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;
113 unsigned int ctts_count;
115 unsigned int edit_count; /* number of 'edit' (elst atom) */
116 unsigned int sample_to_chunk_sz;
117 MOVStsc *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;
152 int itunes_metadata; ///< metadata are itunes style
156 /* XXX: it's the first time I make a recursive parser I think... sorry if it's ugly :P */
158 /* those functions parse an atom */
160 0: continue to parse next atom
161 <0: error occurred, exit
163 /* links atom IDs to parse functions */
164 typedef struct MOVParseTableEntry {
166 int (*parse)(MOVContext *ctx, ByteIOContext *pb, MOVAtom atom);
167 } MOVParseTableEntry;
169 static const MOVParseTableEntry mov_default_parse_table[];
171 static int mov_read_default(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
173 int64_t total_size = 0;
178 a.offset = atom.offset;
181 atom.size = INT64_MAX;
182 while(((total_size + 8) < atom.size) && !url_feof(pb) && !err) {
186 a.size = get_be32(pb);
187 a.type = get_le32(pb);
191 dprintf(c->fc, "type: %08x %.4s sz: %"PRIx64" %"PRIx64" %"PRIx64"\n",
192 a.type, (char*)&a.type, a.size, atom.size, total_size);
193 if (a.size == 1) { /* 64 bit extended size */
194 a.size = get_be64(pb) - 8;
199 a.size = atom.size - total_size;
206 a.size = FFMIN(a.size, atom.size - total_size);
208 for (i = 0; mov_default_parse_table[i].type != 0
209 && mov_default_parse_table[i].type != a.type; i++)
212 if (mov_default_parse_table[i].type == 0) { /* skip leaf atoms data */
213 url_fskip(pb, a.size);
215 int64_t start_pos = url_ftell(pb);
217 err = mov_default_parse_table[i].parse(c, pb, a);
218 if (url_is_streamed(pb) && c->found_moov && c->found_mdat)
220 left = a.size - url_ftell(pb) + start_pos;
221 if (left > 0) /* skip garbage at atom end */
226 total_size += a.size;
229 if (!err && total_size < atom.size && atom.size < 0x7ffff)
230 url_fskip(pb, atom.size - total_size);
235 static int mov_read_dref(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
237 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
238 MOVStreamContext *sc = st->priv_data;
241 get_be32(pb); // version + flags
242 entries = get_be32(pb);
243 if (entries >= UINT_MAX / sizeof(*sc->drefs))
245 sc->drefs_count = entries;
246 sc->drefs = av_mallocz(entries * sizeof(*sc->drefs));
248 for (i = 0; i < sc->drefs_count; i++) {
249 MOVDref *dref = &sc->drefs[i];
250 uint32_t size = get_be32(pb);
251 int64_t next = url_ftell(pb) + size - 4;
253 dref->type = get_le32(pb);
254 get_be32(pb); // version + flags
255 dprintf(c->fc, "type %.4s size %d\n", (char*)&dref->type, size);
257 if (dref->type == MKTAG('a','l','i','s') && size > 150) {
258 /* macintosh alias record */
259 uint16_t volume_len, len;
265 volume_len = get_byte(pb);
266 volume_len = FFMIN(volume_len, 27);
267 get_buffer(pb, volume, 27);
268 volume[volume_len] = 0;
269 av_log(c->fc, AV_LOG_DEBUG, "volume %s, len %d\n", volume, volume_len);
273 for (type = 0; type != -1 && url_ftell(pb) < next; ) {
276 av_log(c->fc, AV_LOG_DEBUG, "type %d, len %d\n", type, len);
279 if (type == 2) { // absolute path
281 dref->path = av_mallocz(len+1);
283 return AVERROR(ENOMEM);
284 get_buffer(pb, dref->path, len);
285 if (len > volume_len && !strncmp(dref->path, volume, volume_len)) {
287 memmove(dref->path, dref->path+volume_len, len);
290 for (j = 0; j < len; j++)
291 if (dref->path[j] == ':')
293 av_log(c->fc, AV_LOG_DEBUG, "path %s\n", dref->path);
298 url_fseek(pb, next, SEEK_SET);
303 static int mov_read_hdlr(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
305 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
309 get_byte(pb); /* version */
310 get_be24(pb); /* flags */
313 ctype = get_le32(pb);
314 type = get_le32(pb); /* component subtype */
316 dprintf(c->fc, "ctype= %c%c%c%c (0x%08x)\n", *((char *)&ctype), ((char *)&ctype)[1],
317 ((char *)&ctype)[2], ((char *)&ctype)[3], (int) ctype);
318 dprintf(c->fc, "stype= %c%c%c%c\n",
319 *((char *)&type), ((char *)&type)[1], ((char *)&type)[2], ((char *)&type)[3]);
322 if (type == MKTAG('v','i','d','e'))
323 st->codec->codec_type = CODEC_TYPE_VIDEO;
324 else if(type == MKTAG('s','o','u','n'))
325 st->codec->codec_type = CODEC_TYPE_AUDIO;
326 else if(type == MKTAG('m','1','a',' '))
327 st->codec->codec_id = CODEC_ID_MP2;
328 else if(type == MKTAG('s','u','b','p')) {
329 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
331 get_be32(pb); /* component manufacture */
332 get_be32(pb); /* component flags */
333 get_be32(pb); /* component flags mask */
336 return 0; /* nothing left to read */
338 url_fskip(pb, atom.size - (url_ftell(pb) - atom.offset));
342 static int mp4_read_descr_len(ByteIOContext *pb)
347 int c = get_byte(pb);
348 len = (len << 7) | (c & 0x7f);
355 static int mp4_read_descr(MOVContext *c, ByteIOContext *pb, int *tag)
359 len = mp4_read_descr_len(pb);
360 dprintf(c->fc, "MPEG4 description: tag=0x%02x len=%d\n", *tag, len);
364 #define MP4ESDescrTag 0x03
365 #define MP4DecConfigDescrTag 0x04
366 #define MP4DecSpecificDescrTag 0x05
368 static const AVCodecTag mp4_audio_types[] = {
369 { CODEC_ID_MP3ON4, 29 }, /* old mp3on4 draft */
370 { CODEC_ID_MP3ON4, 32 }, /* layer 1 */
371 { CODEC_ID_MP3ON4, 33 }, /* layer 2 */
372 { CODEC_ID_MP3ON4, 34 }, /* layer 3 */
373 { CODEC_ID_NONE, 0 },
376 static int mov_read_esds(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
378 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
381 get_be32(pb); /* version + flags */
382 len = mp4_read_descr(c, pb, &tag);
383 if (tag == MP4ESDescrTag) {
384 get_be16(pb); /* ID */
385 get_byte(pb); /* priority */
387 get_be16(pb); /* ID */
389 len = mp4_read_descr(c, pb, &tag);
390 if (tag == MP4DecConfigDescrTag) {
391 int object_type_id = get_byte(pb);
392 get_byte(pb); /* stream type */
393 get_be24(pb); /* buffer size db */
394 get_be32(pb); /* max bitrate */
395 get_be32(pb); /* avg bitrate */
397 st->codec->codec_id= codec_get_id(ff_mp4_obj_type, object_type_id);
398 dprintf(c->fc, "esds object type id %d\n", object_type_id);
399 len = mp4_read_descr(c, pb, &tag);
400 if (tag == MP4DecSpecificDescrTag) {
401 dprintf(c->fc, "Specific MPEG4 header len=%d\n", len);
402 if((uint64_t)len > (1<<30))
404 st->codec->extradata = av_mallocz(len + FF_INPUT_BUFFER_PADDING_SIZE);
405 if (!st->codec->extradata)
406 return AVERROR(ENOMEM);
407 get_buffer(pb, st->codec->extradata, len);
408 st->codec->extradata_size = len;
409 if (st->codec->codec_id == CODEC_ID_AAC) {
410 MPEG4AudioConfig cfg;
411 ff_mpeg4audio_get_config(&cfg, st->codec->extradata,
412 st->codec->extradata_size);
413 if (cfg.chan_config > 7)
415 st->codec->channels = ff_mpeg4audio_channels[cfg.chan_config];
416 if (cfg.object_type == 29 && cfg.sampling_index < 3) // old mp3on4
417 st->codec->sample_rate = ff_mpa_freq_tab[cfg.sampling_index];
419 st->codec->sample_rate = cfg.sample_rate; // ext sample rate ?
