3 * Copyright (c) 2001 Fabrice Bellard
4 * Copyright (c) 2009 Baptiste Coudurier <baptiste dot coudurier at gmail dot com>
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
26 //#define DEBUG_METADATA
27 //#define MOV_EXPORT_ALL_METADATA
29 #include "libavutil/intreadwrite.h"
30 #include "libavutil/avstring.h"
34 #include "libavcodec/mpeg4audio.h"
35 #include "libavcodec/mpegaudiodata.h"
36 #include "libavcodec/get_bits.h"
43 * First version by Francois Revol revol@free.fr
44 * Seek function by Gael Chardon gael.dev@4now.net
46 * Features and limitations:
47 * - reads most of the QT files I have (at least the structure),
48 * Sample QuickTime files with mp3 audio can be found at: http://www.3ivx.com/showcase.html
49 * - the code is quite ugly... maybe I won't do it recursive next time :-)
51 * Funny I didn't know about http://sourceforge.net/projects/qt-ffmpeg/
52 * when coding this :) (it's a writer anyway)
54 * Reference documents:
55 * http://www.geocities.com/xhelmboyx/quicktime/formats/qtm-layout.txt
57 * http://developer.apple.com/documentation/QuickTime/QTFF/
58 * http://developer.apple.com/documentation/QuickTime/QTFF/qtff.pdf
59 * QuickTime is a trademark of Apple (AFAIK :))
62 #include "qtpalette.h"
68 /* XXX: it's the first time I make a recursive parser I think... sorry if it's ugly :P */
70 /* those functions parse an atom */
72 0: continue to parse next atom
73 <0: error occurred, exit
75 /* links atom IDs to parse functions */
76 typedef struct MOVParseTableEntry {
78 int (*parse)(MOVContext *ctx, ByteIOContext *pb, MOVAtom atom);
81 static const MOVParseTableEntry mov_default_parse_table[];
83 static int mov_metadata_trkn(MOVContext *c, ByteIOContext *pb, unsigned len)
87 get_be16(pb); // unknown
88 snprintf(buf, sizeof(buf), "%d", get_be16(pb));
89 av_metadata_set2(&c->fc->metadata, "track", buf, 0);
91 get_be16(pb); // total tracks
96 static const uint32_t mac_to_unicode[128] = {
97 0x00C4,0x00C5,0x00C7,0x00C9,0x00D1,0x00D6,0x00DC,0x00E1,
98 0x00E0,0x00E2,0x00E4,0x00E3,0x00E5,0x00E7,0x00E9,0x00E8,
99 0x00EA,0x00EB,0x00ED,0x00EC,0x00EE,0x00EF,0x00F1,0x00F3,
100 0x00F2,0x00F4,0x00F6,0x00F5,0x00FA,0x00F9,0x00FB,0x00FC,
101 0x2020,0x00B0,0x00A2,0x00A3,0x00A7,0x2022,0x00B6,0x00DF,
102 0x00AE,0x00A9,0x2122,0x00B4,0x00A8,0x2260,0x00C6,0x00D8,
103 0x221E,0x00B1,0x2264,0x2265,0x00A5,0x00B5,0x2202,0x2211,
104 0x220F,0x03C0,0x222B,0x00AA,0x00BA,0x03A9,0x00E6,0x00F8,
105 0x00BF,0x00A1,0x00AC,0x221A,0x0192,0x2248,0x2206,0x00AB,
106 0x00BB,0x2026,0x00A0,0x00C0,0x00C3,0x00D5,0x0152,0x0153,
107 0x2013,0x2014,0x201C,0x201D,0x2018,0x2019,0x00F7,0x25CA,
108 0x00FF,0x0178,0x2044,0x20AC,0x2039,0x203A,0xFB01,0xFB02,
109 0x2021,0x00B7,0x201A,0x201E,0x2030,0x00C2,0x00CA,0x00C1,
110 0x00CB,0x00C8,0x00CD,0x00CE,0x00CF,0x00CC,0x00D3,0x00D4,
111 0xF8FF,0x00D2,0x00DA,0x00DB,0x00D9,0x0131,0x02C6,0x02DC,
112 0x00AF,0x02D8,0x02D9,0x02DA,0x00B8,0x02DD,0x02DB,0x02C7,
115 static int mov_read_mac_string(MOVContext *c, ByteIOContext *pb, int len,
116 char *dst, int dstlen)
119 char *end = dst+dstlen-1;
122 for (i = 0; i < len; i++) {
123 uint8_t t, c = get_byte(pb);
124 if (c < 0x80 && p < end)
127 PUT_UTF8(mac_to_unicode[c-0x80], t, if (p < end) *p++ = t;);
133 static int mov_read_udta_string(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
135 #ifdef MOV_EXPORT_ALL_METADATA
138 char str[1024], key2[16], language[4] = {0};
139 const char *key = NULL;
140 uint16_t str_size, langcode = 0;
141 uint32_t data_type = 0;
142 int (*parse)(MOVContext*, ByteIOContext*, unsigned) = NULL;
145 case MKTAG(0xa9,'n','a','m'): key = "title"; break;
146 case MKTAG(0xa9,'a','u','t'):
147 case MKTAG(0xa9,'A','R','T'): key = "artist"; break;
148 case MKTAG(0xa9,'w','r','t'): key = "composer"; break;
149 case MKTAG( 'c','p','r','t'):
150 case MKTAG(0xa9,'c','p','y'): key = "copyright"; break;
151 case MKTAG(0xa9,'c','m','t'):
152 case MKTAG(0xa9,'i','n','f'): key = "comment"; break;
153 case MKTAG(0xa9,'a','l','b'): key = "album"; break;
154 case MKTAG(0xa9,'d','a','y'): key = "date"; break;
155 case MKTAG(0xa9,'g','e','n'): key = "genre"; break;
156 case MKTAG(0xa9,'t','o','o'):
157 case MKTAG(0xa9,'e','n','c'): key = "encoder"; break;
158 case MKTAG( 'd','e','s','c'): key = "description";break;
159 case MKTAG( 'l','d','e','s'): key = "synopsis"; break;
160 case MKTAG( 't','v','s','h'): key = "show"; break;
161 case MKTAG( 't','v','e','n'): key = "episode_id";break;
162 case MKTAG( 't','v','n','n'): key = "network"; break;
163 case MKTAG( 't','r','k','n'): key = "track";
164 parse = mov_metadata_trkn; break;
167 if (c->itunes_metadata && atom.size > 8) {
168 int data_size = get_be32(pb);
169 int tag = get_le32(pb);
170 if (tag == MKTAG('d','a','t','a')) {
171 data_type = get_be32(pb); // type
172 get_be32(pb); // unknown
173 str_size = data_size - 16;
176 } else if (atom.size > 4 && key && !c->itunes_metadata) {
177 str_size = get_be16(pb); // string length
178 langcode = get_be16(pb);
179 ff_mov_lang_to_iso639(langcode, language);
182 str_size = atom.size;
184 #ifdef MOV_EXPORT_ALL_METADATA
186 snprintf(tmp_key, 5, "%.4s", (char*)&atom.type);
196 str_size = FFMIN3(sizeof(str)-1, str_size, atom.size);
199 parse(c, pb, str_size);
201 if (data_type == 3 || (data_type == 0 && langcode < 0x800)) { // MAC Encoded
202 mov_read_mac_string(c, pb, str_size, str, sizeof(str));
204 get_buffer(pb, str, str_size);
207 av_metadata_set2(&c->fc->metadata, key, str, 0);
208 if (*language && strcmp(language, "und")) {
209 snprintf(key2, sizeof(key2), "%s-%s", key, language);
210 av_metadata_set2(&c->fc->metadata, key2, str, 0);
213 #ifdef DEBUG_METADATA
214 av_log(c->fc, AV_LOG_DEBUG, "lang \"%3s\" ", language);
215 av_log(c->fc, AV_LOG_DEBUG, "tag \"%s\" value \"%s\" atom \"%.4s\" %d %lld\n",
216 key, str, (char*)&atom.type, str_size, atom.size);
222 static int mov_read_chpl(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
225 int i, nb_chapters, str_len, version;
228 if ((atom.size -= 5) < 0)
231 version = get_byte(pb);
235 nb_chapters = get_byte(pb);
237 for (i = 0; i < nb_chapters; i++) {
241 start = get_be64(pb);
242 str_len = get_byte(pb);
244 if ((atom.size -= 9+str_len) < 0)
247 get_buffer(pb, str, str_len);
249 ff_new_chapter(c->fc, i, (AVRational){1,10000000}, start, AV_NOPTS_VALUE, str);
254 static int mov_read_default(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
256 int64_t total_size = 0;
261 atom.size = INT64_MAX;
262 while (total_size + 8 < atom.size && !url_feof(pb)) {
263 int (*parse)(MOVContext*, ByteIOContext*, MOVAtom) = NULL;
267 a.size = get_be32(pb);
268 a.type = get_le32(pb);
270 dprintf(c->fc, "type: %08x '%.4s' parent:'%.4s' sz: %"PRId64" %"PRId64" %"PRId64"\n",
271 a.type, (char*)&a.type, (char*)&atom.type, a.size, total_size, atom.size);
273 if (a.size == 1) { /* 64 bit extended size */
274 a.size = get_be64(pb) - 8;
278 a.size = atom.size - total_size;
285 a.size = FFMIN(a.size, atom.size - total_size);
287 for (i = 0; mov_default_parse_table[i].type; i++)
288 if (mov_default_parse_table[i].type == a.type) {
289 parse = mov_default_parse_table[i].parse;
293 // container is user data
294 if (!parse && (atom.type == MKTAG('u','d','t','a') ||
295 atom.type == MKTAG('i','l','s','t')))
296 parse = mov_read_udta_string;
298 if (!parse) { /* skip leaf atoms data */
299 url_fskip(pb, a.size);
301 int64_t start_pos = url_ftell(pb);
303 int err = parse(c, pb, a);
306 if (c->found_moov && c->found_mdat &&
307 (url_is_streamed(pb) || start_pos + a.size == url_fsize(pb)))
309 left = a.size - url_ftell(pb) + start_pos;
310 if (left > 0) /* skip garbage at atom end */
314 total_size += a.size;
317 if (total_size < atom.size && atom.size < 0x7ffff)
318 url_fskip(pb, atom.size - total_size);
323 static int mov_read_dref(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
326 MOVStreamContext *sc;
329 if (c->fc->nb_streams < 1)
331 st = c->fc->streams[c->fc->nb_streams-1];
334 get_be32(pb); // version + flags
335 entries = get_be32(pb);
336 if (entries >= UINT_MAX / sizeof(*sc->drefs))
338 sc->drefs = av_mallocz(entries * sizeof(*sc->drefs));
340 return AVERROR(ENOMEM);
341 sc->drefs_count = entries;
343 for (i = 0; i < sc->drefs_count; i++) {
344 MOVDref *dref = &sc->drefs[i];
345 uint32_t size = get_be32(pb);
346 int64_t next = url_ftell(pb) + size - 4;
351 dref->type = get_le32(pb);
352 get_be32(pb); // version + flags
353 dprintf(c->fc, "type %.4s size %d\n", (char*)&dref->type, size);
355 if (dref->type == MKTAG('a','l','i','s') && size > 150) {
356 /* macintosh alias record */
357 uint16_t volume_len, len;
362 volume_len = get_byte(pb);
363 volume_len = FFMIN(volume_len, 27);
364 get_buffer(pb, dref->volume, 27);
365 dref->volume[volume_len] = 0;
366 av_log(c->fc, AV_LOG_DEBUG, "volume %s, len %d\n", dref->volume, volume_len);
