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 static int mov_read_pasp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
435 const int num = get_be32(pb);
436 const int den = get_be32(pb);
437 AVStream * const st = c->fc->streams[c->fc->nb_streams-1];
439 if ((st->sample_aspect_ratio.den && den != st->sample_aspect_ratio.den) ||
440 (st->sample_aspect_ratio.num && num != st->sample_aspect_ratio.num))
441 av_log(c->fc, AV_LOG_WARNING,
442 "sample aspect ratio already set, overriding by 'pasp' atom\n");
443 st->sample_aspect_ratio.num = num;
444 st->sample_aspect_ratio.den = den;
449 /* this atom contains actual media data */
450 static int mov_read_mdat(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
452 if(atom.size == 0) /* wrong one (MP4) */
455 return 0; /* now go for moov */
458 static int mov_read_ftyp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
460 uint32_t type = get_le32(pb);
462 if (type != MKTAG('q','t',' ',' '))
464 av_log(c->fc, AV_LOG_DEBUG, "ISO: File Type Major Brand: %.4s\n",(char *)&type);
465 get_be32(pb); /* minor version */
466 url_fskip(pb, atom.size - 8);
470 /* this atom should contain all header atoms */
471 static int mov_read_moov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
473 if (mov_read_default(c, pb, atom) < 0)
475 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
476 /* so we don't parse the whole file if over a network */
478 return 0; /* now go for mdat */
481 static int mov_read_moof(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
483 c->fragment.moof_offset = url_ftell(pb) - 8;
484 dprintf(c->fc, "moof offset %llx\n", c->fragment.moof_offset);
485 return mov_read_default(c, pb, atom);
488 static int mov_read_mdhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
490 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
491 MOVStreamContext *sc = st->priv_data;
492 int version = get_byte(pb);
496 return -1; /* unsupported */
498 get_be24(pb); /* flags */
503 get_be32(pb); /* creation time */
504 get_be32(pb); /* modification time */
507 sc->time_scale = get_be32(pb);
508 st->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
510 lang = get_be16(pb); /* language */
511 ff_mov_lang_to_iso639(lang, st->language);
512 get_be16(pb); /* quality */
517 static int mov_read_mvhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
519 int version = get_byte(pb); /* version */
520 get_be24(pb); /* flags */
526 get_be32(pb); /* creation time */
527 get_be32(pb); /* modification time */
529 c->time_scale = get_be32(pb); /* time scale */
531 dprintf(c->fc, "time scale = %i\n", c->time_scale);
533 c->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
534 get_be32(pb); /* preferred scale */
536 get_be16(pb); /* preferred volume */
538 url_fskip(pb, 10); /* reserved */
540 url_fskip(pb, 36); /* display matrix */
542 get_be32(pb); /* preview time */
543 get_be32(pb); /* preview duration */
544 get_be32(pb); /* poster time */
545 get_be32(pb); /* selection time */
546 get_be32(pb); /* selection duration */
547 get_be32(pb); /* current time */
548 get_be32(pb); /* next track ID */
553 static int mov_read_smi(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
555 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
557 if((uint64_t)atom.size > (1<<30))
560 // currently SVQ3 decoder expect full STSD header - so let's fake it
561 // this should be fixed and just SMI header should be passed
562 av_free(st->codec->extradata);
563 st->codec->extradata = av_mallocz(atom.size + 0x5a + FF_INPUT_BUFFER_PADDING_SIZE);
564 if (!st->codec->extradata)
565 return AVERROR(ENOMEM);
566 st->codec->extradata_size = 0x5a + atom.size;
567 memcpy(st->codec->extradata, "SVQ3", 4); // fake
568 get_buffer(pb, st->codec->extradata + 0x5a, atom.size);
569 dprintf(c->fc, "Reading SMI %"PRId64" %s\n", atom.size, st->codec->extradata + 0x5a);
573 static int mov_read_enda(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
575 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
576 int little_endian = get_be16(pb);
579 switch (st->codec->codec_id) {
580 case CODEC_ID_PCM_S24BE:
581 st->codec->codec_id = CODEC_ID_PCM_S24LE;
583 case CODEC_ID_PCM_S32BE:
584 st->codec->codec_id = CODEC_ID_PCM_S32LE;
586 case CODEC_ID_PCM_F32BE:
587 st->codec->codec_id = CODEC_ID_PCM_F32LE;
589 case CODEC_ID_PCM_F64BE:
590 st->codec->codec_id = CODEC_ID_PCM_F64LE;
599 /* FIXME modify qdm2/svq3/h264 decoders to take full atom as extradata */
600 static int mov_read_extradata(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
606 if (c->fc->nb_streams < 1) // will happen with jp2 files
608 st= c->fc->streams[c->fc->nb_streams-1];
609 size= (uint64_t)st->codec->extradata_size + atom.size + 8 + FF_INPUT_BUFFER_PADDING_SIZE;
610 if(size > INT_MAX || (uint64_t)atom.size > INT_MAX)
612 buf= av_realloc(st->codec->extradata, size);
615 st->codec->extradata= buf;
616 buf+= st->codec->extradata_size;
617 st->codec->extradata_size= size - FF_INPUT_BUFFER_PADDING_SIZE;
618 AV_WB32( buf , atom.size + 8);
619 AV_WL32( buf + 4, atom.type);
620 get_buffer(pb, buf + 8, atom.size);
624 static int mov_read_wave(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
626 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
628 if((uint64_t)atom.size > (1<<30))
631 if (st->codec->codec_id == CODEC_ID_QDM2) {
632 // pass all frma atom to codec, needed at least for QDM2
633 av_free(st->codec->extradata);
634 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
635 if (!st->codec->extradata)
636 return AVERROR(ENOMEM);
637 st->codec->extradata_size = atom.size;
638 get_buffer(pb, st->codec->extradata, atom.size);
639 } else if (atom.size > 8) { /* to read frma, esds atoms */
640 if (mov_read_default(c, pb, atom) < 0)
643 url_fskip(pb, atom.size);
648 * This function reads atom content and puts data in extradata without tag
649 * nor size unlike mov_read_extradata.
651 static int mov_read_glbl(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
653 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
655 if((uint64_t)atom.size > (1<<30))
658 av_free(st->codec->extradata);
659 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
660 if (!st->codec->extradata)
661 return AVERROR(ENOMEM);
662 st->codec->extradata_size = atom.size;
663 get_buffer(pb, st->codec->extradata, atom.size);
667 static int mov_read_stco(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
669 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
670 MOVStreamContext *sc = st->priv_data;
671 unsigned int i, entries;
673 get_byte(pb); /* version */
674 get_be24(pb); /* flags */
676 entries = get_be32(pb);
678 if(entries >= UINT_MAX/sizeof(int64_t))
681 sc->chunk_count = entries;
682 sc->chunk_offsets = av_malloc(entries * sizeof(int64_t));
683 if (!sc->chunk_offsets)
685 if (atom.type == MKTAG('s','t','c','o'))
686 for(i=0; i<entries; i++)
687 sc->chunk_offsets[i] = get_be32(pb);
688 else if (atom.type == MKTAG('c','o','6','4'))
689 for(i=0; i<entries; i++)
690 sc->chunk_offsets[i] = get_be64(pb);
698 * Compute codec id for 'lpcm' tag.