420 dprintf(c->fc, "mp4a config channels %d obj %d ext obj %d "
421 "sample rate %d ext sample rate %d\n", st->codec->channels,
422 cfg.object_type, cfg.ext_object_type,
423 cfg.sample_rate, cfg.ext_sample_rate);
424 if (!(st->codec->codec_id = codec_get_id(mp4_audio_types,
426 st->codec->codec_id = CODEC_ID_AAC;
433 /* this atom contains actual media data */
434 static int mov_read_mdat(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
436 if(atom.size == 0) /* wrong one (MP4) */
439 return 0; /* now go for moov */
442 static int mov_read_ftyp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
444 uint32_t type = get_le32(pb);
446 if (type != MKTAG('q','t',' ',' '))
448 av_log(c->fc, AV_LOG_DEBUG, "ISO: File Type Major Brand: %.4s\n",(char *)&type);
449 get_be32(pb); /* minor version */
450 url_fskip(pb, atom.size - 8);
454 /* this atom should contain all header atoms */
455 static int mov_read_moov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
457 if (mov_read_default(c, pb, atom) < 0)
459 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
460 /* so we don't parse the whole file if over a network */
462 return 0; /* now go for mdat */
465 static int mov_read_moof(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
467 c->fragment.moof_offset = url_ftell(pb) - 8;
468 dprintf(c->fc, "moof offset %llx\n", c->fragment.moof_offset);
469 return mov_read_default(c, pb, atom);
472 static int mov_read_mdhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
474 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
475 MOVStreamContext *sc = st->priv_data;
476 int version = get_byte(pb);
480 return -1; /* unsupported */
482 get_be24(pb); /* flags */
487 get_be32(pb); /* creation time */
488 get_be32(pb); /* modification time */
491 sc->time_scale = get_be32(pb);
492 st->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
494 lang = get_be16(pb); /* language */
495 ff_mov_lang_to_iso639(lang, st->language);
496 get_be16(pb); /* quality */
501 static int mov_read_mvhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
503 int version = get_byte(pb); /* version */
504 get_be24(pb); /* flags */
510 get_be32(pb); /* creation time */
511 get_be32(pb); /* modification time */
513 c->time_scale = get_be32(pb); /* time scale */
515 dprintf(c->fc, "time scale = %i\n", c->time_scale);
517 c->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
518 get_be32(pb); /* preferred scale */
520 get_be16(pb); /* preferred volume */
522 url_fskip(pb, 10); /* reserved */
524 url_fskip(pb, 36); /* display matrix */
526 get_be32(pb); /* preview time */
527 get_be32(pb); /* preview duration */
528 get_be32(pb); /* poster time */
529 get_be32(pb); /* selection time */
530 get_be32(pb); /* selection duration */
531 get_be32(pb); /* current time */
532 get_be32(pb); /* next track ID */
537 static int mov_read_smi(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
539 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
541 if((uint64_t)atom.size > (1<<30))
544 // currently SVQ3 decoder expect full STSD header - so let's fake it
545 // this should be fixed and just SMI header should be passed
546 av_free(st->codec->extradata);
547 st->codec->extradata = av_mallocz(atom.size + 0x5a + FF_INPUT_BUFFER_PADDING_SIZE);
548 if (!st->codec->extradata)
549 return AVERROR(ENOMEM);
550 st->codec->extradata_size = 0x5a + atom.size;
551 memcpy(st->codec->extradata, "SVQ3", 4); // fake
552 get_buffer(pb, st->codec->extradata + 0x5a, atom.size);
553 dprintf(c->fc, "Reading SMI %"PRId64" %s\n", atom.size, st->codec->extradata + 0x5a);
557 static int mov_read_enda(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
559 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
560 int little_endian = get_be16(pb);
563 switch (st->codec->codec_id) {
564 case CODEC_ID_PCM_S24BE:
565 st->codec->codec_id = CODEC_ID_PCM_S24LE;
567 case CODEC_ID_PCM_S32BE:
568 st->codec->codec_id = CODEC_ID_PCM_S32LE;
570 case CODEC_ID_PCM_F32BE:
571 st->codec->codec_id = CODEC_ID_PCM_F32LE;
573 case CODEC_ID_PCM_F64BE:
574 st->codec->codec_id = CODEC_ID_PCM_F64LE;
583 /* FIXME modify qdm2/svq3/h264 decoders to take full atom as extradata */
584 static int mov_read_extradata(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
586 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
587 uint64_t size= (uint64_t)st->codec->extradata_size + atom.size + 8 + FF_INPUT_BUFFER_PADDING_SIZE;
589 if(size > INT_MAX || (uint64_t)atom.size > INT_MAX)
591 buf= av_realloc(st->codec->extradata, size);
594 st->codec->extradata= buf;
595 buf+= st->codec->extradata_size;
596 st->codec->extradata_size= size - FF_INPUT_BUFFER_PADDING_SIZE;
597 AV_WB32( buf , atom.size + 8);
598 AV_WL32( buf + 4, atom.type);
599 get_buffer(pb, buf + 8, atom.size);
603 static int mov_read_wave(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
605 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
607 if((uint64_t)atom.size > (1<<30))
610 if (st->codec->codec_id == CODEC_ID_QDM2) {
611 // pass all frma atom to codec, needed at least for QDM2
612 av_free(st->codec->extradata);
613 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
614 if (!st->codec->extradata)
615 return AVERROR(ENOMEM);
616 st->codec->extradata_size = atom.size;
617 get_buffer(pb, st->codec->extradata, atom.size);
618 } else if (atom.size > 8) { /* to read frma, esds atoms */
619 if (mov_read_default(c, pb, atom) < 0)
622 url_fskip(pb, atom.size);
627 * This function reads atom content and puts data in extradata without tag
628 * nor size unlike mov_read_extradata.
630 static int mov_read_glbl(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
632 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
634 if((uint64_t)atom.size > (1<<30))
637 av_free(st->codec->extradata);
638 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
639 if (!st->codec->extradata)
640 return AVERROR(ENOMEM);
641 st->codec->extradata_size = atom.size;
642 get_buffer(pb, st->codec->extradata, atom.size);
646 static int mov_read_stco(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
648 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
649 MOVStreamContext *sc = st->priv_data;
650 unsigned int i, entries;
652 get_byte(pb); /* version */
653 get_be24(pb); /* flags */
655 entries = get_be32(pb);
657 if(entries >= UINT_MAX/sizeof(int64_t))
660 sc->chunk_count = entries;
661 sc->chunk_offsets = av_malloc(entries * sizeof(int64_t));
662 if (!sc->chunk_offsets)
664 if (atom.type == MKTAG('s','t','c','o'))
665 for(i=0; i<entries; i++)
666 sc->chunk_offsets[i] = get_be32(pb);
667 else if (atom.type == MKTAG('c','o','6','4'))
668 for(i=0; i<entries; i++)
669 sc->chunk_offsets[i] = get_be64(pb);
677 * Compute codec id for 'lpcm' tag.
