alacenc: add a fixed LPC coefficient mode as compression level 1. old
[FFMpeg-mirror/lagarith.git] / libavformat / mov.c
blobfdf921cb2aadb591adb31665cb558e72c5939878
1 /*
2 * MOV demuxer
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
23 #include <limits.h>
25 //#define DEBUG
26 //#define DEBUG_METADATA
27 //#define MOV_EXPORT_ALL_METADATA
29 #include "libavutil/intreadwrite.h"
30 #include "libavutil/avstring.h"
31 #include "avformat.h"
32 #include "riff.h"
33 #include "isom.h"
34 #include "libavcodec/mpeg4audio.h"
35 #include "libavcodec/mpegaudiodata.h"
36 #include "libavcodec/get_bits.h"
38 #if CONFIG_ZLIB
39 #include <zlib.h>
40 #endif
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
56 * Apple:
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"
65 #undef NDEBUG
66 #include <assert.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 */
71 /* return code:
72 0: continue to parse next atom
73 <0: error occurred, exit
75 /* links atom IDs to parse functions */
76 typedef struct MOVParseTableEntry {
77 uint32_t type;
78 int (*parse)(MOVContext *ctx, ByteIOContext *pb, MOVAtom atom);
79 } MOVParseTableEntry;
81 static const MOVParseTableEntry mov_default_parse_table[];
83 static int mov_metadata_trkn(MOVContext *c, ByteIOContext *pb, unsigned len)
85 char buf[16];
87 get_be16(pb); // unknown
88 snprintf(buf, sizeof(buf), "%d", get_be16(pb));
89 av_metadata_set(&c->fc->metadata, "track", buf);
91 get_be16(pb); // total tracks
93 return 0;
96 static int mov_read_udta_string(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
98 #ifdef MOV_EXPORT_ALL_METADATA
99 char tmp_key[5];
100 #endif
101 char str[1024], key2[16], language[4] = {0};
102 const char *key = NULL;
103 uint16_t str_size;
104 int (*parse)(MOVContext*, ByteIOContext*, unsigned) = NULL;
106 switch (atom.type) {
107 case MKTAG(0xa9,'n','a','m'): key = "title"; break;
108 case MKTAG(0xa9,'a','u','t'):
109 case MKTAG(0xa9,'A','R','T'):
110 case MKTAG(0xa9,'w','r','t'): key = "author"; break;
111 case MKTAG(0xa9,'c','p','y'): key = "copyright"; break;
112 case MKTAG(0xa9,'c','m','t'):
113 case MKTAG(0xa9,'i','n','f'): key = "comment"; break;
114 case MKTAG(0xa9,'a','l','b'): key = "album"; break;
115 case MKTAG(0xa9,'d','a','y'): key = "year"; break;
116 case MKTAG(0xa9,'g','e','n'): key = "genre"; break;
117 case MKTAG(0xa9,'t','o','o'):
118 case MKTAG(0xa9,'e','n','c'): key = "muxer"; break;
119 case MKTAG( 't','r','k','n'): key = "track";
120 parse = mov_metadata_trkn; break;
123 if (c->itunes_metadata && atom.size > 8) {
124 int data_size = get_be32(pb);
125 int tag = get_le32(pb);
126 if (tag == MKTAG('d','a','t','a')) {
127 get_be32(pb); // type
128 get_be32(pb); // unknown
129 str_size = data_size - 16;
130 atom.size -= 16;
131 } else return 0;
132 } else if (atom.size > 4 && key && !c->itunes_metadata) {
133 str_size = get_be16(pb); // string length
134 ff_mov_lang_to_iso639(get_be16(pb), language);
135 atom.size -= 4;
136 } else
137 str_size = atom.size;
139 #ifdef MOV_EXPORT_ALL_METADATA
140 if (!key) {
141 snprintf(tmp_key, 5, "%.4s", (char*)&atom.type);
142 key = tmp_key;
144 #endif
146 if (!key)
147 return 0;
148 if (atom.size < 0)
149 return -1;
151 str_size = FFMIN3(sizeof(str)-1, str_size, atom.size);
153 if (parse)
154 parse(c, pb, str_size);
155 else {
156 get_buffer(pb, str, str_size);
157 str[str_size] = 0;
