3 * Copyright (c) 2001 Fabrice Bellard.
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
30 #include "libavcodec/mpeg4audio.h"
31 #include "libavcodec/mpegaudiodata.h"
38 * First version by Francois Revol revol@free.fr
39 * Seek function by Gael Chardon gael.dev@4now.net
41 * Features and limitations:
42 * - reads most of the QT files I have (at least the structure),
43 * Sample QuickTime files with mp3 audio can be found at: http://www.3ivx.com/showcase.html
44 * - the code is quite ugly... maybe I won't do it recursive next time :-)
46 * Funny I didn't know about http://sourceforge.net/projects/qt-ffmpeg/
47 * when coding this :) (it's a writer anyway)
49 * Reference documents:
50 * http://www.geocities.com/xhelmboyx/quicktime/formats/qtm-layout.txt
52 * http://developer.apple.com/documentation/QuickTime/QTFF/
53 * http://developer.apple.com/documentation/QuickTime/QTFF/qtff.pdf
54 * QuickTime is a trademark of Apple (AFAIK :))
57 #include "qtpalette.h"
63 /* the QuickTime file format is quite convoluted...
64 * it has lots of index tables, each indexing something in another one...
65 * Here we just use what is needed to read the chunks
82 int64_t size; /* total size (excluding the size and type fields) */
85 struct MOVParseTableEntry;
89 uint64_t base_data_offset;
105 typedef struct MOVStreamContext {
107 int ffindex; /* the ffmpeg stream id */
109 unsigned int chunk_count;
110 int64_t *chunk_offsets;
111 unsigned int stts_count;
112 MOV_stts_t *stts_data;
113 unsigned int ctts_count;
114 MOV_stts_t *ctts_data;
115 unsigned int edit_count; /* number of 'edit' (elst atom) */
116 unsigned int sample_to_chunk_sz;
117 MOV_stsc_t *sample_to_chunk;
118 int sample_to_ctime_index;
119 int sample_to_ctime_sample;
120 unsigned int sample_size;
121 unsigned int sample_count;
123 unsigned int keyframe_count;
128 unsigned int bytes_per_frame;
129 unsigned int samples_per_frame;
130 int dv_audio_container;
131 int pseudo_stream_id; ///< -1 means demux all ids
132 int16_t audio_cid; ///< stsd audio compression id
133 unsigned drefs_count;
138 typedef struct MOVContext {
141 int64_t duration; /* duration of the longest track */
142 int found_moov; /* when both 'moov' and 'mdat' sections has been found */
143 int found_mdat; /* we suppose we have enough data to read the file */
144 AVPaletteControl palette_control;
145 DVDemuxContext *dv_demux;
146 AVFormatContext *dv_fctx;
147 int isom; /* 1 if file is ISO Media (mp4/3gp) */
148 MOVFragment fragment; ///< current fragment in moof atom
149 MOVTrackExt *trex_data;
154 /* XXX: it's the first time I make a recursive parser I think... sorry if it's ugly :P */
156 /* those functions parse an atom */
158 0: continue to parse next atom
159 <0: error occurred, exit
161 /* links atom IDs to parse functions */
162 typedef struct MOVParseTableEntry {
164 int (*parse)(MOVContext *ctx, ByteIOContext *pb, MOV_atom_t atom);
165 } MOVParseTableEntry;
167 static const MOVParseTableEntry mov_default_parse_table[];
169 static int mov_read_default(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
171 int64_t total_size = 0;
176 a.offset = atom.offset;
179 atom.size = INT64_MAX;
180 while(((total_size + 8) < atom.size) && !url_feof(pb) && !err) {
184 a.size = get_be32(pb);
185 a.type = get_le32(pb);
189 dprintf(c->fc, "type: %08x %.4s sz: %"PRIx64" %"PRIx64" %"PRIx64"\n",
190 a.type, (char*)&a.type, a.size, atom.size, total_size);
191 if (a.size == 1) { /* 64 bit extended size */
192 a.size = get_be64(pb) - 8;
197 a.size = atom.size - total_size;
204 a.size = FFMIN(a.size, atom.size - total_size);
206 for (i = 0; mov_default_parse_table[i].type != 0
207 && mov_default_parse_table[i].type != a.type; i++)
210 if (mov_default_parse_table[i].type == 0) { /* skip leaf atoms data */
211 url_fskip(pb, a.size);
213 offset_t start_pos = url_ftell(pb);
215 err = mov_default_parse_table[i].parse(c, pb, a);
216 if (url_is_streamed(pb) && c->found_moov && c->found_mdat)
218 left = a.size - url_ftell(pb) + start_pos;
219 if (left > 0) /* skip garbage at atom end */
224 total_size += a.size;
227 if (!err && total_size < atom.size && atom.size < 0x7ffff)
228 url_fskip(pb, atom.size - total_size);
233 static int mov_read_dref(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
235 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
236 MOVStreamContext *sc = st->priv_data;
239 get_be32(pb); // version + flags
240 entries = get_be32(pb);
241 if (entries >= UINT_MAX / sizeof(*sc->drefs))
243 sc->drefs_count = entries;
244 sc->drefs = av_mallocz(entries * sizeof(*sc->drefs));
246 for (i = 0; i < sc->drefs_count; i++) {
247 MOV_dref_t *dref = &sc->drefs[i];
248 uint32_t size = get_be32(pb);
249 offset_t next = url_ftell(pb) + size - 4;
251 dref->type = get_le32(pb);
252 get_be32(pb); // version + flags
253 dprintf(c->fc, "type %.4s size %d\n", (char*)&dref->type, size);
255 if (dref->type == MKTAG('a','l','i','s') && size > 150) {
256 /* macintosh alias record */
257 uint16_t volume_len, len;
263 volume_len = get_byte(pb);
264 volume_len = FFMIN(volume_len, 27);
265 get_buffer(pb, volume, 27);
266 volume[volume_len] = 0;
267 av_log(c->fc, AV_LOG_DEBUG, "volume %s, len %d\n", volume, volume_len);
271 for (type = 0; type != -1 && url_ftell(pb) < next; ) {
274 av_log(c->fc, AV_LOG_DEBUG, "type %d, len %d\n", type, len);
277 if (type == 2) { // absolute path
279 dref->path = av_mallocz(len+1);
281 return AVERROR(ENOMEM);
282 get_buffer(pb, dref->path, len);
283 if (len > volume_len && !strncmp(dref->path, volume, volume_len)) {
285 memmove(dref->path, dref->path+volume_len, len);
288 for (j = 0; j < len; j++)
289 if (dref->path[j] == ':')
291 av_log(c->fc, AV_LOG_DEBUG, "path %s\n", dref->path);
296 url_fseek(pb, next, SEEK_SET);
301 static int mov_read_hdlr(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
303 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
307 get_byte(pb); /* version */
308 get_be24(pb); /* flags */
311 ctype = get_le32(pb);
312 type = get_le32(pb); /* component subtype */
314 dprintf(c->fc, "ctype= %c%c%c%c (0x%08x)\n", *((char *)&ctype), ((char *)&ctype)[1],
315 ((char *)&ctype)[2], ((char *)&ctype)[3], (int) ctype);
316 dprintf(c->fc, "stype= %c%c%c%c\n",
317 *((char *)&type), ((char *)&type)[1], ((char *)&type)[2], ((char *)&type)[3]);
320 if (type == MKTAG('v','i','d','e'))
321 st->codec->codec_type = CODEC_TYPE_VIDEO;
322 else if(type == MKTAG('s','o','u','n'))
323 st->codec->codec_type = CODEC_TYPE_AUDIO;
324 else if(type == MKTAG('m','1','a',' '))
325 st->codec->codec_id = CODEC_ID_MP2;
326 else if(type == MKTAG('s','u','b','p')) {
327 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
329 get_be32(pb); /* component manufacture */
330 get_be32(pb); /* component flags */
331 get_be32(pb); /* component flags mask */
334 return 0; /* nothing left to read */
336 url_fskip(pb, atom.size - (url_ftell(pb) - atom.offset));
340 static int mp4_read_descr_len(ByteIOContext *pb)
345 int c = get_byte(pb);
346 len = (len << 7) | (c & 0x7f);
353 static int mp4_read_descr(MOVContext *c, ByteIOContext *pb, int *tag)
357 len = mp4_read_descr_len(pb);
358 dprintf(c->fc, "MPEG4 description: tag=0x%02x len=%d\n", *tag, len);
362 #define MP4ESDescrTag 0x03
363 #define MP4DecConfigDescrTag 0x04
364 #define MP4DecSpecificDescrTag 0x05