678 * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
680 static enum CodecID mov_get_lpcm_codec_id(int bps, int flags)
682 if (flags & 1) { // floating point
683 if (flags & 2) { // big endian
684 if (bps == 32) return CODEC_ID_PCM_F32BE;
685 else if (bps == 64) return CODEC_ID_PCM_F64BE;
687 if (bps == 32) return CODEC_ID_PCM_F32LE;
688 else if (bps == 64) return CODEC_ID_PCM_F64LE;
694 if (flags & 4) return CODEC_ID_PCM_S8;
695 else return CODEC_ID_PCM_U8;
696 else if (bps == 16) return CODEC_ID_PCM_S16BE;
697 else if (bps == 24) return CODEC_ID_PCM_S24BE;
698 else if (bps == 32) return CODEC_ID_PCM_S32BE;
701 if (flags & 4) return CODEC_ID_PCM_S8;
702 else return CODEC_ID_PCM_U8;
703 else if (bps == 16) return CODEC_ID_PCM_S16LE;
704 else if (bps == 24) return CODEC_ID_PCM_S24LE;
705 else if (bps == 32) return CODEC_ID_PCM_S32LE;
708 return CODEC_ID_NONE;
711 static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
713 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
714 MOVStreamContext *sc = st->priv_data;
715 int j, entries, pseudo_stream_id;
717 get_byte(pb); /* version */
718 get_be24(pb); /* flags */
720 entries = get_be32(pb);
722 for(pseudo_stream_id=0; pseudo_stream_id<entries; pseudo_stream_id++) {
723 //Parsing Sample description table
726 MOVAtom a = { 0, 0, 0 };
727 int64_t start_pos = url_ftell(pb);
728 int size = get_be32(pb); /* size */
729 uint32_t format = get_le32(pb); /* data format */
731 get_be32(pb); /* reserved */
732 get_be16(pb); /* reserved */
733 dref_id = get_be16(pb);
735 if (st->codec->codec_tag &&
736 st->codec->codec_tag != format &&
737 (c->fc->video_codec_id ? codec_get_id(codec_movvideo_tags, format) != c->fc->video_codec_id
738 : st->codec->codec_tag != MKTAG('j','p','e','g'))
740 /* Multiple fourcc, we skip JPEG. This is not correct, we should
741 * export it as a separate AVStream but this needs a few changes
742 * in the MOV demuxer, patch welcome. */
743 av_log(c->fc, AV_LOG_WARNING, "multiple fourcc not supported\n");
744 url_fskip(pb, size - (url_ftell(pb) - start_pos));
747 sc->pseudo_stream_id = st->codec->codec_tag ? -1 : pseudo_stream_id;
748 sc->dref_id= dref_id;
750 st->codec->codec_tag = format;
751 id = codec_get_id(codec_movaudio_tags, format);
752 if (id<=0 && (format&0xFFFF) == 'm'+('s'<<8))
753 id = codec_get_id(codec_wav_tags, bswap_32(format)&0xFFFF);
755 if (st->codec->codec_type != CODEC_TYPE_VIDEO && id > 0) {
756 st->codec->codec_type = CODEC_TYPE_AUDIO;
757 } else if (st->codec->codec_type != CODEC_TYPE_AUDIO && /* do not overwrite codec type */
758 format && format != MKTAG('m','p','4','s')) { /* skip old asf mpeg4 tag */
759 id = codec_get_id(codec_movvideo_tags, format);
761 id = codec_get_id(codec_bmp_tags, format);
763 st->codec->codec_type = CODEC_TYPE_VIDEO;
764 else if(st->codec->codec_type == CODEC_TYPE_DATA){
765 id = codec_get_id(ff_codec_movsubtitle_tags, format);
767 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
771 dprintf(c->fc, "size=%d 4CC= %c%c%c%c codec_type=%d\n", size,
772 (format >> 0) & 0xff, (format >> 8) & 0xff, (format >> 16) & 0xff,
773 (format >> 24) & 0xff, st->codec->codec_type);
775 if(st->codec->codec_type==CODEC_TYPE_VIDEO) {
776 uint8_t codec_name[32];
777 unsigned int color_depth;
780 st->codec->codec_id = id;
781 get_be16(pb); /* version */
782 get_be16(pb); /* revision level */
783 get_be32(pb); /* vendor */
784 get_be32(pb); /* temporal quality */
785 get_be32(pb); /* spatial quality */
787 st->codec->width = get_be16(pb); /* width */
788 st->codec->height = get_be16(pb); /* height */
790 get_be32(pb); /* horiz resolution */
791 get_be32(pb); /* vert resolution */
792 get_be32(pb); /* data size, always 0 */
793 get_be16(pb); /* frames per samples */
795 get_buffer(pb, codec_name, 32); /* codec name, pascal string */
796 if (codec_name[0] <= 31) {
797 memcpy(st->codec->codec_name, &codec_name[1],codec_name[0]);
798 st->codec->codec_name[codec_name[0]] = 0;
801 st->codec->bits_per_coded_sample = get_be16(pb); /* depth */
802 st->codec->color_table_id = get_be16(pb); /* colortable id */
803 dprintf(c->fc, "depth %d, ctab id %d\n",
804 st->codec->bits_per_coded_sample, st->codec->color_table_id);
805 /* figure out the palette situation */
806 color_depth = st->codec->bits_per_coded_sample & 0x1F;
807 color_greyscale = st->codec->bits_per_coded_sample & 0x20;
809 /* if the depth is 2, 4, or 8 bpp, file is palettized */
810 if ((color_depth == 2) || (color_depth == 4) ||
811 (color_depth == 8)) {
812 /* for palette traversal */
813 unsigned int color_start, color_count, color_end;
814 unsigned char r, g, b;
816 if (color_greyscale) {
817 int color_index, color_dec;
818 /* compute the greyscale palette */
819 st->codec->bits_per_coded_sample = color_depth;
820 color_count = 1 << color_depth;
822 color_dec = 256 / (color_count - 1);
823 for (j = 0; j < color_count; j++) {
824 r = g = b = color_index;
825 c->palette_control.palette[j] =
826 (r << 16) | (g << 8) | (b);
827 color_index -= color_dec;
831 } else if (st->codec->color_table_id) {
832 const uint8_t *color_table;
833 /* if flag bit 3 is set, use the default palette */
834 color_count = 1 << color_depth;
835 if (color_depth == 2)
836 color_table = ff_qt_default_palette_4;
837 else if (color_depth == 4)
838 color_table = ff_qt_default_palette_16;
840 color_table = ff_qt_default_palette_256;
842 for (j = 0; j < color_count; j++) {
843 r = color_table[j * 4 + 0];
844 g = color_table[j * 4 + 1];
845 b = color_table[j * 4 + 2];
846 c->palette_control.palette[j] =
847 (r << 16) | (g << 8) | (b);
850 /* load the palette from the file */
851 color_start = get_be32(pb);
852 color_count = get_be16(pb);
853 color_end = get_be16(pb);
854 if ((color_start <= 255) &&
855 (color_end <= 255)) {
856 for (j = color_start; j <= color_end; j++) {
857 /* each R, G, or B component is 16 bits;
858 * only use the top 8 bits; skip alpha bytes
868 c->palette_control.palette[j] =
869 (r << 16) | (g << 8) | (b);
873 st->codec->palctrl = &c->palette_control;
874 st->codec->palctrl->palette_changed = 1;
876 st->codec->palctrl = NULL;
877 } else if(st->codec->codec_type==CODEC_TYPE_AUDIO) {
878 int bits_per_sample, flags;
879 uint16_t version = get_be16(pb);
881 st->codec->codec_id = id;
882 get_be16(pb); /* revision level */
883 get_be32(pb); /* vendor */
885 st->codec->channels = get_be16(pb); /* channel count */
886 dprintf(c->fc, "audio channels %d\n", st->codec->channels);
887 st->codec->bits_per_coded_sample = get_be16(pb); /* sample size */
889 sc->audio_cid = get_be16(pb);
890 get_be16(pb); /* packet size = 0 */
892 st->codec->sample_rate = ((get_be32(pb) >> 16));
894 //Read QT version 1 fields. In version 0 these do not exist.