371 len = FFMIN(len, 63);
372 get_buffer(pb, dref->filename, 63);
373 dref->filename[len] = 0;
374 av_log(c->fc, AV_LOG_DEBUG, "filename %s, len %d\n", dref->filename, len);
378 /* read next level up_from_alias/down_to_target */
379 dref->nlvl_from = get_be16(pb);
380 dref->nlvl_to = get_be16(pb);
381 av_log(c->fc, AV_LOG_DEBUG, "nlvl from %d, nlvl to %d\n",
382 dref->nlvl_from, dref->nlvl_to);
386 for (type = 0; type != -1 && url_ftell(pb) < next; ) {
389 av_log(c->fc, AV_LOG_DEBUG, "type %d, len %d\n", type, len);
392 if (type == 2) { // absolute path
394 dref->path = av_mallocz(len+1);
396 return AVERROR(ENOMEM);
397 get_buffer(pb, dref->path, len);
398 if (len > volume_len && !strncmp(dref->path, dref->volume, volume_len)) {
400 memmove(dref->path, dref->path+volume_len, len);
403 for (j = 0; j < len; j++)
404 if (dref->path[j] == ':')
406 av_log(c->fc, AV_LOG_DEBUG, "path %s\n", dref->path);
407 } else if (type == 0) { // directory name
409 dref->dir = av_malloc(len+1);
411 return AVERROR(ENOMEM);
412 get_buffer(pb, dref->dir, len);
414 for (j = 0; j < len; j++)
415 if (dref->dir[j] == ':')
417 av_log(c->fc, AV_LOG_DEBUG, "dir %s\n", dref->dir);
422 url_fseek(pb, next, SEEK_SET);
427 static int mov_read_hdlr(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
433 if (c->fc->nb_streams < 1) // meta before first trak
436 st = c->fc->streams[c->fc->nb_streams-1];
438 get_byte(pb); /* version */
439 get_be24(pb); /* flags */
442 ctype = get_le32(pb);
443 type = get_le32(pb); /* component subtype */
445 dprintf(c->fc, "ctype= %.4s (0x%08x)\n", (char*)&ctype, ctype);
446 dprintf(c->fc, "stype= %.4s\n", (char*)&type);
448 if (type == MKTAG('v','i','d','e'))
449 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
450 else if(type == MKTAG('s','o','u','n'))
451 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
452 else if(type == MKTAG('m','1','a',' '))
453 st->codec->codec_id = CODEC_ID_MP2;
454 else if(type == MKTAG('s','u','b','p'))
455 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
457 get_be32(pb); /* component manufacture */
458 get_be32(pb); /* component flags */
459 get_be32(pb); /* component flags mask */
464 int ff_mp4_read_descr_len(ByteIOContext *pb)
469 int c = get_byte(pb);
470 len = (len << 7) | (c & 0x7f);
477 static int mp4_read_descr(AVFormatContext *fc, ByteIOContext *pb, int *tag)
481 len = ff_mp4_read_descr_len(pb);
482 dprintf(fc, "MPEG4 description: tag=0x%02x len=%d\n", *tag, len);
486 #define MP4ESDescrTag 0x03
487 #define MP4DecConfigDescrTag 0x04
488 #define MP4DecSpecificDescrTag 0x05
490 static const AVCodecTag mp4_audio_types[] = {
491 { CODEC_ID_MP3ON4, AOT_PS }, /* old mp3on4 draft */
492 { CODEC_ID_MP3ON4, AOT_L1 }, /* layer 1 */
493 { CODEC_ID_MP3ON4, AOT_L2 }, /* layer 2 */
494 { CODEC_ID_MP3ON4, AOT_L3 }, /* layer 3 */
495 { CODEC_ID_MP4ALS, AOT_ALS }, /* MPEG-4 ALS */
496 { CODEC_ID_NONE, AOT_NULL },
499 int ff_mov_read_esds(AVFormatContext *fc, ByteIOContext *pb, MOVAtom atom)
504 if (fc->nb_streams < 1)
506 st = fc->streams[fc->nb_streams-1];
508 get_be32(pb); /* version + flags */
509 len = mp4_read_descr(fc, pb, &tag);
510 if (tag == MP4ESDescrTag) {
511 get_be16(pb); /* ID */
512 get_byte(pb); /* priority */
514 get_be16(pb); /* ID */
516 len = mp4_read_descr(fc, pb, &tag);
517 if (tag == MP4DecConfigDescrTag) {
518 int object_type_id = get_byte(pb);
519 get_byte(pb); /* stream type */
520 get_be24(pb); /* buffer size db */
521 get_be32(pb); /* max bitrate */
522 get_be32(pb); /* avg bitrate */
524 st->codec->codec_id= ff_codec_get_id(ff_mp4_obj_type, object_type_id);
525 dprintf(fc, "esds object type id 0x%02x\n", object_type_id);
526 len = mp4_read_descr(fc, pb, &tag);
527 if (tag == MP4DecSpecificDescrTag) {
528 dprintf(fc, "Specific MPEG4 header len=%d\n", len);
529 if((uint64_t)len > (1<<30))
531 av_free(st->codec->extradata);
532 st->codec->extradata = av_mallocz(len + FF_INPUT_BUFFER_PADDING_SIZE);
533 if (!st->codec->extradata)
534 return AVERROR(ENOMEM);
535 get_buffer(pb, st->codec->extradata, len);
536 st->codec->extradata_size = len;
537 if (st->codec->codec_id == CODEC_ID_AAC) {
538 MPEG4AudioConfig cfg;
539 ff_mpeg4audio_get_config(&cfg, st->codec->extradata,
540 st->codec->extradata_size);
541 st->codec->channels = cfg.channels;
542 if (cfg.object_type == 29 && cfg.sampling_index < 3) // old mp3on4
543 st->codec->sample_rate = ff_mpa_freq_tab[cfg.sampling_index];
544 else if (cfg.ext_sample_rate)
545 st->codec->sample_rate = cfg.ext_sample_rate;
547 st->codec->sample_rate = cfg.sample_rate;
548 dprintf(fc, "mp4a config channels %d obj %d ext obj %d "
549 "sample rate %d ext sample rate %d\n", st->codec->channels,
550 cfg.object_type, cfg.ext_object_type,
551 cfg.sample_rate, cfg.ext_sample_rate);
552 if (!(st->codec->codec_id = ff_codec_get_id(mp4_audio_types,
554 st->codec->codec_id = CODEC_ID_AAC;
561 static int mov_read_esds(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
563 return ff_mov_read_esds(c->fc, pb, atom);
566 static int mov_read_dac3(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
569 int ac3info, acmod, lfeon;
571 if (c->fc->nb_streams < 1)
573 st = c->fc->streams[c->fc->nb_streams-1];
575 ac3info = get_be24(pb);
576 acmod = (ac3info >> 11) & 0x7;
577 lfeon = (ac3info >> 10) & 0x1;
578 st->codec->channels = ((int[]){2,1,2,3,3,4,4,5})[acmod] + lfeon;
583 static int mov_read_pasp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
585 const int num = get_be32(pb);
586 const int den = get_be32(pb);
589 if (c->fc->nb_streams < 1)
591 st = c->fc->streams[c->fc->nb_streams-1];
593 if ((st->sample_aspect_ratio.den != 1 || st->sample_aspect_ratio.num) && // default
594 (den != st->sample_aspect_ratio.den || num != st->sample_aspect_ratio.num)) {
595 av_log(c->fc, AV_LOG_WARNING,
596 "sample aspect ratio already set to %d:%d, ignoring 'pasp' atom (%d:%d)\n",
597 st->sample_aspect_ratio.num, st->sample_aspect_ratio.den,
599 } else if (den != 0) {
600 st->sample_aspect_ratio.num = num;
601 st->sample_aspect_ratio.den = den;
606 /* this atom contains actual media data */
607 static int mov_read_mdat(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
609 if(atom.size == 0) /* wrong one (MP4) */
612 return 0; /* now go for moov */
615 /* read major brand, minor version and compatible brands and store them as metadata */
616 static int mov_read_ftyp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
620 char minor_ver_str[11]; /* 32 bit integer -> 10 digits + null */
621 char* comp_brands_str;
622 uint8_t type[5] = {0};
624 get_buffer(pb, type, 4);
625 if (strcmp(type, "qt "))
627 av_log(c->fc, AV_LOG_DEBUG, "ISO: File Type Major Brand: %.4s\n",(char *)&type);
628 av_metadata_set2(&c->fc->metadata, "major_brand", type, 0);
629 minor_ver = get_be32(pb); /* minor version */
630 snprintf(minor_ver_str, sizeof(minor_ver_str), "%d", minor_ver);
631 av_metadata_set2(&c->fc->metadata, "minor_version", minor_ver_str, 0);
633 comp_brand_size = atom.size - 8;
634 if (comp_brand_size < 0)
636 comp_brands_str = av_malloc(comp_brand_size + 1); /* Add null terminator */
637 if (!comp_brands_str)
638 return AVERROR(ENOMEM);
639 get_buffer(pb, comp_brands_str, comp_brand_size);
640 comp_brands_str[comp_brand_size] = 0;
641 av_metadata_set2(&c->fc->metadata, "compatible_brands", comp_brands_str, 0);
642 av_freep(&comp_brands_str);
647 /* this atom should contain all header atoms */
648 static int mov_read_moov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
650 if (mov_read_default(c, pb, atom) < 0)
652 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
653 /* so we don't parse the whole file if over a network */
655 return 0; /* now go for mdat */
658 static int mov_read_moof(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
660 c->fragment.moof_offset = url_ftell(pb) - 8;
661 dprintf(c->fc, "moof offset %llx\n", c->fragment.moof_offset);
662 return mov_read_default(c, pb, atom);
665 static int mov_read_mdhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
668 MOVStreamContext *sc;
670 char language[4] = {0};
673 if (c->fc->nb_streams < 1)
675 st = c->fc->streams[c->fc->nb_streams-1];
678 version = get_byte(pb);
680 return -1; /* unsupported */
682 get_be24(pb); /* flags */
687 get_be32(pb); /* creation time */
688 get_be32(pb); /* modification time */
691 sc->time_scale = get_be32(pb);
692 st->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
694 lang = get_be16(pb); /* language */
695 if (ff_mov_lang_to_iso639(lang, language))
696 av_metadata_set2(&st->metadata, "language", language, 0);
697 get_be16(pb); /* quality */
702 static int mov_read_mvhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
704 int version = get_byte(pb); /* version */
