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 st->codec->extradata = av_mallocz(len + FF_INPUT_BUFFER_PADDING_SIZE);
532 if (!st->codec->extradata)
533 return AVERROR(ENOMEM);
534 get_buffer(pb, st->codec->extradata, len);
535 st->codec->extradata_size = len;
536 if (st->codec->codec_id == CODEC_ID_AAC) {
537 MPEG4AudioConfig cfg;
538 ff_mpeg4audio_get_config(&cfg, st->codec->extradata,
539 st->codec->extradata_size);
540 st->codec->channels = cfg.channels;
541 if (cfg.object_type == 29 && cfg.sampling_index < 3) // old mp3on4
542 st->codec->sample_rate = ff_mpa_freq_tab[cfg.sampling_index];
543 else if (cfg.ext_sample_rate)
544 st->codec->sample_rate = cfg.ext_sample_rate;
546 st->codec->sample_rate = cfg.sample_rate;
547 dprintf(fc, "mp4a config channels %d obj %d ext obj %d "
548 "sample rate %d ext sample rate %d\n", st->codec->channels,
549 cfg.object_type, cfg.ext_object_type,
550 cfg.sample_rate, cfg.ext_sample_rate);
551 if (!(st->codec->codec_id = ff_codec_get_id(mp4_audio_types,
553 st->codec->codec_id = CODEC_ID_AAC;
560 static int mov_read_esds(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
562 return ff_mov_read_esds(c->fc, pb, atom);
565 static int mov_read_dac3(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
568 int ac3info, acmod, lfeon;
570 st = c->fc->streams[c->fc->nb_streams-1];
572 ac3info = get_be24(pb);
573 acmod = (ac3info >> 11) & 0x7;
574 lfeon = (ac3info >> 10) & 0x1;
575 st->codec->channels = ((int[]){2,1,2,3,3,4,4,5})[acmod] + lfeon;
580 static int mov_read_pasp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
582 const int num = get_be32(pb);
583 const int den = get_be32(pb);
586 if (c->fc->nb_streams < 1)
588 st = c->fc->streams[c->fc->nb_streams-1];
590 if ((st->sample_aspect_ratio.den != 1 || st->sample_aspect_ratio.num) && // default
591 (den != st->sample_aspect_ratio.den || num != st->sample_aspect_ratio.num)) {
592 av_log(c->fc, AV_LOG_WARNING,
593 "sample aspect ratio already set to %d:%d, ignoring 'pasp' atom (%d:%d)\n",
594 st->sample_aspect_ratio.num, st->sample_aspect_ratio.den,
596 } else if (den != 0) {
597 st->sample_aspect_ratio.num = num;
598 st->sample_aspect_ratio.den = den;
603 /* this atom contains actual media data */
604 static int mov_read_mdat(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
606 if(atom.size == 0) /* wrong one (MP4) */
609 return 0; /* now go for moov */
612 /* read major brand, minor version and compatible brands and store them as metadata */
613 static int mov_read_ftyp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
617 char minor_ver_str[11]; /* 32 bit integer -> 10 digits + null */
618 char* comp_brands_str;
619 uint8_t type[5] = {0};
621 get_buffer(pb, type, 4);
622 if (strcmp(type, "qt "))
624 av_log(c->fc, AV_LOG_DEBUG, "ISO: File Type Major Brand: %.4s\n",(char *)&type);
625 av_metadata_set2(&c->fc->metadata, "major_brand", type, 0);
626 minor_ver = get_be32(pb); /* minor version */
627 snprintf(minor_ver_str, sizeof(minor_ver_str), "%d", minor_ver);
628 av_metadata_set2(&c->fc->metadata, "minor_version", minor_ver_str, 0);
630 comp_brand_size = atom.size - 8;
631 if (comp_brand_size < 0)
633 comp_brands_str = av_malloc(comp_brand_size + 1); /* Add null terminator */
634 if (!comp_brands_str)
635 return AVERROR(ENOMEM);
636 get_buffer(pb, comp_brands_str, comp_brand_size);
637 comp_brands_str[comp_brand_size] = 0;
638 av_metadata_set2(&c->fc->metadata, "compatible_brands", comp_brands_str, 0);
639 av_freep(&comp_brands_str);
644 /* this atom should contain all header atoms */
645 static int mov_read_moov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
647 if (mov_read_default(c, pb, atom) < 0)
649 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
650 /* so we don't parse the whole file if over a network */
652 return 0; /* now go for mdat */
655 static int mov_read_moof(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
657 c->fragment.moof_offset = url_ftell(pb) - 8;
658 dprintf(c->fc, "moof offset %llx\n", c->fragment.moof_offset);
659 return mov_read_default(c, pb, atom);
662 static int mov_read_mdhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
665 MOVStreamContext *sc;
667 char language[4] = {0};
670 if (c->fc->nb_streams < 1)
672 st = c->fc->streams[c->fc->nb_streams-1];
675 version = get_byte(pb);
677 return -1; /* unsupported */
679 get_be24(pb); /* flags */
684 get_be32(pb); /* creation time */
685 get_be32(pb); /* modification time */
688 sc->time_scale = get_be32(pb);
689 st->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
691 lang = get_be16(pb); /* language */
692 if (ff_mov_lang_to_iso639(lang, language))
693 av_metadata_set2(&st->metadata, "language", language, 0);
694 get_be16(pb); /* quality */
699 static int mov_read_mvhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
701 int version = get_byte(pb); /* version */
702 get_be24(pb); /* flags */
708 get_be32(pb); /* creation time */
709 get_be32(pb); /* modification time */
711 c->time_scale = get_be32(pb); /* time scale */
713 dprintf(c->fc, "time scale = %i\n", c->time_scale);
715 c->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
716 get_be32(pb); /* preferred scale */
718 get_be16(pb); /* preferred volume */
720 url_fskip(pb, 10); /* reserved */
722 url_fskip(pb, 36); /* display matrix */
724 get_be32(pb); /* preview time */
725 get_be32(pb); /* preview duration */
726 get_be32(pb); /* poster time */
727 get_be32(pb); /* selection time */
728 get_be32(pb); /* selection duration */
729 get_be32(pb); /* current time */
730 get_be32(pb); /* next track ID */
735 static int mov_read_smi(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
739 if (c->fc->nb_streams < 1)
741 st = c->fc->streams[c->fc->nb_streams-1];
743 if((uint64_t)atom.size > (1<<30))
746 // currently SVQ3 decoder expect full STSD header - so let's fake it
747 // this should be fixed and just SMI header should be passed
748 av_free(st->codec->extradata);
749 st->codec->extradata = av_mallocz(atom.size + 0x5a + FF_INPUT_BUFFER_PADDING_SIZE);
750 if (!st->codec->extradata)
751 return AVERROR(ENOMEM);
752 st->codec->extradata_size = 0x5a + atom.size;
753 memcpy(st->codec->extradata, "SVQ3", 4); // fake
754 get_buffer(pb, st->codec->extradata + 0x5a, atom.size);
755 dprintf(c->fc, "Reading SMI %"PRId64" %s\n", atom.size, st->codec->extradata + 0x5a);
759 static int mov_read_enda(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
764 if (c->fc->nb_streams < 1)
766 st = c->fc->streams[c->fc->nb_streams-1];
768 little_endian = get_be16(pb);
769 dprintf(c->fc, "enda %d\n", little_endian);
770 if (little_endian == 1) {
771 switch (st->codec->codec_id) {
772 case CODEC_ID_PCM_S24BE:
773 st->codec->codec_id = CODEC_ID_PCM_S24LE;
775 case CODEC_ID_PCM_S32BE:
776 st->codec->codec_id = CODEC_ID_PCM_S32LE;
778 case CODEC_ID_PCM_F32BE:
779 st->codec->codec_id = CODEC_ID_PCM_F32LE;
781 case CODEC_ID_PCM_F64BE:
782 st->codec->codec_id = CODEC_ID_PCM_F64LE;
791 /* FIXME modify qdm2/svq3/h264 decoders to take full atom as extradata */
792 static int mov_read_extradata(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
798 if (c->fc->nb_streams < 1) // will happen with jp2 files
800 st= c->fc->streams[c->fc->nb_streams-1];
801 size= (uint64_t)st->codec->extradata_size + atom.size + 8 + FF_INPUT_BUFFER_PADDING_SIZE;
802 if(size > INT_MAX || (uint64_t)atom.size > INT_MAX)
804 buf= av_realloc(st->codec->extradata, size);
807 st->codec->extradata= buf;
808 buf+= st->codec->extradata_size;
809 st->codec->extradata_size= size - FF_INPUT_BUFFER_PADDING_SIZE;
810 AV_WB32( buf , atom.size + 8);
811 AV_WL32( buf + 4, atom.type);
812 get_buffer(pb, buf + 8, atom.size);
816 static int mov_read_wave(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
820 if (c->fc->nb_streams < 1)
822 st = c->fc->streams[c->fc->nb_streams-1];
824 if((uint64_t)atom.size > (1<<30))
827 if (st->codec->codec_id == CODEC_ID_QDM2) {
828 // pass all frma atom to codec, needed at least for QDM2
829 av_free(st->codec->extradata);
830 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
831 if (!st->codec->extradata)
832 return AVERROR(ENOMEM);
833 st->codec->extradata_size = atom.size;
834 get_buffer(pb, st->codec->extradata, atom.size);
835 } else if (atom.size > 8) { /* to read frma, esds atoms */
836 if (mov_read_default(c, pb, atom) < 0)
839 url_fskip(pb, atom.size);
844 * This function reads atom content and puts data in extradata without tag
845 * nor size unlike mov_read_extradata.