366 static const AVCodecTag mp4_audio_types[] = {
367 { CODEC_ID_MP3ON4, 29 }, /* old mp3on4 draft */
368 { CODEC_ID_MP3ON4, 32 }, /* layer 1 */
369 { CODEC_ID_MP3ON4, 33 }, /* layer 2 */
370 { CODEC_ID_MP3ON4, 34 }, /* layer 3 */
371 { CODEC_ID_NONE, 0 },
374 static int mov_read_esds(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
376 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
379 get_be32(pb); /* version + flags */
380 len = mp4_read_descr(c, pb, &tag);
381 if (tag == MP4ESDescrTag) {
382 get_be16(pb); /* ID */
383 get_byte(pb); /* priority */
385 get_be16(pb); /* ID */
387 len = mp4_read_descr(c, pb, &tag);
388 if (tag == MP4DecConfigDescrTag) {
389 int object_type_id = get_byte(pb);
390 get_byte(pb); /* stream type */
391 get_be24(pb); /* buffer size db */
392 get_be32(pb); /* max bitrate */
393 get_be32(pb); /* avg bitrate */
395 st->codec->codec_id= codec_get_id(ff_mp4_obj_type, object_type_id);
396 dprintf(c->fc, "esds object type id %d\n", object_type_id);
397 len = mp4_read_descr(c, pb, &tag);
398 if (tag == MP4DecSpecificDescrTag) {
399 dprintf(c->fc, "Specific MPEG4 header len=%d\n", len);
400 if((uint64_t)len > (1<<30))
402 st->codec->extradata = av_mallocz(len + FF_INPUT_BUFFER_PADDING_SIZE);
403 if (!st->codec->extradata)
404 return AVERROR(ENOMEM);
405 get_buffer(pb, st->codec->extradata, len);
406 st->codec->extradata_size = len;
407 if (st->codec->codec_id == CODEC_ID_AAC) {
408 MPEG4AudioConfig cfg;
409 ff_mpeg4audio_get_config(&cfg, st->codec->extradata,
410 st->codec->extradata_size);
411 if (cfg.chan_config > 7)
413 st->codec->channels = ff_mpeg4audio_channels[cfg.chan_config];
414 if (cfg.object_type == 29 && cfg.sampling_index < 3) // old mp3on4
415 st->codec->sample_rate = ff_mpa_freq_tab[cfg.sampling_index];
417 st->codec->sample_rate = cfg.sample_rate; // ext sample rate ?
418 dprintf(c->fc, "mp4a config channels %d obj %d ext obj %d "
419 "sample rate %d ext sample rate %d\n", st->codec->channels,
420 cfg.object_type, cfg.ext_object_type,
421 cfg.sample_rate, cfg.ext_sample_rate);
422 if (!(st->codec->codec_id = codec_get_id(mp4_audio_types,
424 st->codec->codec_id = CODEC_ID_AAC;
431 /* this atom contains actual media data */
432 static int mov_read_mdat(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
434 if(atom.size == 0) /* wrong one (MP4) */
437 return 0; /* now go for moov */
440 static int mov_read_ftyp(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
442 uint32_t type = get_le32(pb);
444 if (type != MKTAG('q','t',' ',' '))
446 av_log(c->fc, AV_LOG_DEBUG, "ISO: File Type Major Brand: %.4s\n",(char *)&type);
447 get_be32(pb); /* minor version */
448 url_fskip(pb, atom.size - 8);
452 /* this atom should contain all header atoms */
453 static int mov_read_moov(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
455 if (mov_read_default(c, pb, atom) < 0)
457 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
458 /* so we don't parse the whole file if over a network */
460 return 0; /* now go for mdat */
463 static int mov_read_moof(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
465 c->fragment.moof_offset = url_ftell(pb) - 8;
466 dprintf(c->fc, "moof offset %llx\n", c->fragment.moof_offset);
467 return mov_read_default(c, pb, atom);
470 static int mov_read_mdhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
472 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
473 MOVStreamContext *sc = st->priv_data;
474 int version = get_byte(pb);
478 return -1; /* unsupported */
480 get_be24(pb); /* flags */
485 get_be32(pb); /* creation time */
486 get_be32(pb); /* modification time */
489 sc->time_scale = get_be32(pb);
490 st->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
492 lang = get_be16(pb); /* language */
493 ff_mov_lang_to_iso639(lang, st->language);
494 get_be16(pb); /* quality */
499 static int mov_read_mvhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
501 int version = get_byte(pb); /* version */
502 get_be24(pb); /* flags */
508 get_be32(pb); /* creation time */
509 get_be32(pb); /* modification time */
511 c->time_scale = get_be32(pb); /* time scale */
513 dprintf(c->fc, "time scale = %i\n", c->time_scale);
515 c->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
516 get_be32(pb); /* preferred scale */
518 get_be16(pb); /* preferred volume */
520 url_fskip(pb, 10); /* reserved */
522 url_fskip(pb, 36); /* display matrix */
524 get_be32(pb); /* preview time */
525 get_be32(pb); /* preview duration */
526 get_be32(pb); /* poster time */
527 get_be32(pb); /* selection time */
528 get_be32(pb); /* selection duration */
529 get_be32(pb); /* current time */
530 get_be32(pb); /* next track ID */
535 static int mov_read_smi(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
537 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
539 if((uint64_t)atom.size > (1<<30))
542 // currently SVQ3 decoder expect full STSD header - so let's fake it
543 // this should be fixed and just SMI header should be passed
544 av_free(st->codec->extradata);
545 st->codec->extradata = av_mallocz(atom.size + 0x5a + FF_INPUT_BUFFER_PADDING_SIZE);
546 if (!st->codec->extradata)
547 return AVERROR(ENOMEM);
548 st->codec->extradata_size = 0x5a + atom.size;
549 memcpy(st->codec->extradata, "SVQ3", 4); // fake
550 get_buffer(pb, st->codec->extradata + 0x5a, atom.size);
551 dprintf(c->fc, "Reading SMI %"PRId64" %s\n", atom.size, st->codec->extradata + 0x5a);
555 static int mov_read_enda(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
557 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
558 int little_endian = get_be16(pb);
561 switch (st->codec->codec_id) {
562 case CODEC_ID_PCM_S24BE:
563 st->codec->codec_id = CODEC_ID_PCM_S24LE;
565 case CODEC_ID_PCM_S32BE:
566 st->codec->codec_id = CODEC_ID_PCM_S32LE;
575 /* FIXME modify qdm2/svq3/h264 decoders to take full atom as extradata */
576 static int mov_read_extradata(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
578 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
579 uint64_t size= (uint64_t)st->codec->extradata_size + atom.size + 8 + FF_INPUT_BUFFER_PADDING_SIZE;
581 if(size > INT_MAX || (uint64_t)atom.size > INT_MAX)
583 buf= av_realloc(st->codec->extradata, size);
586 st->codec->extradata= buf;
587 buf+= st->codec->extradata_size;
588 st->codec->extradata_size= size - FF_INPUT_BUFFER_PADDING_SIZE;
589 AV_WB32( buf , atom.size + 8);
590 AV_WL32( buf + 4, atom.type);
591 get_buffer(pb, buf + 8, atom.size);
595 static int mov_read_wave(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
597 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
599 if((uint64_t)atom.size > (1<<30))
602 if (st->codec->codec_id == CODEC_ID_QDM2) {
603 // pass all frma atom to codec, needed at least for QDM2
604 av_free(st->codec->extradata);
605 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
606 if (!st->codec->extradata)
607 return AVERROR(ENOMEM);
608 st->codec->extradata_size = atom.size;
609 get_buffer(pb, st->codec->extradata, atom.size);
610 } else if (atom.size > 8) { /* to read frma, esds atoms */
611 if (mov_read_default(c, pb, atom) < 0)
614 url_fskip(pb, atom.size);
619 * This function reads atom content and puts data in extradata without tag
620 * nor size unlike mov_read_extradata.