699 * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
701 static enum CodecID mov_get_lpcm_codec_id(int bps, int flags)
703 if (flags & 1) { // floating point
704 if (flags & 2) { // big endian
705 if (bps == 32) return CODEC_ID_PCM_F32BE;
706 else if (bps == 64) return CODEC_ID_PCM_F64BE;
708 if (bps == 32) return CODEC_ID_PCM_F32LE;
709 else if (bps == 64) return CODEC_ID_PCM_F64LE;
715 if (flags & 4) return CODEC_ID_PCM_S8;
716 else return CODEC_ID_PCM_U8;
717 else if (bps == 16) return CODEC_ID_PCM_S16BE;
718 else if (bps == 24) return CODEC_ID_PCM_S24BE;
719 else if (bps == 32) return CODEC_ID_PCM_S32BE;
722 if (flags & 4) return CODEC_ID_PCM_S8;
723 else return CODEC_ID_PCM_U8;
724 else if (bps == 16) return CODEC_ID_PCM_S16LE;
725 else if (bps == 24) return CODEC_ID_PCM_S24LE;
726 else if (bps == 32) return CODEC_ID_PCM_S32LE;
729 return CODEC_ID_NONE;
732 static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
734 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
735 MOVStreamContext *sc = st->priv_data;
736 int j, entries, pseudo_stream_id;
738 get_byte(pb); /* version */
739 get_be24(pb); /* flags */
741 entries = get_be32(pb);
743 for(pseudo_stream_id=0; pseudo_stream_id<entries; pseudo_stream_id++) {
744 //Parsing Sample description table
747 MOVAtom a = { 0, 0, 0 };
748 int64_t start_pos = url_ftell(pb);
749 int size = get_be32(pb); /* size */
750 uint32_t format = get_le32(pb); /* data format */
752 get_be32(pb); /* reserved */
753 get_be16(pb); /* reserved */
754 dref_id = get_be16(pb);
756 if (st->codec->codec_tag &&
757 st->codec->codec_tag != format &&
758 (c->fc->video_codec_id ? codec_get_id(codec_movvideo_tags, format) != c->fc->video_codec_id
759 : st->codec->codec_tag != MKTAG('j','p','e','g'))
761 /* Multiple fourcc, we skip JPEG. This is not correct, we should
762 * export it as a separate AVStream but this needs a few changes
763 * in the MOV demuxer, patch welcome. */
764 av_log(c->fc, AV_LOG_WARNING, "multiple fourcc not supported\n");
765 url_fskip(pb, size - (url_ftell(pb) - start_pos));
768 sc->pseudo_stream_id = st->codec->codec_tag ? -1 : pseudo_stream_id;
769 sc->dref_id= dref_id;
771 st->codec->codec_tag = format;
772 id = codec_get_id(codec_movaudio_tags, format);
773 if (id<=0 && (format&0xFFFF) == 'm'+('s'<<8))
774 id = codec_get_id(codec_wav_tags, bswap_32(format)&0xFFFF);
776 if (st->codec->codec_type != CODEC_TYPE_VIDEO && id > 0) {
777 st->codec->codec_type = CODEC_TYPE_AUDIO;
778 } else if (st->codec->codec_type != CODEC_TYPE_AUDIO && /* do not overwrite codec type */
779 format && format != MKTAG('m','p','4','s')) { /* skip old asf mpeg4 tag */
780 id = codec_get_id(codec_movvideo_tags, format);
782 id = codec_get_id(codec_bmp_tags, format);
784 st->codec->codec_type = CODEC_TYPE_VIDEO;
785 else if(st->codec->codec_type == CODEC_TYPE_DATA){
786 id = codec_get_id(ff_codec_movsubtitle_tags, format);
788 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
792 dprintf(c->fc, "size=%d 4CC= %c%c%c%c codec_type=%d\n", size,
793 (format >> 0) & 0xff, (format >> 8) & 0xff, (format >> 16) & 0xff,
794 (format >> 24) & 0xff, st->codec->codec_type);
796 if(st->codec->codec_type==CODEC_TYPE_VIDEO) {
797 uint8_t codec_name[32];
798 unsigned int color_depth;
801 st->codec->codec_id = id;
802 get_be16(pb); /* version */
803 get_be16(pb); /* revision level */
804 get_be32(pb); /* vendor */
805 get_be32(pb); /* temporal quality */
806 get_be32(pb); /* spatial quality */
808 st->codec->width = get_be16(pb); /* width */
809 st->codec->height = get_be16(pb); /* height */
811 get_be32(pb); /* horiz resolution */
812 get_be32(pb); /* vert resolution */
813 get_be32(pb); /* data size, always 0 */
814 get_be16(pb); /* frames per samples */
816 get_buffer(pb, codec_name, 32); /* codec name, pascal string */
817 if (codec_name[0] <= 31) {
818 memcpy(st->codec->codec_name, &codec_name[1],codec_name[0]);
819 st->codec->codec_name[codec_name[0]] = 0;
822 st->codec->bits_per_coded_sample = get_be16(pb); /* depth */
823 st->codec->color_table_id = get_be16(pb); /* colortable id */
824 dprintf(c->fc, "depth %d, ctab id %d\n",
825 st->codec->bits_per_coded_sample, st->codec->color_table_id);
826 /* figure out the palette situation */
827 color_depth = st->codec->bits_per_coded_sample & 0x1F;
828 color_greyscale = st->codec->bits_per_coded_sample & 0x20;
830 /* if the depth is 2, 4, or 8 bpp, file is palettized */
831 if ((color_depth == 2) || (color_depth == 4) ||
832 (color_depth == 8)) {
833 /* for palette traversal */
834 unsigned int color_start, color_count, color_end;
835 unsigned char r, g, b;
837 if (color_greyscale) {
838 int color_index, color_dec;
839 /* compute the greyscale palette */
840 st->codec->bits_per_coded_sample = color_depth;
841 color_count = 1 << color_depth;
843 color_dec = 256 / (color_count - 1);
844 for (j = 0; j < color_count; j++) {
845 r = g = b = color_index;
846 c->palette_control.palette[j] =
847 (r << 16) | (g << 8) | (b);
848 color_index -= color_dec;
852 } else if (st->codec->color_table_id) {
853 const uint8_t *color_table;
854 /* if flag bit 3 is set, use the default palette */
855 color_count = 1 << color_depth;
856 if (color_depth == 2)
857 color_table = ff_qt_default_palette_4;
858 else if (color_depth == 4)
859 color_table = ff_qt_default_palette_16;
861 color_table = ff_qt_default_palette_256;
863 for (j = 0; j < color_count; j++) {
864 r = color_table[j * 4 + 0];
865 g = color_table[j * 4 + 1];
866 b = color_table[j * 4 + 2];
867 c->palette_control.palette[j] =
868 (r << 16) | (g << 8) | (b);
871 /* load the palette from the file */
872 color_start = get_be32(pb);
873 color_count = get_be16(pb);
874 color_end = get_be16(pb);
875 if ((color_start <= 255) &&
876 (color_end <= 255)) {
877 for (j = color_start; j <= color_end; j++) {
878 /* each R, G, or B component is 16 bits;
879 * only use the top 8 bits; skip alpha bytes
889 c->palette_control.palette[j] =
890 (r << 16) | (g << 8) | (b);
894 st->codec->palctrl = &c->palette_control;
895 st->codec->palctrl->palette_changed = 1;
897 st->codec->palctrl = NULL;
898 } else if(st->codec->codec_type==CODEC_TYPE_AUDIO) {
899 int bits_per_sample, flags;
900 uint16_t version = get_be16(pb);
902 st->codec->codec_id = id;
903 get_be16(pb); /* revision level */
904 get_be32(pb); /* vendor */
906 st->codec->channels = get_be16(pb); /* channel count */
907 dprintf(c->fc, "audio channels %d\n", st->codec->channels);
908 st->codec->bits_per_coded_sample = get_be16(pb); /* sample size */
910 sc->audio_cid = get_be16(pb);
911 get_be16(pb); /* packet size = 0 */
913 st->codec->sample_rate = ((get_be32(pb) >> 16));
915 //Read QT version 1 fields. In version 0 these do not exist.