847 static int mov_read_glbl(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
851 if (c->fc->nb_streams < 1)
853 st = c->fc->streams[c->fc->nb_streams-1];
855 if((uint64_t)atom.size > (1<<30))
858 av_free(st->codec->extradata);
859 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
860 if (!st->codec->extradata)
861 return AVERROR(ENOMEM);
862 st->codec->extradata_size = atom.size;
863 get_buffer(pb, st->codec->extradata, atom.size);
868 * An strf atom is a BITMAPINFOHEADER struct. This struct is 40 bytes itself,
869 * but can have extradata appended at the end after the 40 bytes belonging
872 static int mov_read_strf(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
876 if (c->fc->nb_streams < 1)
880 st = c->fc->streams[c->fc->nb_streams-1];
882 if((uint64_t)atom.size > (1<<30))
885 av_free(st->codec->extradata);
886 st->codec->extradata = av_mallocz(atom.size - 40 + FF_INPUT_BUFFER_PADDING_SIZE);
887 if (!st->codec->extradata)
888 return AVERROR(ENOMEM);
889 st->codec->extradata_size = atom.size - 40;
891 get_buffer(pb, st->codec->extradata, atom.size - 40);
895 static int mov_read_stco(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
898 MOVStreamContext *sc;
899 unsigned int i, entries;
901 if (c->fc->nb_streams < 1)
903 st = c->fc->streams[c->fc->nb_streams-1];
906 get_byte(pb); /* version */
907 get_be24(pb); /* flags */
909 entries = get_be32(pb);
911 if(entries >= UINT_MAX/sizeof(int64_t))
914 sc->chunk_offsets = av_malloc(entries * sizeof(int64_t));
915 if (!sc->chunk_offsets)
916 return AVERROR(ENOMEM);
917 sc->chunk_count = entries;
919 if (atom.type == MKTAG('s','t','c','o'))
920 for(i=0; i<entries; i++)
921 sc->chunk_offsets[i] = get_be32(pb);
922 else if (atom.type == MKTAG('c','o','6','4'))
923 for(i=0; i<entries; i++)
924 sc->chunk_offsets[i] = get_be64(pb);
932 * Compute codec id for 'lpcm' tag.
933 * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
935 enum CodecID ff_mov_get_lpcm_codec_id(int bps, int flags)
937 if (flags & 1) { // floating point
938 if (flags & 2) { // big endian
939 if (bps == 32) return CODEC_ID_PCM_F32BE;
940 else if (bps == 64) return CODEC_ID_PCM_F64BE;
942 if (bps == 32) return CODEC_ID_PCM_F32LE;
943 else if (bps == 64) return CODEC_ID_PCM_F64LE;
949 if (flags & 4) return CODEC_ID_PCM_S8;
950 else return CODEC_ID_PCM_U8;
951 else if (bps == 16) return CODEC_ID_PCM_S16BE;
952 else if (bps == 24) return CODEC_ID_PCM_S24BE;
953 else if (bps == 32) return CODEC_ID_PCM_S32BE;
956 if (flags & 4) return CODEC_ID_PCM_S8;
957 else return CODEC_ID_PCM_U8;
958 else if (bps == 16) return CODEC_ID_PCM_S16LE;
959 else if (bps == 24) return CODEC_ID_PCM_S24LE;
960 else if (bps == 32) return CODEC_ID_PCM_S32LE;
963 return CODEC_ID_NONE;
966 static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
969 MOVStreamContext *sc;
970 int j, entries, pseudo_stream_id;
972 if (c->fc->nb_streams < 1)
974 st = c->fc->streams[c->fc->nb_streams-1];
977 get_byte(pb); /* version */
978 get_be24(pb); /* flags */
980 entries = get_be32(pb);
982 for(pseudo_stream_id=0; pseudo_stream_id<entries; pseudo_stream_id++) {
983 //Parsing Sample description table
986 MOVAtom a = { AV_RL32("stsd") };
987 int64_t start_pos = url_ftell(pb);
988 int size = get_be32(pb); /* size */
989 uint32_t format = get_le32(pb); /* data format */
992 get_be32(pb); /* reserved */
993 get_be16(pb); /* reserved */
994 dref_id = get_be16(pb);
997 if (st->codec->codec_tag &&
998 st->codec->codec_tag != format &&
999 (c->fc->video_codec_id ? ff_codec_get_id(codec_movvideo_tags, format) != c->fc->video_codec_id
1000 : st->codec->codec_tag != MKTAG('j','p','e','g'))
1002 /* Multiple fourcc, we skip JPEG. This is not correct, we should
1003 * export it as a separate AVStream but this needs a few changes
1004 * in the MOV demuxer, patch welcome. */
1005 av_log(c->fc, AV_LOG_WARNING, "multiple fourcc not supported\n");
1006 url_fskip(pb, size - (url_ftell(pb) - start_pos));
1009 sc->pseudo_stream_id = st->codec->codec_tag ? -1 : pseudo_stream_id;
1010 sc->dref_id= dref_id;
1012 st->codec->codec_tag = format;
1013 id = ff_codec_get_id(codec_movaudio_tags, format);
1014 if (id<=0 && ((format&0xFFFF) == 'm'+('s'<<8) || (format&0xFFFF) == 'T'+('S'<<8)))
1015 id = ff_codec_get_id(ff_codec_wav_tags, av_bswap32(format)&0xFFFF);
1017 if (st->codec->codec_type != AVMEDIA_TYPE_VIDEO && id > 0) {
1018 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1019 } else if (st->codec->codec_type != AVMEDIA_TYPE_AUDIO && /* do not overwrite codec type */
1020 format && format != MKTAG('m','p','4','s')) { /* skip old asf mpeg4 tag */
1021 id = ff_codec_get_id(codec_movvideo_tags, format);
1023 id = ff_codec_get_id(ff_codec_bmp_tags, format);
1025 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1026 else if(st->codec->codec_type == AVMEDIA_TYPE_DATA){
1027 id = ff_codec_get_id(ff_codec_movsubtitle_tags, format);
1029 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1033 dprintf(c->fc, "size=%d 4CC= %c%c%c%c codec_type=%d\n", size,
1034 (format >> 0) & 0xff, (format >> 8) & 0xff, (format >> 16) & 0xff,
1035 (format >> 24) & 0xff, st->codec->codec_type);
1037 if(st->codec->codec_type==AVMEDIA_TYPE_VIDEO) {
1038 unsigned int color_depth, len;
1039 int color_greyscale;
1041 st->codec->codec_id = id;
1042 get_be16(pb); /* version */
1043 get_be16(pb); /* revision level */
1044 get_be32(pb); /* vendor */
1045 get_be32(pb); /* temporal quality */
1046 get_be32(pb); /* spatial quality */
1048 st->codec->width = get_be16(pb); /* width */
1049 st->codec->height = get_be16(pb); /* height */
1051 get_be32(pb); /* horiz resolution */
1052 get_be32(pb); /* vert resolution */
1053 get_be32(pb); /* data size, always 0 */
1054 get_be16(pb); /* frames per samples */
1056 len = get_byte(pb); /* codec name, pascal string */
1059 mov_read_mac_string(c, pb, len, st->codec->codec_name, 32);
1061 url_fskip(pb, 31 - len);
1062 /* codec_tag YV12 triggers an UV swap in rawdec.c */
1063 if (!memcmp(st->codec->codec_name, "Planar Y'CbCr 8-bit 4:2:0", 25))
1064 st->codec->codec_tag=MKTAG('I', '4', '2', '0');
1066 st->codec->bits_per_coded_sample = get_be16(pb); /* depth */
1067 st->codec->color_table_id = get_be16(pb); /* colortable id */
1068 dprintf(c->fc, "depth %d, ctab id %d\n",
1069 st->codec->bits_per_coded_sample, st->codec->color_table_id);
1070 /* figure out the palette situation */
1071 color_depth = st->codec->bits_per_coded_sample & 0x1F;
1072 color_greyscale = st->codec->bits_per_coded_sample & 0x20;
1074 /* if the depth is 2, 4, or 8 bpp, file is palettized */
1075 if ((color_depth == 2) || (color_depth == 4) ||
1076 (color_depth == 8)) {
1077 /* for palette traversal */
1078 unsigned int color_start, color_count, color_end;
1079 unsigned char r, g, b;
1081 st->codec->palctrl = av_malloc(sizeof(*st->codec->palctrl));
1082 if (color_greyscale) {
1083 int color_index, color_dec;
1084 /* compute the greyscale palette */
1085 st->codec->bits_per_coded_sample = color_depth;