158 av_metadata_set(&c->fc->metadata, key, str);
159 if (*language && strcmp(language, "und")) {
160 snprintf(key2, sizeof(key2), "%s-%s", key, language);
161 av_metadata_set(&c->fc->metadata, key2, str);
164 #ifdef DEBUG_METADATA
165 av_log(c->fc, AV_LOG_DEBUG, "lang \"%3s\" ", language);
166 av_log(c->fc, AV_LOG_DEBUG, "tag \"%s\" value \"%s\" atom \"%.4s\" %d %lld\n",
167 key, str, (char*)&atom.type, str_size, atom.size);
168 #endif
170 return 0;
173 static int mov_read_default(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
175 int64_t total_size = 0;
176 MOVAtom a;
177 int i;
178 int err = 0;
180 a.offset = atom.offset;
182 if (atom.size < 0)
183 atom.size = INT64_MAX;
184 while(((total_size + 8) < atom.size) && !url_feof(pb) && !err) {
185 int (*parse)(MOVContext*, ByteIOContext*, MOVAtom) = NULL;
186 a.size = atom.size;
187 a.type=0;
188 if(atom.size >= 8) {
189 a.size = get_be32(pb);
190 a.type = get_le32(pb);
192 total_size += 8;
193 a.offset += 8;
194 dprintf(c->fc, "type: %08x %.4s sz: %"PRIx64" %"PRIx64" %"PRIx64"\n",
195 a.type, (char*)&a.type, a.size, atom.size, total_size);
196 if (a.size == 1) { /* 64 bit extended size */
197 a.size = get_be64(pb) - 8;
198 a.offset += 8;
199 total_size += 8;
201 if (a.size == 0) {
202 a.size = atom.size - total_size;
203 if (a.size <= 8)
204 break;
206 a.size -= 8;
207 if(a.size < 0)
208 break;
209 a.size = FFMIN(a.size, atom.size - total_size);
211 for (i = 0; mov_default_parse_table[i].type; i++)
212 if (mov_default_parse_table[i].type == a.type) {
213 parse = mov_default_parse_table[i].parse;
214 break;
217 // container is user data
218 if (!parse && (atom.type == MKTAG('u','d','t','a') ||
219 atom.type == MKTAG('i','l','s','t')))
220 parse = mov_read_udta_string;
222 if (!parse) { /* skip leaf atoms data */
223 url_fskip(pb, a.size);
224 } else {
225 int64_t start_pos = url_ftell(pb);
226 int64_t left;
227 err = parse(c, pb, a);
228 if (url_is_streamed(pb) && c->found_moov && c->found_mdat)
229 break;
230 left = a.size - url_ftell(pb) + start_pos;
231 if (left > 0) /* skip garbage at atom end */
232 url_fskip(pb, left);
235 a.offset += a.size;
236 total_size += a.size;
239 if (!err && total_size < atom.size && atom.size < 0x7ffff)
240 url_fskip(pb, atom.size - total_size);
242 return err;
245 static int mov_read_dref(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
247 AVStream *st;
248 MOVStreamContext *sc;
249 int entries, i, j;
251 if (c->fc->nb_streams < 1)
252 return 0;
253 st = c->fc->streams[c->fc->nb_streams-1];
254 sc = st->priv_data;
256 get_be32(pb); // version + flags
257 entries = get_be32(pb);
258 if (entries >= UINT_MAX / sizeof(*sc->drefs))
259 return -1;
260 sc->drefs = av_mallocz(entries * sizeof(*sc->drefs));
261 if (!sc->drefs)
262 return AVERROR(ENOMEM);
263 sc->drefs_count = entries;
265 for (i = 0; i < sc->drefs_count; i++) {
266 MOVDref *dref = &sc->drefs[i];
267 uint32_t size = get_be32(pb);
268 int64_t next = url_ftell(pb) + size - 4;
270 dref->type = get_le32(pb);
271 get_be32(pb); // version + flags
272 dprintf(c->fc, "type %.4s size %d\n", (char*)&dref->type, size);
274 if (dref->type == MKTAG('a','l','i','s') && size > 150) {
275 /* macintosh alias record */
276 uint16_t volume_len, len;
277 char volume[28];
278 int16_t type;
280 url_fskip(pb, 10);
282 volume_len = get_byte(pb);
283 volume_len = FFMIN(volume_len, 27);