916 dprintf(c->fc, "version =%d, isom =%d\n",version,c->isom);
919 sc->samples_per_frame = get_be32(pb);
920 get_be32(pb); /* bytes per packet */
921 sc->bytes_per_frame = get_be32(pb);
922 get_be32(pb); /* bytes per sample */
923 } else if(version==2) {
924 get_be32(pb); /* sizeof struct only */
925 st->codec->sample_rate = av_int2dbl(get_be64(pb)); /* float 64 */
926 st->codec->channels = get_be32(pb);
927 get_be32(pb); /* always 0x7F000000 */
928 st->codec->bits_per_coded_sample = get_be32(pb); /* bits per channel if sound is uncompressed */
929 flags = get_be32(pb); /* lcpm format specific flag */
930 sc->bytes_per_frame = get_be32(pb); /* bytes per audio packet if constant */
931 sc->samples_per_frame = get_be32(pb); /* lpcm frames per audio packet if constant */
932 if (format == MKTAG('l','p','c','m'))
933 st->codec->codec_id = mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, flags);
937 switch (st->codec->codec_id) {
938 case CODEC_ID_PCM_S8:
939 case CODEC_ID_PCM_U8:
940 if (st->codec->bits_per_coded_sample == 16)
941 st->codec->codec_id = CODEC_ID_PCM_S16BE;
943 case CODEC_ID_PCM_S16LE:
944 case CODEC_ID_PCM_S16BE:
945 if (st->codec->bits_per_coded_sample == 8)
946 st->codec->codec_id = CODEC_ID_PCM_S8;
947 else if (st->codec->bits_per_coded_sample == 24)
948 st->codec->codec_id =
949 st->codec->codec_id == CODEC_ID_PCM_S16BE ?
950 CODEC_ID_PCM_S24BE : CODEC_ID_PCM_S24LE;
952 /* set values for old format before stsd version 1 appeared */
954 sc->samples_per_frame = 6;
955 sc->bytes_per_frame = 2*st->codec->channels;
958 sc->samples_per_frame = 6;
959 sc->bytes_per_frame = 1*st->codec->channels;
961 case CODEC_ID_ADPCM_IMA_QT:
962 sc->samples_per_frame = 64;
963 sc->bytes_per_frame = 34*st->codec->channels;
966 sc->samples_per_frame = 160;
967 sc->bytes_per_frame = 33;
973 bits_per_sample = av_get_bits_per_sample(st->codec->codec_id);
974 if (bits_per_sample) {
975 st->codec->bits_per_coded_sample = bits_per_sample;
976 sc->sample_size = (bits_per_sample >> 3) * st->codec->channels;
978 } else if(st->codec->codec_type==CODEC_TYPE_SUBTITLE){
979 st->codec->codec_id= id;
981 /* other codec type, just skip (rtp, mp4s, tmcd ...) */
982 url_fskip(pb, size - (url_ftell(pb) - start_pos));
984 /* this will read extra atoms at the end (wave, alac, damr, avcC, SMI ...) */
985 a.size = size - (url_ftell(pb) - start_pos);
987 if (mov_read_default(c, pb, a) < 0)
989 } else if (a.size > 0)
990 url_fskip(pb, a.size);
993 if(st->codec->codec_type==CODEC_TYPE_AUDIO && st->codec->sample_rate==0 && sc->time_scale>1)
994 st->codec->sample_rate= sc->time_scale;
996 /* special codec parameters handling */
997 switch (st->codec->codec_id) {
998 #ifdef CONFIG_DV_DEMUXER
999 case CODEC_ID_DVAUDIO:
1000 c->dv_fctx = av_alloc_format_context();
1001 c->dv_demux = dv_init_demux(c->dv_fctx);
1003 av_log(c->fc, AV_LOG_ERROR, "dv demux context init error\n");
1006 sc->dv_audio_container = 1;
1007 st->codec->codec_id = CODEC_ID_PCM_S16LE;
1010 /* no ifdef since parameters are always those */
1011 case CODEC_ID_QCELP:
1012 st->codec->frame_size= 160;
1013 st->codec->channels= 1; /* really needed */
1015 case CODEC_ID_AMR_NB:
1016 case CODEC_ID_AMR_WB:
1017 st->codec->frame_size= sc->samples_per_frame;
1018 st->codec->channels= 1; /* really needed */
1019 /* force sample rate for amr, stsd in 3gp does not store sample rate */
1020 if (st->codec->codec_id == CODEC_ID_AMR_NB)
1021 st->codec->sample_rate = 8000;
1022 else if (st->codec->codec_id == CODEC_ID_AMR_WB)
1023 st->codec->sample_rate = 16000;
1027 st->codec->codec_type = CODEC_TYPE_AUDIO; /* force type after stsd for m1a hdlr */
1028 st->need_parsing = AVSTREAM_PARSE_FULL;
1031 case CODEC_ID_ADPCM_MS:
1032 case CODEC_ID_ADPCM_IMA_WAV:
1033 st->codec->block_align = sc->bytes_per_frame;
1036 if (st->codec->extradata_size == 36) {
1037 st->codec->frame_size = AV_RB32((st->codec->extradata+12));
1038 st->codec->channels = AV_RB8 (st->codec->extradata+21);
1048 static int mov_read_stsc(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1050 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1051 MOVStreamContext *sc = st->priv_data;
1052 unsigned int i, entries;
1054 get_byte(pb); /* version */
1055 get_be24(pb); /* flags */
1057 entries = get_be32(pb);
1059 if(entries >= UINT_MAX / sizeof(*sc->sample_to_chunk))
1062 dprintf(c->fc, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
1064 sc->sample_to_chunk_sz = entries;
1065 sc->sample_to_chunk = av_malloc(entries * sizeof(*sc->sample_to_chunk));
1066 if (!sc->sample_to_chunk)
1068 for(i=0; i<entries; i++) {
1069 sc->sample_to_chunk[i].first = get_be32(pb);
1070 sc->sample_to_chunk[i].count = get_be32(pb);
1071 sc->sample_to_chunk[i].id = get_be32(pb);
1076 static int mov_read_stss(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1078 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1079 MOVStreamContext *sc = st->priv_data;
1080 unsigned int i, entries;
1082 get_byte(pb); /* version */
1083 get_be24(pb); /* flags */
1085 entries = get_be32(pb);
1087 if(entries >= UINT_MAX / sizeof(int))
1090 sc->keyframe_count = entries;
1092 dprintf(c->fc, "keyframe_count = %d\n", sc->keyframe_count);
1094 sc->keyframes = av_malloc(entries * sizeof(int));
1097 for(i=0; i<entries; i++) {
1098 sc->keyframes[i] = get_be32(pb);
1099 //dprintf(c->fc, "keyframes[]=%d\n", sc->keyframes[i]);
1104 static int mov_read_stsz(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1106 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1107 MOVStreamContext *sc = st->priv_data;
1108 unsigned int i, entries, sample_size;
1110 get_byte(pb); /* version */