622 static int mov_read_glbl(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
624 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
626 if((uint64_t)atom.size > (1<<30))
629 av_free(st->codec->extradata);
630 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
631 if (!st->codec->extradata)
632 return AVERROR(ENOMEM);
633 st->codec->extradata_size = atom.size;
634 get_buffer(pb, st->codec->extradata, atom.size);
638 static int mov_read_stco(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
640 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
641 MOVStreamContext *sc = st->priv_data;
642 unsigned int i, entries;
644 get_byte(pb); /* version */
645 get_be24(pb); /* flags */
647 entries = get_be32(pb);
649 if(entries >= UINT_MAX/sizeof(int64_t))
652 sc->chunk_count = entries;
653 sc->chunk_offsets = av_malloc(entries * sizeof(int64_t));
654 if (!sc->chunk_offsets)
656 if (atom.type == MKTAG('s','t','c','o'))
657 for(i=0; i<entries; i++)
658 sc->chunk_offsets[i] = get_be32(pb);
659 else if (atom.type == MKTAG('c','o','6','4'))
660 for(i=0; i<entries; i++)
661 sc->chunk_offsets[i] = get_be64(pb);
669 * Compute codec id for 'lpcm' tag.
670 * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
672 static int mov_get_lpcm_codec_id(int bps, int flags)
674 if (flags & 1) { // floating point
675 if (flags & 2) { // big endian
676 if (bps == 32) return CODEC_ID_PCM_F32BE;
677 //else if (bps == 64) return CODEC_ID_PCM_F64BE;
679 //if (bps == 32) return CODEC_ID_PCM_F32LE;
680 //else if (bps == 64) return CODEC_ID_PCM_F64LE;
686 if (flags & 4) return CODEC_ID_PCM_S8;
687 else return CODEC_ID_PCM_U8;
688 else if (bps == 16) return CODEC_ID_PCM_S16BE;
689 else if (bps == 24) return CODEC_ID_PCM_S24BE;
690 else if (bps == 32) return CODEC_ID_PCM_S32BE;
693 if (flags & 4) return CODEC_ID_PCM_S8;
694 else return CODEC_ID_PCM_U8;
695 else if (bps == 16) return CODEC_ID_PCM_S16LE;
696 else if (bps == 24) return CODEC_ID_PCM_S24LE;
697 else if (bps == 32) return CODEC_ID_PCM_S32LE;
703 static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
705 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
706 MOVStreamContext *sc = st->priv_data;
707 int j, entries, pseudo_stream_id;
709 get_byte(pb); /* version */
710 get_be24(pb); /* flags */
712 entries = get_be32(pb);
714 for(pseudo_stream_id=0; pseudo_stream_id<entries; pseudo_stream_id++) {
715 //Parsing Sample description table
718 MOV_atom_t a = { 0, 0, 0 };
719 offset_t start_pos = url_ftell(pb);
720 int size = get_be32(pb); /* size */
721 uint32_t format = get_le32(pb); /* data format */
723 get_be32(pb); /* reserved */
724 get_be16(pb); /* reserved */
725 dref_id = get_be16(pb);
727 if (st->codec->codec_tag &&
728 st->codec->codec_tag != format &&
729 (c->fc->video_codec_id ? codec_get_id(codec_movvideo_tags, format) != c->fc->video_codec_id
730 : st->codec->codec_tag != MKTAG('j','p','e','g'))
732 /* Multiple fourcc, we skip JPEG. This is not correct, we should
733 * export it as a separate AVStream but this needs a few changes
734 * in the MOV demuxer, patch welcome. */
735 av_log(c->fc, AV_LOG_WARNING, "multiple fourcc not supported\n");
736 url_fskip(pb, size - (url_ftell(pb) - start_pos));
739 sc->pseudo_stream_id = st->codec->codec_tag ? -1 : pseudo_stream_id;
740 sc->dref_id= dref_id;
742 st->codec->codec_tag = format;
743 id = codec_get_id(codec_movaudio_tags, format);
744 if (id<=0 && (format&0xFFFF) == 'm'+('s'<<8))
745 id = codec_get_id(codec_wav_tags, bswap_32(format)&0xFFFF);
747 if (st->codec->codec_type != CODEC_TYPE_VIDEO && id > 0) {
748 st->codec->codec_type = CODEC_TYPE_AUDIO;
749 } else if (st->codec->codec_type != CODEC_TYPE_AUDIO && /* do not overwrite codec type */
750 format && format != MKTAG('m','p','4','s')) { /* skip old asf mpeg4 tag */
751 id = codec_get_id(codec_movvideo_tags, format);
753 id = codec_get_id(codec_bmp_tags, format);
755 st->codec->codec_type = CODEC_TYPE_VIDEO;
756 else if(st->codec->codec_type == CODEC_TYPE_DATA){
757 id = codec_get_id(ff_codec_movsubtitle_tags, format);
759 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
763 dprintf(c->fc, "size=%d 4CC= %c%c%c%c codec_type=%d\n", size,
764 (format >> 0) & 0xff, (format >> 8) & 0xff, (format >> 16) & 0xff,
765 (format >> 24) & 0xff, st->codec->codec_type);
767 if(st->codec->codec_type==CODEC_TYPE_VIDEO) {
768 uint8_t codec_name[32];
769 unsigned int color_depth;
772 st->codec->codec_id = id;
773 get_be16(pb); /* version */
774 get_be16(pb); /* revision level */
775 get_be32(pb); /* vendor */
776 get_be32(pb); /* temporal quality */
777 get_be32(pb); /* spatial quality */
779 st->codec->width = get_be16(pb); /* width */
780 st->codec->height = get_be16(pb); /* height */
782 get_be32(pb); /* horiz resolution */
783 get_be32(pb); /* vert resolution */
784 get_be32(pb); /* data size, always 0 */
785 get_be16(pb); /* frames per samples */
787 get_buffer(pb, codec_name, 32); /* codec name, pascal string */
788 if (codec_name[0] <= 31) {
789 memcpy(st->codec->codec_name, &codec_name[1],codec_name[0]);
790 st->codec->codec_name[codec_name[0]] = 0;
793 st->codec->bits_per_sample = get_be16(pb); /* depth */
794 st->codec->color_table_id = get_be16(pb); /* colortable id */
795 dprintf(c->fc, "depth %d, ctab id %d\n",
796 st->codec->bits_per_sample, st->codec->color_table_id);
797 /* figure out the palette situation */
798 color_depth = st->codec->bits_per_sample & 0x1F;
799 color_greyscale = st->codec->bits_per_sample & 0x20;
801 /* if the depth is 2, 4, or 8 bpp, file is palettized */
802 if ((color_depth == 2) || (color_depth == 4) ||
803 (color_depth == 8)) {
804 /* for palette traversal */
805 unsigned int color_start, color_count, color_end;
806 unsigned char r, g, b;
808 if (color_greyscale) {
809 int color_index, color_dec;
810 /* compute the greyscale palette */
811 st->codec->bits_per_sample = color_depth;
812 color_count = 1 << color_depth;
814 color_dec = 256 / (color_count - 1);
815 for (j = 0; j < color_count; j++) {
816 r = g = b = color_index;
817 c->palette_control.palette[j] =
818 (r << 16) | (g << 8) | (b);
819 color_index -= color_dec;
823 } else if (st->codec->color_table_id) {
824 const uint8_t *color_table;
825 /* if flag bit 3 is set, use the default palette */
826 color_count = 1 << color_depth;
827 if (color_depth == 2)
828 color_table = ff_qt_default_palette_4;
829 else if (color_depth == 4)
830 color_table = ff_qt_default_palette_16;
832 color_table = ff_qt_default_palette_256;
834 for (j = 0; j < color_count; j++) {
835 r = color_table[j * 4 + 0];
836 g = color_table[j * 4 + 1];
837 b = color_table[j * 4 + 2];
838 c->palette_control.palette[j] =
839 (r << 16) | (g << 8) | (b);
842 /* load the palette from the file */
843 color_start = get_be32(pb);
844 color_count = get_be16(pb);
845 color_end = get_be16(pb);
846 if ((color_start <= 255) &&
847 (color_end <= 255)) {
848 for (j = color_start; j <= color_end; j++) {
849 /* each R, G, or B component is 16 bits;
850 * only use the top 8 bits; skip alpha bytes
860 c->palette_control.palette[j] =
861 (r << 16) | (g << 8) | (b);
865 st->codec->palctrl = &c->palette_control;
866 st->codec->palctrl->palette_changed = 1;
868 st->codec->palctrl = NULL;
869 } else if(st->codec->codec_type==CODEC_TYPE_AUDIO) {
870 int bits_per_sample, flags;
871 uint16_t version = get_be16(pb);
873 st->codec->codec_id = id;
874 get_be16(pb); /* revision level */
875 get_be32(pb); /* vendor */
877 st->codec->channels = get_be16(pb); /* channel count */
878 dprintf(c->fc, "audio channels %d\n", st->codec->channels);
879 st->codec->bits_per_sample = get_be16(pb); /* sample size */
881 sc->audio_cid = get_be16(pb);
882 get_be16(pb); /* packet size = 0 */
884 st->codec->sample_rate = ((get_be32(pb) >> 16));
886 //Read QT version 1 fields. In version 0 these do not exist.