895 dprintf(c->fc, "version =%d, isom =%d\n",version,c->isom);
898 sc->samples_per_frame = get_be32(pb);
899 get_be32(pb); /* bytes per packet */
900 sc->bytes_per_frame = get_be32(pb);
901 get_be32(pb); /* bytes per sample */
902 } else if(version==2) {
903 get_be32(pb); /* sizeof struct only */
904 st->codec->sample_rate = av_int2dbl(get_be64(pb)); /* float 64 */
905 st->codec->channels = get_be32(pb);
906 get_be32(pb); /* always 0x7F000000 */
907 st->codec->bits_per_coded_sample = get_be32(pb); /* bits per channel if sound is uncompressed */
908 flags = get_be32(pb); /* lcpm format specific flag */
909 sc->bytes_per_frame = get_be32(pb); /* bytes per audio packet if constant */
910 sc->samples_per_frame = get_be32(pb); /* lpcm frames per audio packet if constant */
911 if (format == MKTAG('l','p','c','m'))
912 st->codec->codec_id = mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, flags);
916 switch (st->codec->codec_id) {
917 case CODEC_ID_PCM_S8:
918 case CODEC_ID_PCM_U8:
919 if (st->codec->bits_per_coded_sample == 16)
920 st->codec->codec_id = CODEC_ID_PCM_S16BE;
922 case CODEC_ID_PCM_S16LE:
923 case CODEC_ID_PCM_S16BE:
924 if (st->codec->bits_per_coded_sample == 8)
925 st->codec->codec_id = CODEC_ID_PCM_S8;
926 else if (st->codec->bits_per_coded_sample == 24)
927 st->codec->codec_id =
928 st->codec->codec_id == CODEC_ID_PCM_S16BE ?
929 CODEC_ID_PCM_S24BE : CODEC_ID_PCM_S24LE;
931 /* set values for old format before stsd version 1 appeared */
933 sc->samples_per_frame = 6;
934 sc->bytes_per_frame = 2*st->codec->channels;
937 sc->samples_per_frame = 6;
938 sc->bytes_per_frame = 1*st->codec->channels;
940 case CODEC_ID_ADPCM_IMA_QT:
941 sc->samples_per_frame = 64;
942 sc->bytes_per_frame = 34*st->codec->channels;
945 sc->samples_per_frame = 160;
946 sc->bytes_per_frame = 33;
952 bits_per_sample = av_get_bits_per_sample(st->codec->codec_id);
953 if (bits_per_sample) {
954 st->codec->bits_per_coded_sample = bits_per_sample;
955 sc->sample_size = (bits_per_sample >> 3) * st->codec->channels;
957 } else if(st->codec->codec_type==CODEC_TYPE_SUBTITLE){
958 st->codec->codec_id= id;
960 /* other codec type, just skip (rtp, mp4s, tmcd ...) */
961 url_fskip(pb, size - (url_ftell(pb) - start_pos));
963 /* this will read extra atoms at the end (wave, alac, damr, avcC, SMI ...) */
964 a.size = size - (url_ftell(pb) - start_pos);
966 if (mov_read_default(c, pb, a) < 0)
968 } else if (a.size > 0)
969 url_fskip(pb, a.size);
972 if(st->codec->codec_type==CODEC_TYPE_AUDIO && st->codec->sample_rate==0 && sc->time_scale>1)
973 st->codec->sample_rate= sc->time_scale;
975 /* special codec parameters handling */
976 switch (st->codec->codec_id) {
977 #ifdef CONFIG_DV_DEMUXER
978 case CODEC_ID_DVAUDIO:
979 c->dv_fctx = av_alloc_format_context();
980 c->dv_demux = dv_init_demux(c->dv_fctx);
982 av_log(c->fc, AV_LOG_ERROR, "dv demux context init error\n");
985 sc->dv_audio_container = 1;
986 st->codec->codec_id = CODEC_ID_PCM_S16LE;
989 /* no ifdef since parameters are always those */
991 st->codec->frame_size= 160;
992 st->codec->channels= 1; /* really needed */
994 case CODEC_ID_AMR_NB:
995 case CODEC_ID_AMR_WB:
996 st->codec->frame_size= sc->samples_per_frame;
997 st->codec->channels= 1; /* really needed */
998 /* force sample rate for amr, stsd in 3gp does not store sample rate */
999 if (st->codec->codec_id == CODEC_ID_AMR_NB)
1000 st->codec->sample_rate = 8000;
1001 else if (st->codec->codec_id == CODEC_ID_AMR_WB)
1002 st->codec->sample_rate = 16000;
1006 st->codec->codec_type = CODEC_TYPE_AUDIO; /* force type after stsd for m1a hdlr */
1007 st->need_parsing = AVSTREAM_PARSE_FULL;
1010 case CODEC_ID_ADPCM_MS:
1011 case CODEC_ID_ADPCM_IMA_WAV:
1012 st->codec->block_align = sc->bytes_per_frame;
1015 if (st->codec->extradata_size == 36)
1016 st->codec->frame_size = AV_RB32((st->codec->extradata+12));
1025 static int mov_read_stsc(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1027 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1028 MOVStreamContext *sc = st->priv_data;
1029 unsigned int i, entries;
1031 get_byte(pb); /* version */
1032 get_be24(pb); /* flags */
1034 entries = get_be32(pb);
1036 if(entries >= UINT_MAX / sizeof(*sc->sample_to_chunk))
1039 dprintf(c->fc, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
1041 sc->sample_to_chunk_sz = entries;
1042 sc->sample_to_chunk = av_malloc(entries * sizeof(*sc->sample_to_chunk));
1043 if (!sc->sample_to_chunk)
1045 for(i=0; i<entries; i++) {
1046 sc->sample_to_chunk[i].first = get_be32(pb);
1047 sc->sample_to_chunk[i].count = get_be32(pb);
1048 sc->sample_to_chunk[i].id = get_be32(pb);
1053 static int mov_read_stss(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1055 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1056 MOVStreamContext *sc = st->priv_data;
1057 unsigned int i, entries;
1059 get_byte(pb); /* version */
1060 get_be24(pb); /* flags */
1062 entries = get_be32(pb);
1064 if(entries >= UINT_MAX / sizeof(int))
1067 sc->keyframe_count = entries;
1069 dprintf(c->fc, "keyframe_count = %d\n", sc->keyframe_count);
1071 sc->keyframes = av_malloc(entries * sizeof(int));
1074 for(i=0; i<entries; i++) {
1075 sc->keyframes[i] = get_be32(pb);
1076 //dprintf(c->fc, "keyframes[]=%d\n", sc->keyframes[i]);
1081 static int mov_read_stsz(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1083 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1084 MOVStreamContext *sc = st->priv_data;
1085 unsigned int i, entries, sample_size;
1087 get_byte(pb); /* version */
1088 get_be24(pb); /* flags */
1090 sample_size = get_be32(pb);
1091 if (!sc->sample_size) /* do not overwrite value computed in stsd */
1092 sc->sample_size = sample_size;
1093 entries = get_be32(pb);
1094 if(entries >= UINT_MAX / sizeof(int))
1097 sc->sample_count = entries;
1101 dprintf(c->fc, "sample_size = %d sample_count = %d\n", sc->sample_size, sc->sample_count);
1103 sc->sample_sizes = av_malloc(entries * sizeof(int));
1104 if (!sc->sample_sizes)
1106 for(i=0; i<entries; i++)
1107 sc->sample_sizes[i] = get_be32(pb);
1111 static int mov_read_stts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1113 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1114 MOVStreamContext *sc = st->priv_data;
1115 unsigned int i, entries;
1117 int64_t total_sample_count=0;
1119 get_byte(pb); /* version */
1120 get_be24(pb); /* flags */
1121 entries = get_be32(pb);
1122 if(entries >= UINT_MAX / sizeof(*sc->stts_data))
1125 sc->stts_count = entries;
1126 sc->stts_data = av_malloc(entries * sizeof(*sc->stts_data));
1129 dprintf(c->fc, "track[%i].stts.entries = %i\n", c->fc->nb_streams-1, entries);
1133 for(i=0; i<entries; i++) {
1134 int sample_duration;
1137 sample_count=get_be32(pb);