1111 get_be24(pb); /* flags */
1113 sample_size = get_be32(pb);
1114 if (!sc->sample_size) /* do not overwrite value computed in stsd */
1115 sc->sample_size = sample_size;
1116 entries = get_be32(pb);
1117 if(entries >= UINT_MAX / sizeof(int))
1120 sc->sample_count = entries;
1124 dprintf(c->fc, "sample_size = %d sample_count = %d\n", sc->sample_size, sc->sample_count);
1126 sc->sample_sizes = av_malloc(entries * sizeof(int));
1127 if (!sc->sample_sizes)
1129 for(i=0; i<entries; i++)
1130 sc->sample_sizes[i] = get_be32(pb);
1134 static int mov_read_stts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1136 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1137 MOVStreamContext *sc = st->priv_data;
1138 unsigned int i, entries;
1140 int64_t total_sample_count=0;
1142 get_byte(pb); /* version */
1143 get_be24(pb); /* flags */
1144 entries = get_be32(pb);
1145 if(entries >= UINT_MAX / sizeof(*sc->stts_data))
1148 sc->stts_count = entries;
1149 sc->stts_data = av_malloc(entries * sizeof(*sc->stts_data));
1152 dprintf(c->fc, "track[%i].stts.entries = %i\n", c->fc->nb_streams-1, entries);
1156 for(i=0; i<entries; i++) {
1157 int sample_duration;
1160 sample_count=get_be32(pb);
1161 sample_duration = get_be32(pb);
1162 sc->stts_data[i].count= sample_count;
1163 sc->stts_data[i].duration= sample_duration;
1165 sc->time_rate= ff_gcd(sc->time_rate, sample_duration);
1167 dprintf(c->fc, "sample_count=%d, sample_duration=%d\n",sample_count,sample_duration);
1169 duration+=(int64_t)sample_duration*sample_count;
1170 total_sample_count+=sample_count;
1173 st->nb_frames= total_sample_count;
1175 st->duration= duration;
1179 static int mov_read_ctts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1181 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1182 MOVStreamContext *sc = st->priv_data;
1183 unsigned int i, entries;
1185 get_byte(pb); /* version */
1186 get_be24(pb); /* flags */
1187 entries = get_be32(pb);
1188 if(entries >= UINT_MAX / sizeof(*sc->ctts_data))
1191 sc->ctts_count = entries;
1192 sc->ctts_data = av_malloc(entries * sizeof(*sc->ctts_data));
1195 dprintf(c->fc, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
1197 for(i=0; i<entries; i++) {
1198 int count =get_be32(pb);
1199 int duration =get_be32(pb);
1203 st->codec->has_b_frames = 1;
1205 sc->ctts_data[i].count = count;
1206 sc->ctts_data[i].duration= duration;
1208 sc->time_rate= ff_gcd(sc->time_rate, FFABS(duration));
1213 static void mov_build_index(MOVContext *mov, AVStream *st)
1215 MOVStreamContext *sc = st->priv_data;
1216 int64_t current_offset;
1217 int64_t current_dts = 0;
1218 unsigned int stts_index = 0;
1219 unsigned int stsc_index = 0;
1220 unsigned int stss_index = 0;
1223 /* only use old uncompressed audio chunk demuxing when stts specifies it */
1224 if (!(st->codec->codec_type == CODEC_TYPE_AUDIO &&
1225 sc->stts_count == 1 && sc->stts_data[0].duration == 1)) {
1226 unsigned int current_sample = 0;
1227 unsigned int stts_sample = 0;
1228 unsigned int keyframe, sample_size;
1229 unsigned int distance = 0;
1230 int key_off = sc->keyframes && sc->keyframes[0] == 1;
1232 st->nb_frames = sc->sample_count;
1233 for (i = 0; i < sc->chunk_count; i++) {
1234 current_offset = sc->chunk_offsets[i];
1235 if (stsc_index + 1 < sc->sample_to_chunk_sz &&
1236 i + 1 == sc->sample_to_chunk[stsc_index + 1].first)
1238 for (j = 0; j < sc->sample_to_chunk[stsc_index].count; j++) {
1239 if (current_sample >= sc->sample_count) {
1240 av_log(mov->fc, AV_LOG_ERROR, "wrong sample count\n");
1243 keyframe = !sc->keyframe_count || current_sample+key_off == sc->keyframes[stss_index];
1246 if (stss_index + 1 < sc->keyframe_count)
1249 sample_size = sc->sample_size > 0 ? sc->sample_size : sc->sample_sizes[current_sample];
1250 if(sc->pseudo_stream_id == -1 ||
1251 sc->sample_to_chunk[stsc_index].id - 1 == sc->pseudo_stream_id) {
1252 av_add_index_entry(st, current_offset, current_dts, sample_size, distance,
1253 keyframe ? AVINDEX_KEYFRAME : 0);
1254 dprintf(mov->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1255 "size %d, distance %d, keyframe %d\n", st->index, current_sample,
1256 current_offset, current_dts, sample_size, distance, keyframe);
1258 current_offset += sample_size;
1259 assert(sc->stts_data[stts_index].duration % sc->time_rate == 0);
1260 current_dts += sc->stts_data[stts_index].duration / sc->time_rate;
1264 if (stts_index + 1 < sc->stts_count && stts_sample == sc->stts_data[stts_index].count) {
1270 } else { /* read whole chunk */
1271 unsigned int chunk_samples, chunk_size, chunk_duration;
1272 unsigned int frames = 1;
1273 for (i = 0; i < sc->chunk_count; i++) {
1274 current_offset = sc->chunk_offsets[i];
1275 if (stsc_index + 1 < sc->sample_to_chunk_sz &&
1276 i + 1 == sc->sample_to_chunk[stsc_index + 1].first)
1278 chunk_samples = sc->sample_to_chunk[stsc_index].count;
1279 /* get chunk size, beware of alaw/ulaw/mace */
1280 if (sc->samples_per_frame > 0 &&
1281 (chunk_samples * sc->bytes_per_frame % sc->samples_per_frame == 0)) {
1282 if (sc->samples_per_frame < 160)
1283 chunk_size = chunk_samples * sc->bytes_per_frame / sc->samples_per_frame;
1285 chunk_size = sc->bytes_per_frame;
1286 frames = chunk_samples / sc->samples_per_frame;
1287 chunk_samples = sc->samples_per_frame;
1290 chunk_size = chunk_samples * sc->sample_size;
1291 for (j = 0; j < frames; j++) {
1292 av_add_index_entry(st, current_offset, current_dts, chunk_size, 0, AVINDEX_KEYFRAME);
1293 /* get chunk duration */
1295 while (chunk_samples > 0) {
1296 if (chunk_samples < sc->stts_data[stts_index].count) {