887 dprintf(c->fc, "version =%d, isom =%d\n",version,c->isom);
890 sc->samples_per_frame = get_be32(pb);
891 get_be32(pb); /* bytes per packet */
892 sc->bytes_per_frame = get_be32(pb);
893 get_be32(pb); /* bytes per sample */
894 } else if(version==2) {
895 get_be32(pb); /* sizeof struct only */
896 st->codec->sample_rate = av_int2dbl(get_be64(pb)); /* float 64 */
897 st->codec->channels = get_be32(pb);
898 get_be32(pb); /* always 0x7F000000 */
899 st->codec->bits_per_sample = get_be32(pb); /* bits per channel if sound is uncompressed */
900 flags = get_be32(pb); /* lcpm format specific flag */
901 sc->bytes_per_frame = get_be32(pb); /* bytes per audio packet if constant */
902 sc->samples_per_frame = get_be32(pb); /* lpcm frames per audio packet if constant */
903 if (format == MKTAG('l','p','c','m'))
904 st->codec->codec_id = mov_get_lpcm_codec_id(st->codec->bits_per_sample, flags);
908 switch (st->codec->codec_id) {
909 case CODEC_ID_PCM_S8:
910 case CODEC_ID_PCM_U8:
911 if (st->codec->bits_per_sample == 16)
912 st->codec->codec_id = CODEC_ID_PCM_S16BE;
914 case CODEC_ID_PCM_S16LE:
915 case CODEC_ID_PCM_S16BE:
916 if (st->codec->bits_per_sample == 8)
917 st->codec->codec_id = CODEC_ID_PCM_S8;
918 else if (st->codec->bits_per_sample == 24)
919 st->codec->codec_id =
920 st->codec->codec_id == CODEC_ID_PCM_S16BE ?
921 CODEC_ID_PCM_S24BE : CODEC_ID_PCM_S24LE;
923 /* set values for old format before stsd version 1 appeared */
925 sc->samples_per_frame = 6;
926 sc->bytes_per_frame = 2*st->codec->channels;
929 sc->samples_per_frame = 6;
930 sc->bytes_per_frame = 1*st->codec->channels;
932 case CODEC_ID_ADPCM_IMA_QT:
933 sc->samples_per_frame = 64;
934 sc->bytes_per_frame = 34*st->codec->channels;
937 sc->samples_per_frame = 160;
938 sc->bytes_per_frame = 33;
944 bits_per_sample = av_get_bits_per_sample(st->codec->codec_id);
945 if (bits_per_sample) {
946 st->codec->bits_per_sample = bits_per_sample;
947 sc->sample_size = (bits_per_sample >> 3) * st->codec->channels;
949 } else if(st->codec->codec_type==CODEC_TYPE_SUBTITLE){
950 st->codec->codec_id= id;
952 /* other codec type, just skip (rtp, mp4s, tmcd ...) */
953 url_fskip(pb, size - (url_ftell(pb) - start_pos));
955 /* this will read extra atoms at the end (wave, alac, damr, avcC, SMI ...) */
956 a.size = size - (url_ftell(pb) - start_pos);
958 if (mov_read_default(c, pb, a) < 0)
960 } else if (a.size > 0)
961 url_fskip(pb, a.size);
964 if(st->codec->codec_type==CODEC_TYPE_AUDIO && st->codec->sample_rate==0 && sc->time_scale>1)
965 st->codec->sample_rate= sc->time_scale;
967 /* special codec parameters handling */
968 switch (st->codec->codec_id) {
969 #ifdef CONFIG_DV_DEMUXER
970 case CODEC_ID_DVAUDIO:
971 c->dv_fctx = av_alloc_format_context();
972 c->dv_demux = dv_init_demux(c->dv_fctx);
974 av_log(c->fc, AV_LOG_ERROR, "dv demux context init error\n");
977 sc->dv_audio_container = 1;
978 st->codec->codec_id = CODEC_ID_PCM_S16LE;
981 /* no ifdef since parameters are always those */
982 case CODEC_ID_AMR_WB:
983 st->codec->sample_rate= 16000;
984 st->codec->channels= 1; /* really needed */
987 case CODEC_ID_AMR_NB:
988 st->codec->frame_size= sc->samples_per_frame;
989 st->codec->sample_rate= 8000;
990 st->codec->channels= 1; /* really needed */
994 st->codec->codec_type = CODEC_TYPE_AUDIO; /* force type after stsd for m1a hdlr */
995 st->need_parsing = AVSTREAM_PARSE_FULL;
998 case CODEC_ID_ADPCM_MS:
999 case CODEC_ID_ADPCM_IMA_WAV:
1000 st->codec->block_align = sc->bytes_per_frame;
1003 if (st->codec->extradata_size == 36)
1004 st->codec->frame_size = AV_RB32((st->codec->extradata+12));
1013 static int mov_read_stsc(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1015 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1016 MOVStreamContext *sc = st->priv_data;
1017 unsigned int i, entries;
1019 get_byte(pb); /* version */
1020 get_be24(pb); /* flags */
1022 entries = get_be32(pb);
1024 if(entries >= UINT_MAX / sizeof(MOV_stsc_t))
1027 dprintf(c->fc, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
1029 sc->sample_to_chunk_sz = entries;
1030 sc->sample_to_chunk = av_malloc(entries * sizeof(MOV_stsc_t));
1031 if (!sc->sample_to_chunk)
1033 for(i=0; i<entries; i++) {
1034 sc->sample_to_chunk[i].first = get_be32(pb);
1035 sc->sample_to_chunk[i].count = get_be32(pb);
1036 sc->sample_to_chunk[i].id = get_be32(pb);
1041 static int mov_read_stss(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1043 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1044 MOVStreamContext *sc = st->priv_data;
1045 unsigned int i, entries;
1047 get_byte(pb); /* version */
1048 get_be24(pb); /* flags */
1050 entries = get_be32(pb);
1052 if(entries >= UINT_MAX / sizeof(int))
1055 sc->keyframe_count = entries;
1057 dprintf(c->fc, "keyframe_count = %d\n", sc->keyframe_count);
1059 sc->keyframes = av_malloc(entries * sizeof(int));
1062 for(i=0; i<entries; i++) {
1063 sc->keyframes[i] = get_be32(pb);
1064 //dprintf(c->fc, "keyframes[]=%d\n", sc->keyframes[i]);
1069 static int mov_read_stsz(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1071 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1072 MOVStreamContext *sc = st->priv_data;
1073 unsigned int i, entries, sample_size;
1075 get_byte(pb); /* version */
1076 get_be24(pb); /* flags */
1078 sample_size = get_be32(pb);
1079 if (!sc->sample_size) /* do not overwrite value computed in stsd */
1080 sc->sample_size = sample_size;
1081 entries = get_be32(pb);
1082 if(entries >= UINT_MAX / sizeof(int))
1085 sc->sample_count = entries;
1089 dprintf(c->fc, "sample_size = %d sample_count = %d\n", sc->sample_size, sc->sample_count);
1091 sc->sample_sizes = av_malloc(entries * sizeof(int));
1092 if (!sc->sample_sizes)
1094 for(i=0; i<entries; i++)
1095 sc->sample_sizes[i] = get_be32(pb);
1099 static int mov_read_stts(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1101 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1102 MOVStreamContext *sc = st->priv_data;
1103 unsigned int i, entries;
1105 int64_t total_sample_count=0;
1107 get_byte(pb); /* version */
1108 get_be24(pb); /* flags */
1109 entries = get_be32(pb);
1110 if(entries >= UINT_MAX / sizeof(MOV_stts_t))
1113 sc->stts_count = entries;
1114 sc->stts_data = av_malloc(entries * sizeof(MOV_stts_t));
1117 dprintf(c->fc, "track[%i].stts.entries = %i\n", c->fc->nb_streams-1, entries);
1121 for(i=0; i<entries; i++) {
1122 int sample_duration;
1125 sample_count=get_be32(pb);