1138 sample_duration = get_be32(pb);
1139 sc->stts_data[i].count= sample_count;
1140 sc->stts_data[i].duration= sample_duration;
1142 sc->time_rate= ff_gcd(sc->time_rate, sample_duration);
1144 dprintf(c->fc, "sample_count=%d, sample_duration=%d\n",sample_count,sample_duration);
1146 duration+=(int64_t)sample_duration*sample_count;
1147 total_sample_count+=sample_count;
1150 st->nb_frames= total_sample_count;
1152 st->duration= duration;
1156 static int mov_read_ctts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1158 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1159 MOVStreamContext *sc = st->priv_data;
1160 unsigned int i, entries;
1162 get_byte(pb); /* version */
1163 get_be24(pb); /* flags */
1164 entries = get_be32(pb);
1165 if(entries >= UINT_MAX / sizeof(*sc->ctts_data))
1168 sc->ctts_count = entries;
1169 sc->ctts_data = av_malloc(entries * sizeof(*sc->ctts_data));
1172 dprintf(c->fc, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
1174 for(i=0; i<entries; i++) {
1175 int count =get_be32(pb);
1176 int duration =get_be32(pb);
1180 st->codec->has_b_frames = 1;
1182 sc->ctts_data[i].count = count;
1183 sc->ctts_data[i].duration= duration;
1185 sc->time_rate= ff_gcd(sc->time_rate, FFABS(duration));
1190 static void mov_build_index(MOVContext *mov, AVStream *st)
1192 MOVStreamContext *sc = st->priv_data;
1193 int64_t current_offset;
1194 int64_t current_dts = 0;
1195 unsigned int stts_index = 0;
1196 unsigned int stsc_index = 0;
1197 unsigned int stss_index = 0;
1200 /* only use old uncompressed audio chunk demuxing when stts specifies it */
1201 if (!(st->codec->codec_type == CODEC_TYPE_AUDIO &&
1202 sc->stts_count == 1 && sc->stts_data[0].duration == 1)) {
1203 unsigned int current_sample = 0;
1204 unsigned int stts_sample = 0;
1205 unsigned int keyframe, sample_size;
1206 unsigned int distance = 0;
1207 int key_off = sc->keyframes && sc->keyframes[0] == 1;
1209 st->nb_frames = sc->sample_count;
1210 for (i = 0; i < sc->chunk_count; i++) {
1211 current_offset = sc->chunk_offsets[i];
1212 if (stsc_index + 1 < sc->sample_to_chunk_sz &&
1213 i + 1 == sc->sample_to_chunk[stsc_index + 1].first)
1215 for (j = 0; j < sc->sample_to_chunk[stsc_index].count; j++) {
1216 if (current_sample >= sc->sample_count) {
1217 av_log(mov->fc, AV_LOG_ERROR, "wrong sample count\n");
1220 keyframe = !sc->keyframe_count || current_sample+key_off == sc->keyframes[stss_index];
1223 if (stss_index + 1 < sc->keyframe_count)
1226 sample_size = sc->sample_size > 0 ? sc->sample_size : sc->sample_sizes[current_sample];
1227 if(sc->pseudo_stream_id == -1 ||
1228 sc->sample_to_chunk[stsc_index].id - 1 == sc->pseudo_stream_id) {
1229 av_add_index_entry(st, current_offset, current_dts, sample_size, distance,
1230 keyframe ? AVINDEX_KEYFRAME : 0);
1231 dprintf(mov->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1232 "size %d, distance %d, keyframe %d\n", st->index, current_sample,
1233 current_offset, current_dts, sample_size, distance, keyframe);
1235 current_offset += sample_size;
1236 assert(sc->stts_data[stts_index].duration % sc->time_rate == 0);
1237 current_dts += sc->stts_data[stts_index].duration / sc->time_rate;
1241 if (stts_index + 1 < sc->stts_count && stts_sample == sc->stts_data[stts_index].count) {
1247 } else { /* read whole chunk */
1248 unsigned int chunk_samples, chunk_size, chunk_duration;
1249 unsigned int frames = 1;
1250 for (i = 0; i < sc->chunk_count; i++) {
1251 current_offset = sc->chunk_offsets[i];
1252 if (stsc_index + 1 < sc->sample_to_chunk_sz &&
1253 i + 1 == sc->sample_to_chunk[stsc_index + 1].first)
1255 chunk_samples = sc->sample_to_chunk[stsc_index].count;
1256 /* get chunk size, beware of alaw/ulaw/mace */
1257 if (sc->samples_per_frame > 0 &&
1258 (chunk_samples * sc->bytes_per_frame % sc->samples_per_frame == 0)) {
1259 if (sc->samples_per_frame < 160)
1260 chunk_size = chunk_samples * sc->bytes_per_frame / sc->samples_per_frame;
1262 chunk_size = sc->bytes_per_frame;
1263 frames = chunk_samples / sc->samples_per_frame;
1264 chunk_samples = sc->samples_per_frame;
1267 chunk_size = chunk_samples * sc->sample_size;
1268 for (j = 0; j < frames; j++) {
1269 av_add_index_entry(st, current_offset, current_dts, chunk_size, 0, AVINDEX_KEYFRAME);
1270 /* get chunk duration */
1272 while (chunk_samples > 0) {
1273 if (chunk_samples < sc->stts_data[stts_index].count) {
1274 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1275 sc->stts_data[stts_index].count -= chunk_samples;
1278 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1279 chunk_samples -= sc->stts_data[stts_index].count;
1280 if (stts_index + 1 < sc->stts_count)
1284 current_offset += sc->bytes_per_frame;
1285 dprintf(mov->fc, "AVIndex stream %d, chunk %d, offset %"PRIx64", dts %"PRId64", "
1286 "size %d, duration %d\n", st->index, i, current_offset, current_dts,
1287 chunk_size, chunk_duration);
1288 assert(chunk_duration % sc->time_rate == 0);
1289 current_dts += chunk_duration / sc->time_rate;
1294 /* adjust sample count to avindex entries */
1295 sc->sample_count = st->nb_index_entries;
1298 static int mov_read_trak(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1301 MOVStreamContext *sc;
1304 st = av_new_stream(c->fc, c->fc->nb_streams);
1305 if (!st) return AVERROR(ENOMEM);
1306 sc = av_mallocz(sizeof(MOVStreamContext));
1307 if (!sc) return AVERROR(ENOMEM);
1310 st->codec->codec_type = CODEC_TYPE_DATA;
1311 st->start_time = 0; /* XXX: check */
1313 if ((ret = mov_read_default(c, pb, atom)) < 0)
1317 if(sc->chunk_count && (!sc->stts_count || !sc->sample_to_chunk_sz ||
1318 (!sc->sample_size && !sc->sample_count))){
1319 av_log(c->fc, AV_LOG_ERROR, "stream %d, missing mandatory atoms, broken header\n",
1321 sc->sample_count = 0; //ignore track
1327 sc->time_scale= c->time_scale;
1328 av_set_pts_info(st, 64, sc->time_rate, sc->time_scale);
1330 if (st->codec->codec_type == CODEC_TYPE_AUDIO &&
1331 !st->codec->frame_size && sc->stts_count == 1)
1332 st->codec->frame_size = av_rescale(sc->time_rate, st->codec->sample_rate, sc->time_scale);
1334 if(st->duration != AV_NOPTS_VALUE){
1335 assert(st->duration % sc->time_rate == 0);
1336 st->duration /= sc->time_rate;
1338 sc->ffindex = st->index;
1339 mov_build_index(c, st);
1341 if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
1342 if (url_fopen(&sc->pb, sc->drefs[sc->dref_id-1].path, URL_RDONLY) < 0)
1343 av_log(c->fc, AV_LOG_ERROR, "stream %d, error opening file %s: %s\n",
1344 st->index, sc->drefs[sc->dref_id-1].path, strerror(errno));
1348 switch (st->codec->codec_id) {
1349 #ifdef CONFIG_H261_DECODER
1352 #ifdef CONFIG_H263_DECODER
1355 #ifdef CONFIG_MPEG4_DECODER
1356 case CODEC_ID_MPEG4:
1358 st->codec->width= 0; /* let decoder init width/height */
1359 st->codec->height= 0;
1363 /* Do not need those anymore. */
1364 av_freep(&sc->chunk_offsets);
1365 av_freep(&sc->sample_to_chunk);
1366 av_freep(&sc->sample_sizes);
1367 av_freep(&sc->keyframes);
1368 av_freep(&sc->stts_data);
1373 static int mov_read_ilst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1376 c->itunes_metadata = 1;
1377 ret = mov_read_default(c, pb, atom);
1378 c->itunes_metadata = 0;
1382 static int mov_read_meta(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1384 url_fskip(pb, 4); // version + flags
1386 return mov_read_default(c, pb, atom);
1389 static int mov_read_trkn(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1391 get_be32(pb); // type
1392 get_be32(pb); // unknown
1393 c->fc->track = get_be32(pb);
1394 dprintf(c->fc, "%.4s %d\n", (char*)&atom.type, c->fc->track);
1398 static int mov_read_udta_string(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1404 if (c->itunes_metadata) {
1405 int data_size = get_be32(pb);
1406 int tag = get_le32(pb);
1407 if (tag == MKTAG('d','a','t','a')) {
1408 get_be32(pb); // type
1409 get_be32(pb); // unknown
1410 str_size = data_size - 16;
1413 str_size = get_be16(pb); // string length
1414 get_be16(pb); // language
1416 switch (atom.type) {
1417 case MKTAG(0xa9,'n','a','m'):
1418 str = c->fc->title; size = sizeof(c->fc->title); break;
1419 case MKTAG(0xa9,'A','R','T'):
1420 case MKTAG(0xa9,'w','r','t'):
1421 str = c->fc->author; size = sizeof(c->fc->author); break;
1422 case MKTAG(0xa9,'c','p','y'):
1423 str = c->fc->copyright; size = sizeof(c->fc->copyright); break;
1424 case MKTAG(0xa9,'c','m','t'):
1425 case MKTAG(0xa9,'i','n','f'):
1426 str = c->fc->comment; size = sizeof(c->fc->comment); break;
1427 case MKTAG(0xa9,'a','l','b'):
1428 str = c->fc->album; size = sizeof(c->fc->album); break;
1432 get_buffer(pb, str, FFMIN(size, str_size));
1433 dprintf(c->fc, "%.4s %s\n", (char*)&atom.type, str);
1437 static int mov_read_tkhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1442 int64_t disp_transform[2];
1443 int display_matrix[3][2];
1444 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1445 int version = get_byte(pb);
1447 get_be24(pb); /* flags */
1449 MOV_TRACK_ENABLED 0x0001
1450 MOV_TRACK_IN_MOVIE 0x0002
1451 MOV_TRACK_IN_PREVIEW 0x0004
1452 MOV_TRACK_IN_POSTER 0x0008
1459 get_be32(pb); /* creation time */
1460 get_be32(pb); /* modification time */
1462 st->id = (int)get_be32(pb); /* track id (NOT 0 !)*/
1463 get_be32(pb); /* reserved */
1464 st->start_time = 0; /* check */
1465 /* highlevel (considering edits) duration in movie timebase */
1466 (version == 1) ? get_be64(pb) : get_be32(pb);
1467 get_be32(pb); /* reserved */
1468 get_be32(pb); /* reserved */
1470 get_be16(pb); /* layer */
1471 get_be16(pb); /* alternate group */
1472 get_be16(pb); /* volume */
1473 get_be16(pb); /* reserved */
1475 //read in the display matrix (outlined in ISO 14496-12, Section 6.2.2)
1476 // they're kept in fixed point format through all calculations
1477 // ignore u,v,z b/c we don't need the scale factor to calc aspect ratio
1478 for (i = 0; i < 3; i++) {
1479 display_matrix[i][0] = get_be32(pb); // 16.16 fixed point
1480 display_matrix[i][1] = get_be32(pb); // 16.16 fixed point
1481 get_be32(pb); // 2.30 fixed point (not used)
1484 width = get_be32(pb); // 16.16 fixed point track width
1485 height = get_be32(pb); // 16.16 fixed point track height
1487 //transform the display width/height according to the matrix
1488 // skip this if the display matrix is the default identity matrix
1489 // to keep the same scale, use [width height 1<<16]
1490 if (width && height &&
1491 (display_matrix[0][0] != 65536 || display_matrix[0][1] ||
1492 display_matrix[1][0] || display_matrix[1][1] != 65536 ||
1493 display_matrix[2][0] || display_matrix[2][1])) {
1494 for (i = 0; i < 2; i++)
1496 (int64_t) width * display_matrix[0][i] +
1497 (int64_t) height * display_matrix[1][i] +
1498 ((int64_t) display_matrix[2][i] << 16);
1500 //sample aspect ratio is new width/height divided by old width/height
1501 st->sample_aspect_ratio = av_d2q(
1502 ((double) disp_transform[0] * height) /
1503 ((double) disp_transform[1] * width), INT_MAX);
1508 static int mov_read_tfhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1510 MOVFragment *frag = &c->fragment;
1511 MOVTrackExt *trex = NULL;
1512 int flags, track_id, i;
1514 get_byte(pb); /* version */
1515 flags = get_be24(pb);
1517 track_id = get_be32(pb);
1518 if (!track_id || track_id > c->fc->nb_streams)
1520 frag->track_id = track_id;
1521 for (i = 0; i < c->trex_count; i++)
1522 if (c->trex_data[i].track_id == frag->track_id) {
1523 trex = &c->trex_data[i];
1527 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding trex\n");
1531 if (flags & 0x01) frag->base_data_offset = get_be64(pb);
1532 else frag->base_data_offset = frag->moof_offset;
1533 if (flags & 0x02) frag->stsd_id = get_be32(pb);
1534 else frag->stsd_id = trex->stsd_id;
1536 frag->duration = flags & 0x08 ? get_be32(pb) : trex->duration;
1537 frag->size = flags & 0x10 ? get_be32(pb) : trex->size;
1538 frag->flags = flags & 0x20 ? get_be32(pb) : trex->flags;
1539 dprintf(c->fc, "frag flags 0x%x\n", frag->flags);
1543 static int mov_read_trex(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1547 if ((uint64_t)c->trex_count+1 >= UINT_MAX / sizeof(*c->trex_data))
1549 c->trex_data = av_realloc(c->trex_data, (c->trex_count+1)*sizeof(*c->trex_data));
1551 return AVERROR(ENOMEM);
1552 trex = &c->trex_data[c->trex_count++];
1553 get_byte(pb); /* version */
1554 get_be24(pb); /* flags */
1555 trex->track_id = get_be32(pb);
1556 trex->stsd_id = get_be32(pb);
1557 trex->duration = get_be32(pb);
1558 trex->size = get_be32(pb);
1559 trex->flags = get_be32(pb);
1563 static int mov_read_trun(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1565 MOVFragment *frag = &c->fragment;
1567 MOVStreamContext *sc;
1570 int data_offset = 0;
1571 unsigned entries, first_sample_flags = frag->flags;
1572 int flags, distance, i;
1574 if (!frag->track_id || frag->track_id > c->fc->nb_streams)
1576 st = c->fc->streams[frag->track_id-1];
1578 if (sc->pseudo_stream_id+1 != frag->stsd_id)
1580 get_byte(pb); /* version */
1581 flags = get_be24(pb);
1582 entries = get_be32(pb);
1583 dprintf(c->fc, "flags 0x%x entries %d\n", flags, entries);
1584 if (flags & 0x001) data_offset = get_be32(pb);
1585 if (flags & 0x004) first_sample_flags = get_be32(pb);
1586 if (flags & 0x800) {
1587 if ((uint64_t)entries+sc->ctts_count >= UINT_MAX/sizeof(*sc->ctts_data))
1589 sc->ctts_data = av_realloc(sc->ctts_data,
1590 (entries+sc->ctts_count)*sizeof(*sc->ctts_data));
1592 return AVERROR(ENOMEM);