1297 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1298 sc->stts_data[stts_index].count -= chunk_samples;
1301 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1302 chunk_samples -= sc->stts_data[stts_index].count;
1303 if (stts_index + 1 < sc->stts_count)
1307 current_offset += sc->bytes_per_frame;
1308 dprintf(mov->fc, "AVIndex stream %d, chunk %d, offset %"PRIx64", dts %"PRId64", "
1309 "size %d, duration %d\n", st->index, i, current_offset, current_dts,
1310 chunk_size, chunk_duration);
1311 assert(chunk_duration % sc->time_rate == 0);
1312 current_dts += chunk_duration / sc->time_rate;
1317 /* adjust sample count to avindex entries */
1318 sc->sample_count = st->nb_index_entries;
1321 static int mov_read_trak(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1324 MOVStreamContext *sc;
1327 st = av_new_stream(c->fc, c->fc->nb_streams);
1328 if (!st) return AVERROR(ENOMEM);
1329 sc = av_mallocz(sizeof(MOVStreamContext));
1330 if (!sc) return AVERROR(ENOMEM);
1333 st->codec->codec_type = CODEC_TYPE_DATA;
1334 st->start_time = 0; /* XXX: check */
1336 if ((ret = mov_read_default(c, pb, atom)) < 0)
1340 if(sc->chunk_count && (!sc->stts_count || !sc->sample_to_chunk_sz ||
1341 (!sc->sample_size && !sc->sample_count))){
1342 av_log(c->fc, AV_LOG_ERROR, "stream %d, missing mandatory atoms, broken header\n",
1344 sc->sample_count = 0; //ignore track
1350 sc->time_scale= c->time_scale;
1351 av_set_pts_info(st, 64, sc->time_rate, sc->time_scale);
1353 if (st->codec->codec_type == CODEC_TYPE_AUDIO &&
1354 !st->codec->frame_size && sc->stts_count == 1)
1355 st->codec->frame_size = av_rescale(sc->time_rate, st->codec->sample_rate, sc->time_scale);
1357 if(st->duration != AV_NOPTS_VALUE){
1358 assert(st->duration % sc->time_rate == 0);
1359 st->duration /= sc->time_rate;
1361 sc->ffindex = st->index;
1362 mov_build_index(c, st);
1364 if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
1365 if (url_fopen(&sc->pb, sc->drefs[sc->dref_id-1].path, URL_RDONLY) < 0)
1366 av_log(c->fc, AV_LOG_ERROR, "stream %d, error opening file %s: %s\n",
1367 st->index, sc->drefs[sc->dref_id-1].path, strerror(errno));
1371 switch (st->codec->codec_id) {
1372 #ifdef CONFIG_H261_DECODER
1375 #ifdef CONFIG_H263_DECODER
1378 #ifdef CONFIG_MPEG4_DECODER
1379 case CODEC_ID_MPEG4:
1381 st->codec->width= 0; /* let decoder init width/height */
1382 st->codec->height= 0;
1386 /* Do not need those anymore. */
1387 av_freep(&sc->chunk_offsets);
1388 av_freep(&sc->sample_to_chunk);
1389 av_freep(&sc->sample_sizes);
1390 av_freep(&sc->keyframes);
1391 av_freep(&sc->stts_data);
1396 static int mov_read_ilst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1399 c->itunes_metadata = 1;
1400 ret = mov_read_default(c, pb, atom);
1401 c->itunes_metadata = 0;
1405 static int mov_read_meta(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1407 url_fskip(pb, 4); // version + flags
1409 return mov_read_default(c, pb, atom);
1412 static int mov_read_trkn(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1414 get_be32(pb); // type
1415 get_be32(pb); // unknown
1416 c->fc->track = get_be32(pb);
1417 dprintf(c->fc, "%.4s %d\n", (char*)&atom.type, c->fc->track);
1421 static int mov_read_udta_string(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1427 if (c->itunes_metadata) {
1428 int data_size = get_be32(pb);
1429 int tag = get_le32(pb);
1430 if (tag == MKTAG('d','a','t','a')) {
1431 get_be32(pb); // type
1432 get_be32(pb); // unknown
1433 str_size = data_size - 16;
1436 str_size = get_be16(pb); // string length
1437 get_be16(pb); // language
1439 switch (atom.type) {
1440 case MKTAG(0xa9,'n','a','m'):
1441 str = c->fc->title; size = sizeof(c->fc->title); break;
1442 case MKTAG(0xa9,'A','R','T'):
1443 case MKTAG(0xa9,'w','r','t'):
1444 str = c->fc->author; size = sizeof(c->fc->author); break;
1445 case MKTAG(0xa9,'c','p','y'):
1446 str = c->fc->copyright; size = sizeof(c->fc->copyright); break;
1447 case MKTAG(0xa9,'c','m','t'):
1448 case MKTAG(0xa9,'i','n','f'):
1449 str = c->fc->comment; size = sizeof(c->fc->comment); break;
1450 case MKTAG(0xa9,'a','l','b'):
1451 str = c->fc->album; size = sizeof(c->fc->album); break;
1455 get_buffer(pb, str, FFMIN(size, str_size));
1456 dprintf(c->fc, "%.4s %s\n", (char*)&atom.type, str);
1460 static int mov_read_tkhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1465 int64_t disp_transform[2];
1466 int display_matrix[3][2];
1467 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1468 int version = get_byte(pb);
1470 get_be24(pb); /* flags */
1472 MOV_TRACK_ENABLED 0x0001
1473 MOV_TRACK_IN_MOVIE 0x0002
1474 MOV_TRACK_IN_PREVIEW 0x0004
1475 MOV_TRACK_IN_POSTER 0x0008
1482 get_be32(pb); /* creation time */
1483 get_be32(pb); /* modification time */
1485 st->id = (int)get_be32(pb); /* track id (NOT 0 !)*/
1486 get_be32(pb); /* reserved */
1487 st->start_time = 0; /* check */
1488 /* highlevel (considering edits) duration in movie timebase */
1489 (version == 1) ? get_be64(pb) : get_be32(pb);
1490 get_be32(pb); /* reserved */
1491 get_be32(pb); /* reserved */
1493 get_be16(pb); /* layer */
1494 get_be16(pb); /* alternate group */
1495 get_be16(pb); /* volume */
1496 get_be16(pb); /* reserved */
1498 //read in the display matrix (outlined in ISO 14496-12, Section 6.2.2)
1499 // they're kept in fixed point format through all calculations
1500 // ignore u,v,z b/c we don't need the scale factor to calc aspect ratio
1501 for (i = 0; i < 3; i++) {
1502 display_matrix[i][0] = get_be32(pb); // 16.16 fixed point
1503 display_matrix[i][1] = get_be32(pb); // 16.16 fixed point
1504 get_be32(pb); // 2.30 fixed point (not used)