1126 sample_duration = get_be32(pb);
1127 sc->stts_data[i].count= sample_count;
1128 sc->stts_data[i].duration= sample_duration;
1130 sc->time_rate= ff_gcd(sc->time_rate, sample_duration);
1132 dprintf(c->fc, "sample_count=%d, sample_duration=%d\n",sample_count,sample_duration);
1134 duration+=(int64_t)sample_duration*sample_count;
1135 total_sample_count+=sample_count;
1138 st->nb_frames= total_sample_count;
1140 st->duration= duration;
1144 static int mov_read_ctts(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1146 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1147 MOVStreamContext *sc = st->priv_data;
1148 unsigned int i, entries;
1150 get_byte(pb); /* version */
1151 get_be24(pb); /* flags */
1152 entries = get_be32(pb);
1153 if(entries >= UINT_MAX / sizeof(MOV_stts_t))
1156 sc->ctts_count = entries;
1157 sc->ctts_data = av_malloc(entries * sizeof(MOV_stts_t));
1160 dprintf(c->fc, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
1162 for(i=0; i<entries; i++) {
1163 int count =get_be32(pb);
1164 int duration =get_be32(pb);
1167 av_log(c->fc, AV_LOG_WARNING, "negative ctts, ignoring\n");
1169 url_fskip(pb, 8 * (entries - i - 1));
1172 sc->ctts_data[i].count = count;
1173 sc->ctts_data[i].duration= duration;
1175 sc->time_rate= ff_gcd(sc->time_rate, duration);
1180 static void mov_build_index(MOVContext *mov, AVStream *st)
1182 MOVStreamContext *sc = st->priv_data;
1183 offset_t current_offset;
1184 int64_t current_dts = 0;
1185 unsigned int stts_index = 0;
1186 unsigned int stsc_index = 0;
1187 unsigned int stss_index = 0;
1190 /* only use old uncompressed audio chunk demuxing when stts specifies it */
1191 if (!(st->codec->codec_type == CODEC_TYPE_AUDIO &&
1192 sc->stts_count == 1 && sc->stts_data[0].duration == 1)) {
1193 unsigned int current_sample = 0;
1194 unsigned int stts_sample = 0;
1195 unsigned int keyframe, sample_size;
1196 unsigned int distance = 0;
1197 int key_off = sc->keyframes && sc->keyframes[0] == 1;
1199 st->nb_frames = sc->sample_count;
1200 for (i = 0; i < sc->chunk_count; i++) {
1201 current_offset = sc->chunk_offsets[i];
1202 if (stsc_index + 1 < sc->sample_to_chunk_sz &&
1203 i + 1 == sc->sample_to_chunk[stsc_index + 1].first)
1205 for (j = 0; j < sc->sample_to_chunk[stsc_index].count; j++) {
1206 if (current_sample >= sc->sample_count) {
1207 av_log(mov->fc, AV_LOG_ERROR, "wrong sample count\n");
1210 keyframe = !sc->keyframe_count || current_sample+key_off == sc->keyframes[stss_index];
1213 if (stss_index + 1 < sc->keyframe_count)
1216 sample_size = sc->sample_size > 0 ? sc->sample_size : sc->sample_sizes[current_sample];
1217 if(sc->pseudo_stream_id == -1 ||
1218 sc->sample_to_chunk[stsc_index].id - 1 == sc->pseudo_stream_id) {
1219 av_add_index_entry(st, current_offset, current_dts, sample_size, distance,
1220 keyframe ? AVINDEX_KEYFRAME : 0);
1221 dprintf(mov->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1222 "size %d, distance %d, keyframe %d\n", st->index, current_sample,
1223 current_offset, current_dts, sample_size, distance, keyframe);
1225 current_offset += sample_size;
1226 assert(sc->stts_data[stts_index].duration % sc->time_rate == 0);
1227 current_dts += sc->stts_data[stts_index].duration / sc->time_rate;
1231 if (stts_index + 1 < sc->stts_count && stts_sample == sc->stts_data[stts_index].count) {
1237 } else { /* read whole chunk */
1238 unsigned int chunk_samples, chunk_size, chunk_duration;
1239 unsigned int frames = 1;
1240 for (i = 0; i < sc->chunk_count; i++) {
1241 current_offset = sc->chunk_offsets[i];
1242 if (stsc_index + 1 < sc->sample_to_chunk_sz &&
1243 i + 1 == sc->sample_to_chunk[stsc_index + 1].first)
1245 chunk_samples = sc->sample_to_chunk[stsc_index].count;
1246 /* get chunk size, beware of alaw/ulaw/mace */
1247 if (sc->samples_per_frame > 0 &&
1248 (chunk_samples * sc->bytes_per_frame % sc->samples_per_frame == 0)) {
1249 if (sc->samples_per_frame < 160)
1250 chunk_size = chunk_samples * sc->bytes_per_frame / sc->samples_per_frame;
1252 chunk_size = sc->bytes_per_frame;
1253 frames = chunk_samples / sc->samples_per_frame;
1254 chunk_samples = sc->samples_per_frame;
1257 chunk_size = chunk_samples * sc->sample_size;
1258 for (j = 0; j < frames; j++) {
1259 av_add_index_entry(st, current_offset, current_dts, chunk_size, 0, AVINDEX_KEYFRAME);
1260 /* get chunk duration */
1262 while (chunk_samples > 0) {
1263 if (chunk_samples < sc->stts_data[stts_index].count) {
1264 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1265 sc->stts_data[stts_index].count -= chunk_samples;
1268 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1269 chunk_samples -= sc->stts_data[stts_index].count;
1270 if (stts_index + 1 < sc->stts_count)
1274 current_offset += sc->bytes_per_frame;
1275 dprintf(mov->fc, "AVIndex stream %d, chunk %d, offset %"PRIx64", dts %"PRId64", "
1276 "size %d, duration %d\n", st->index, i, current_offset, current_dts,
1277 chunk_size, chunk_duration);
1278 assert(chunk_duration % sc->time_rate == 0);
1279 current_dts += chunk_duration / sc->time_rate;
1284 /* adjust sample count to avindex entries */
1285 sc->sample_count = st->nb_index_entries;
1288 static int mov_read_trak(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1291 MOVStreamContext *sc;
1294 st = av_new_stream(c->fc, c->fc->nb_streams);
1295 if (!st) return AVERROR(ENOMEM);
1296 sc = av_mallocz(sizeof(MOVStreamContext));
1297 if (!sc) return AVERROR(ENOMEM);
1300 st->codec->codec_type = CODEC_TYPE_DATA;
1301 st->start_time = 0; /* XXX: check */
1303 if ((ret = mov_read_default(c, pb, atom)) < 0)
1307 if(sc->chunk_count && (!sc->stts_count || !sc->sample_to_chunk_sz ||
1308 (!sc->sample_size && !sc->sample_count))){
1309 av_log(c->fc, AV_LOG_ERROR, "stream %d, missing mandatory atoms, broken header\n",
1311 sc->sample_count = 0; //ignore track
1317 sc->time_scale= c->time_scale;
1318 av_set_pts_info(st, 64, sc->time_rate, sc->time_scale);
1320 if (st->codec->codec_type == CODEC_TYPE_AUDIO &&
1321 !st->codec->frame_size && sc->stts_count == 1)
1322 st->codec->frame_size = av_rescale(sc->time_rate, st->codec->sample_rate, sc->time_scale);
1324 if(st->duration != AV_NOPTS_VALUE){
1325 assert(st->duration % sc->time_rate == 0);
1326 st->duration /= sc->time_rate;
1328 sc->ffindex = st->index;
1329 mov_build_index(c, st);
1331 if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
1332 if (url_fopen(&sc->pb, sc->drefs[sc->dref_id-1].path, URL_RDONLY) < 0)