705 get_be24(pb); /* flags */
711 get_be32(pb); /* creation time */
712 get_be32(pb); /* modification time */
714 c->time_scale = get_be32(pb); /* time scale */
716 dprintf(c->fc, "time scale = %i\n", c->time_scale);
718 c->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
719 get_be32(pb); /* preferred scale */
721 get_be16(pb); /* preferred volume */
723 url_fskip(pb, 10); /* reserved */
725 url_fskip(pb, 36); /* display matrix */
727 get_be32(pb); /* preview time */
728 get_be32(pb); /* preview duration */
729 get_be32(pb); /* poster time */
730 get_be32(pb); /* selection time */
731 get_be32(pb); /* selection duration */
732 get_be32(pb); /* current time */
733 get_be32(pb); /* next track ID */
738 static int mov_read_smi(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
742 if (c->fc->nb_streams < 1)
744 st = c->fc->streams[c->fc->nb_streams-1];
746 if((uint64_t)atom.size > (1<<30))
749 // currently SVQ3 decoder expect full STSD header - so let's fake it
750 // this should be fixed and just SMI header should be passed
751 av_free(st->codec->extradata);
752 st->codec->extradata = av_mallocz(atom.size + 0x5a + FF_INPUT_BUFFER_PADDING_SIZE);
753 if (!st->codec->extradata)
754 return AVERROR(ENOMEM);
755 st->codec->extradata_size = 0x5a + atom.size;
756 memcpy(st->codec->extradata, "SVQ3", 4); // fake
757 get_buffer(pb, st->codec->extradata + 0x5a, atom.size);
758 dprintf(c->fc, "Reading SMI %"PRId64" %s\n", atom.size, st->codec->extradata + 0x5a);
762 static int mov_read_enda(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
767 if (c->fc->nb_streams < 1)
769 st = c->fc->streams[c->fc->nb_streams-1];
771 little_endian = get_be16(pb);
772 dprintf(c->fc, "enda %d\n", little_endian);
773 if (little_endian == 1) {
774 switch (st->codec->codec_id) {
775 case CODEC_ID_PCM_S24BE:
776 st->codec->codec_id = CODEC_ID_PCM_S24LE;
778 case CODEC_ID_PCM_S32BE:
779 st->codec->codec_id = CODEC_ID_PCM_S32LE;
781 case CODEC_ID_PCM_F32BE:
782 st->codec->codec_id = CODEC_ID_PCM_F32LE;
784 case CODEC_ID_PCM_F64BE:
785 st->codec->codec_id = CODEC_ID_PCM_F64LE;
794 /* FIXME modify qdm2/svq3/h264 decoders to take full atom as extradata */
795 static int mov_read_extradata(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
801 if (c->fc->nb_streams < 1) // will happen with jp2 files
803 st= c->fc->streams[c->fc->nb_streams-1];
804 size= (uint64_t)st->codec->extradata_size + atom.size + 8 + FF_INPUT_BUFFER_PADDING_SIZE;
805 if(size > INT_MAX || (uint64_t)atom.size > INT_MAX)
807 buf= av_realloc(st->codec->extradata, size);
810 st->codec->extradata= buf;
811 buf+= st->codec->extradata_size;
812 st->codec->extradata_size= size - FF_INPUT_BUFFER_PADDING_SIZE;
813 AV_WB32( buf , atom.size + 8);
814 AV_WL32( buf + 4, atom.type);
815 get_buffer(pb, buf + 8, atom.size);
819 static int mov_read_wave(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
823 if (c->fc->nb_streams < 1)
825 st = c->fc->streams[c->fc->nb_streams-1];
827 if((uint64_t)atom.size > (1<<30))
830 if (st->codec->codec_id == CODEC_ID_QDM2) {
831 // pass all frma atom to codec, needed at least for QDM2
832 av_free(st->codec->extradata);
833 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
834 if (!st->codec->extradata)
835 return AVERROR(ENOMEM);
836 st->codec->extradata_size = atom.size;
837 get_buffer(pb, st->codec->extradata, atom.size);
838 } else if (atom.size > 8) { /* to read frma, esds atoms */
839 if (mov_read_default(c, pb, atom) < 0)
842 url_fskip(pb, atom.size);
847 * This function reads atom content and puts data in extradata without tag
848 * nor size unlike mov_read_extradata.
850 static int mov_read_glbl(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
854 if (c->fc->nb_streams < 1)
856 st = c->fc->streams[c->fc->nb_streams-1];
858 if((uint64_t)atom.size > (1<<30))
861 av_free(st->codec->extradata);
862 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
863 if (!st->codec->extradata)
864 return AVERROR(ENOMEM);
865 st->codec->extradata_size = atom.size;
866 get_buffer(pb, st->codec->extradata, atom.size);
871 * An strf atom is a BITMAPINFOHEADER struct. This struct is 40 bytes itself,
872 * but can have extradata appended at the end after the 40 bytes belonging
875 static int mov_read_strf(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
879 if (c->fc->nb_streams < 1)
883 st = c->fc->streams[c->fc->nb_streams-1];
885 if((uint64_t)atom.size > (1<<30))
888 av_free(st->codec->extradata);
889 st->codec->extradata = av_mallocz(atom.size - 40 + FF_INPUT_BUFFER_PADDING_SIZE);
890 if (!st->codec->extradata)
891 return AVERROR(ENOMEM);
892 st->codec->extradata_size = atom.size - 40;
894 get_buffer(pb, st->codec->extradata, atom.size - 40);
898 static int mov_read_stco(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
901 MOVStreamContext *sc;
902 unsigned int i, entries;
904 if (c->fc->nb_streams < 1)
906 st = c->fc->streams[c->fc->nb_streams-1];
909 get_byte(pb); /* version */
910 get_be24(pb); /* flags */
912 entries = get_be32(pb);
914 if(entries >= UINT_MAX/sizeof(int64_t))
917 sc->chunk_offsets = av_malloc(entries * sizeof(int64_t));
918 if (!sc->chunk_offsets)
919 return AVERROR(ENOMEM);
920 sc->chunk_count = entries;
922 if (atom.type == MKTAG('s','t','c','o'))
923 for(i=0; i<entries; i++)
924 sc->chunk_offsets[i] = get_be32(pb);
925 else if (atom.type == MKTAG('c','o','6','4'))
926 for(i=0; i<entries; i++)
927 sc->chunk_offsets[i] = get_be64(pb);
935 * Compute codec id for 'lpcm' tag.
936 * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
938 enum CodecID ff_mov_get_lpcm_codec_id(int bps, int flags)
940 if (flags & 1) { // floating point
941 if (flags & 2) { // big endian
942 if (bps == 32) return CODEC_ID_PCM_F32BE;
943 else if (bps == 64) return CODEC_ID_PCM_F64BE;
945 if (bps == 32) return CODEC_ID_PCM_F32LE;
946 else if (bps == 64) return CODEC_ID_PCM_F64LE;
952 if (flags & 4) return CODEC_ID_PCM_S8;
953 else return CODEC_ID_PCM_U8;
954 else if (bps == 16) return CODEC_ID_PCM_S16BE;
955 else if (bps == 24) return CODEC_ID_PCM_S24BE;
956 else if (bps == 32) return CODEC_ID_PCM_S32BE;
959 if (flags & 4) return CODEC_ID_PCM_S8;
960 else return CODEC_ID_PCM_U8;
961 else if (bps == 16) return CODEC_ID_PCM_S16LE;
962 else if (bps == 24) return CODEC_ID_PCM_S24LE;
963 else if (bps == 32) return CODEC_ID_PCM_S32LE;
966 return CODEC_ID_NONE;
969 int ff_mov_read_stsd_entries(MOVContext *c, ByteIOContext *pb, int entries)
972 MOVStreamContext *sc;
973 int j, pseudo_stream_id;
975 if (c->fc->nb_streams < 1)
977 st = c->fc->streams[c->fc->nb_streams-1];
980 for(pseudo_stream_id=0; pseudo_stream_id<entries; pseudo_stream_id++) {
981 //Parsing Sample description table
984 MOVAtom a = { AV_RL32("stsd") };
985 int64_t start_pos = url_ftell(pb);
986 int size = get_be32(pb); /* size */
987 uint32_t format = get_le32(pb); /* data format */
990 get_be32(pb); /* reserved */
991 get_be16(pb); /* reserved */
992 dref_id = get_be16(pb);
995 if (st->codec->codec_tag &&
996 st->codec->codec_tag != format &&
997 (c->fc->video_codec_id ? ff_codec_get_id(codec_movvideo_tags, format) != c->fc->video_codec_id
998 : st->codec->codec_tag != MKTAG('j','p','e','g'))
1000 /* Multiple fourcc, we skip JPEG. This is not correct, we should
1001 * export it as a separate AVStream but this needs a few changes
1002 * in the MOV demuxer, patch welcome. */
1003 av_log(c->fc, AV_LOG_WARNING, "multiple fourcc not supported\n");
1004 url_fskip(pb, size - (url_ftell(pb) - start_pos));
1007 sc->pseudo_stream_id = st->codec->codec_tag ? -1 : pseudo_stream_id;
1008 sc->dref_id= dref_id;
1010 st->codec->codec_tag = format;
1011 id = ff_codec_get_id(codec_movaudio_tags, format);
1012 if (id<=0 && ((format&0xFFFF) == 'm'+('s'<<8) || (format&0xFFFF) == 'T'+('S'<<8)))
1013 id = ff_codec_get_id(ff_codec_wav_tags, av_bswap32(format)&0xFFFF);
1015 if (st->codec->codec_type != AVMEDIA_TYPE_VIDEO && id > 0) {
1016 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1017 } else if (st->codec->codec_type != AVMEDIA_TYPE_AUDIO && /* do not overwrite codec type */
1018 format && format != MKTAG('m','p','4','s')) { /* skip old asf mpeg4 tag */
1019 id = ff_codec_get_id(codec_movvideo_tags, format);
1021 id = ff_codec_get_id(ff_codec_bmp_tags, format);
1023 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1024 else if(st->codec->codec_type == AVMEDIA_TYPE_DATA){
1025 id = ff_codec_get_id(ff_codec_movsubtitle_tags, format);
1027 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1031 dprintf(c->fc, "size=%d 4CC= %c%c%c%c codec_type=%d\n", size,
1032 (format >> 0) & 0xff, (format >> 8) & 0xff, (format >> 16) & 0xff,
1033 (format >> 24) & 0xff, st->codec->codec_type);
1035 if(st->codec->codec_type==AVMEDIA_TYPE_VIDEO) {
1036 unsigned int color_depth, len;