1086 color_count = 1 << color_depth;
1088 color_dec = 256 / (color_count - 1);
1089 for (j = 0; j < color_count; j++) {
1090 r = g = b = color_index;
1091 st->codec->palctrl->palette[j] =
1092 (r << 16) | (g << 8) | (b);
1093 color_index -= color_dec;
1094 if (color_index < 0)
1097 } else if (st->codec->color_table_id) {
1098 const uint8_t *color_table;
1099 /* if flag bit 3 is set, use the default palette */
1100 color_count = 1 << color_depth;
1101 if (color_depth == 2)
1102 color_table = ff_qt_default_palette_4;
1103 else if (color_depth == 4)
1104 color_table = ff_qt_default_palette_16;
1106 color_table = ff_qt_default_palette_256;
1108 for (j = 0; j < color_count; j++) {
1109 r = color_table[j * 3 + 0];
1110 g = color_table[j * 3 + 1];
1111 b = color_table[j * 3 + 2];
1112 st->codec->palctrl->palette[j] =
1113 (r << 16) | (g << 8) | (b);
1116 /* load the palette from the file */
1117 color_start = get_be32(pb);
1118 color_count = get_be16(pb);
1119 color_end = get_be16(pb);
1120 if ((color_start <= 255) &&
1121 (color_end <= 255)) {
1122 for (j = color_start; j <= color_end; j++) {
1123 /* each R, G, or B component is 16 bits;
1124 * only use the top 8 bits; skip alpha bytes
1134 st->codec->palctrl->palette[j] =
1135 (r << 16) | (g << 8) | (b);
1139 st->codec->palctrl->palette_changed = 1;
1141 } else if(st->codec->codec_type==AVMEDIA_TYPE_AUDIO) {
1142 int bits_per_sample, flags;
1143 uint16_t version = get_be16(pb);
1145 st->codec->codec_id = id;
1146 get_be16(pb); /* revision level */
1147 get_be32(pb); /* vendor */
1149 st->codec->channels = get_be16(pb); /* channel count */
1150 dprintf(c->fc, "audio channels %d\n", st->codec->channels);
1151 st->codec->bits_per_coded_sample = get_be16(pb); /* sample size */
1153 sc->audio_cid = get_be16(pb);
1154 get_be16(pb); /* packet size = 0 */
1156 st->codec->sample_rate = ((get_be32(pb) >> 16));
1158 //Read QT version 1 fields. In version 0 these do not exist.
1159 dprintf(c->fc, "version =%d, isom =%d\n",version,c->isom);
1162 sc->samples_per_frame = get_be32(pb);
1163 get_be32(pb); /* bytes per packet */
1164 sc->bytes_per_frame = get_be32(pb);
1165 get_be32(pb); /* bytes per sample */
1166 } else if(version==2) {
1167 get_be32(pb); /* sizeof struct only */
1168 st->codec->sample_rate = av_int2dbl(get_be64(pb)); /* float 64 */
1169 st->codec->channels = get_be32(pb);
1170 get_be32(pb); /* always 0x7F000000 */
1171 st->codec->bits_per_coded_sample = get_be32(pb); /* bits per channel if sound is uncompressed */
1172 flags = get_be32(pb); /* lpcm format specific flag */
1173 sc->bytes_per_frame = get_be32(pb); /* bytes per audio packet if constant */
1174 sc->samples_per_frame = get_be32(pb); /* lpcm frames per audio packet if constant */
1175 if (format == MKTAG('l','p','c','m'))
1176 st->codec->codec_id = ff_mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, flags);
1180 switch (st->codec->codec_id) {
1181 case CODEC_ID_PCM_S8:
1182 case CODEC_ID_PCM_U8:
1183 if (st->codec->bits_per_coded_sample == 16)
1184 st->codec->codec_id = CODEC_ID_PCM_S16BE;
1186 case CODEC_ID_PCM_S16LE:
1187 case CODEC_ID_PCM_S16BE:
1188 if (st->codec->bits_per_coded_sample == 8)
1189 st->codec->codec_id = CODEC_ID_PCM_S8;
1190 else if (st->codec->bits_per_coded_sample == 24)
1191 st->codec->codec_id =
1192 st->codec->codec_id == CODEC_ID_PCM_S16BE ?
1193 CODEC_ID_PCM_S24BE : CODEC_ID_PCM_S24LE;
1195 /* set values for old format before stsd version 1 appeared */
1196 case CODEC_ID_MACE3:
1197 sc->samples_per_frame = 6;
1198 sc->bytes_per_frame = 2*st->codec->channels;
1200 case CODEC_ID_MACE6:
1201 sc->samples_per_frame = 6;
1202 sc->bytes_per_frame = 1*st->codec->channels;
1204 case CODEC_ID_ADPCM_IMA_QT:
1205 sc->samples_per_frame = 64;
1206 sc->bytes_per_frame = 34*st->codec->channels;
1209 sc->samples_per_frame = 160;
1210 sc->bytes_per_frame = 33;
1216 bits_per_sample = av_get_bits_per_sample(st->codec->codec_id);
1217 if (bits_per_sample) {
1218 st->codec->bits_per_coded_sample = bits_per_sample;
1219 sc->sample_size = (bits_per_sample >> 3) * st->codec->channels;
1221 } else if(st->codec->codec_type==AVMEDIA_TYPE_SUBTITLE){
1222 // ttxt stsd contains display flags, justification, background
1223 // color, fonts, and default styles, so fake an atom to read it
1224 MOVAtom fake_atom = { .size = size - (url_ftell(pb) - start_pos) };
1225 if (format != AV_RL32("mp4s")) // mp4s contains a regular esds atom
1226 mov_read_glbl(c, pb, fake_atom);
1227 st->codec->codec_id= id;
1228 st->codec->width = sc->width;
1229 st->codec->height = sc->height;
1231 /* other codec type, just skip (rtp, mp4s, tmcd ...) */
1232 url_fskip(pb, size - (url_ftell(pb) - start_pos));
1234 /* this will read extra atoms at the end (wave, alac, damr, avcC, SMI ...) */
1235 a.size = size - (url_ftell(pb) - start_pos);
1237 if (mov_read_default(c, pb, a) < 0)
1239 } else if (a.size > 0)
1240 url_fskip(pb, a.size);
1243 if(st->codec->codec_type==AVMEDIA_TYPE_AUDIO && st->codec->sample_rate==0 && sc->time_scale>1)
1244 st->codec->sample_rate= sc->time_scale;
1246 /* special codec parameters handling */
1247 switch (st->codec->codec_id) {
1248 #if CONFIG_DV_DEMUXER
1249 case CODEC_ID_DVAUDIO:
1250 c->dv_fctx = avformat_alloc_context();
1251 c->dv_demux = dv_init_demux(c->dv_fctx);
1253 av_log(c->fc, AV_LOG_ERROR, "dv demux context init error\n");
1256 sc->dv_audio_container = 1;
1257 st->codec->codec_id = CODEC_ID_PCM_S16LE;
1260 /* no ifdef since parameters are always those */
1261 case CODEC_ID_QCELP:
1262 // force sample rate for qcelp when not stored in mov
1263 if (st->codec->codec_tag != MKTAG('Q','c','l','p'))
1264 st->codec->sample_rate = 8000;
1265 st->codec->frame_size= 160;
1266 st->codec->channels= 1; /* really needed */
1268 case CODEC_ID_AMR_NB:
1269 case CODEC_ID_AMR_WB:
1270 st->codec->frame_size= sc->samples_per_frame;
1271 st->codec->channels= 1; /* really needed */
1272 /* force sample rate for amr, stsd in 3gp does not store sample rate */
1273 if (st->codec->codec_id == CODEC_ID_AMR_NB)
1274 st->codec->sample_rate = 8000;
1275 else if (st->codec->codec_id == CODEC_ID_AMR_WB)
1276 st->codec->sample_rate = 16000;
1280 st->codec->codec_type = AVMEDIA_TYPE_AUDIO; /* force type after stsd for m1a hdlr */
1281 st->need_parsing = AVSTREAM_PARSE_FULL;
1284 case CODEC_ID_ADPCM_MS:
1285 case CODEC_ID_ADPCM_IMA_WAV:
1286 st->codec->block_align = sc->bytes_per_frame;
1289 if (st->codec->extradata_size == 36) {
1290 st->codec->frame_size = AV_RB32(st->codec->extradata+12);
1291 st->codec->channels = AV_RB8 (st->codec->extradata+21);
1301 static int mov_read_stsc(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1304 MOVStreamContext *sc;
1305 unsigned int i, entries;
1307 if (c->fc->nb_streams < 1)
1309 st = c->fc->streams[c->fc->nb_streams-1];
1312 get_byte(pb); /* version */
1313 get_be24(pb); /* flags */
1315 entries = get_be32(pb);
1317 dprintf(c->fc, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