1333 av_log(c->fc, AV_LOG_ERROR, "stream %d, error opening file %s: %s\n",
1334 st->index, sc->drefs[sc->dref_id-1].path, strerror(errno));
1338 switch (st->codec->codec_id) {
1339 #ifdef CONFIG_H261_DECODER
1342 #ifdef CONFIG_H263_DECODER
1345 #ifdef CONFIG_MPEG4_DECODER
1346 case CODEC_ID_MPEG4:
1348 st->codec->width= 0; /* let decoder init width/height */
1349 st->codec->height= 0;
1351 #ifdef CONFIG_VORBIS_DECODER
1352 case CODEC_ID_VORBIS:
1354 st->codec->sample_rate= 0; /* let decoder init parameters properly */
1358 /* Do not need those anymore. */
1359 av_freep(&sc->chunk_offsets);
1360 av_freep(&sc->sample_to_chunk);
1361 av_freep(&sc->sample_sizes);
1362 av_freep(&sc->keyframes);
1363 av_freep(&sc->stts_data);
1368 static void mov_parse_udta_string(ByteIOContext *pb, char *str, int size)
1370 uint16_t str_size = get_be16(pb); /* string length */;
1372 get_be16(pb); /* skip language */
1373 get_buffer(pb, str, FFMIN(size, str_size));
1376 static int mov_read_udta(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1378 uint64_t end = url_ftell(pb) + atom.size;
1380 while (url_ftell(pb) + 8 < end) {
1381 uint32_t tag_size = get_be32(pb);
1382 uint32_t tag = get_le32(pb);
1383 uint64_t next = url_ftell(pb) + tag_size - 8;
1385 if (next > end) // stop if tag_size is wrong
1389 case MKTAG(0xa9,'n','a','m'):
1390 mov_parse_udta_string(pb, c->fc->title, sizeof(c->fc->title));
1392 case MKTAG(0xa9,'w','r','t'):
1393 mov_parse_udta_string(pb, c->fc->author, sizeof(c->fc->author));
1395 case MKTAG(0xa9,'c','p','y'):
1396 mov_parse_udta_string(pb, c->fc->copyright, sizeof(c->fc->copyright));
1398 case MKTAG(0xa9,'i','n','f'):
1399 mov_parse_udta_string(pb, c->fc->comment, sizeof(c->fc->comment));
1405 url_fseek(pb, next, SEEK_SET);
1411 static int mov_read_tkhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1416 int64_t disp_transform[2];
1417 int display_matrix[3][2];
1418 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1419 int version = get_byte(pb);
1421 get_be24(pb); /* flags */
1423 MOV_TRACK_ENABLED 0x0001
1424 MOV_TRACK_IN_MOVIE 0x0002
1425 MOV_TRACK_IN_PREVIEW 0x0004
1426 MOV_TRACK_IN_POSTER 0x0008
1433 get_be32(pb); /* creation time */
1434 get_be32(pb); /* modification time */
1436 st->id = (int)get_be32(pb); /* track id (NOT 0 !)*/
1437 get_be32(pb); /* reserved */
1438 st->start_time = 0; /* check */
1439 /* highlevel (considering edits) duration in movie timebase */
1440 (version == 1) ? get_be64(pb) : get_be32(pb);
1441 get_be32(pb); /* reserved */
1442 get_be32(pb); /* reserved */
1444 get_be16(pb); /* layer */
1445 get_be16(pb); /* alternate group */
1446 get_be16(pb); /* volume */
1447 get_be16(pb); /* reserved */
1449 //read in the display matrix (outlined in ISO 14496-12, Section 6.2.2)
1450 // they're kept in fixed point format through all calculations
1451 // ignore u,v,z b/c we don't need the scale factor to calc aspect ratio
1452 for (i = 0; i < 3; i++) {
1453 display_matrix[i][0] = get_be32(pb); // 16.16 fixed point
1454 display_matrix[i][1] = get_be32(pb); // 16.16 fixed point
1455 get_be32(pb); // 2.30 fixed point (not used)
1458 width = get_be32(pb); // 16.16 fixed point track width
1459 height = get_be32(pb); // 16.16 fixed point track height
1461 //transform the display width/height according to the matrix
1462 // skip this if the display matrix is the default identity matrix
1463 // to keep the same scale, use [width height 1<<16]
1464 if (width && height &&
1465 (display_matrix[0][0] != 65536 || display_matrix[0][1] ||
1466 display_matrix[1][0] || display_matrix[1][1] != 65536 ||
1467 display_matrix[2][0] || display_matrix[2][1])) {
1468 for (i = 0; i < 2; i++)
1470 (int64_t) width * display_matrix[0][i] +
1471 (int64_t) height * display_matrix[1][i] +
1472 ((int64_t) display_matrix[2][i] << 16);
1474 //sample aspect ratio is new width/height divided by old width/height
1475 st->codec->sample_aspect_ratio = av_d2q(
1476 ((double) disp_transform[0] * height) /
1477 ((double) disp_transform[1] * width), INT_MAX);
1482 static int mov_read_tfhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1484 MOVFragment *frag = &c->fragment;
1485 MOVTrackExt *trex = NULL;
1486 int flags, track_id, i;
1488 get_byte(pb); /* version */
1489 flags = get_be24(pb);
1491 track_id = get_be32(pb);
1492 if (!track_id || track_id > c->fc->nb_streams)
1494 frag->track_id = track_id;
1495 for (i = 0; i < c->trex_count; i++)
1496 if (c->trex_data[i].track_id == frag->track_id) {
1497 trex = &c->trex_data[i];
1501 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding trex\n");
1505 if (flags & 0x01) frag->base_data_offset = get_be64(pb);
1506 else frag->base_data_offset = frag->moof_offset;
1507 if (flags & 0x02) frag->stsd_id = get_be32(pb);
1508 else frag->stsd_id = trex->stsd_id;
1510 frag->duration = flags & 0x08 ? get_be32(pb) : trex->duration;
1511 frag->size = flags & 0x10 ? get_be32(pb) : trex->size;
1512 frag->flags = flags & 0x20 ? get_be32(pb) : trex->flags;
1513 dprintf(c->fc, "frag flags 0x%x\n", frag->flags);
1517 static int mov_read_trex(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1521 if ((uint64_t)c->trex_count+1 >= UINT_MAX / sizeof(*c->trex_data))
1523 c->trex_data = av_realloc(c->trex_data, (c->trex_count+1)*sizeof(*c->trex_data));
1525 return AVERROR(ENOMEM);
1526 trex = &c->trex_data[c->trex_count++];
1527 get_byte(pb); /* version */
1528 get_be24(pb); /* flags */
1529 trex->track_id = get_be32(pb);
1530 trex->stsd_id = get_be32(pb);
1531 trex->duration = get_be32(pb);
1532 trex->size = get_be32(pb);
1533 trex->flags = get_be32(pb);
1537 static int mov_read_trun(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1539 MOVFragment *frag = &c->fragment;
1541 MOVStreamContext *sc;
1544 int data_offset = 0;
1545 unsigned entries, first_sample_flags = frag->flags;
1546 int flags, distance, i;
1548 if (!frag->track_id || frag->track_id > c->fc->nb_streams)
1550 st = c->fc->streams[frag->track_id-1];
1552 if (sc->pseudo_stream_id+1 != frag->stsd_id)
1554 get_byte(pb); /* version */
1555 flags = get_be24(pb);
1556 entries = get_be32(pb);
1557 dprintf(c->fc, "flags 0x%x entries %d\n", flags, entries);
1558 if (flags & 0x001) data_offset = get_be32(pb);
1559 if (flags & 0x004) first_sample_flags = get_be32(pb);
1560 if (flags & 0x800) {
1561 if ((uint64_t)entries+sc->ctts_count >= UINT_MAX/sizeof(*sc->ctts_data))
1563 sc->ctts_data = av_realloc(sc->ctts_data,