1037 int color_greyscale;
1039 st->codec->codec_id = id;
1040 get_be16(pb); /* version */
1041 get_be16(pb); /* revision level */
1042 get_be32(pb); /* vendor */
1043 get_be32(pb); /* temporal quality */
1044 get_be32(pb); /* spatial quality */
1046 st->codec->width = get_be16(pb); /* width */
1047 st->codec->height = get_be16(pb); /* height */
1049 get_be32(pb); /* horiz resolution */
1050 get_be32(pb); /* vert resolution */
1051 get_be32(pb); /* data size, always 0 */
1052 get_be16(pb); /* frames per samples */
1054 len = get_byte(pb); /* codec name, pascal string */
1057 mov_read_mac_string(c, pb, len, st->codec->codec_name, 32);
1059 url_fskip(pb, 31 - len);
1060 /* codec_tag YV12 triggers an UV swap in rawdec.c */
1061 if (!memcmp(st->codec->codec_name, "Planar Y'CbCr 8-bit 4:2:0", 25))
1062 st->codec->codec_tag=MKTAG('I', '4', '2', '0');
1064 st->codec->bits_per_coded_sample = get_be16(pb); /* depth */
1065 st->codec->color_table_id = get_be16(pb); /* colortable id */
1066 dprintf(c->fc, "depth %d, ctab id %d\n",
1067 st->codec->bits_per_coded_sample, st->codec->color_table_id);
1068 /* figure out the palette situation */
1069 color_depth = st->codec->bits_per_coded_sample & 0x1F;
1070 color_greyscale = st->codec->bits_per_coded_sample & 0x20;
1072 /* if the depth is 2, 4, or 8 bpp, file is palettized */
1073 if ((color_depth == 2) || (color_depth == 4) ||
1074 (color_depth == 8)) {
1075 /* for palette traversal */
1076 unsigned int color_start, color_count, color_end;
1077 unsigned char r, g, b;
1079 st->codec->palctrl = av_malloc(sizeof(*st->codec->palctrl));
1080 if (color_greyscale) {
1081 int color_index, color_dec;
1082 /* compute the greyscale palette */
1083 st->codec->bits_per_coded_sample = color_depth;
1084 color_count = 1 << color_depth;
1086 color_dec = 256 / (color_count - 1);
1087 for (j = 0; j < color_count; j++) {
1088 r = g = b = color_index;
1089 st->codec->palctrl->palette[j] =
1090 (r << 16) | (g << 8) | (b);
1091 color_index -= color_dec;
1092 if (color_index < 0)
1095 } else if (st->codec->color_table_id) {
1096 const uint8_t *color_table;
1097 /* if flag bit 3 is set, use the default palette */
1098 color_count = 1 << color_depth;
1099 if (color_depth == 2)
1100 color_table = ff_qt_default_palette_4;
1101 else if (color_depth == 4)
1102 color_table = ff_qt_default_palette_16;
1104 color_table = ff_qt_default_palette_256;
1106 for (j = 0; j < color_count; j++) {
1107 r = color_table[j * 3 + 0];
1108 g = color_table[j * 3 + 1];
1109 b = color_table[j * 3 + 2];
1110 st->codec->palctrl->palette[j] =
1111 (r << 16) | (g << 8) | (b);
1114 /* load the palette from the file */
1115 color_start = get_be32(pb);
1116 color_count = get_be16(pb);
1117 color_end = get_be16(pb);
1118 if ((color_start <= 255) &&
1119 (color_end <= 255)) {
1120 for (j = color_start; j <= color_end; j++) {
1121 /* each R, G, or B component is 16 bits;
1122 * only use the top 8 bits; skip alpha bytes
1132 st->codec->palctrl->palette[j] =
1133 (r << 16) | (g << 8) | (b);
1137 st->codec->palctrl->palette_changed = 1;
1139 } else if(st->codec->codec_type==AVMEDIA_TYPE_AUDIO) {
1140 int bits_per_sample, flags;
1141 uint16_t version = get_be16(pb);
1143 st->codec->codec_id = id;
1144 get_be16(pb); /* revision level */
1145 get_be32(pb); /* vendor */
1147 st->codec->channels = get_be16(pb); /* channel count */
1148 dprintf(c->fc, "audio channels %d\n", st->codec->channels);
1149 st->codec->bits_per_coded_sample = get_be16(pb); /* sample size */
1151 sc->audio_cid = get_be16(pb);
1152 get_be16(pb); /* packet size = 0 */
1154 st->codec->sample_rate = ((get_be32(pb) >> 16));
1156 //Read QT version 1 fields. In version 0 these do not exist.
1157 dprintf(c->fc, "version =%d, isom =%d\n",version,c->isom);
1160 sc->samples_per_frame = get_be32(pb);
1161 get_be32(pb); /* bytes per packet */
1162 sc->bytes_per_frame = get_be32(pb);
1163 get_be32(pb); /* bytes per sample */
1164 } else if(version==2) {
1165 get_be32(pb); /* sizeof struct only */
1166 st->codec->sample_rate = av_int2dbl(get_be64(pb)); /* float 64 */
1167 st->codec->channels = get_be32(pb);
1168 get_be32(pb); /* always 0x7F000000 */
1169 st->codec->bits_per_coded_sample = get_be32(pb); /* bits per channel if sound is uncompressed */
1170 flags = get_be32(pb); /* lpcm format specific flag */
1171 sc->bytes_per_frame = get_be32(pb); /* bytes per audio packet if constant */
1172 sc->samples_per_frame = get_be32(pb); /* lpcm frames per audio packet if constant */
1173 if (format == MKTAG('l','p','c','m'))
1174 st->codec->codec_id = ff_mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, flags);
1178 switch (st->codec->codec_id) {
1179 case CODEC_ID_PCM_S8:
1180 case CODEC_ID_PCM_U8:
1181 if (st->codec->bits_per_coded_sample == 16)
1182 st->codec->codec_id = CODEC_ID_PCM_S16BE;
1184 case CODEC_ID_PCM_S16LE:
1185 case CODEC_ID_PCM_S16BE:
1186 if (st->codec->bits_per_coded_sample == 8)
1187 st->codec->codec_id = CODEC_ID_PCM_S8;
1188 else if (st->codec->bits_per_coded_sample == 24)
1189 st->codec->codec_id =
1190 st->codec->codec_id == CODEC_ID_PCM_S16BE ?
1191 CODEC_ID_PCM_S24BE : CODEC_ID_PCM_S24LE;
1193 /* set values for old format before stsd version 1 appeared */
1194 case CODEC_ID_MACE3:
1195 sc->samples_per_frame = 6;
1196 sc->bytes_per_frame = 2*st->codec->channels;
1198 case CODEC_ID_MACE6:
1199 sc->samples_per_frame = 6;
1200 sc->bytes_per_frame = 1*st->codec->channels;
1202 case CODEC_ID_ADPCM_IMA_QT:
1203 sc->samples_per_frame = 64;
1204 sc->bytes_per_frame = 34*st->codec->channels;
1207 sc->samples_per_frame = 160;
1208 sc->bytes_per_frame = 33;
1214 bits_per_sample = av_get_bits_per_sample(st->codec->codec_id);
1215 if (bits_per_sample) {
1216 st->codec->bits_per_coded_sample = bits_per_sample;
1217 sc->sample_size = (bits_per_sample >> 3) * st->codec->channels;
1219 } else if(st->codec->codec_type==AVMEDIA_TYPE_SUBTITLE){
1220 // ttxt stsd contains display flags, justification, background
1221 // color, fonts, and default styles, so fake an atom to read it
1222 MOVAtom fake_atom = { .size = size - (url_ftell(pb) - start_pos) };
1223 if (format != AV_RL32("mp4s")) // mp4s contains a regular esds atom
1224 mov_read_glbl(c, pb, fake_atom);
1225 st->codec->codec_id= id;
1226 st->codec->width = sc->width;
1227 st->codec->height = sc->height;
1229 /* other codec type, just skip (rtp, mp4s, tmcd ...) */
1230 url_fskip(pb, size - (url_ftell(pb) - start_pos));
1232 /* this will read extra atoms at the end (wave, alac, damr, avcC, SMI ...) */
1233 a.size = size - (url_ftell(pb) - start_pos);
1235 if (mov_read_default(c, pb, a) < 0)
1237 } else if (a.size > 0)
1238 url_fskip(pb, a.size);
1241 if(st->codec->codec_type==AVMEDIA_TYPE_AUDIO && st->codec->sample_rate==0 && sc->time_scale>1)
1242 st->codec->sample_rate= sc->time_scale;
1244 /* special codec parameters handling */
1245 switch (st->codec->codec_id) {
1246 #if CONFIG_DV_DEMUXER
1247 case CODEC_ID_DVAUDIO:
1248 c->dv_fctx = avformat_alloc_context();
1249 c->dv_demux = dv_init_demux(c->dv_fctx);
1251 av_log(c->fc, AV_LOG_ERROR, "dv demux context init error\n");
1254 sc->dv_audio_container = 1;
1255 st->codec->codec_id = CODEC_ID_PCM_S16LE;
1258 /* no ifdef since parameters are always those */
1259 case CODEC_ID_QCELP:
1260 // force sample rate for qcelp when not stored in mov
1261 if (st->codec->codec_tag != MKTAG('Q','c','l','p'))
1262 st->codec->sample_rate = 8000;
1263 st->codec->frame_size= 160;
1264 st->codec->channels= 1; /* really needed */
1266 case CODEC_ID_AMR_NB:
1267 case CODEC_ID_AMR_WB:
1268 st->codec->frame_size= sc->samples_per_frame;
1269 st->codec->channels= 1; /* really needed */
1270 /* force sample rate for amr, stsd in 3gp does not store sample rate */
1271 if (st->codec->codec_id == CODEC_ID_AMR_NB)
1272 st->codec->sample_rate = 8000;
1273 else if (st->codec->codec_id == CODEC_ID_AMR_WB)
1274 st->codec->sample_rate = 16000;
1278 st->codec->codec_type = AVMEDIA_TYPE_AUDIO; /* force type after stsd for m1a hdlr */
1279 st->need_parsing = AVSTREAM_PARSE_FULL;
1282 case CODEC_ID_ADPCM_MS:
1283 case CODEC_ID_ADPCM_IMA_WAV:
1284 st->codec->block_align = sc->bytes_per_frame;
1287 if (st->codec->extradata_size == 36) {
1288 st->codec->frame_size = AV_RB32(st->codec->extradata+12);
1289 st->codec->channels = AV_RB8 (st->codec->extradata+21);
1299 static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1303 get_byte(pb); /* version */
1304 get_be24(pb); /* flags */
1305 entries = get_be32(pb);
1307 return ff_mov_read_stsd_entries(c, pb, entries);
1310 static int mov_read_stsc(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1313 MOVStreamContext *sc;
1314 unsigned int i, entries;
1316 if (c->fc->nb_streams < 1)
1318 st = c->fc->streams[c->fc->nb_streams-1];