1507 width = get_be32(pb); // 16.16 fixed point track width
1508 height = get_be32(pb); // 16.16 fixed point track height
1510 //transform the display width/height according to the matrix
1511 // skip this if the display matrix is the default identity matrix
1512 // to keep the same scale, use [width height 1<<16]
1513 if (width && height &&
1514 (display_matrix[0][0] != 65536 || display_matrix[0][1] ||
1515 display_matrix[1][0] || display_matrix[1][1] != 65536 ||
1516 display_matrix[2][0] || display_matrix[2][1])) {
1517 for (i = 0; i < 2; i++)
1519 (int64_t) width * display_matrix[0][i] +
1520 (int64_t) height * display_matrix[1][i] +
1521 ((int64_t) display_matrix[2][i] << 16);
1523 //sample aspect ratio is new width/height divided by old width/height
1524 st->sample_aspect_ratio = av_d2q(
1525 ((double) disp_transform[0] * height) /
1526 ((double) disp_transform[1] * width), INT_MAX);
1531 static int mov_read_tfhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1533 MOVFragment *frag = &c->fragment;
1534 MOVTrackExt *trex = NULL;
1535 int flags, track_id, i;
1537 get_byte(pb); /* version */
1538 flags = get_be24(pb);
1540 track_id = get_be32(pb);
1541 if (!track_id || track_id > c->fc->nb_streams)
1543 frag->track_id = track_id;
1544 for (i = 0; i < c->trex_count; i++)
1545 if (c->trex_data[i].track_id == frag->track_id) {
1546 trex = &c->trex_data[i];
1550 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding trex\n");
1554 if (flags & 0x01) frag->base_data_offset = get_be64(pb);
1555 else frag->base_data_offset = frag->moof_offset;
1556 if (flags & 0x02) frag->stsd_id = get_be32(pb);
1557 else frag->stsd_id = trex->stsd_id;
1559 frag->duration = flags & 0x08 ? get_be32(pb) : trex->duration;
1560 frag->size = flags & 0x10 ? get_be32(pb) : trex->size;
1561 frag->flags = flags & 0x20 ? get_be32(pb) : trex->flags;
1562 dprintf(c->fc, "frag flags 0x%x\n", frag->flags);
1566 static int mov_read_trex(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1570 if ((uint64_t)c->trex_count+1 >= UINT_MAX / sizeof(*c->trex_data))
1572 c->trex_data = av_realloc(c->trex_data, (c->trex_count+1)*sizeof(*c->trex_data));
1574 return AVERROR(ENOMEM);
1575 trex = &c->trex_data[c->trex_count++];
1576 get_byte(pb); /* version */
1577 get_be24(pb); /* flags */
1578 trex->track_id = get_be32(pb);
1579 trex->stsd_id = get_be32(pb);
1580 trex->duration = get_be32(pb);
1581 trex->size = get_be32(pb);
1582 trex->flags = get_be32(pb);
1586 static int mov_read_trun(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1588 MOVFragment *frag = &c->fragment;
1590 MOVStreamContext *sc;
1593 int data_offset = 0;
1594 unsigned entries, first_sample_flags = frag->flags;
1595 int flags, distance, i;
1597 if (!frag->track_id || frag->track_id > c->fc->nb_streams)
1599 st = c->fc->streams[frag->track_id-1];
1601 if (sc->pseudo_stream_id+1 != frag->stsd_id)
1603 get_byte(pb); /* version */
1604 flags = get_be24(pb);
1605 entries = get_be32(pb);
1606 dprintf(c->fc, "flags 0x%x entries %d\n", flags, entries);
1607 if (flags & 0x001) data_offset = get_be32(pb);
1608 if (flags & 0x004) first_sample_flags = get_be32(pb);
1609 if (flags & 0x800) {
1610 if ((uint64_t)entries+sc->ctts_count >= UINT_MAX/sizeof(*sc->ctts_data))
1612 sc->ctts_data = av_realloc(sc->ctts_data,
1613 (entries+sc->ctts_count)*sizeof(*sc->ctts_data));
1615 return AVERROR(ENOMEM);
1618 offset = frag->base_data_offset + data_offset;
1620 dprintf(c->fc, "first sample flags 0x%x\n", first_sample_flags);
1621 for (i = 0; i < entries; i++) {
1622 unsigned sample_size = frag->size;
1623 int sample_flags = i ? frag->flags : first_sample_flags;
1624 unsigned sample_duration = frag->duration;
1627 if (flags & 0x100) sample_duration = get_be32(pb);
1628 if (flags & 0x200) sample_size = get_be32(pb);
1629 if (flags & 0x400) sample_flags = get_be32(pb);
1630 if (flags & 0x800) {
1631 sc->ctts_data[sc->ctts_count].count = 1;
1632 sc->ctts_data[sc->ctts_count].duration = get_be32(pb);
1635 if ((keyframe = st->codec->codec_type == CODEC_TYPE_AUDIO ||
1636 (flags & 0x004 && !i && !sample_flags) || sample_flags & 0x2000000))
1638 av_add_index_entry(st, offset, dts, sample_size, distance,
1639 keyframe ? AVINDEX_KEYFRAME : 0);
1640 dprintf(c->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1641 "size %d, distance %d, keyframe %d\n", st->index, sc->sample_count+i,
1642 offset, dts, sample_size, distance, keyframe);
1644 assert(sample_duration % sc->time_rate == 0);
1645 dts += sample_duration / sc->time_rate;
1646 offset += sample_size;
1648 frag->moof_offset = offset;
1649 sc->sample_count = st->nb_index_entries;
1654 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
1655 /* like the files created with Adobe Premiere 5.0, for samples see */
1656 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
1657 static int mov_read_wide(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1662 return 0; /* continue */
1663 if (get_be32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
1664 url_fskip(pb, atom.size - 4);
1667 atom.type = get_le32(pb);
1670 if (atom.type != MKTAG('m','d','a','t')) {
1671 url_fskip(pb, atom.size);
1674 err = mov_read_mdat(c, pb, atom);
1678 static int mov_read_cmov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1683 uint8_t *moov_data; /* uncompressed data */
1684 long cmov_len, moov_len;
1687 get_be32(pb); /* dcom atom */
1688 if (get_le32(pb) != MKTAG('d','c','o','m'))
1690 if (get_le32(pb) != MKTAG('z','l','i','b')) {
1691 av_log(NULL, AV_LOG_ERROR, "unknown compression for cmov atom !");
1694 get_be32(pb); /* cmvd atom */
1695 if (get_le32(pb) != MKTAG('c','m','v','d'))