1564 (entries+sc->ctts_count)*sizeof(*sc->ctts_data));
1566 return AVERROR(ENOMEM);
1569 offset = frag->base_data_offset + data_offset;
1571 dprintf(c->fc, "first sample flags 0x%x\n", first_sample_flags);
1572 for (i = 0; i < entries; i++) {
1573 unsigned sample_size = frag->size;
1574 int sample_flags = i ? frag->flags : first_sample_flags;
1575 unsigned sample_duration = frag->duration;
1578 if (flags & 0x100) sample_duration = get_be32(pb);
1579 if (flags & 0x200) sample_size = get_be32(pb);
1580 if (flags & 0x400) sample_flags = get_be32(pb);
1581 if (flags & 0x800) {
1582 sc->ctts_data[sc->ctts_count].count = 1;
1583 sc->ctts_data[sc->ctts_count].duration = get_be32(pb);
1586 if ((keyframe = st->codec->codec_type == CODEC_TYPE_AUDIO ||
1587 (flags & 0x004 && !i && !sample_flags) || sample_flags & 0x2000000))
1589 av_add_index_entry(st, offset, dts, sample_size, distance,
1590 keyframe ? AVINDEX_KEYFRAME : 0);
1591 dprintf(c->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1592 "size %d, distance %d, keyframe %d\n", st->index, sc->sample_count+i,
1593 offset, dts, sample_size, distance, keyframe);
1595 assert(sample_duration % sc->time_rate == 0);
1596 dts += sample_duration / sc->time_rate;
1597 offset += sample_size;
1599 frag->moof_offset = offset;
1600 sc->sample_count = st->nb_index_entries;
1605 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
1606 /* like the files created with Adobe Premiere 5.0, for samples see */
1607 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
1608 static int mov_read_wide(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1613 return 0; /* continue */
1614 if (get_be32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
1615 url_fskip(pb, atom.size - 4);
1618 atom.type = get_le32(pb);
1621 if (atom.type != MKTAG('m','d','a','t')) {
1622 url_fskip(pb, atom.size);
1625 err = mov_read_mdat(c, pb, atom);
1629 static int mov_read_cmov(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1634 uint8_t *moov_data; /* uncompressed data */
1635 long cmov_len, moov_len;
1638 get_be32(pb); /* dcom atom */
1639 if (get_le32(pb) != MKTAG('d','c','o','m'))
1641 if (get_le32(pb) != MKTAG('z','l','i','b')) {
1642 av_log(NULL, AV_LOG_ERROR, "unknown compression for cmov atom !");
1645 get_be32(pb); /* cmvd atom */
1646 if (get_le32(pb) != MKTAG('c','m','v','d'))
1648 moov_len = get_be32(pb); /* uncompressed size */
1649 cmov_len = atom.size - 6 * 4;
1651 cmov_data = av_malloc(cmov_len);
1654 moov_data = av_malloc(moov_len);
1659 get_buffer(pb, cmov_data, cmov_len);
1660 if(uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
1661 goto free_and_return;
1662 if(init_put_byte(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL, NULL) != 0)
1663 goto free_and_return;
1664 atom.type = MKTAG('m','o','o','v');
1666 atom.size = moov_len;
1668 // { int fd = open("/tmp/uncompheader.mov", O_WRONLY | O_CREAT); write(fd, moov_data, moov_len); close(fd); }
1670 ret = mov_read_default(c, &ctx, atom);
1676 av_log(c->fc, AV_LOG_ERROR, "this file requires zlib support compiled in\n");
1681 /* edit list atom */
1682 static int mov_read_elst(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1684 MOVStreamContext *sc = c->fc->streams[c->fc->nb_streams-1]->priv_data;
1687 get_byte(pb); /* version */
1688 get_be24(pb); /* flags */
1689 edit_count= sc->edit_count = get_be32(pb); /* entries */
1691 for(i=0; i<edit_count; i++){
1693 get_be32(pb); /* Track duration */
1694 time = get_be32(pb); /* Media time */
1695 get_be32(pb); /* Media rate */
1697 av_log(c->fc, AV_LOG_WARNING, "edit list not starting at 0, "
1698 "a/v desync might occur, patch welcome\n");
1700 dprintf(c->fc, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, sc->edit_count);
1704 static const MOVParseTableEntry mov_default_parse_table[] = {
1705 { MKTAG('a','v','s','s'), mov_read_extradata },
1706 { MKTAG('c','o','6','4'), mov_read_stco },
1707 { MKTAG('c','t','t','s'), mov_read_ctts }, /* composition time to sample */
1708 { MKTAG('d','i','n','f'), mov_read_default },
1709 { MKTAG('d','r','e','f'), mov_read_dref },
1710 { MKTAG('e','d','t','s'), mov_read_default },
1711 { MKTAG('e','l','s','t'), mov_read_elst },
1712 { MKTAG('e','n','d','a'), mov_read_enda },
1713 { MKTAG('f','i','e','l'), mov_read_extradata },
1714 { MKTAG('f','t','y','p'), mov_read_ftyp },
1715 { MKTAG('g','l','b','l'), mov_read_glbl },
1716 { MKTAG('h','d','l','r'), mov_read_hdlr },
1717 { MKTAG('j','p','2','h'), mov_read_extradata },
1718 { MKTAG('m','d','a','t'), mov_read_mdat },
1719 { MKTAG('m','d','h','d'), mov_read_mdhd },
1720 { MKTAG('m','d','i','a'), mov_read_default },
1721 { MKTAG('m','i','n','f'), mov_read_default },
1722 { MKTAG('m','o','o','f'), mov_read_moof },
1723 { MKTAG('m','o','o','v'), mov_read_moov },
1724 { MKTAG('m','v','e','x'), mov_read_default },
1725 { MKTAG('m','v','h','d'), mov_read_mvhd },
1726 { MKTAG('S','M','I',' '), mov_read_smi }, /* Sorenson extension ??? */
1727 { MKTAG('a','l','a','c'), mov_read_extradata }, /* alac specific atom */
1728 { MKTAG('a','v','c','C'), mov_read_glbl },
1729 { MKTAG('s','t','b','l'), mov_read_default },
1730 { MKTAG('s','t','c','o'), mov_read_stco },
1731 { MKTAG('s','t','s','c'), mov_read_stsc },
1732 { MKTAG('s','t','s','d'), mov_read_stsd }, /* sample description */
1733 { MKTAG('s','t','s','s'), mov_read_stss }, /* sync sample */
1734 { MKTAG('s','t','s','z'), mov_read_stsz }, /* sample size */
1735 { MKTAG('s','t','t','s'), mov_read_stts },
1736 { MKTAG('t','k','h','d'), mov_read_tkhd }, /* track header */
1737 { MKTAG('t','f','h','d'), mov_read_tfhd }, /* track fragment header */
1738 { MKTAG('t','r','a','k'), mov_read_trak },
1739 { MKTAG('t','r','a','f'), mov_read_default },
1740 { MKTAG('t','r','e','x'), mov_read_trex },
1741 { MKTAG('t','r','u','n'), mov_read_trun },
1742 { MKTAG('u','d','t','a'), mov_read_udta },
1743 { MKTAG('w','a','v','e'), mov_read_wave },
1744 { MKTAG('e','s','d','s'), mov_read_esds },
1745 { MKTAG('w','i','d','e'), mov_read_wide }, /* place holder */
1746 { MKTAG('c','m','o','v'), mov_read_cmov },
1750 static int mov_probe(AVProbeData *p)
1752 unsigned int offset;
1756 /* check file header */
1759 /* ignore invalid offset */
1760 if ((offset + 8) > (unsigned int)p->buf_size)
1762 tag = AV_RL32(p->buf + offset + 4);
1764 /* check for obvious tags */
1765 case MKTAG('j','P',' ',' '): /* jpeg 2000 signature */
1766 case MKTAG('m','o','o','v'):
1767 case MKTAG('m','d','a','t'):