1321 get_byte(pb); /* version */
1322 get_be24(pb); /* flags */
1324 entries = get_be32(pb);
1326 dprintf(c->fc, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
1328 if(entries >= UINT_MAX / sizeof(*sc->stsc_data))
1330 sc->stsc_data = av_malloc(entries * sizeof(*sc->stsc_data));
1332 return AVERROR(ENOMEM);
1333 sc->stsc_count = entries;
1335 for(i=0; i<entries; i++) {
1336 sc->stsc_data[i].first = get_be32(pb);
1337 sc->stsc_data[i].count = get_be32(pb);
1338 sc->stsc_data[i].id = get_be32(pb);
1343 static int mov_read_stps(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1346 MOVStreamContext *sc;
1347 unsigned i, entries;
1349 if (c->fc->nb_streams < 1)
1351 st = c->fc->streams[c->fc->nb_streams-1];
1354 get_be32(pb); // version + flags
1356 entries = get_be32(pb);
1357 if (entries >= UINT_MAX / sizeof(*sc->stps_data))
1359 sc->stps_data = av_malloc(entries * sizeof(*sc->stps_data));
1361 return AVERROR(ENOMEM);
1362 sc->stps_count = entries;
1364 for (i = 0; i < entries; i++) {
1365 sc->stps_data[i] = get_be32(pb);
1366 //dprintf(c->fc, "stps %d\n", sc->stps_data[i]);
1372 static int mov_read_stss(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1375 MOVStreamContext *sc;
1376 unsigned int i, entries;
1378 if (c->fc->nb_streams < 1)
1380 st = c->fc->streams[c->fc->nb_streams-1];
1383 get_byte(pb); /* version */
1384 get_be24(pb); /* flags */
1386 entries = get_be32(pb);
1388 dprintf(c->fc, "keyframe_count = %d\n", entries);
1390 if(entries >= UINT_MAX / sizeof(int))
1392 sc->keyframes = av_malloc(entries * sizeof(int));
1394 return AVERROR(ENOMEM);
1395 sc->keyframe_count = entries;
1397 for(i=0; i<entries; i++) {
1398 sc->keyframes[i] = get_be32(pb);
1399 //dprintf(c->fc, "keyframes[]=%d\n", sc->keyframes[i]);
1404 static int mov_read_stsz(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1407 MOVStreamContext *sc;
1408 unsigned int i, entries, sample_size, field_size, num_bytes;
1412 if (c->fc->nb_streams < 1)
1414 st = c->fc->streams[c->fc->nb_streams-1];
1417 get_byte(pb); /* version */
1418 get_be24(pb); /* flags */
1420 if (atom.type == MKTAG('s','t','s','z')) {
1421 sample_size = get_be32(pb);
1422 if (!sc->sample_size) /* do not overwrite value computed in stsd */
1423 sc->sample_size = sample_size;
1427 get_be24(pb); /* reserved */
1428 field_size = get_byte(pb);
1430 entries = get_be32(pb);
1432 dprintf(c->fc, "sample_size = %d sample_count = %d\n", sc->sample_size, entries);
1434 sc->sample_count = entries;
1438 if (field_size != 4 && field_size != 8 && field_size != 16 && field_size != 32) {
1439 av_log(c->fc, AV_LOG_ERROR, "Invalid sample field size %d\n", field_size);
1443 if (entries >= UINT_MAX / sizeof(int) || entries >= (UINT_MAX - 4) / field_size)
1445 sc->sample_sizes = av_malloc(entries * sizeof(int));
1446 if (!sc->sample_sizes)
1447 return AVERROR(ENOMEM);
1449 num_bytes = (entries*field_size+4)>>3;
1451 buf = av_malloc(num_bytes+FF_INPUT_BUFFER_PADDING_SIZE);
1453 av_freep(&sc->sample_sizes);
1454 return AVERROR(ENOMEM);
1457 if (get_buffer(pb, buf, num_bytes) < num_bytes) {
1458 av_freep(&sc->sample_sizes);
1463 init_get_bits(&gb, buf, 8*num_bytes);
1465 for(i=0; i<entries; i++)
1466 sc->sample_sizes[i] = get_bits_long(&gb, field_size);
1472 static int mov_read_stts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1475 MOVStreamContext *sc;
1476 unsigned int i, entries;
1478 int64_t total_sample_count=0;
1480 if (c->fc->nb_streams < 1)
1482 st = c->fc->streams[c->fc->nb_streams-1];
1485 get_byte(pb); /* version */
1486 get_be24(pb); /* flags */
1487 entries = get_be32(pb);
1489 dprintf(c->fc, "track[%i].stts.entries = %i\n", c->fc->nb_streams-1, entries);
1491 if(entries >= UINT_MAX / sizeof(*sc->stts_data))
1493 sc->stts_data = av_malloc(entries * sizeof(*sc->stts_data));
1495 return AVERROR(ENOMEM);
1496 sc->stts_count = entries;
1498 for(i=0; i<entries; i++) {
1499 int sample_duration;
1502 sample_count=get_be32(pb);
1503 sample_duration = get_be32(pb);
1504 sc->stts_data[i].count= sample_count;
1505 sc->stts_data[i].duration= sample_duration;
1507 dprintf(c->fc, "sample_count=%d, sample_duration=%d\n",sample_count,sample_duration);
1509 duration+=(int64_t)sample_duration*sample_count;
1510 total_sample_count+=sample_count;
1513 st->nb_frames= total_sample_count;
1515 st->duration= duration;
1519 static int mov_read_ctts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1522 MOVStreamContext *sc;
1523 unsigned int i, entries;
1525 if (c->fc->nb_streams < 1)
1527 st = c->fc->streams[c->fc->nb_streams-1];
1530 get_byte(pb); /* version */
1531 get_be24(pb); /* flags */
1532 entries = get_be32(pb);
1534 dprintf(c->fc, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
1536 if(entries >= UINT_MAX / sizeof(*sc->ctts_data))
1538 sc->ctts_data = av_malloc(entries * sizeof(*sc->ctts_data));
1540 return AVERROR(ENOMEM);
1541 sc->ctts_count = entries;
1543 for(i=0; i<entries; i++) {
1544 int count =get_be32(pb);
1545 int duration =get_be32(pb);
1547 sc->ctts_data[i].count = count;
1548 sc->ctts_data[i].duration= duration;
1550 sc->dts_shift = FFMAX(sc->dts_shift, -duration);
1553 dprintf(c->fc, "dts shift %d\n", sc->dts_shift);
1558 static void mov_build_index(MOVContext *mov, AVStream *st)
1560 MOVStreamContext *sc = st->priv_data;
1561 int64_t current_offset;
1562 int64_t current_dts = 0;
1563 unsigned int stts_index = 0;
1564 unsigned int stsc_index = 0;
1565 unsigned int stss_index = 0;
1566 unsigned int stps_index = 0;
1568 uint64_t stream_size = 0;
1570 /* adjust first dts according to edit list */
1571 if (sc->time_offset) {
1572 int rescaled = sc->time_offset < 0 ? av_rescale(sc->time_offset, sc->time_scale, mov->time_scale) : sc->time_offset;
1573 current_dts = -rescaled;
1574 if (sc->ctts_data && sc->ctts_data[0].duration / sc->stts_data[0].duration > 16) {
1575 /* more than 16 frames delay, dts are likely wrong
1576 this happens with files created by iMovie */
1578 st->codec->has_b_frames = 1;
1582 /* only use old uncompressed audio chunk demuxing when stts specifies it */
1583 if (!(st->codec->codec_type == AVMEDIA_TYPE_AUDIO &&
1584 sc->stts_count == 1 && sc->stts_data[0].duration == 1)) {
1585 unsigned int current_sample = 0;
1586 unsigned int stts_sample = 0;
1587 unsigned int sample_size;
1588 unsigned int distance = 0;
1589 int key_off = sc->keyframes && sc->keyframes[0] == 1;
1591 current_dts -= sc->dts_shift;
1593 if (sc->sample_count >= UINT_MAX / sizeof(*st->index_entries))
1595 st->index_entries = av_malloc(sc->sample_count*sizeof(*st->index_entries));
1596 if (!st->index_entries)
1598 st->index_entries_allocated_size = sc->sample_count*sizeof(*st->index_entries);
1600 for (i = 0; i < sc->chunk_count; i++) {
1601 current_offset = sc->chunk_offsets[i];
1602 if (stsc_index + 1 < sc->stsc_count &&
1603 i + 1 == sc->stsc_data[stsc_index + 1].first)
1605 for (j = 0; j < sc->stsc_data[stsc_index].count; j++) {
1607 if (current_sample >= sc->sample_count) {
1608 av_log(mov->fc, AV_LOG_ERROR, "wrong sample count\n");
1612 if (!sc->keyframe_count || current_sample+key_off == sc->keyframes[stss_index]) {
1614 if (stss_index + 1 < sc->keyframe_count)
1616 } else if (sc->stps_count && current_sample+key_off == sc->stps_data[stps_index]) {
1618 if (stps_index + 1 < sc->stps_count)
1623 sample_size = sc->sample_size > 0 ? sc->sample_size : sc->sample_sizes[current_sample];
1624 if(sc->pseudo_stream_id == -1 ||
1625 sc->stsc_data[stsc_index].id - 1 == sc->pseudo_stream_id) {
1626 AVIndexEntry *e = &st->index_entries[st->nb_index_entries++];
1627 e->pos = current_offset;
1628 e->timestamp = current_dts;
1629 e->size = sample_size;
1630 e->min_distance = distance;
1631 e->flags = keyframe ? AVINDEX_KEYFRAME : 0;
1632 dprintf(mov->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1633 "size %d, distance %d, keyframe %d\n", st->index, current_sample,
1634 current_offset, current_dts, sample_size, distance, keyframe);
1637 current_offset += sample_size;
1638 stream_size += sample_size;
1639 current_dts += sc->stts_data[stts_index].duration;
1643 if (stts_index + 1 < sc->stts_count && stts_sample == sc->stts_data[stts_index].count) {
1649 if (st->duration > 0)
1650 st->codec->bit_rate = stream_size*8*sc->time_scale/st->duration;
1652 unsigned chunk_samples, total = 0;
1654 // compute total chunk count
1655 for (i = 0; i < sc->stsc_count; i++) {
1656 unsigned count, chunk_count;
1658 chunk_samples = sc->stsc_data[i].count;
1659 if (sc->samples_per_frame && chunk_samples % sc->samples_per_frame) {
1660 av_log(mov->fc, AV_LOG_ERROR, "error unaligned chunk\n");
1664 if (sc->samples_per_frame >= 160) { // gsm
1665 count = chunk_samples / sc->samples_per_frame;
1666 } else if (sc->samples_per_frame > 1) {