1319 if(entries >= UINT_MAX / sizeof(*sc->stsc_data))
1321 sc->stsc_data = av_malloc(entries * sizeof(*sc->stsc_data));
1323 return AVERROR(ENOMEM);
1324 sc->stsc_count = entries;
1326 for(i=0; i<entries; i++) {
1327 sc->stsc_data[i].first = get_be32(pb);
1328 sc->stsc_data[i].count = get_be32(pb);
1329 sc->stsc_data[i].id = get_be32(pb);
1334 static int mov_read_stps(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1337 MOVStreamContext *sc;
1338 unsigned i, entries;
1340 if (c->fc->nb_streams < 1)
1342 st = c->fc->streams[c->fc->nb_streams-1];
1345 get_be32(pb); // version + flags
1347 entries = get_be32(pb);
1348 if (entries >= UINT_MAX / sizeof(*sc->stps_data))
1350 sc->stps_data = av_malloc(entries * sizeof(*sc->stps_data));
1352 return AVERROR(ENOMEM);
1353 sc->stps_count = entries;
1355 for (i = 0; i < entries; i++) {
1356 sc->stps_data[i] = get_be32(pb);
1357 //dprintf(c->fc, "stps %d\n", sc->stps_data[i]);
1363 static int mov_read_stss(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1366 MOVStreamContext *sc;
1367 unsigned int i, entries;
1369 if (c->fc->nb_streams < 1)
1371 st = c->fc->streams[c->fc->nb_streams-1];
1374 get_byte(pb); /* version */
1375 get_be24(pb); /* flags */
1377 entries = get_be32(pb);
1379 dprintf(c->fc, "keyframe_count = %d\n", entries);
1381 if(entries >= UINT_MAX / sizeof(int))
1383 sc->keyframes = av_malloc(entries * sizeof(int));
1385 return AVERROR(ENOMEM);
1386 sc->keyframe_count = entries;
1388 for(i=0; i<entries; i++) {
1389 sc->keyframes[i] = get_be32(pb);
1390 //dprintf(c->fc, "keyframes[]=%d\n", sc->keyframes[i]);
1395 static int mov_read_stsz(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1398 MOVStreamContext *sc;
1399 unsigned int i, entries, sample_size, field_size, num_bytes;
1403 if (c->fc->nb_streams < 1)
1405 st = c->fc->streams[c->fc->nb_streams-1];
1408 get_byte(pb); /* version */
1409 get_be24(pb); /* flags */
1411 if (atom.type == MKTAG('s','t','s','z')) {
1412 sample_size = get_be32(pb);
1413 if (!sc->sample_size) /* do not overwrite value computed in stsd */
1414 sc->sample_size = sample_size;
1418 get_be24(pb); /* reserved */
1419 field_size = get_byte(pb);
1421 entries = get_be32(pb);
1423 dprintf(c->fc, "sample_size = %d sample_count = %d\n", sc->sample_size, entries);
1425 sc->sample_count = entries;
1429 if (field_size != 4 && field_size != 8 && field_size != 16 && field_size != 32) {
1430 av_log(c->fc, AV_LOG_ERROR, "Invalid sample field size %d\n", field_size);
1434 if (entries >= UINT_MAX / sizeof(int) || entries >= (UINT_MAX - 4) / field_size)
1436 sc->sample_sizes = av_malloc(entries * sizeof(int));
1437 if (!sc->sample_sizes)
1438 return AVERROR(ENOMEM);
1440 num_bytes = (entries*field_size+4)>>3;
1442 buf = av_malloc(num_bytes+FF_INPUT_BUFFER_PADDING_SIZE);
1444 av_freep(&sc->sample_sizes);
1445 return AVERROR(ENOMEM);
1448 if (get_buffer(pb, buf, num_bytes) < num_bytes) {
1449 av_freep(&sc->sample_sizes);
1454 init_get_bits(&gb, buf, 8*num_bytes);
1456 for(i=0; i<entries; i++)
1457 sc->sample_sizes[i] = get_bits_long(&gb, field_size);
1463 static int mov_read_stts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1466 MOVStreamContext *sc;
1467 unsigned int i, entries;
1469 int64_t total_sample_count=0;
1471 if (c->fc->nb_streams < 1)
1473 st = c->fc->streams[c->fc->nb_streams-1];
1476 get_byte(pb); /* version */
1477 get_be24(pb); /* flags */
1478 entries = get_be32(pb);
1480 dprintf(c->fc, "track[%i].stts.entries = %i\n", c->fc->nb_streams-1, entries);
1482 if(entries >= UINT_MAX / sizeof(*sc->stts_data))
1484 sc->stts_data = av_malloc(entries * sizeof(*sc->stts_data));
1486 return AVERROR(ENOMEM);
1487 sc->stts_count = entries;
1489 for(i=0; i<entries; i++) {
1490 int sample_duration;
1493 sample_count=get_be32(pb);
1494 sample_duration = get_be32(pb);
1495 sc->stts_data[i].count= sample_count;
1496 sc->stts_data[i].duration= sample_duration;
1498 dprintf(c->fc, "sample_count=%d, sample_duration=%d\n",sample_count,sample_duration);
1500 duration+=(int64_t)sample_duration*sample_count;
1501 total_sample_count+=sample_count;
1504 st->nb_frames= total_sample_count;
1506 st->duration= duration;
1510 static int mov_read_ctts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1513 MOVStreamContext *sc;
1514 unsigned int i, entries;
1516 if (c->fc->nb_streams < 1)
1518 st = c->fc->streams[c->fc->nb_streams-1];
1521 get_byte(pb); /* version */
1522 get_be24(pb); /* flags */
1523 entries = get_be32(pb);
1525 dprintf(c->fc, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
1527 if(entries >= UINT_MAX / sizeof(*sc->ctts_data))
1529 sc->ctts_data = av_malloc(entries * sizeof(*sc->ctts_data));
1531 return AVERROR(ENOMEM);
1532 sc->ctts_count = entries;
1534 for(i=0; i<entries; i++) {
1535 int count =get_be32(pb);
1536 int duration =get_be32(pb);
1538 sc->ctts_data[i].count = count;
1539 sc->ctts_data[i].duration= duration;
1541 sc->dts_shift = FFMAX(sc->dts_shift, -duration);
1544 dprintf(c->fc, "dts shift %d\n", sc->dts_shift);
1549 static void mov_build_index(MOVContext *mov, AVStream *st)
1551 MOVStreamContext *sc = st->priv_data;
1552 int64_t current_offset;
1553 int64_t current_dts = 0;
1554 unsigned int stts_index = 0;
1555 unsigned int stsc_index = 0;
1556 unsigned int stss_index = 0;
1557 unsigned int stps_index = 0;
1559 uint64_t stream_size = 0;
1561 /* adjust first dts according to edit list */
1562 if (sc->time_offset) {
1563 int rescaled = sc->time_offset < 0 ? av_rescale(sc->time_offset, sc->time_scale, mov->time_scale) : sc->time_offset;
1564 current_dts = -rescaled;
1565 if (sc->ctts_data && sc->ctts_data[0].duration / sc->stts_data[0].duration > 16) {
1566 /* more than 16 frames delay, dts are likely wrong
1567 this happens with files created by iMovie */
1569 st->codec->has_b_frames = 1;
1573 /* only use old uncompressed audio chunk demuxing when stts specifies it */
1574 if (!(st->codec->codec_type == AVMEDIA_TYPE_AUDIO &&
1575 sc->stts_count == 1 && sc->stts_data[0].duration == 1)) {
1576 unsigned int current_sample = 0;
1577 unsigned int stts_sample = 0;
1578 unsigned int sample_size;
1579 unsigned int distance = 0;
1580 int key_off = sc->keyframes && sc->keyframes[0] == 1;
1582 current_dts -= sc->dts_shift;
1584 if (sc->sample_count >= UINT_MAX / sizeof(*st->index_entries))
1586 st->index_entries = av_malloc(sc->sample_count*sizeof(*st->index_entries));
1587 if (!st->index_entries)
1589 st->index_entries_allocated_size = sc->sample_count*sizeof(*st->index_entries);
1591 for (i = 0; i < sc->chunk_count; i++) {
1592 current_offset = sc->chunk_offsets[i];
1593 if (stsc_index + 1 < sc->stsc_count &&
1594 i + 1 == sc->stsc_data[stsc_index + 1].first)
1596 for (j = 0; j < sc->stsc_data[stsc_index].count; j++) {
1598 if (current_sample >= sc->sample_count) {
1599 av_log(mov->fc, AV_LOG_ERROR, "wrong sample count\n");
1603 if (!sc->keyframe_count || current_sample+key_off == sc->keyframes[stss_index]) {
1605 if (stss_index + 1 < sc->keyframe_count)