1595 offset = frag->base_data_offset + data_offset;
1597 dprintf(c->fc, "first sample flags 0x%x\n", first_sample_flags);
1598 for (i = 0; i < entries; i++) {
1599 unsigned sample_size = frag->size;
1600 int sample_flags = i ? frag->flags : first_sample_flags;
1601 unsigned sample_duration = frag->duration;
1604 if (flags & 0x100) sample_duration = get_be32(pb);
1605 if (flags & 0x200) sample_size = get_be32(pb);
1606 if (flags & 0x400) sample_flags = get_be32(pb);
1607 if (flags & 0x800) {
1608 sc->ctts_data[sc->ctts_count].count = 1;
1609 sc->ctts_data[sc->ctts_count].duration = get_be32(pb);
1612 if ((keyframe = st->codec->codec_type == CODEC_TYPE_AUDIO ||
1613 (flags & 0x004 && !i && !sample_flags) || sample_flags & 0x2000000))
1615 av_add_index_entry(st, offset, dts, sample_size, distance,
1616 keyframe ? AVINDEX_KEYFRAME : 0);
1617 dprintf(c->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1618 "size %d, distance %d, keyframe %d\n", st->index, sc->sample_count+i,
1619 offset, dts, sample_size, distance, keyframe);
1621 assert(sample_duration % sc->time_rate == 0);
1622 dts += sample_duration / sc->time_rate;
1623 offset += sample_size;
1625 frag->moof_offset = offset;
1626 sc->sample_count = st->nb_index_entries;
1631 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
1632 /* like the files created with Adobe Premiere 5.0, for samples see */
1633 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
1634 static int mov_read_wide(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1639 return 0; /* continue */
1640 if (get_be32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
1641 url_fskip(pb, atom.size - 4);
1644 atom.type = get_le32(pb);
1647 if (atom.type != MKTAG('m','d','a','t')) {
1648 url_fskip(pb, atom.size);
1651 err = mov_read_mdat(c, pb, atom);
1655 static int mov_read_cmov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1660 uint8_t *moov_data; /* uncompressed data */
1661 long cmov_len, moov_len;
1664 get_be32(pb); /* dcom atom */
1665 if (get_le32(pb) != MKTAG('d','c','o','m'))
1667 if (get_le32(pb) != MKTAG('z','l','i','b')) {
1668 av_log(NULL, AV_LOG_ERROR, "unknown compression for cmov atom !");
1671 get_be32(pb); /* cmvd atom */
1672 if (get_le32(pb) != MKTAG('c','m','v','d'))
1674 moov_len = get_be32(pb); /* uncompressed size */
1675 cmov_len = atom.size - 6 * 4;
1677 cmov_data = av_malloc(cmov_len);
1680 moov_data = av_malloc(moov_len);
1685 get_buffer(pb, cmov_data, cmov_len);
1686 if(uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
1687 goto free_and_return;
1688 if(init_put_byte(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL, NULL) != 0)
1689 goto free_and_return;
1690 atom.type = MKTAG('m','o','o','v');
1692 atom.size = moov_len;
1694 // { int fd = open("/tmp/uncompheader.mov", O_WRONLY | O_CREAT); write(fd, moov_data, moov_len); close(fd); }
1696 ret = mov_read_default(c, &ctx, atom);
1702 av_log(c->fc, AV_LOG_ERROR, "this file requires zlib support compiled in\n");
1707 /* edit list atom */
1708 static int mov_read_elst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1710 MOVStreamContext *sc = c->fc->streams[c->fc->nb_streams-1]->priv_data;
1713 get_byte(pb); /* version */
1714 get_be24(pb); /* flags */
1715 edit_count= sc->edit_count = get_be32(pb); /* entries */
1717 for(i=0; i<edit_count; i++){
1719 get_be32(pb); /* Track duration */
1720 time = get_be32(pb); /* Media time */
1721 get_be32(pb); /* Media rate */
1723 av_log(c->fc, AV_LOG_WARNING, "edit list not starting at 0, "
1724 "a/v desync might occur, patch welcome\n");
1726 dprintf(c->fc, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, sc->edit_count);
1730 static const MOVParseTableEntry mov_default_parse_table[] = {
1731 { MKTAG('a','v','s','s'), mov_read_extradata },
1732 { MKTAG('c','o','6','4'), mov_read_stco },
1733 { MKTAG('c','t','t','s'), mov_read_ctts }, /* composition time to sample */
1734 { MKTAG('d','i','n','f'), mov_read_default },
1735 { MKTAG('d','r','e','f'), mov_read_dref },
1736 { MKTAG('e','d','t','s'), mov_read_default },
1737 { MKTAG('e','l','s','t'), mov_read_elst },
1738 { MKTAG('e','n','d','a'), mov_read_enda },
1739 { MKTAG('f','i','e','l'), mov_read_extradata },
1740 { MKTAG('f','t','y','p'), mov_read_ftyp },
1741 { MKTAG('g','l','b','l'), mov_read_glbl },
1742 { MKTAG('h','d','l','r'), mov_read_hdlr },
1743 { MKTAG('i','l','s','t'), mov_read_ilst },
1744 { MKTAG('j','p','2','h'), mov_read_extradata },
1745 { MKTAG('m','d','a','t'), mov_read_mdat },
1746 { MKTAG('m','d','h','d'), mov_read_mdhd },
1747 { MKTAG('m','d','i','a'), mov_read_default },
1748 { MKTAG('m','e','t','a'), mov_read_meta },
1749 { MKTAG('m','i','n','f'), mov_read_default },
1750 { MKTAG('m','o','o','f'), mov_read_moof },
1751 { MKTAG('m','o','o','v'), mov_read_moov },
1752 { MKTAG('m','v','e','x'), mov_read_default },
1753 { MKTAG('m','v','h','d'), mov_read_mvhd },
1754 { MKTAG('S','M','I',' '), mov_read_smi }, /* Sorenson extension ??? */
1755 { MKTAG('a','l','a','c'), mov_read_extradata }, /* alac specific atom */
1756 { MKTAG('a','v','c','C'), mov_read_glbl },
1757 { MKTAG('s','t','b','l'), mov_read_default },
1758 { MKTAG('s','t','c','o'), mov_read_stco },
1759 { MKTAG('s','t','s','c'), mov_read_stsc },
1760 { MKTAG('s','t','s','d'), mov_read_stsd }, /* sample description */
1761 { MKTAG('s','t','s','s'), mov_read_stss }, /* sync sample */
1762 { MKTAG('s','t','s','z'), mov_read_stsz }, /* sample size */
1763 { MKTAG('s','t','t','s'), mov_read_stts },
1764 { MKTAG('t','k','h','d'), mov_read_tkhd }, /* track header */
1765 { MKTAG('t','f','h','d'), mov_read_tfhd }, /* track fragment header */
1766 { MKTAG('t','r','a','k'), mov_read_trak },
1767 { MKTAG('t','r','a','f'), mov_read_default },
1768 { MKTAG('t','r','e','x'), mov_read_trex },
1769 { MKTAG('t','r','k','n'), mov_read_trkn },
1770 { MKTAG('t','r','u','n'), mov_read_trun },
1771 { MKTAG('u','d','t','a'), mov_read_default },
1772 { MKTAG('w','a','v','e'), mov_read_wave },
1773 { MKTAG('e','s','d','s'), mov_read_esds },
1774 { MKTAG('w','i','d','e'), mov_read_wide }, /* place holder */
1775 { MKTAG('c','m','o','v'), mov_read_cmov },
1776 { MKTAG(0xa9,'n','a','m'), mov_read_udta_string },
1777 { MKTAG(0xa9,'w','r','t'), mov_read_udta_string },
1778 { MKTAG(0xa9,'c','p','y'), mov_read_udta_string },
1779 { MKTAG(0xa9,'i','n','f'), mov_read_udta_string },
1780 { MKTAG(0xa9,'i','n','f'), mov_read_udta_string },
1781 { MKTAG(0xa9,'A','R','T'), mov_read_udta_string },
1782 { MKTAG(0xa9,'a','l','b'), mov_read_udta_string },
1783 { MKTAG(0xa9,'c','m','t'), mov_read_udta_string },
1787 static int mov_probe(AVProbeData *p)
1789 unsigned int offset;
1793 /* check file header */
1796 /* ignore invalid offset */