1697 moov_len = get_be32(pb); /* uncompressed size */
1698 cmov_len = atom.size - 6 * 4;
1700 cmov_data = av_malloc(cmov_len);
1703 moov_data = av_malloc(moov_len);
1708 get_buffer(pb, cmov_data, cmov_len);
1709 if(uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
1710 goto free_and_return;
1711 if(init_put_byte(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL, NULL) != 0)
1712 goto free_and_return;
1713 atom.type = MKTAG('m','o','o','v');
1715 atom.size = moov_len;
1717 // { int fd = open("/tmp/uncompheader.mov", O_WRONLY | O_CREAT); write(fd, moov_data, moov_len); close(fd); }
1719 ret = mov_read_default(c, &ctx, atom);
1725 av_log(c->fc, AV_LOG_ERROR, "this file requires zlib support compiled in\n");
1730 /* edit list atom */
1731 static int mov_read_elst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1733 MOVStreamContext *sc = c->fc->streams[c->fc->nb_streams-1]->priv_data;
1736 get_byte(pb); /* version */
1737 get_be24(pb); /* flags */
1738 edit_count= sc->edit_count = get_be32(pb); /* entries */
1740 for(i=0; i<edit_count; i++){
1742 get_be32(pb); /* Track duration */
1743 time = get_be32(pb); /* Media time */
1744 get_be32(pb); /* Media rate */
1746 av_log(c->fc, AV_LOG_WARNING, "edit list not starting at 0, "
1747 "a/v desync might occur, patch welcome\n");
1749 dprintf(c->fc, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, sc->edit_count);
1753 static const MOVParseTableEntry mov_default_parse_table[] = {
1754 { MKTAG('a','v','s','s'), mov_read_extradata },
1755 { MKTAG('c','o','6','4'), mov_read_stco },
1756 { MKTAG('c','t','t','s'), mov_read_ctts }, /* composition time to sample */
1757 { MKTAG('d','i','n','f'), mov_read_default },
1758 { MKTAG('d','r','e','f'), mov_read_dref },
1759 { MKTAG('e','d','t','s'), mov_read_default },
1760 { MKTAG('e','l','s','t'), mov_read_elst },
1761 { MKTAG('e','n','d','a'), mov_read_enda },
1762 { MKTAG('f','i','e','l'), mov_read_extradata },
1763 { MKTAG('f','t','y','p'), mov_read_ftyp },
1764 { MKTAG('g','l','b','l'), mov_read_glbl },
1765 { MKTAG('h','d','l','r'), mov_read_hdlr },
1766 { MKTAG('i','l','s','t'), mov_read_ilst },
1767 { MKTAG('j','p','2','h'), mov_read_extradata },
1768 { MKTAG('m','d','a','t'), mov_read_mdat },
1769 { MKTAG('m','d','h','d'), mov_read_mdhd },
1770 { MKTAG('m','d','i','a'), mov_read_default },
1771 { MKTAG('m','e','t','a'), mov_read_meta },
1772 { MKTAG('m','i','n','f'), mov_read_default },
1773 { MKTAG('m','o','o','f'), mov_read_moof },
1774 { MKTAG('m','o','o','v'), mov_read_moov },
1775 { MKTAG('m','v','e','x'), mov_read_default },
1776 { MKTAG('m','v','h','d'), mov_read_mvhd },
1777 { MKTAG('S','M','I',' '), mov_read_smi }, /* Sorenson extension ??? */
1778 { MKTAG('a','l','a','c'), mov_read_extradata }, /* alac specific atom */
1779 { MKTAG('a','v','c','C'), mov_read_glbl },
1780 { MKTAG('p','a','s','p'), mov_read_pasp },
1781 { MKTAG('s','t','b','l'), mov_read_default },
1782 { MKTAG('s','t','c','o'), mov_read_stco },
1783 { MKTAG('s','t','s','c'), mov_read_stsc },
1784 { MKTAG('s','t','s','d'), mov_read_stsd }, /* sample description */
1785 { MKTAG('s','t','s','s'), mov_read_stss }, /* sync sample */
1786 { MKTAG('s','t','s','z'), mov_read_stsz }, /* sample size */
1787 { MKTAG('s','t','t','s'), mov_read_stts },
1788 { MKTAG('t','k','h','d'), mov_read_tkhd }, /* track header */
1789 { MKTAG('t','f','h','d'), mov_read_tfhd }, /* track fragment header */
1790 { MKTAG('t','r','a','k'), mov_read_trak },
1791 { MKTAG('t','r','a','f'), mov_read_default },
1792 { MKTAG('t','r','e','x'), mov_read_trex },
1793 { MKTAG('t','r','k','n'), mov_read_trkn },
1794 { MKTAG('t','r','u','n'), mov_read_trun },
1795 { MKTAG('u','d','t','a'), mov_read_default },
1796 { MKTAG('w','a','v','e'), mov_read_wave },
1797 { MKTAG('e','s','d','s'), mov_read_esds },
1798 { MKTAG('w','i','d','e'), mov_read_wide }, /* place holder */
1799 { MKTAG('c','m','o','v'), mov_read_cmov },
1800 { MKTAG(0xa9,'n','a','m'), mov_read_udta_string },
1801 { MKTAG(0xa9,'w','r','t'), mov_read_udta_string },
1802 { MKTAG(0xa9,'c','p','y'), mov_read_udta_string },
1803 { MKTAG(0xa9,'i','n','f'), mov_read_udta_string },
1804 { MKTAG(0xa9,'i','n','f'), mov_read_udta_string },
1805 { MKTAG(0xa9,'A','R','T'), mov_read_udta_string },
1806 { MKTAG(0xa9,'a','l','b'), mov_read_udta_string },
1807 { MKTAG(0xa9,'c','m','t'), mov_read_udta_string },
1811 static int mov_probe(AVProbeData *p)
1813 unsigned int offset;
1817 /* check file header */
1820 /* ignore invalid offset */
1821 if ((offset + 8) > (unsigned int)p->buf_size)
1823 tag = AV_RL32(p->buf + offset + 4);
1825 /* check for obvious tags */
1826 case MKTAG('j','P',' ',' '): /* jpeg 2000 signature */
1827 case MKTAG('m','o','o','v'):
1828 case MKTAG('m','d','a','t'):
1829 case MKTAG('p','n','o','t'): /* detect movs with preview pics like ew.mov and april.mov */
1830 case MKTAG('u','d','t','a'): /* Packet Video PVAuthor adds this and a lot of more junk */
1831 case MKTAG('f','t','y','p'):
1832 return AVPROBE_SCORE_MAX;
1833 /* those are more common words, so rate then a bit less */
1834 case MKTAG('e','d','i','w'): /* xdcam files have reverted first tags */
1835 case MKTAG('w','i','d','e'):
1836 case MKTAG('f','r','e','e'):
1837 case MKTAG('j','u','n','k'):
1838 case MKTAG('p','i','c','t'):
1839 return AVPROBE_SCORE_MAX - 5;
1840 case MKTAG(0x82,0x82,0x7f,0x7d):
1841 case MKTAG('s','k','i','p'):
1842 case MKTAG('u','u','i','d'):
1843 case MKTAG('p','r','f','l'):
1844 offset = AV_RB32(p->buf+offset) + offset;
1845 /* if we only find those cause probedata is too small at least rate them */