1607 } else if (sc->stps_count && current_sample+key_off == sc->stps_data[stps_index]) {
1609 if (stps_index + 1 < sc->stps_count)
1614 sample_size = sc->sample_size > 0 ? sc->sample_size : sc->sample_sizes[current_sample];
1615 if(sc->pseudo_stream_id == -1 ||
1616 sc->stsc_data[stsc_index].id - 1 == sc->pseudo_stream_id) {
1617 AVIndexEntry *e = &st->index_entries[st->nb_index_entries++];
1618 e->pos = current_offset;
1619 e->timestamp = current_dts;
1620 e->size = sample_size;
1621 e->min_distance = distance;
1622 e->flags = keyframe ? AVINDEX_KEYFRAME : 0;
1623 dprintf(mov->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1624 "size %d, distance %d, keyframe %d\n", st->index, current_sample,
1625 current_offset, current_dts, sample_size, distance, keyframe);
1628 current_offset += sample_size;
1629 stream_size += sample_size;
1630 current_dts += sc->stts_data[stts_index].duration;
1634 if (stts_index + 1 < sc->stts_count && stts_sample == sc->stts_data[stts_index].count) {
1640 if (st->duration > 0)
1641 st->codec->bit_rate = stream_size*8*sc->time_scale/st->duration;
1643 unsigned chunk_samples, total = 0;
1645 // compute total chunk count
1646 for (i = 0; i < sc->stsc_count; i++) {
1647 unsigned count, chunk_count;
1649 chunk_samples = sc->stsc_data[i].count;
1650 if (sc->samples_per_frame && chunk_samples % sc->samples_per_frame) {
1651 av_log(mov->fc, AV_LOG_ERROR, "error unaligned chunk\n");
1655 if (sc->samples_per_frame >= 160) { // gsm
1656 count = chunk_samples / sc->samples_per_frame;
1657 } else if (sc->samples_per_frame > 1) {
1658 unsigned samples = (1024/sc->samples_per_frame)*sc->samples_per_frame;
1659 count = (chunk_samples+samples-1) / samples;
1661 count = (chunk_samples+1023) / 1024;
1664 if (i < sc->stsc_count - 1)
1665 chunk_count = sc->stsc_data[i+1].first - sc->stsc_data[i].first;
1667 chunk_count = sc->chunk_count - (sc->stsc_data[i].first - 1);
1668 total += chunk_count * count;
1671 dprintf(mov->fc, "chunk count %d\n", total);
1672 if (total >= UINT_MAX / sizeof(*st->index_entries))
1674 st->index_entries = av_malloc(total*sizeof(*st->index_entries));
1675 if (!st->index_entries)
1677 st->index_entries_allocated_size = total*sizeof(*st->index_entries);
1680 for (i = 0; i < sc->chunk_count; i++) {
1681 current_offset = sc->chunk_offsets[i];
1682 if (stsc_index + 1 < sc->stsc_count &&
1683 i + 1 == sc->stsc_data[stsc_index + 1].first)
1685 chunk_samples = sc->stsc_data[stsc_index].count;
1687 while (chunk_samples > 0) {
1689 unsigned size, samples;
1691 if (sc->samples_per_frame >= 160) { // gsm
1692 samples = sc->samples_per_frame;
1693 size = sc->bytes_per_frame;
1695 if (sc->samples_per_frame > 1) {
1696 samples = FFMIN((1024 / sc->samples_per_frame)*
1697 sc->samples_per_frame, chunk_samples);
1698 size = (samples / sc->samples_per_frame) * sc->bytes_per_frame;
1700 samples = FFMIN(1024, chunk_samples);
1701 size = samples * sc->sample_size;
1705 if (st->nb_index_entries >= total) {
1706 av_log(mov->fc, AV_LOG_ERROR, "wrong chunk count %d\n", total);
1709 e = &st->index_entries[st->nb_index_entries++];
1710 e->pos = current_offset;
1711 e->timestamp = current_dts;
1713 e->min_distance = 0;
1714 e->flags = AVINDEX_KEYFRAME;
1715 dprintf(mov->fc, "AVIndex stream %d, chunk %d, offset %"PRIx64", dts %"PRId64", "
1716 "size %d, duration %d\n", st->index, i, current_offset, current_dts,
1719 current_offset += size;
1720 current_dts += samples;
1721 chunk_samples -= samples;
1727 static int mov_open_dref(ByteIOContext **pb, char *src, MOVDref *ref)
1729 /* try relative path, we do not try the absolute because it can leak information about our
1730 system to an attacker */
1731 if (ref->nlvl_to > 0 && ref->nlvl_from > 0) {
1732 char filename[1024];
1736 /* find a source dir */
1737 src_path = strrchr(src, '/');
1743 /* find a next level down to target */
1744 for (i = 0, l = strlen(ref->path) - 1; l >= 0; l--)
1745 if (ref->path[l] == '/') {
1746 if (i == ref->nlvl_to - 1)
1752 /* compose filename if next level down to target was found */
1753 if (i == ref->nlvl_to - 1 && src_path - src < sizeof(filename)) {
1754 memcpy(filename, src, src_path - src);
1755 filename[src_path - src] = 0;
1757 for (i = 1; i < ref->nlvl_from; i++)
1758 av_strlcat(filename, "../", 1024);
1760 av_strlcat(filename, ref->path + l + 1, 1024);
1762 if (!url_fopen(pb, filename, URL_RDONLY))
1767 return AVERROR(ENOENT);
1770 static int mov_read_trak(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1773 MOVStreamContext *sc;
1776 st = av_new_stream(c->fc, c->fc->nb_streams);
1777 if (!st) return AVERROR(ENOMEM);
1778 sc = av_mallocz(sizeof(MOVStreamContext));
1779 if (!sc) return AVERROR(ENOMEM);
1782 st->codec->codec_type = AVMEDIA_TYPE_DATA;
1783 sc->ffindex = st->index;
1785 if ((ret = mov_read_default(c, pb, atom)) < 0)
1789 if (sc->chunk_count && (!sc->stts_count || !sc->stsc_count ||
1790 (!sc->sample_size && !sc->sample_count))) {
1791 av_log(c->fc, AV_LOG_ERROR, "stream %d, missing mandatory atoms, broken header\n",
1796 if (!sc->time_scale) {
1797 av_log(c->fc, AV_LOG_WARNING, "stream %d, timescale not set\n", st->index);
1798 sc->time_scale = c->time_scale;
1799 if (!sc->time_scale)
1803 av_set_pts_info(st, 64, 1, sc->time_scale);
1805 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO &&
1806 !st->codec->frame_size && sc->stts_count == 1) {
1807 st->codec->frame_size = av_rescale(sc->stts_data[0].duration,
1808 st->codec->sample_rate, sc->time_scale);
1809 dprintf(c->fc, "frame size %d\n", st->codec->frame_size);
1812 mov_build_index(c, st);
1814 if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
1815 MOVDref *dref = &sc->drefs[sc->dref_id - 1];
1816 if (mov_open_dref(&sc->pb, c->fc->filename, dref) < 0)
1817 av_log(c->fc, AV_LOG_ERROR,
1818 "stream %d, error opening alias: path='%s', dir='%s', "
1819 "filename='%s', volume='%s', nlvl_from=%d, nlvl_to=%d\n",
1820 st->index, dref->path, dref->dir, dref->filename,
1821 dref->volume, dref->nlvl_from, dref->nlvl_to);
1825 if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
1826 if (!st->sample_aspect_ratio.num &&
1827 (st->codec->width != sc->width || st->codec->height != sc->height)) {
1828 st->sample_aspect_ratio = av_d2q(((double)st->codec->height * sc->width) /
1829 ((double)st->codec->width * sc->height), INT_MAX);
1832 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1833 sc->time_scale*st->nb_frames, st->duration, INT_MAX);
1836 switch (st->codec->codec_id) {
1837 #if CONFIG_H261_DECODER
1840 #if CONFIG_H263_DECODER
1843 #if CONFIG_H264_DECODER
1846 #if CONFIG_MPEG4_DECODER
1847 case CODEC_ID_MPEG4:
1849 st->codec->width = 0; /* let decoder init width/height */
1850 st->codec->height= 0;
1854 /* Do not need those anymore. */
1855 av_freep(&sc->chunk_offsets);
1856 av_freep(&sc->stsc_data);
1857 av_freep(&sc->sample_sizes);
1858 av_freep(&sc->keyframes);
1859 av_freep(&sc->stts_data);
1860 av_freep(&sc->stps_data);