1768 case MKTAG('p','n','o','t'): /* detect movs with preview pics like ew.mov and april.mov */
1769 case MKTAG('u','d','t','a'): /* Packet Video PVAuthor adds this and a lot of more junk */
1770 case MKTAG('f','t','y','p'):
1771 return AVPROBE_SCORE_MAX;
1772 /* those are more common words, so rate then a bit less */
1773 case MKTAG('e','d','i','w'): /* xdcam files have reverted first tags */
1774 case MKTAG('w','i','d','e'):
1775 case MKTAG('f','r','e','e'):
1776 case MKTAG('j','u','n','k'):
1777 case MKTAG('p','i','c','t'):
1778 return AVPROBE_SCORE_MAX - 5;
1779 case MKTAG(0x82,0x82,0x7f,0x7d):
1780 case MKTAG('s','k','i','p'):
1781 case MKTAG('u','u','i','d'):
1782 case MKTAG('p','r','f','l'):
1783 offset = AV_RB32(p->buf+offset) + offset;
1784 /* if we only find those cause probedata is too small at least rate them */
1785 score = AVPROBE_SCORE_MAX - 50;
1788 /* unrecognized tag */
1795 static int mov_read_header(AVFormatContext *s, AVFormatParameters *ap)
1797 MOVContext *mov = s->priv_data;
1798 ByteIOContext *pb = s->pb;
1800 MOV_atom_t atom = { 0, 0, 0 };
1803 /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
1804 if(!url_is_streamed(pb))
1805 atom.size = url_fsize(pb);
1807 atom.size = INT64_MAX;
1809 /* check MOV header */
1810 if ((err = mov_read_default(mov, pb, atom)) < 0) {
1811 av_log(s, AV_LOG_ERROR, "error reading header: %d\n", err);
1814 if (!mov->found_moov) {
1815 av_log(s, AV_LOG_ERROR, "moov atom not found\n");
1818 dprintf(mov->fc, "on_parse_exit_offset=%lld\n", url_ftell(pb));
1823 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
1825 MOVContext *mov = s->priv_data;
1826 MOVStreamContext *sc = 0;
1827 AVIndexEntry *sample = 0;
1828 int64_t best_dts = INT64_MAX;
1831 for (i = 0; i < s->nb_streams; i++) {
1832 AVStream *st = s->streams[i];
1833 MOVStreamContext *msc = st->priv_data;
1834 if (st->discard != AVDISCARD_ALL && msc->pb && msc->current_sample < msc->sample_count) {
1835 AVIndexEntry *current_sample = &st->index_entries[msc->current_sample];
1836 int64_t dts = av_rescale(current_sample->timestamp * (int64_t)msc->time_rate,
1837 AV_TIME_BASE, msc->time_scale);
1838 dprintf(s, "stream %d, sample %d, dts %"PRId64"\n", i, msc->current_sample, dts);
1839 if (!sample || (url_is_streamed(s->pb) && current_sample->pos < sample->pos) ||
1840 (!url_is_streamed(s->pb) &&
1841 ((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
1842 ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
1843 (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
1844 sample = current_sample;
1851 mov->found_mdat = 0;
1852 if (!url_is_streamed(s->pb) ||
1853 mov_read_default(mov, s->pb, (MOV_atom_t){ 0, 0, INT64_MAX }) < 0 ||
1856 dprintf(s, "read fragments, offset 0x%llx\n", url_ftell(s->pb));
1859 /* must be done just before reading, to avoid infinite loop on sample */
1860 sc->current_sample++;
1861 if (url_fseek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
1862 av_log(mov->fc, AV_LOG_ERROR, "stream %d, offset 0x%"PRIx64": partial file\n",
1863 sc->ffindex, sample->pos);
1866 av_get_packet(sc->pb, pkt, sample->size);
1867 #ifdef CONFIG_DV_DEMUXER
1868 if (mov->dv_demux && sc->dv_audio_container) {
1869 dv_produce_packet(mov->dv_demux, pkt, pkt->data, pkt->size);
1872 if (dv_get_packet(mov->dv_demux, pkt) < 0)
1876 pkt->stream_index = sc->ffindex;
1877 pkt->dts = sample->timestamp;
1878 if (sc->ctts_data) {
1879 assert(sc->ctts_data[sc->sample_to_ctime_index].duration % sc->time_rate == 0);
1880 pkt->pts = pkt->dts + sc->ctts_data[sc->sample_to_ctime_index].duration / sc->time_rate;
1881 /* update ctts context */
1882 sc->sample_to_ctime_sample++;
1883 if (sc->sample_to_ctime_index < sc->ctts_count &&
1884 sc->ctts_data[sc->sample_to_ctime_index].count == sc->sample_to_ctime_sample) {
1885 sc->sample_to_ctime_index++;
1886 sc->sample_to_ctime_sample = 0;
1889 AVStream *st = s->streams[sc->ffindex];
1890 int64_t next_dts = (sc->current_sample < sc->sample_count) ?
1891 st->index_entries[sc->current_sample].timestamp : st->duration;
1892 pkt->duration = next_dts - pkt->dts;
1893 pkt->pts = pkt->dts;
1895 pkt->flags |= sample->flags & AVINDEX_KEYFRAME ? PKT_FLAG_KEY : 0;
1896 pkt->pos = sample->pos;
1897 dprintf(s, "stream %d, pts %"PRId64", dts %"PRId64", pos 0x%"PRIx64", duration %d\n",
1898 pkt->stream_index, pkt->pts, pkt->dts, pkt->pos, pkt->duration);
1902 static int mov_seek_stream(AVStream *st, int64_t timestamp, int flags)
1904 MOVStreamContext *sc = st->priv_data;
1905 int sample, time_sample;
1908 sample = av_index_search_timestamp(st, timestamp, flags);
1909 dprintf(st->codec, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
1910 if (sample < 0) /* not sure what to do */
1912 sc->current_sample = sample;
1913 dprintf(st->codec, "stream %d, found sample %d\n", st->index, sc->current_sample);
1914 /* adjust ctts index */
1915 if (sc->ctts_data) {
1917 for (i = 0; i < sc->ctts_count; i++) {
1918 int next = time_sample + sc->ctts_data[i].count;
1919 if (next > sc->current_sample) {
1920 sc->sample_to_ctime_index = i;
1921 sc->sample_to_ctime_sample = sc->current_sample - time_sample;
1930 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
1933 int64_t seek_timestamp, timestamp;
1937 if (stream_index >= s->nb_streams)
1940 st = s->streams[stream_index];
1941 sample = mov_seek_stream(st, sample_time, flags);
1945 /* adjust seek timestamp to found sample timestamp */
1946 seek_timestamp = st->index_entries[sample].timestamp;
1948 for (i = 0; i < s->nb_streams; i++) {
1950 if (stream_index == i || st->discard == AVDISCARD_ALL)
1953 timestamp = av_rescale_q(seek_timestamp, s->streams[stream_index]->time_base, st->time_base);
1954 mov_seek_stream(st, timestamp, flags);
1959 static int mov_read_close(AVFormatContext *s)
1962 MOVContext *mov = s->priv_data;
1963 for(i=0; i<s->nb_streams; i++) {
1964 MOVStreamContext *sc = s->streams[i]->priv_data;
1965 av_freep(&sc->ctts_data);
1966 for (j=0; j<sc->drefs_count; j++)
1967 av_freep(&sc->drefs[j].path);
1968 av_freep(&sc->drefs);
1969 if (sc->pb && sc->pb != s->pb)
1973 for(i=0; i<mov->dv_fctx->nb_streams; i++){
1974 av_freep(&mov->dv_fctx->streams[i]->codec);
1975 av_freep(&mov->dv_fctx->streams[i]);
1977 av_freep(&mov->dv_fctx);
1978 av_freep(&mov->dv_demux);
1980 av_freep(&mov->trex_data);
1984 AVInputFormat mov_demuxer = {
1985 "mov,mp4,m4a,3gp,3g2,mj2",
1986 NULL_IF_CONFIG_SMALL("QuickTime/MPEG-4/Motion JPEG 2000 format"),