1667 unsigned samples = (1024/sc->samples_per_frame)*sc->samples_per_frame;
1668 count = (chunk_samples+samples-1) / samples;
1670 count = (chunk_samples+1023) / 1024;
1673 if (i < sc->stsc_count - 1)
1674 chunk_count = sc->stsc_data[i+1].first - sc->stsc_data[i].first;
1676 chunk_count = sc->chunk_count - (sc->stsc_data[i].first - 1);
1677 total += chunk_count * count;
1680 dprintf(mov->fc, "chunk count %d\n", total);
1681 if (total >= UINT_MAX / sizeof(*st->index_entries))
1683 st->index_entries = av_malloc(total*sizeof(*st->index_entries));
1684 if (!st->index_entries)
1686 st->index_entries_allocated_size = total*sizeof(*st->index_entries);
1689 for (i = 0; i < sc->chunk_count; i++) {
1690 current_offset = sc->chunk_offsets[i];
1691 if (stsc_index + 1 < sc->stsc_count &&
1692 i + 1 == sc->stsc_data[stsc_index + 1].first)
1694 chunk_samples = sc->stsc_data[stsc_index].count;
1696 while (chunk_samples > 0) {
1698 unsigned size, samples;
1700 if (sc->samples_per_frame >= 160) { // gsm
1701 samples = sc->samples_per_frame;
1702 size = sc->bytes_per_frame;
1704 if (sc->samples_per_frame > 1) {
1705 samples = FFMIN((1024 / sc->samples_per_frame)*
1706 sc->samples_per_frame, chunk_samples);
1707 size = (samples / sc->samples_per_frame) * sc->bytes_per_frame;
1709 samples = FFMIN(1024, chunk_samples);
1710 size = samples * sc->sample_size;
1714 if (st->nb_index_entries >= total) {
1715 av_log(mov->fc, AV_LOG_ERROR, "wrong chunk count %d\n", total);
1718 e = &st->index_entries[st->nb_index_entries++];
1719 e->pos = current_offset;
1720 e->timestamp = current_dts;
1722 e->min_distance = 0;
1723 e->flags = AVINDEX_KEYFRAME;
1724 dprintf(mov->fc, "AVIndex stream %d, chunk %d, offset %"PRIx64", dts %"PRId64", "
1725 "size %d, duration %d\n", st->index, i, current_offset, current_dts,
1728 current_offset += size;
1729 current_dts += samples;
1730 chunk_samples -= samples;
1736 static int mov_open_dref(ByteIOContext **pb, char *src, MOVDref *ref)
1738 /* try relative path, we do not try the absolute because it can leak information about our
1739 system to an attacker */
1740 if (ref->nlvl_to > 0 && ref->nlvl_from > 0) {
1741 char filename[1024];
1745 /* find a source dir */
1746 src_path = strrchr(src, '/');
1752 /* find a next level down to target */
1753 for (i = 0, l = strlen(ref->path) - 1; l >= 0; l--)
1754 if (ref->path[l] == '/') {
1755 if (i == ref->nlvl_to - 1)
1761 /* compose filename if next level down to target was found */
1762 if (i == ref->nlvl_to - 1 && src_path - src < sizeof(filename)) {
1763 memcpy(filename, src, src_path - src);
1764 filename[src_path - src] = 0;
1766 for (i = 1; i < ref->nlvl_from; i++)
1767 av_strlcat(filename, "../", 1024);
1769 av_strlcat(filename, ref->path + l + 1, 1024);
1771 if (!url_fopen(pb, filename, URL_RDONLY))
1776 return AVERROR(ENOENT);
1779 static int mov_read_trak(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1782 MOVStreamContext *sc;
1785 st = av_new_stream(c->fc, c->fc->nb_streams);
1786 if (!st) return AVERROR(ENOMEM);
1787 sc = av_mallocz(sizeof(MOVStreamContext));
1788 if (!sc) return AVERROR(ENOMEM);
1791 st->codec->codec_type = AVMEDIA_TYPE_DATA;
1792 sc->ffindex = st->index;
1794 if ((ret = mov_read_default(c, pb, atom)) < 0)
1798 if (sc->chunk_count && (!sc->stts_count || !sc->stsc_count ||
1799 (!sc->sample_size && !sc->sample_count))) {
1800 av_log(c->fc, AV_LOG_ERROR, "stream %d, missing mandatory atoms, broken header\n",
1805 if (!sc->time_scale) {
1806 av_log(c->fc, AV_LOG_WARNING, "stream %d, timescale not set\n", st->index);
1807 sc->time_scale = c->time_scale;
1808 if (!sc->time_scale)
1812 av_set_pts_info(st, 64, 1, sc->time_scale);
1814 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO &&
1815 !st->codec->frame_size && sc->stts_count == 1) {
1816 st->codec->frame_size = av_rescale(sc->stts_data[0].duration,
1817 st->codec->sample_rate, sc->time_scale);
1818 dprintf(c->fc, "frame size %d\n", st->codec->frame_size);
1821 mov_build_index(c, st);
1823 if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
1824 MOVDref *dref = &sc->drefs[sc->dref_id - 1];
1825 if (mov_open_dref(&sc->pb, c->fc->filename, dref) < 0)
1826 av_log(c->fc, AV_LOG_ERROR,
1827 "stream %d, error opening alias: path='%s', dir='%s', "
1828 "filename='%s', volume='%s', nlvl_from=%d, nlvl_to=%d\n",
1829 st->index, dref->path, dref->dir, dref->filename,
1830 dref->volume, dref->nlvl_from, dref->nlvl_to);
1834 if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
1835 if (!st->sample_aspect_ratio.num &&
1836 (st->codec->width != sc->width || st->codec->height != sc->height)) {
1837 st->sample_aspect_ratio = av_d2q(((double)st->codec->height * sc->width) /
1838 ((double)st->codec->width * sc->height), INT_MAX);
1841 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1842 sc->time_scale*st->nb_frames, st->duration, INT_MAX);
1845 switch (st->codec->codec_id) {
1846 #if CONFIG_H261_DECODER
1849 #if CONFIG_H263_DECODER
1852 #if CONFIG_H264_DECODER
1855 #if CONFIG_MPEG4_DECODER
1856 case CODEC_ID_MPEG4:
1858 st->codec->width = 0; /* let decoder init width/height */
1859 st->codec->height= 0;
1863 /* Do not need those anymore. */
1864 av_freep(&sc->chunk_offsets);
1865 av_freep(&sc->stsc_data);
1866 av_freep(&sc->sample_sizes);
1867 av_freep(&sc->keyframes);
1868 av_freep(&sc->stts_data);
1869 av_freep(&sc->stps_data);
1874 static int mov_read_ilst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1877 c->itunes_metadata = 1;
1878 ret = mov_read_default(c, pb, atom);
1879 c->itunes_metadata = 0;
1883 static int mov_read_meta(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1885 while (atom.size > 8) {
1886 uint32_t tag = get_le32(pb);
1888 if (tag == MKTAG('h','d','l','r')) {
1889 url_fseek(pb, -8, SEEK_CUR);
1891 return mov_read_default(c, pb, atom);
1897 static int mov_read_tkhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1902 int64_t disp_transform[2];
1903 int display_matrix[3][2];
1905 MOVStreamContext *sc;
1908 if (c->fc->nb_streams < 1)
1910 st = c->fc->streams[c->fc->nb_streams-1];
1913 version = get_byte(pb);
1914 get_be24(pb); /* flags */
1916 MOV_TRACK_ENABLED 0x0001
1917 MOV_TRACK_IN_MOVIE 0x0002
1918 MOV_TRACK_IN_PREVIEW 0x0004
1919 MOV_TRACK_IN_POSTER 0x0008
1926 get_be32(pb); /* creation time */
1927 get_be32(pb); /* modification time */
1929 st->id = (int)get_be32(pb); /* track id (NOT 0 !)*/
1930 get_be32(pb); /* reserved */
1932 /* highlevel (considering edits) duration in movie timebase */
1933 (version == 1) ? get_be64(pb) : get_be32(pb);
1934 get_be32(pb); /* reserved */
1935 get_be32(pb); /* reserved */
1937 get_be16(pb); /* layer */
1938 get_be16(pb); /* alternate group */
1939 get_be16(pb); /* volume */
1940 get_be16(pb); /* reserved */
1942 //read in the display matrix (outlined in ISO 14496-12, Section 6.2.2)
1943 // they're kept in fixed point format through all calculations
1944 // ignore u,v,z b/c we don't need the scale factor to calc aspect ratio
1945 for (i = 0; i < 3; i++) {
1946 display_matrix[i][0] = get_be32(pb); // 16.16 fixed point
1947 display_matrix[i][1] = get_be32(pb); // 16.16 fixed point
1948 get_be32(pb); // 2.30 fixed point (not used)
1951 width = get_be32(pb); // 16.16 fixed point track width
1952 height = get_be32(pb); // 16.16 fixed point track height
1953 sc->width = width >> 16;
1954 sc->height = height >> 16;
1956 // transform the display width/height according to the matrix
1957 // skip this if the display matrix is the default identity matrix
1958 // or if it is rotating the picture, ex iPhone 3GS
1959 // to keep the same scale, use [width height 1<<16]
1960 if (width && height &&
1961 ((display_matrix[0][0] != 65536 ||
1962 display_matrix[1][1] != 65536) &&
1963 !display_matrix[0][1] &&
1964 !display_matrix[1][0] &&
1965 !display_matrix[2][0] && !display_matrix[2][1])) {
1966 for (i = 0; i < 2; i++)
1968 (int64_t) width * display_matrix[0][i] +
1969 (int64_t) height * display_matrix[1][i] +
1970 ((int64_t) display_matrix[2][i] << 16);
1972 //sample aspect ratio is new width/height divided by old width/height
1973 st->sample_aspect_ratio = av_d2q(
1974 ((double) disp_transform[0] * height) /
1975 ((double) disp_transform[1] * width), INT_MAX);
1980 static int mov_read_tfhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1982 MOVFragment *frag = &c->fragment;
1983 MOVTrackExt *trex = NULL;
1984 int flags, track_id, i;
1986 get_byte(pb); /* version */
1987 flags = get_be24(pb);
1989 track_id = get_be32(pb);
1992 frag->track_id = track_id;
1993 for (i = 0; i < c->trex_count; i++)
1994 if (c->trex_data[i].track_id == frag->track_id) {
1995 trex = &c->trex_data[i];
1999 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding trex\n");
2003 if (flags & 0x01) frag->base_data_offset = get_be64(pb);