1865 static int mov_read_ilst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1868 c->itunes_metadata = 1;
1869 ret = mov_read_default(c, pb, atom);
1870 c->itunes_metadata = 0;
1874 static int mov_read_meta(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1876 while (atom.size > 8) {
1877 uint32_t tag = get_le32(pb);
1879 if (tag == MKTAG('h','d','l','r')) {
1880 url_fseek(pb, -8, SEEK_CUR);
1882 return mov_read_default(c, pb, atom);
1888 static int mov_read_tkhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1893 int64_t disp_transform[2];
1894 int display_matrix[3][2];
1896 MOVStreamContext *sc;
1899 if (c->fc->nb_streams < 1)
1901 st = c->fc->streams[c->fc->nb_streams-1];
1904 version = get_byte(pb);
1905 get_be24(pb); /* flags */
1907 MOV_TRACK_ENABLED 0x0001
1908 MOV_TRACK_IN_MOVIE 0x0002
1909 MOV_TRACK_IN_PREVIEW 0x0004
1910 MOV_TRACK_IN_POSTER 0x0008
1917 get_be32(pb); /* creation time */
1918 get_be32(pb); /* modification time */
1920 st->id = (int)get_be32(pb); /* track id (NOT 0 !)*/
1921 get_be32(pb); /* reserved */
1923 /* highlevel (considering edits) duration in movie timebase */
1924 (version == 1) ? get_be64(pb) : get_be32(pb);
1925 get_be32(pb); /* reserved */
1926 get_be32(pb); /* reserved */
1928 get_be16(pb); /* layer */
1929 get_be16(pb); /* alternate group */
1930 get_be16(pb); /* volume */
1931 get_be16(pb); /* reserved */
1933 //read in the display matrix (outlined in ISO 14496-12, Section 6.2.2)
1934 // they're kept in fixed point format through all calculations
1935 // ignore u,v,z b/c we don't need the scale factor to calc aspect ratio
1936 for (i = 0; i < 3; i++) {
1937 display_matrix[i][0] = get_be32(pb); // 16.16 fixed point
1938 display_matrix[i][1] = get_be32(pb); // 16.16 fixed point
1939 get_be32(pb); // 2.30 fixed point (not used)
1942 width = get_be32(pb); // 16.16 fixed point track width
1943 height = get_be32(pb); // 16.16 fixed point track height
1944 sc->width = width >> 16;
1945 sc->height = height >> 16;
1947 // transform the display width/height according to the matrix
1948 // skip this if the display matrix is the default identity matrix
1949 // or if it is rotating the picture, ex iPhone 3GS
1950 // to keep the same scale, use [width height 1<<16]
1951 if (width && height &&
1952 ((display_matrix[0][0] != 65536 ||
1953 display_matrix[1][1] != 65536) &&
1954 !display_matrix[0][1] &&
1955 !display_matrix[1][0] &&
1956 !display_matrix[2][0] && !display_matrix[2][1])) {
1957 for (i = 0; i < 2; i++)
1959 (int64_t) width * display_matrix[0][i] +
1960 (int64_t) height * display_matrix[1][i] +
1961 ((int64_t) display_matrix[2][i] << 16);
1963 //sample aspect ratio is new width/height divided by old width/height
1964 st->sample_aspect_ratio = av_d2q(
1965 ((double) disp_transform[0] * height) /
1966 ((double) disp_transform[1] * width), INT_MAX);
1971 static int mov_read_tfhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1973 MOVFragment *frag = &c->fragment;
1974 MOVTrackExt *trex = NULL;
1975 int flags, track_id, i;
1977 get_byte(pb); /* version */
1978 flags = get_be24(pb);
1980 track_id = get_be32(pb);
1983 frag->track_id = track_id;
1984 for (i = 0; i < c->trex_count; i++)
1985 if (c->trex_data[i].track_id == frag->track_id) {
1986 trex = &c->trex_data[i];
1990 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding trex\n");
1994 if (flags & 0x01) frag->base_data_offset = get_be64(pb);
1995 else frag->base_data_offset = frag->moof_offset;
1996 if (flags & 0x02) frag->stsd_id = get_be32(pb);
1997 else frag->stsd_id = trex->stsd_id;
1999 frag->duration = flags & 0x08 ? get_be32(pb) : trex->duration;
2000 frag->size = flags & 0x10 ? get_be32(pb) : trex->size;
2001 frag->flags = flags & 0x20 ? get_be32(pb) : trex->flags;
2002 dprintf(c->fc, "frag flags 0x%x\n", frag->flags);
2006 static int mov_read_chap(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
2008 c->chapter_track = get_be32(pb);
2012 static int mov_read_trex(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
2016 if ((uint64_t)c->trex_count+1 >= UINT_MAX / sizeof(*c->trex_data))
2018 trex = av_realloc(c->trex_data, (c->trex_count+1)*sizeof(*c->trex_data));
2020 return AVERROR(ENOMEM);
2021 c->trex_data = trex;
2022 trex = &c->trex_data[c->trex_count++];
2023 get_byte(pb); /* version */
2024 get_be24(pb); /* flags */
2025 trex->track_id = get_be32(pb);
2026 trex->stsd_id = get_be32(pb);
2027 trex->duration = get_be32(pb);
2028 trex->size = get_be32(pb);
2029 trex->flags = get_be32(pb);
2033 static int mov_read_trun(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
2035 MOVFragment *frag = &c->fragment;
2036 AVStream *st = NULL;
2037 MOVStreamContext *sc;
2040 int data_offset = 0;
2041 unsigned entries, first_sample_flags = frag->flags;
2042 int flags, distance, i;
2044 for (i = 0; i < c->fc->nb_streams; i++) {
2045 if (c->fc->streams[i]->id == frag->track_id) {
2046 st = c->fc->streams[i];
2051 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding track id %d\n", frag->track_id);
2055 if (sc->pseudo_stream_id+1 != frag->stsd_id)
2057 get_byte(pb); /* version */
2058 flags = get_be24(pb);
2059 entries = get_be32(pb);
2060 dprintf(c->fc, "flags 0x%x entries %d\n", flags, entries);
2061 if (flags & 0x001) data_offset = get_be32(pb);
2062 if (flags & 0x004) first_sample_flags = get_be32(pb);
2063 if (flags & 0x800) {
2065 if ((uint64_t)entries+sc->ctts_count >= UINT_MAX/sizeof(*sc->ctts_data))
2067 ctts_data = av_realloc(sc->ctts_data,
2068 (entries+sc->ctts_count)*sizeof(*sc->ctts_data));
2070 return AVERROR(ENOMEM);
2071 sc->ctts_data = ctts_data;
2074 offset = frag->base_data_offset + data_offset;
2076 dprintf(c->fc, "first sample flags 0x%x\n", first_sample_flags);
2077 for (i = 0; i < entries; i++) {
2078 unsigned sample_size = frag->size;
2079 int sample_flags = i ? frag->flags : first_sample_flags;
2080 unsigned sample_duration = frag->duration;
2083 if (flags & 0x100) sample_duration = get_be32(pb);
2084 if (flags & 0x200) sample_size = get_be32(pb);
2085 if (flags & 0x400) sample_flags = get_be32(pb);
2086 if (flags & 0x800) {
2087 sc->ctts_data[sc->ctts_count].count = 1;
2088 sc->ctts_data[sc->ctts_count].duration = get_be32(pb);
2091 if ((keyframe = st->codec->codec_type == AVMEDIA_TYPE_AUDIO ||
2092 (flags & 0x004 && !i && !sample_flags) || sample_flags & 0x2000000))
2094 av_add_index_entry(st, offset, dts, sample_size, distance,
2095 keyframe ? AVINDEX_KEYFRAME : 0);
2096 dprintf(c->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
2097 "size %d, distance %d, keyframe %d\n", st->index, sc->sample_count+i,
2098 offset, dts, sample_size, distance, keyframe);
2100 dts += sample_duration;
2101 offset += sample_size;
2103 frag->moof_offset = offset;
2108 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
2109 /* like the files created with Adobe Premiere 5.0, for samples see */
2110 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
2111 static int mov_read_wide(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