1797 if ((offset + 8) > (unsigned int)p->buf_size)
1799 tag = AV_RL32(p->buf + offset + 4);
1801 /* check for obvious tags */
1802 case MKTAG('j','P',' ',' '): /* jpeg 2000 signature */
1803 case MKTAG('m','o','o','v'):
1804 case MKTAG('m','d','a','t'):
1805 case MKTAG('p','n','o','t'): /* detect movs with preview pics like ew.mov and april.mov */
1806 case MKTAG('u','d','t','a'): /* Packet Video PVAuthor adds this and a lot of more junk */
1807 case MKTAG('f','t','y','p'):
1808 return AVPROBE_SCORE_MAX;
1809 /* those are more common words, so rate then a bit less */
1810 case MKTAG('e','d','i','w'): /* xdcam files have reverted first tags */
1811 case MKTAG('w','i','d','e'):
1812 case MKTAG('f','r','e','e'):
1813 case MKTAG('j','u','n','k'):
1814 case MKTAG('p','i','c','t'):
1815 return AVPROBE_SCORE_MAX - 5;
1816 case MKTAG(0x82,0x82,0x7f,0x7d):
1817 case MKTAG('s','k','i','p'):
1818 case MKTAG('u','u','i','d'):
1819 case MKTAG('p','r','f','l'):
1820 offset = AV_RB32(p->buf+offset) + offset;
1821 /* if we only find those cause probedata is too small at least rate them */
1822 score = AVPROBE_SCORE_MAX - 50;
1825 /* unrecognized tag */
1832 static int mov_read_header(AVFormatContext *s, AVFormatParameters *ap)
1834 MOVContext *mov = s->priv_data;
1835 ByteIOContext *pb = s->pb;
1837 MOVAtom atom = { 0, 0, 0 };
1840 /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
1841 if(!url_is_streamed(pb))
1842 atom.size = url_fsize(pb);
1844 atom.size = INT64_MAX;
1846 /* check MOV header */
1847 if ((err = mov_read_default(mov, pb, atom)) < 0) {
1848 av_log(s, AV_LOG_ERROR, "error reading header: %d\n", err);
1851 if (!mov->found_moov) {
1852 av_log(s, AV_LOG_ERROR, "moov atom not found\n");
1855 dprintf(mov->fc, "on_parse_exit_offset=%lld\n", url_ftell(pb));
1860 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
1862 MOVContext *mov = s->priv_data;
1863 MOVStreamContext *sc = 0;
1864 AVIndexEntry *sample = 0;
1865 int64_t best_dts = INT64_MAX;
1868 for (i = 0; i < s->nb_streams; i++) {
1869 AVStream *st = s->streams[i];
1870 MOVStreamContext *msc = st->priv_data;
1871 if (st->discard != AVDISCARD_ALL && msc->pb && msc->current_sample < msc->sample_count) {
1872 AVIndexEntry *current_sample = &st->index_entries[msc->current_sample];
1873 int64_t dts = av_rescale(current_sample->timestamp * (int64_t)msc->time_rate,
1874 AV_TIME_BASE, msc->time_scale);
1875 dprintf(s, "stream %d, sample %d, dts %"PRId64"\n", i, msc->current_sample, dts);
1876 if (!sample || (url_is_streamed(s->pb) && current_sample->pos < sample->pos) ||
1877 (!url_is_streamed(s->pb) &&
1878 ((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
1879 ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
1880 (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
1881 sample = current_sample;
1888 mov->found_mdat = 0;
1889 if (!url_is_streamed(s->pb) ||
1890 mov_read_default(mov, s->pb, (MOVAtom){ 0, 0, INT64_MAX }) < 0 ||
1893 dprintf(s, "read fragments, offset 0x%llx\n", url_ftell(s->pb));
1896 /* must be done just before reading, to avoid infinite loop on sample */
1897 sc->current_sample++;
1898 if (url_fseek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
1899 av_log(mov->fc, AV_LOG_ERROR, "stream %d, offset 0x%"PRIx64": partial file\n",
1900 sc->ffindex, sample->pos);
1903 av_get_packet(sc->pb, pkt, sample->size);
1904 #ifdef CONFIG_DV_DEMUXER
1905 if (mov->dv_demux && sc->dv_audio_container) {
1906 dv_produce_packet(mov->dv_demux, pkt, pkt->data, pkt->size);
1909 if (dv_get_packet(mov->dv_demux, pkt) < 0)
1913 pkt->stream_index = sc->ffindex;
1914 pkt->dts = sample->timestamp;
1915 if (sc->ctts_data) {
1916 assert(sc->ctts_data[sc->sample_to_ctime_index].duration % sc->time_rate == 0);
1917 pkt->pts = pkt->dts + sc->ctts_data[sc->sample_to_ctime_index].duration / sc->time_rate;
1918 /* update ctts context */
1919 sc->sample_to_ctime_sample++;
1920 if (sc->sample_to_ctime_index < sc->ctts_count &&
1921 sc->ctts_data[sc->sample_to_ctime_index].count == sc->sample_to_ctime_sample) {
1922 sc->sample_to_ctime_index++;
1923 sc->sample_to_ctime_sample = 0;
1926 pkt->dts = AV_NOPTS_VALUE;
1928 AVStream *st = s->streams[sc->ffindex];
1929 int64_t next_dts = (sc->current_sample < sc->sample_count) ?
1930 st->index_entries[sc->current_sample].timestamp : st->duration;
1931 pkt->duration = next_dts - pkt->dts;
1932 pkt->pts = pkt->dts;
1934 pkt->flags |= sample->flags & AVINDEX_KEYFRAME ? PKT_FLAG_KEY : 0;
1935 pkt->pos = sample->pos;
1936 dprintf(s, "stream %d, pts %"PRId64", dts %"PRId64", pos 0x%"PRIx64", duration %d\n",
1937 pkt->stream_index, pkt->pts, pkt->dts, pkt->pos, pkt->duration);
1941 static int mov_seek_stream(AVStream *st, int64_t timestamp, int flags)
1943 MOVStreamContext *sc = st->priv_data;
1944 int sample, time_sample;
1947 sample = av_index_search_timestamp(st, timestamp, flags);
1948 dprintf(st->codec, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
1949 if (sample < 0) /* not sure what to do */
1951 sc->current_sample = sample;
1952 dprintf(st->codec, "stream %d, found sample %d\n", st->index, sc->current_sample);
1953 /* adjust ctts index */
1954 if (sc->ctts_data) {
1956 for (i = 0; i < sc->ctts_count; i++) {
1957 int next = time_sample + sc->ctts_data[i].count;
1958 if (next > sc->current_sample) {
1959 sc->sample_to_ctime_index = i;
1960 sc->sample_to_ctime_sample = sc->current_sample - time_sample;
1969 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
1972 int64_t seek_timestamp, timestamp;
1976 if (stream_index >= s->nb_streams)
1979 st = s->streams[stream_index];
1980 sample = mov_seek_stream(st, sample_time, flags);
1984 /* adjust seek timestamp to found sample timestamp */
1985 seek_timestamp = st->index_entries[sample].timestamp;
1987 for (i = 0; i < s->nb_streams; i++) {
1989 if (stream_index == i || st->discard == AVDISCARD_ALL)
1992 timestamp = av_rescale_q(seek_timestamp, s->streams[stream_index]->time_base, st->time_base);
1993 mov_seek_stream(st, timestamp, flags);
1998 static int mov_read_close(AVFormatContext *s)
2001 MOVContext *mov = s->priv_data;
2002 for(i=0; i<s->nb_streams; i++) {
2003 MOVStreamContext *sc = s->streams[i]->priv_data;
2004 av_freep(&sc->ctts_data);
2005 for (j=0; j<sc->drefs_count; j++)
2006 av_freep(&sc->drefs[j].path);
2007 av_freep(&sc->drefs);
2008 if (sc->pb && sc->pb != s->pb)
2012 for(i=0; i<mov->dv_fctx->nb_streams; i++){
2013 av_freep(&mov->dv_fctx->streams[i]->codec);
2014 av_freep(&mov->dv_fctx->streams[i]);
2016 av_freep(&mov->dv_fctx);
2017 av_freep(&mov->dv_demux);
2019 av_freep(&mov->trex_data);
2023 AVInputFormat mov_demuxer = {
2024 "mov,mp4,m4a,3gp,3g2,mj2",
2025 NULL_IF_CONFIG_SMALL("QuickTime/MPEG-4/Motion JPEG 2000 format"),