MAX;
1808 /* check MOV header */
1809 if ((err = mov_read_default(mov, pb, atom)) < 0) {
1810 av_log(s, AV_LOG_ERROR, "error reading header: %d\n", err);
1813 if (!mov->found_moov) {
1814 av_log(s, AV_LOG_ERROR, "moov atom not found\n");
1817 dprintf(mov->fc, "on_parse_exit_offset=%lld\n", url_ftell(pb));
1822 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
1824 MOVContext *mov = s->priv_data;
1825 MOVStreamContext *sc = 0;
1826 AVIndexEntry *sample = 0;
1827 int64_t best_dts = INT64_MAX;
1830 for (i = 0; i < s->nb_streams; i++) {
1831 AVStream *st = s->streams[i];
1832 MOVStreamContext *msc = st->priv_data;
1833 if (st->discard != AVDISCARD_ALL && msc->pb && msc->current_sample < msc->sample_count) {
1834 AVIndexEntry *current_sample = &st->index_entries[msc->current_sample];
1835 int64_t dts = av_rescale(current_sample->timestamp * (int64_t)msc->time_rate,
1836 AV_TIME_BASE, msc->time_scale);
1837 dprintf(s, "stream %d, sample %d, dts %"PRId64"\n", i, msc->current_sample, dts);
1838 if (!sample || (url_is_streamed(s->pb) && current_sample->pos < sample->pos) ||
1839 (!url_is_streamed(s->pb) &&
1840 ((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
1841 ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
1842 (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
1843 sample = current_sample;
1850 mov->found_mdat = 0;
1851 if (!url_is_streamed(s->pb) ||
1852 mov_read_default(mov, s->pb, (MOV_atom_t){ 0, 0, INT64_MAX }) < 0 ||
1855 dprintf(s, "read fragments, offset 0x%llx\n", url_ftell(s->pb));
1858 /* must be done just before reading, to avoid infinite loop on sample */
1859 sc->current_sample++;
1860 if (url_fseek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
1861 av_log(mov->fc, AV_LOG_ERROR, "stream %d, offset 0x%"PRIx64": partial file\n",
1862 sc->ffindex, sample->pos);
1865 av_get_packet(sc->pb, pkt, sample->size);
1866 #ifdef CONFIG_DV_DEMUXER
1867 if (mov->dv_demux && sc->dv_audio_container) {
1868 dv_produce_packet(mov->dv_demux, pkt, pkt->data, pkt->size);
1871 if (dv_get_packet(mov->dv_demux, pkt) < 0)
1875 pkt->stream_index = sc->ffindex;
1876 pkt->dts = sample->timestamp;
1877 if (sc->ctts_data) {
1878 assert(sc->ctts_data[sc->sample_to_ctime_index].duration % sc->time_rate == 0);
1879 pkt->pts = pkt->dts + sc->ctts_data[sc->sample_to_ctime_index].duration / sc->time_rate;
1880 /* update ctts context */
1881 sc->sample_to_ctime_sample++;
1882 if (sc->sample_to_ctime_index < sc->ctts_count &&
1883 sc->ctts_data[sc->sample_to_ctime_index].count == sc->sample_to_ctime_sample) {
1884 sc->sample_to_ctime_index++;
1885 sc->sample_to_ctime_sample = 0;
1888 pkt->dts = AV_NOPTS_VALUE;
1890 AVStream *st = s->streams[sc->ffindex];
1891 int64_t next_dts = (sc->current_sample < sc->sample_count) ?
1892 st->index_entries[sc->current_sample].timestamp : st->duration;
1893 pkt->duration = next_dts - pkt->dts;
1894 pkt->pts = pkt->dts;
1896 pkt->flags |= sample->flags & AVINDEX_KEYFRAME ? PKT_FLAG_KEY : 0;
1897 pkt->pos = sample->pos;
1898 dprintf(s, "stream %d, pts %"PRId64", dts %"PRId64", pos 0x%"PRIx64", duration %d\n",
1899 pkt->stream_index, pkt->pts, pkt->dts, pkt->pos, pkt->duration);
1903 static int mov_seek_stream(AVStream *st, int64_t timestamp, int flags)
1905 MOVStreamContext *sc = st->priv_data;
1906 int sample, time_sample;
1909 sample = av_index_search_timestamp(st, timestamp, flags);
1910 dprintf(st->codec, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
1911 if (sample < 0) /* not sure what to do */
1913 sc->current_sample = sample;
1914 dprintf(st->codec, "stream %d, found sample %d\n", st->index, sc->current_sample);
1915 /* adjust ctts index */
1916 if (sc->ctts_data) {
1918 for (i = 0; i < sc->ctts_count; i++) {
1919 int next = time_sample + sc->ctts_data[i].count;
1920 if (next > sc->current_sample) {
1921 sc->sample_to_ctime_index = i;
1922 sc->sample_to_ctime_sample = sc->current_sample - time_sample;
1931 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
1934 int64_t seek_timestamp, timestamp;
1938 if (stream_index >= s->nb_streams)
1941 st = s->streams[stream_index];
1942 sample = mov_seek_stream(st, sample_time, flags);
1946 /* adjust seek timestamp to found sample timestamp */
1947 seek_timestamp = st->index_entries[sample].timestamp;
1949 for (i = 0; i < s->nb_streams; i++) {
1951 if (stream_index == i || st->discard == AVDISCARD_ALL)
1954 timestamp = av_rescale_q(seek_timestamp, s->streams[stream_index]->time_base, st->time_base);
1955 mov_seek_stream(st, timestamp, flags);
1960 static int mov_read_close(AVFormatContext *s)
1963 MOVContext *mov = s->priv_data;
1964 for(i=0; i<s->nb_streams; i++) {
1965 MOVStreamContext *sc = s->streams[i]->priv_data;
1966 av_freep(&sc->ctts_data);
1967 for (j=0; j<sc->drefs_count; j++)
1968 av_freep(&sc->drefs[j].path);
1969 av_freep(&sc->drefs);
1970 if (sc->pb && sc->pb != s->pb)
1974 for(i=0; i<mov->dv_fctx->nb_streams; i++){
1975 av_freep(&mov->dv_fctx->streams[i]->codec);
1976 av_freep(&mov->dv_fctx->streams[i]);
1978 av_freep(&mov->dv_fctx);
1979 av_freep(&mov->dv_demux);
1981 av_freep(&mov->trex_data);
1985 AVInputFormat mov_demuxer = {
1986 "mov,mp4,m4a,3gp,3g2,mj2",
1987 NULL_IF_CONFIG_SMALL("QuickTime/MPEG-4/Motion JPEG 2000 format"),