2116 return 0; /* continue */
2117 if (get_be32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
2118 url_fskip(pb, atom.size - 4);
2121 atom.type = get_le32(pb);
2123 if (atom.type != MKTAG('m','d','a','t')) {
2124 url_fskip(pb, atom.size);
2127 err = mov_read_mdat(c, pb, atom);
2131 static int mov_read_cmov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
2136 uint8_t *moov_data; /* uncompressed data */
2137 long cmov_len, moov_len;
2140 get_be32(pb); /* dcom atom */
2141 if (get_le32(pb) != MKTAG('d','c','o','m'))
2143 if (get_le32(pb) != MKTAG('z','l','i','b')) {
2144 av_log(c->fc, AV_LOG_ERROR, "unknown compression for cmov atom !");
2147 get_be32(pb); /* cmvd atom */
2148 if (get_le32(pb) != MKTAG('c','m','v','d'))
2150 moov_len = get_be32(pb); /* uncompressed size */
2151 cmov_len = atom.size - 6 * 4;
2153 cmov_data = av_malloc(cmov_len);
2155 return AVERROR(ENOMEM);
2156 moov_data = av_malloc(moov_len);
2159 return AVERROR(ENOMEM);
2161 get_buffer(pb, cmov_data, cmov_len);
2162 if(uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
2163 goto free_and_return;
2164 if(init_put_byte(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL, NULL) != 0)
2165 goto free_and_return;
2166 atom.type = MKTAG('m','o','o','v');
2167 atom.size = moov_len;
2169 // { int fd = open("/tmp/uncompheader.mov", O_WRONLY | O_CREAT); write(fd, moov_data, moov_len); close(fd); }
2171 ret = mov_read_default(c, &ctx, atom);
2177 av_log(c->fc, AV_LOG_ERROR, "this file requires zlib support compiled in\n");
2182 /* edit list atom */
2183 static int mov_read_elst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
2185 MOVStreamContext *sc;
2188 if (c->fc->nb_streams < 1)
2190 sc = c->fc->streams[c->fc->nb_streams-1]->priv_data;
2192 get_byte(pb); /* version */
2193 get_be24(pb); /* flags */
2194 edit_count = get_be32(pb); /* entries */
2196 if((uint64_t)edit_count*12+8 > atom.size)
2199 for(i=0; i<edit_count; i++){
2201 int duration = get_be32(pb); /* Track duration */
2202 time = get_be32(pb); /* Media time */
2203 get_be32(pb); /* Media rate */
2204 if (i == 0 && time >= -1) {
2205 sc->time_offset = time != -1 ? time : -duration;
2210 av_log(c->fc, AV_LOG_WARNING, "multiple edit list entries, "
2211 "a/v desync might occur, patch welcome\n");
2213 dprintf(c->fc, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, edit_count);
2217 static const MOVParseTableEntry mov_default_parse_table[] = {
2218 { MKTAG('a','v','s','s'), mov_read_extradata },
2219 { MKTAG('c','h','p','l'), mov_read_chpl },
2220 { MKTAG('c','o','6','4'), mov_read_stco },
2221 { MKTAG('c','t','t','s'), mov_read_ctts }, /* composition time to sample */
2222 { MKTAG('d','i','n','f'), mov_read_default },
2223 { MKTAG('d','r','e','f'), mov_read_dref },
2224 { MKTAG('e','d','t','s'), mov_read_default },
2225 { MKTAG('e','l','s','t'), mov_read_elst },
2226 { MKTAG('e','n','d','a'), mov_read_enda },
2227 { MKTAG('f','i','e','l'), mov_read_extradata },
2228 { MKTAG('f','t','y','p'), mov_read_ftyp },
2229 { MKTAG('g','l','b','l'), mov_read_glbl },
2230 { MKTAG('h','d','l','r'), mov_read_hdlr },
2231 { MKTAG('i','l','s','t'), mov_read_ilst },
2232 { MKTAG('j','p','2','h'), mov_read_extradata },
2233 { MKTAG('m','d','a','t'), mov_read_mdat },
2234 { MKTAG('m','d','h','d'), mov_read_mdhd },
2235 { MKTAG('m','d','i','a'), mov_read_default },
2236 { MKTAG('m','e','t','a'), mov_read_meta },
2237 { MKTAG('m','i','n','f'), mov_read_default },
2238 { MKTAG('m','o','o','f'), mov_read_moof },
2239 { MKTAG('m','o','o','v'), mov_read_moov },
2240 { MKTAG('m','v','e','x'), mov_read_default },
2241 { MKTAG('m','v','h','d'), mov_read_mvhd },
2242 { MKTAG('S','M','I',' '), mov_read_smi }, /* Sorenson extension ??? */
2243 { MKTAG('a','l','a','c'), mov_read_extradata }, /* alac specific atom */
2244 { MKTAG('a','v','c','C'), mov_read_glbl },
2245 { MKTAG('p','a','s','p'), mov_read_pasp },
2246 { MKTAG('s','t','b','l'), mov_read_default },
2247 { MKTAG('s','t','c','o'), mov_read_stco },
2248 { MKTAG('s','t','p','s'), mov_read_stps },
2249 { MKTAG('s','t','r','f'), mov_read_strf },
2250 { MKTAG('s','t','s','c'), mov_read_stsc },
2251 { MKTAG('s','t','s','d'), mov_read_stsd }, /* sample description */
2252 { MKTAG('s','t','s','s'), mov_read_stss }, /* sync sample */
2253 { MKTAG('s','t','s','z'), mov_read_stsz }, /* sample size */
2254 { MKTAG('s','t','t','s'), mov_read_stts },
2255 { MKTAG('s','t','z','2'), mov_read_stsz }, /* compact sample size */
2256 { MKTAG('t','k','h','d'), mov_read_tkhd }, /* track header */
2257 { MKTAG('t','f','h','d'), mov_read_tfhd }, /* track fragment header */
2258 { MKTAG('t','r','a','k'), mov_read_trak },
2259 { MKTAG('t','r','a','f'), mov_read_default },
2260 { MKTAG('t','r','e','f'), mov_read_default },
2261 { MKTAG('c','h','a','p'), mov_read_chap },
2262 { MKTAG('t','r','e','x'), mov_read_trex },
2263 { MKTAG('t','r','u','n'), mov_read_trun },
2264 { MKTAG('u','d','t','a'), mov_read_default },
2265 { MKTAG('w','a','v','e'), mov_read_wave },
2266 { MKTAG('e','s','d','s'), mov_read_esds },
2267 { MKTAG('d','a','c','3'), mov_read_dac3 }, /* AC-3 info */
2268 { MKTAG('w','i','d','e'), mov_read_wide }, /* place holder */
2269 { MKTAG('c','m','o','v'), mov_read_cmov },
2273 static int mov_probe(AVProbeData *p)
2275 unsigned int offset;
2279 /* check file header */
2282 /* ignore invalid offset */
2283 if ((offset + 8) > (unsigned int)p->buf_size)
2285 tag = AV_RL32(p->buf + offset + 4);
2287 /* check for obvious tags */
2288 case MKTAG('j','P',' ',' '): /* jpeg 2000 signature */
2289 case MKTAG('m','o','o','v'):
2290 case MKTAG('m','d','a','t'):
2291 case MKTAG('p','n','o','t'): /* detect movs with preview pics like ew.mov and april.mov */
2292 case MKTAG('u','d','t','a'): /* Packet Video PVAuthor adds this and a lot of more junk */
2293 case MKTAG('f','t','y','p'):
2294 return AVPROBE_SCORE_MAX;
2295 /* those are more common words, so rate then a bit less */
2296 case MKTAG('e','d','i','w'): /* xdcam files have reverted first tags */
2297 case MKTAG('w','i','d','e'):
2298 case MKTAG('f','r','e','e'):
2299 case MKTAG('j','u','n','k'):
2300 case MKTAG('p','i','c','t'):
2301 return AVPROBE_SCORE_MAX - 5;
2302 case MKTAG(0x82,0x82,0x7f,0x7d):
2303 case MKTAG('s','k','i','p'):
2304 case MKTAG('u','u','i','d'):
2305 case MKTAG('p','r','f','l'):
2306 offset = AV_RB32(p->buf+offset) + offset;
2307 /* if we only find those cause probedata is too small at least rate them */
2308 score = AVPROBE_SCORE_MAX - 50;
2311 /* unrecognized tag */
2318 // must be done after parsing all trak because there's no order requirement
2319 static void mov_read_chapters(AVFormatContext *s)
2321 MOVContext *mov = s->priv_data;
2322 AVStream *st = NULL;
2323 MOVStreamContext *sc;
2325 uint8_t *title = NULL;
2326 int i, len, i8, i16;
2328 for (i = 0; i < s->nb_streams; i++)
2329 if (s->streams[i]->id == mov->chapter_track) {
2334 av_log(s, AV_LOG_ERROR, "Referenced QT chapter track not found\n");