2004 else frag->base_data_offset = frag->moof_offset;
2005 if (flags & 0x02) frag->stsd_id = get_be32(pb);
2006 else frag->stsd_id = trex->stsd_id;
2008 frag->duration = flags & 0x08 ? get_be32(pb) : trex->duration;
2009 frag->size = flags & 0x10 ? get_be32(pb) : trex->size;
2010 frag->flags = flags & 0x20 ? get_be32(pb) : trex->flags;
2011 dprintf(c->fc, "frag flags 0x%x\n", frag->flags);
2015 static int mov_read_chap(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
2017 c->chapter_track = get_be32(pb);
2021 static int mov_read_trex(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
2025 if ((uint64_t)c->trex_count+1 >= UINT_MAX / sizeof(*c->trex_data))
2027 trex = av_realloc(c->trex_data, (c->trex_count+1)*sizeof(*c->trex_data));
2029 return AVERROR(ENOMEM);
2030 c->trex_data = trex;
2031 trex = &c->trex_data[c->trex_count++];
2032 get_byte(pb); /* version */
2033 get_be24(pb); /* flags */
2034 trex->track_id = get_be32(pb);
2035 trex->stsd_id = get_be32(pb);
2036 trex->duration = get_be32(pb);
2037 trex->size = get_be32(pb);
2038 trex->flags = get_be32(pb);
2042 static int mov_read_trun(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
2044 MOVFragment *frag = &c->fragment;
2045 AVStream *st = NULL;
2046 MOVStreamContext *sc;
2049 int data_offset = 0;
2050 unsigned entries, first_sample_flags = frag->flags;
2051 int flags, distance, i;
2053 for (i = 0; i < c->fc->nb_streams; i++) {
2054 if (c->fc->streams[i]->id == frag->track_id) {
2055 st = c->fc->streams[i];
2060 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding track id %d\n", frag->track_id);
2064 if (sc->pseudo_stream_id+1 != frag->stsd_id)
2066 get_byte(pb); /* version */
2067 flags = get_be24(pb);
2068 entries = get_be32(pb);
2069 dprintf(c->fc, "flags 0x%x entries %d\n", flags, entries);
2070 if (flags & 0x001) data_offset = get_be32(pb);
2071 if (flags & 0x004) first_sample_flags = get_be32(pb);
2072 if (flags & 0x800) {
2074 if ((uint64_t)entries+sc->ctts_count >= UINT_MAX/sizeof(*sc->ctts_data))
2076 ctts_data = av_realloc(sc->ctts_data,
2077 (entries+sc->ctts_count)*sizeof(*sc->ctts_data));
2079 return AVERROR(ENOMEM);
2080 sc->ctts_data = ctts_data;
2083 offset = frag->base_data_offset + data_offset;
2085 dprintf(c->fc, "first sample flags 0x%x\n", first_sample_flags);
2086 for (i = 0; i < entries; i++) {
2087 unsigned sample_size = frag->size;
2088 int sample_flags = i ? frag->flags : first_sample_flags;
2089 unsigned sample_duration = frag->duration;
2092 if (flags & 0x100) sample_duration = get_be32(pb);
2093 if (flags & 0x200) sample_size = get_be32(pb);
2094 if (flags & 0x400) sample_flags = get_be32(pb);
2095 if (flags & 0x800) {
2096 sc->ctts_data[sc->ctts_count].count = 1;
2097 sc->ctts_data[sc->ctts_count].duration = get_be32(pb);
2100 if ((keyframe = st->codec->codec_type == AVMEDIA_TYPE_AUDIO ||
2101 (flags & 0x004 && !i && !sample_flags) || sample_flags & 0x2000000))
2103 av_add_index_entry(st, offset, dts, sample_size, distance,
2104 keyframe ? AVINDEX_KEYFRAME : 0);
2105 dprintf(c->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
2106 "size %d, distance %d, keyframe %d\n", st->index, sc->sample_count+i,
2107 offset, dts, sample_size, distance, keyframe);
2109 dts += sample_duration;
2110 offset += sample_size;
2112 frag->moof_offset = offset;
2117 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
2118 /* like the files created with Adobe Premiere 5.0, for samples see */
2119 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
2120 static int mov_read_wide(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
2125 return 0; /* continue */
2126 if (get_be32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
2127 url_fskip(pb, atom.size - 4);
2130 atom.type = get_le32(pb);
2132 if (atom.type != MKTAG('m','d','a','t')) {
2133 url_fskip(pb, atom.size);
2136 err = mov_read_mdat(c, pb, atom);
2140 static int mov_read_cmov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
2145 uint8_t *moov_data; /* uncompressed data */
2146 long cmov_len, moov_len;
2149 get_be32(pb); /* dcom atom */
2150 if (get_le32(pb) != MKTAG('d','c','o','m'))
2152 if (get_le32(pb) != MKTAG('z','l','i','b')) {
2153 av_log(c->fc, AV_LOG_ERROR, "unknown compression for cmov atom !");
2156 get_be32(pb); /* cmvd atom */
2157 if (get_le32(pb) != MKTAG('c','m','v','d'))
2159 moov_len = get_be32(pb); /* uncompressed size */
2160 cmov_len = atom.size - 6 * 4;
2162 cmov_data = av_malloc(cmov_len);
2164 return AVERROR(ENOMEM);
2165 moov_data = av_malloc(moov_len);
2168 return AVERROR(ENOMEM);
2170 get_buffer(pb, cmov_data, cmov_len);
2171 if(uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
2172 goto free_and_return;
2173 if(init_put_byte(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL, NULL) != 0)
2174 goto free_and_return;
2175 atom.type = MKTAG('m','o','o','v');
2176 atom.size = moov_len;
2178 // { int fd = open("/tmp/uncompheader.mov", O_WRONLY | O_CREAT); write(fd, moov_data, moov_len); close(fd); }
2180 ret = mov_read_default(c, &ctx, atom);
2186 av_log(c->fc, AV_LOG_ERROR, "this file requires zlib support compiled in\n");
2191 /* edit list atom */
2192 static int mov_read_elst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
2194 MOVStreamContext *sc;
2197 if (c->fc->nb_streams < 1)
2199 sc = c->fc->streams[c->fc->nb_streams-1]->priv_data;
2201 get_byte(pb); /* version */
2202 get_be24(pb); /* flags */
2203 edit_count = get_be32(pb); /* entries */
2205 if((uint64_t)edit_count*12+8 > atom.size)
2208 for(i=0; i<edit_count; i++){
2210 int duration = get_be32(pb); /* Track duration */
2211 time = get_be32(pb); /* Media time */
2212 get_be32(pb); /* Media rate */
2213 if (i == 0 && time >= -1) {
2214 sc->time_offset = time != -1 ? time : -duration;
2219 av_log(c->fc, AV_LOG_WARNING, "multiple edit list entries, "
2220 "a/v desync might occur, patch welcome\n");
2222 dprintf(c->fc, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, edit_count);
2226 static const MOVParseTableEntry mov_default_parse_table[] = {
2227 { MKTAG('a','v','s','s'), mov_read_extradata },
2228 { MKTAG('c','h','p','l'), mov_read_chpl },
2229 { MKTAG('c','o','6','4'), mov_read_stco },
2230 { MKTAG('c','t','t','s'), mov_read_ctts }, /* composition time to sample */
2231 { MKTAG('d','i','n','f'), mov_read_default },
2232 { MKTAG('d','r','e','f'), mov_read_dref },
2233 { MKTAG('e','d','t','s'), mov_read_default },
2234 { MKTAG('e','l','s','t'), mov_read_elst },
2235 { MKTAG('e','n','d','a'), mov_read_enda },
2236 { MKTAG('f','i','e','l'), mov_read_extradata },
2237 { MKTAG('f','t','y','p'), mov_read_ftyp },
2238 { MKTAG('g','l','b','l'), mov_read_glbl },
2239 { MKTAG('h','d','l','r'), mov_read_hdlr },
2240 { MKTAG('i','l','s','t'), mov_read_ilst },
2241 { MKTAG('j','p','2','h'), mov_read_extradata },
2242 { MKTAG('m','d','a','t'), mov_read_mdat },
2243 { MKTAG('m','d','h','d'), mov_read_mdhd },
2244 { MKTAG('m','d','i','a'), mov_read_default },
2245 { MKTAG('m','e','t','a'), mov_read_meta },
2246 { MKTAG('m','i','n','f'), mov_read_default },
2247 { MKTAG('m','o','o','f'), mov_read_moof },
2248 { MKTAG('m','o','o','v'), mov_read_moov },
2249 { MKTAG('m','v','e','x'), mov_read_default },
2250 { MKTAG('m','v','h','d'), mov_read_mvhd },
2251 { MKTAG('S','M','I',' '), mov_read_smi }, /* Sorenson extension ??? */
2252 { MKTAG('a','l','a','c'), mov_read_extradata }, /* alac specific atom */
2253 { MKTAG('a','v','c','C'), mov_read_glbl },
2254 { MKTAG('p','a','s','p'), mov_read_pasp },
2255 { MKTAG('s','t','b','l'), mov_read_default },
2256 { MKTAG('s','t','c','o'), mov_read_stco },
2257 { MKTAG('s','t','p','s'), mov_read_stps },
2258 { MKTAG('s','t','r','f'), mov_read_strf },
2259 { MKTAG('s','t','s','c'), mov_read_stsc },
2260 { MKTAG('s','t','s','d'), mov_read_stsd }, /* sample description */
2261 { MKTAG('s','t','s','s'), mov_read_stss }, /* sync sample */
2262 { MKTAG('s','t','s','z'), mov_read_stsz }, /* sample size */
2263 { MKTAG('s','t','t','s'), mov_read_stts },
2264 { MKTAG('s','t','z','2'), mov_read_stsz }, /* compact sample size */
2265 { MKTAG('t','k','h','d'), mov_read_tkhd }, /* track header */
2266 { MKTAG('t','f','h','d'), mov_read_tfhd }, /* track fragment header */
2267 { MKTAG('t','r','a','k'), mov_read_trak },
2268 { MKTAG('t','r','a','f'), mov_read_default },
2269 { MKTAG('t','r','e','f'), mov_read_default },
2270 { MKTAG('c','h','a','p'), mov_read_chap },
2271 { MKTAG('t','r','e','x'), mov_read_trex },
2272 { MKTAG('t','r','u','n'), mov_read_trun },
2273 { MKTAG('u','d','t','a'), mov_read_default },
2274 { MKTAG('w','a','v','e'), mov_read_wave },
2275 { MKTAG('e','s','d','s'), mov_read_esds },
2276 { MKTAG('d','a','c','3'), mov_read_dac3 }, /* AC-3 info */
2277 { MKTAG('w','i','d','e'), mov_read_wide }, /* place holder */