1846 score = AVPROBE_SCORE_MAX - 50;
1849 /* unrecognized tag */
1856 static int mov_read_header(AVFormatContext *s, AVFormatParameters *ap)
1858 MOVContext *mov = s->priv_data;
1859 ByteIOContext *pb = s->pb;
1861 MOVAtom atom = { 0, 0, 0 };
1864 /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
1865 if(!url_is_streamed(pb))
1866 atom.size = url_fsize(pb);
1868 atom.size = INT64_MAX;
1870 /* check MOV header */
1871 if ((err = mov_read_default(mov, pb, atom)) < 0) {
1872 av_log(s, AV_LOG_ERROR, "error reading header: %d\n", err);
1875 if (!mov->found_moov) {
1876 av_log(s, AV_LOG_ERROR, "moov atom not found\n");
1879 dprintf(mov->fc, "on_parse_exit_offset=%lld\n", url_ftell(pb));
1884 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
1886 MOVContext *mov = s->priv_data;
1887 MOVStreamContext *sc = 0;
1888 AVIndexEntry *sample = 0;
1889 int64_t best_dts = INT64_MAX;
1892 for (i = 0; i < s->nb_streams; i++) {
1893 AVStream *st = s->streams[i];
1894 MOVStreamContext *msc = st->priv_data;
1895 if (st->discard != AVDISCARD_ALL && msc->pb && msc->current_sample < msc->sample_count) {
1896 AVIndexEntry *current_sample = &st->index_entries[msc->current_sample];
1897 int64_t dts = av_rescale(current_sample->timestamp * (int64_t)msc->time_rate,
1898 AV_TIME_BASE, msc->time_scale);
1899 dprintf(s, "stream %d, sample %d, dts %"PRId64"\n", i, msc->current_sample, dts);
1900 if (!sample || (url_is_streamed(s->pb) && current_sample->pos < sample->pos) ||
1901 (!url_is_streamed(s->pb) &&
1902 ((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
1903 ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
1904 (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
1905 sample = current_sample;
1912 mov->found_mdat = 0;
1913 if (!url_is_streamed(s->pb) ||
1914 mov_read_default(mov, s->pb, (MOVAtom){ 0, 0, INT64_MAX }) < 0 ||
1917 dprintf(s, "read fragments, offset 0x%llx\n", url_ftell(s->pb));
1920 /* must be done just before reading, to avoid infinite loop on sample */
1921 sc->current_sample++;
1922 if (url_fseek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
1923 av_log(mov->fc, AV_LOG_ERROR, "stream %d, offset 0x%"PRIx64": partial file\n",
1924 sc->ffindex, sample->pos);
1927 av_get_packet(sc->pb, pkt, sample->size);
1928 #ifdef CONFIG_DV_DEMUXER
1929 if (mov->dv_demux && sc->dv_audio_container) {
1930 dv_produce_packet(mov->dv_demux, pkt, pkt->data, pkt->size);
1933 if (dv_get_packet(mov->dv_demux, pkt) < 0)
1937 pkt->stream_index = sc->ffindex;
1938 pkt->dts = sample->timestamp;
1939 if (sc->ctts_data) {
1940 assert(sc->ctts_data[sc->sample_to_ctime_index].duration % sc->time_rate == 0);
1941 pkt->pts = pkt->dts + sc->ctts_data[sc->sample_to_ctime_index].duration / sc->time_rate;
1942 /* update ctts context */
1943 sc->sample_to_ctime_sample++;
1944 if (sc->sample_to_ctime_index < sc->ctts_count &&
1945 sc->ctts_data[sc->sample_to_ctime_index].count == sc->sample_to_ctime_sample) {
1946 sc->sample_to_ctime_index++;
1947 sc->sample_to_ctime_sample = 0;
1950 pkt->dts = AV_NOPTS_VALUE;
1952 AVStream *st = s->streams[sc->ffindex];
1953 int64_t next_dts = (sc->current_sample < sc->sample_count) ?
1954 st->index_entries[sc->current_sample].timestamp : st->duration;
1955 pkt->duration = next_dts - pkt->dts;
1956 pkt->pts = pkt->dts;
1958 pkt->flags |= sample->flags & AVINDEX_KEYFRAME ? PKT_FLAG_KEY : 0;
1959 pkt->pos = sample->pos;
1960 dprintf(s, "stream %d, pts %"PRId64", dts %"PRId64", pos 0x%"PRIx64", duration %d\n",
1961 pkt->stream_index, pkt->pts, pkt->dts, pkt->pos, pkt->duration);
1965 static int mov_seek_stream(AVStream *st, int64_t timestamp, int flags)
1967 MOVStreamContext *sc = st->priv_data;
1968 int sample, time_sample;
1971 sample = av_index_search_timestamp(st, timestamp, flags);
1972 dprintf(st->codec, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
1973 if (sample < 0) /* not sure what to do */
1975 sc->current_sample = sample;
1976 dprintf(st->codec, "stream %d, found sample %d\n", st->index, sc->current_sample);
1977 /* adjust ctts index */
1978 if (sc->ctts_data) {
1980 for (i = 0; i < sc->ctts_count; i++) {
1981 int next = time_sample + sc->ctts_data[i].count;
1982 if (next > sc->current_sample) {
1983 sc->sample_to_ctime_index = i;
1984 sc->sample_to_ctime_sample = sc->current_sample - time_sample;
1993 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
1996 int64_t seek_timestamp, timestamp;
2000 if (stream_index >= s->nb_streams)
2003 st = s->streams[stream_index];
2004 sample = mov_seek_stream(st, sample_time, flags);
2008 /* adjust seek timestamp to found sample timestamp */
2009 seek_timestamp = st->index_entries[sample].timestamp;
2011 for (i = 0; i < s->nb_streams; i++) {
2013 if (stream_index == i || st->discard == AVDISCARD_ALL)
2016 timestamp = av_rescale_q(seek_timestamp, s->streams[stream_index]->time_base, st->time_base);
2017 mov_seek_stream(st, timestamp, flags);
2022 static int mov_read_close(AVFormatContext *s)
2025 MOVContext *mov = s->priv_data;
2026 for(i=0; i<s->nb_streams; i++) {
2027 MOVStreamContext *sc = s->streams[i]->priv_data;
2028 av_freep(&sc->ctts_data);
2029 for (j=0; j<sc->drefs_count; j++)
2030 av_freep(&sc->drefs[j].path);
2031 av_freep(&sc->drefs);
2032 if (sc->pb && sc->pb != s->pb)
2036 for(i=0; i<mov->dv_fctx->nb_streams; i++){
2037 av_freep(&mov->dv_fctx->streams[i]->codec);
2038 av_freep(&mov->dv_fctx->streams[i]);
2040 av_freep(&mov->dv_fctx);
2041 av_freep(&mov->dv_demux);
2043 av_freep(&mov->trex_data);
2047 AVInputFormat mov_demuxer = {
2048 "mov,mp4,m4a,3gp,3g2,mj2",
2049 NULL_IF_CONFIG_SMALL("QuickTime/MPEG-4/Motion JPEG 2000 format"),