2338 st->discard = AVDISCARD_ALL;
2340 cur_pos = url_ftell(sc->pb);
2342 for (i = 0; i < st->nb_index_entries; i++) {
2343 AVIndexEntry *sample = &st->index_entries[i];
2344 int64_t end = i+1 < st->nb_index_entries ? st->index_entries[i+1].timestamp : st->duration;
2346 if (url_fseek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
2347 av_log(s, AV_LOG_ERROR, "Chapter %d not found in file\n", i);
2351 title = av_malloc(sample->size+2);
2352 get_buffer(sc->pb, title, sample->size);
2354 // the first two bytes are the length of the title
2355 len = AV_RB16(title);
2356 if (len > sample->size-2)
2359 // The samples could theoretically be in any encoding if there's an encd
2360 // atom following, but in practice are only utf-8 or utf-16, distinguished
2361 // instead by the presence of a BOM
2362 if (AV_RB16(title+2) == 0xfeff) {
2363 uint8_t *utf8 = av_malloc(2*len+3);
2369 GET_UTF16(ch, i16 < len ? AV_RB16(title + (i16+=2)) : 0, break;)
2370 PUT_UTF8(ch, tmp, if (i8 < 2*len) utf8[2+i8++] = tmp;)
2377 ff_new_chapter(s, i, st->time_base, sample->timestamp, end, title+2);
2382 url_fseek(sc->pb, cur_pos, SEEK_SET);
2385 static int mov_read_header(AVFormatContext *s, AVFormatParameters *ap)
2387 MOVContext *mov = s->priv_data;
2388 ByteIOContext *pb = s->pb;
2390 MOVAtom atom = { AV_RL32("root") };
2393 /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
2394 if(!url_is_streamed(pb))
2395 atom.size = url_fsize(pb);
2397 atom.size = INT64_MAX;
2399 /* check MOV header */
2400 if ((err = mov_read_default(mov, pb, atom)) < 0) {
2401 av_log(s, AV_LOG_ERROR, "error reading header: %d\n", err);
2404 if (!mov->found_moov) {
2405 av_log(s, AV_LOG_ERROR, "moov atom not found\n");
2408 dprintf(mov->fc, "on_parse_exit_offset=%lld\n", url_ftell(pb));
2410 if (!url_is_streamed(pb) && mov->chapter_track > 0)
2411 mov_read_chapters(s);
2416 static AVIndexEntry *mov_find_next_sample(AVFormatContext *s, AVStream **st)
2418 AVIndexEntry *sample = NULL;
2419 int64_t best_dts = INT64_MAX;
2421 for (i = 0; i < s->nb_streams; i++) {
2422 AVStream *avst = s->streams[i];
2423 MOVStreamContext *msc = avst->priv_data;
2424 if (msc->pb && msc->current_sample < avst->nb_index_entries) {
2425 AVIndexEntry *current_sample = &avst->index_entries[msc->current_sample];
2426 int64_t dts = av_rescale(current_sample->timestamp, AV_TIME_BASE, msc->time_scale);
2427 dprintf(s, "stream %d, sample %d, dts %"PRId64"\n", i, msc->current_sample, dts);
2428 if (!sample || (url_is_streamed(s->pb) && current_sample->pos < sample->pos) ||
2429 (!url_is_streamed(s->pb) &&
2430 ((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
2431 ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
2432 (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
2433 sample = current_sample;
2442 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
2444 MOVContext *mov = s->priv_data;
2445 MOVStreamContext *sc;
2446 AVIndexEntry *sample;
2447 AVStream *st = NULL;
2450 sample = mov_find_next_sample(s, &st);
2452 mov->found_mdat = 0;
2453 if (!url_is_streamed(s->pb) ||
2454 mov_read_default(mov, s->pb, (MOVAtom){ AV_RL32("root"), INT64_MAX }) < 0 ||
2457 dprintf(s, "read fragments, offset 0x%llx\n", url_ftell(s->pb));
2461 /* must be done just before reading, to avoid infinite loop on sample */
2462 sc->current_sample++;
2464 if (st->discard != AVDISCARD_ALL) {
2465 if (url_fseek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
2466 av_log(mov->fc, AV_LOG_ERROR, "stream %d, offset 0x%"PRIx64": partial file\n",
2467 sc->ffindex, sample->pos);
2470 ret = av_get_packet(sc->pb, pkt, sample->size);
2473 #if CONFIG_DV_DEMUXER
2474 if (mov->dv_demux && sc->dv_audio_container) {
2475 dv_produce_packet(mov->dv_demux, pkt, pkt->data, pkt->size);
2478 ret = dv_get_packet(mov->dv_demux, pkt);
2485 pkt->stream_index = sc->ffindex;
2486 pkt->dts = sample->timestamp;
2487 if (sc->ctts_data) {
2488 pkt->pts = pkt->dts + sc->dts_shift + sc->ctts_data[sc->ctts_index].duration;
2489 /* update ctts context */
2491 if (sc->ctts_index < sc->ctts_count &&
2492 sc->ctts_data[sc->ctts_index].count == sc->ctts_sample) {
2494 sc->ctts_sample = 0;
2497 pkt->dts = AV_NOPTS_VALUE;
2499 int64_t next_dts = (sc->current_sample < st->nb_index_entries) ?
2500 st->index_entries[sc->current_sample].timestamp : st->duration;
2501 pkt->duration = next_dts - pkt->dts;
2502 pkt->pts = pkt->dts;
2504 if (st->discard == AVDISCARD_ALL)
2506 pkt->flags |= sample->flags & AVINDEX_KEYFRAME ? AV_PKT_FLAG_KEY : 0;
2507 pkt->pos = sample->pos;
2508 dprintf(s, "stream %d, pts %"PRId64", dts %"PRId64", pos 0x%"PRIx64", duration %d\n",
2509 pkt->stream_index, pkt->pts, pkt->dts, pkt->pos, pkt->duration);
2513 static int mov_seek_stream(AVFormatContext *s, AVStream *st, int64_t timestamp, int flags)
2515 MOVStreamContext *sc = st->priv_data;
2516 int sample, time_sample;
2519 sample = av_index_search_timestamp(st, timestamp, flags);
2520 dprintf(s, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
2521 if (sample < 0 && st->nb_index_entries && timestamp < st->index_entries[0].timestamp)
2523 if (sample < 0) /* not sure what to do */
2525 sc->current_sample = sample;
2526 dprintf(s, "stream %d, found sample %d\n", st->index, sc->current_sample);
2527 /* adjust ctts index */
2528 if (sc->ctts_data) {
2530 for (i = 0; i < sc->ctts_count; i++) {
2531 int next = time_sample + sc->ctts_data[i].count;
2532 if (next > sc->current_sample) {
2534 sc->ctts_sample = sc->current_sample - time_sample;
2543 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
2546 int64_t seek_timestamp, timestamp;
2550 if (stream_index >= s->nb_streams)
2552 if (sample_time < 0)
2555 st = s->streams[stream_index];
2556 sample = mov_seek_stream(s, st, sample_time, flags);
2560 /* adjust seek timestamp to found sample timestamp */
2561 seek_timestamp = st->index_entries[sample].timestamp;
2563 for (i = 0; i < s->nb_streams; i++) {
2565 if (stream_index == i)
2568 timestamp = av_rescale_q(seek_timestamp, s->streams[stream_index]->time_base, st->time_base);
2569 mov_seek_stream(s, st, timestamp, flags);
2574 static int mov_read_close(AVFormatContext *s)
2576 MOVContext *mov = s->priv_data;
2579 for (i = 0; i < s->nb_streams; i++) {
2580 AVStream *st = s->streams[i];
2581 MOVStreamContext *sc = st->priv_data;
2583 av_freep(&sc->ctts_data);
2584 for (j = 0; j < sc->drefs_count; j++) {
2585 av_freep(&sc->drefs[j].path);
2586 av_freep(&sc->drefs[j].dir);
2588 av_freep(&sc->drefs);
2589 if (sc->pb && sc->pb != s->pb)
2592 av_freep(&st->codec->palctrl);
2595 if (mov->dv_demux) {
2596 for(i = 0; i < mov->dv_fctx->nb_streams; i++) {
2597 av_freep(&mov->dv_fctx->streams[i]->codec);
2598 av_freep(&mov->dv_fctx->streams[i]);
2600 av_freep(&mov->dv_fctx);
2601 av_freep(&mov->dv_demux);
2604 av_freep(&mov->trex_data);
2609 AVInputFormat mov_demuxer = {
2610 "mov,mp4,m4a,3gp,3g2,mj2",
2611 NULL_IF_CONFIG_SMALL("QuickTime/MPEG-4/Motion JPEG 2000 format"),