2278 { MKTAG('c','m','o','v'), mov_read_cmov },
2282 static int mov_probe(AVProbeData *p)
2284 unsigned int offset;
2288 /* check file header */
2291 /* ignore invalid offset */
2292 if ((offset + 8) > (unsigned int)p->buf_size)
2294 tag = AV_RL32(p->buf + offset + 4);
2296 /* check for obvious tags */
2297 case MKTAG('j','P',' ',' '): /* jpeg 2000 signature */
2298 case MKTAG('m','o','o','v'):
2299 case MKTAG('m','d','a','t'):
2300 case MKTAG('p','n','o','t'): /* detect movs with preview pics like ew.mov and april.mov */
2301 case MKTAG('u','d','t','a'): /* Packet Video PVAuthor adds this and a lot of more junk */
2302 case MKTAG('f','t','y','p'):
2303 return AVPROBE_SCORE_MAX;
2304 /* those are more common words, so rate then a bit less */
2305 case MKTAG('e','d','i','w'): /* xdcam files have reverted first tags */
2306 case MKTAG('w','i','d','e'):
2307 case MKTAG('f','r','e','e'):
2308 case MKTAG('j','u','n','k'):
2309 case MKTAG('p','i','c','t'):
2310 return AVPROBE_SCORE_MAX - 5;
2311 case MKTAG(0x82,0x82,0x7f,0x7d):
2312 case MKTAG('s','k','i','p'):
2313 case MKTAG('u','u','i','d'):
2314 case MKTAG('p','r','f','l'):
2315 offset = AV_RB32(p->buf+offset) + offset;
2316 /* if we only find those cause probedata is too small at least rate them */
2317 score = AVPROBE_SCORE_MAX - 50;
2320 /* unrecognized tag */
2327 // must be done after parsing all trak because there's no order requirement
2328 static void mov_read_chapters(AVFormatContext *s)
2330 MOVContext *mov = s->priv_data;
2331 AVStream *st = NULL;
2332 MOVStreamContext *sc;
2334 uint8_t *title = NULL;
2335 int i, len, i8, i16;
2337 for (i = 0; i < s->nb_streams; i++)
2338 if (s->streams[i]->id == mov->chapter_track) {
2343 av_log(s, AV_LOG_ERROR, "Referenced QT chapter track not found\n");
2347 st->discard = AVDISCARD_ALL;
2349 cur_pos = url_ftell(sc->pb);
2351 for (i = 0; i < st->nb_index_entries; i++) {
2352 AVIndexEntry *sample = &st->index_entries[i];
2353 int64_t end = i+1 < st->nb_index_entries ? st->index_entries[i+1].timestamp : st->duration;
2355 if (url_fseek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
2356 av_log(s, AV_LOG_ERROR, "Chapter %d not found in file\n", i);
2360 title = av_malloc(sample->size+2);
2361 get_buffer(sc->pb, title, sample->size);
2363 // the first two bytes are the length of the title
2364 len = AV_RB16(title);
2365 if (len > sample->size-2)
2368 // The samples could theoretically be in any encoding if there's an encd
2369 // atom following, but in practice are only utf-8 or utf-16, distinguished
2370 // instead by the presence of a BOM
2371 if (AV_RB16(title+2) == 0xfeff) {
2372 uint8_t *utf8 = av_malloc(2*len+3);
2378 GET_UTF16(ch, i16 < len ? AV_RB16(title + (i16+=2)) : 0, break;)
2379 PUT_UTF8(ch, tmp, if (i8 < 2*len) utf8[2+i8++] = tmp;)
2386 ff_new_chapter(s, i, st->time_base, sample->timestamp, end, title+2);
2391 url_fseek(sc->pb, cur_pos, SEEK_SET);
2394 static int mov_read_header(AVFormatContext *s, AVFormatParameters *ap)
2396 MOVContext *mov = s->priv_data;
2397 ByteIOContext *pb = s->pb;
2399 MOVAtom atom = { AV_RL32("root") };
2402 /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
2403 if(!url_is_streamed(pb))
2404 atom.size = url_fsize(pb);
2406 atom.size = INT64_MAX;
2408 /* check MOV header */
2409 if ((err = mov_read_default(mov, pb, atom)) < 0) {
2410 av_log(s, AV_LOG_ERROR, "error reading header: %d\n", err);
2413 if (!mov->found_moov) {
2414 av_log(s, AV_LOG_ERROR, "moov atom not found\n");
2417 dprintf(mov->fc, "on_parse_exit_offset=%lld\n", url_ftell(pb));
2419 if (!url_is_streamed(pb) && mov->chapter_track > 0)
2420 mov_read_chapters(s);
2425 static AVIndexEntry *mov_find_next_sample(AVFormatContext *s, AVStream **st)
2427 AVIndexEntry *sample = NULL;
2428 int64_t best_dts = INT64_MAX;
2430 for (i = 0; i < s->nb_streams; i++) {
2431 AVStream *avst = s->streams[i];
2432 MOVStreamContext *msc = avst->priv_data;
2433 if (msc->pb && msc->current_sample < avst->nb_index_entries) {
2434 AVIndexEntry *current_sample = &avst->index_entries[msc->current_sample];
2435 int64_t dts = av_rescale(current_sample->timestamp, AV_TIME_BASE, msc->time_scale);
2436 dprintf(s, "stream %d, sample %d, dts %"PRId64"\n", i, msc->current_sample, dts);
2437 if (!sample || (url_is_streamed(s->pb) && current_sample->pos < sample->pos) ||
2438 (!url_is_streamed(s->pb) &&
2439 ((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
2440 ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
2441 (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
2442 sample = current_sample;
2451 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
2453 MOVContext *mov = s->priv_data;
2454 MOVStreamContext *sc;
2455 AVIndexEntry *sample;
2456 AVStream *st = NULL;
2459 sample = mov_find_next_sample(s, &st);
2461 mov->found_mdat = 0;
2462 if (!url_is_streamed(s->pb) ||
2463 mov_read_default(mov, s->pb, (MOVAtom){ AV_RL32("root"), INT64_MAX }) < 0 ||
2466 dprintf(s, "read fragments, offset 0x%llx\n", url_ftell(s->pb));
2470 /* must be done just before reading, to avoid infinite loop on sample */
2471 sc->current_sample++;
2473 if (st->discard != AVDISCARD_ALL) {
2474 if (url_fseek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
2475 av_log(mov->fc, AV_LOG_ERROR, "stream %d, offset 0x%"PRIx64": partial file\n",
2476 sc->ffindex, sample->pos);
2479 ret = av_get_packet(sc->pb, pkt, sample->size);
2482 #if CONFIG_DV_DEMUXER
2483 if (mov->dv_demux && sc->dv_audio_container) {
2484 dv_produce_packet(mov->dv_demux, pkt, pkt->data, pkt->size);
2487 ret = dv_get_packet(mov->dv_demux, pkt);
2494 pkt->stream_index = sc->ffindex;
2495 pkt->dts = sample->timestamp;
2496 if (sc->ctts_data) {
2497 pkt->pts = pkt->dts + sc->dts_shift + sc->ctts_data[sc->ctts_index].duration;
2498 /* update ctts context */
2500 if (sc->ctts_index < sc->ctts_count &&
2501 sc->ctts_data[sc->ctts_index].count == sc->ctts_sample) {
2503 sc->ctts_sample = 0;
2506 pkt->dts = AV_NOPTS_VALUE;
2508 int64_t next_dts = (sc->current_sample < st->nb_index_entries) ?
2509 st->index_entries[sc->current_sample].timestamp : st->duration;
2510 pkt->duration = next_dts - pkt->dts;
2511 pkt->pts = pkt->dts;
2513 if (st->discard == AVDISCARD_ALL)
2515 pkt->flags |= sample->flags & AVINDEX_KEYFRAME ? AV_PKT_FLAG_KEY : 0;
2516 pkt->pos = sample->pos;
2517 dprintf(s, "stream %d, pts %"PRId64", dts %"PRId64", pos 0x%"PRIx64", duration %d\n",
2518 pkt->stream_index, pkt->pts, pkt->dts, pkt->pos, pkt->duration);
2522 static int mov_seek_stream(AVFormatContext *s, AVStream *st, int64_t timestamp, int flags)
2524 MOVStreamContext *sc = st->priv_data;
2525 int sample, time_sample;
2528 sample = av_index_search_timestamp(st, timestamp, flags);
2529 dprintf(s, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
2530 if (sample < 0 && st->nb_index_entries && timestamp < st->index_entries[0].timestamp)
2532 if (sample < 0) /* not sure what to do */
2534 sc->current_sample = sample;
2535 dprintf(s, "stream %d, found sample %d\n", st->index, sc->current_sample);
2536 /* adjust ctts index */
2537 if (sc->ctts_data) {
2539 for (i = 0; i < sc->ctts_count; i++) {
2540 int next = time_sample + sc->ctts_data[i].count;
2541 if (next > sc->current_sample) {
2543 sc->ctts_sample = sc->current_sample - time_sample;
2552 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
2555 int64_t seek_timestamp, timestamp;
2559 if (stream_index >= s->nb_streams)
2561 if (sample_time < 0)
2564 st = s->streams[stream_index];
2565 sample = mov_seek_stream(s, st, sample_time, flags);
2569 /* adjust seek timestamp to found sample timestamp */
2570 seek_timestamp = st->index_entries[sample].timestamp;
2572 for (i = 0; i < s->nb_streams; i++) {
2574 if (stream_index == i)
2577 timestamp = av_rescale_q(seek_timestamp, s->streams[stream_index]->time_base, st->time_base);
2578 mov_seek_stream(s, st, timestamp, flags);
2583 static int mov_read_close(AVFormatContext *s)
2585 MOVContext *mov = s->priv_data;
2588 for (i = 0; i < s->nb_streams; i++) {
2589 AVStream *st = s->streams[i];
2590 MOVStreamContext *sc = st->priv_data;
2592 av_freep(&sc->ctts_data);
2593 for (j = 0; j < sc->drefs_count; j++) {
2594 av_freep(&sc->drefs[j].path);
2595 av_freep(&sc->drefs[j].dir);
2597 av_freep(&sc->drefs);
2598 if (sc->pb && sc->pb != s->pb)
2601 av_freep(&st->codec->palctrl);
2604 if (mov->dv_demux) {
2605 for(i = 0; i < mov->dv_fctx->nb_streams; i++) {
2606 av_freep(&mov->dv_fctx->streams[i]->codec);
2607 av_freep(&mov->dv_fctx->streams[i]);
2609 av_freep(&mov->dv_fctx);
2610 av_freep(&mov->dv_demux);
2613 av_freep(&mov->trex_data);
2618 AVInputFormat mov_demuxer = {
2619 "mov,mp4,m4a,3gp,3g2,mj2",
2620 NULL_IF_CONFIG_SMALL("QuickTime/MPEG-4/Motion JPEG 2000 format"),