284 get_buffer(pb, volume, 27);
285 volume[volume_len] = 0;
286 av_log(c->fc, AV_LOG_DEBUG, "volume %s, len %d\n", volume, volume_len);
288 url_fskip(pb, 112);
290 for (type = 0; type != -1 && url_ftell(pb) < next; ) {
291 type = get_be16(pb);
292 len = get_be16(pb);
293 av_log(c->fc, AV_LOG_DEBUG, "type %d, len %d\n", type, len);
294 if (len&1)
295 len += 1;
296 if (type == 2) { // absolute path
297 av_free(dref->path);
298 dref->path = av_mallocz(len+1);
299 if (!dref->path)
300 return AVERROR(ENOMEM);
301 get_buffer(pb, dref->path, len);
302 if (len > volume_len && !strncmp(dref->path, volume, volume_len)) {
303 len -= volume_len;
304 memmove(dref->path, dref->path+volume_len, len);
305 dref->path[len] = 0;
307 for (j = 0; j < len; j++)
308 if (dref->path[j] == ':')
309 dref->path[j] = '/';
310 av_log(c->fc, AV_LOG_DEBUG, "path %s\n", dref->path);
311 } else
312 url_fskip(pb, len);
315 url_fseek(pb, next, SEEK_SET);
317 return 0;
320 static int mov_read_hdlr(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
322 AVStream *st;
323 uint32_t type;
324 uint32_t ctype;
326 if (c->fc->nb_streams < 1) // meta before first trak
327 return 0;
329 st = c->fc->streams[c->fc->nb_streams-1];
331 get_byte(pb); /* version */
332 get_be24(pb); /* flags */
334 /* component type */
335 ctype = get_le32(pb);
336 type = get_le32(pb); /* component subtype */
338 dprintf(c->fc, "ctype= %.4s (0x%08x)\n", (char*)&ctype, ctype);
339 dprintf(c->fc, "stype= %.4s\n", (char*)&type);
341 if (type == MKTAG('v','i','d','e'))
342 st->codec->codec_type = CODEC_TYPE_VIDEO;
343 else if(type == MKTAG('s','o','u','n'))
344 st->codec->codec_type = CODEC_TYPE_AUDIO;
345 else if(type == MKTAG('m','1','a',' '))
346 st->codec->codec_id = CODEC_ID_MP2;
347 else if(type == MKTAG('s','u','b','p'))
348 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
350 get_be32(pb); /* component manufacture */
351 get_be32(pb); /* component flags */
352 get_be32(pb); /* component flags mask */
354 if(atom.size <= 24)
355 return 0; /* nothing left to read */
357 url_fskip(pb, atom.size - (url_ftell(pb) - atom.offset));
358 return 0;
361 int ff_mp4_read_descr_len(ByteIOContext *pb)
363 int len = 0;
364 int count = 4;
365 while (count--) {
366 int c = get_byte(pb);
367 len = (len << 7) | (c & 0x7f);
368 if (!(c & 0x80))
369 break;
371 return len;
374 int mp4_read_descr(AVFormatContext *fc, ByteIOContext *pb, int *tag)
376 int len;
377 *tag = get_byte(pb);
378 len = ff_mp4_read_descr_len(pb);
379 dprintf(fc, "MPEG4 description: tag=0x%02x len=%d\n", *tag, len);
380 return len;
383 #define MP4ESDescrTag 0x03
384 #define MP4DecConfigDescrTag 0x04
385 #define MP4DecSpecificDescrTag 0x05
387 static const AVCodecTag mp4_audio_types[] = {
388 { CODEC_ID_MP3ON4, AOT_PS }, /* old mp3on4 draft */
389 { CODEC_ID_MP3ON4, AOT_L1 }, /* layer 1 */
390 { CODEC_ID_MP3ON4, AOT_L2 }, /* layer 2 */
391 { CODEC_ID_MP3ON4, AOT_L3 }, /* layer 3 */
392 { CODEC_ID_MP4ALS, AOT_ALS }, /* MPEG-4 ALS */
393 { CODEC_ID_NONE, AOT_NULL },
396 int ff_mov_read_esds(AVFormatContext *fc, ByteIOContext *pb, MOVAtom atom)
398 AVStream *st;
399 int tag, len;
401 if (fc->nb_streams < 1)
402 return 0;
403 st = fc->streams[fc->nb_streams-1];
405 get_be32(pb); /* version + flags */
406 len = mp4_read_descr(fc, pb, &tag);
407 if (tag == MP4ESDescrTag) {
408 get_be16(pb); /* ID */
409 get_byte(pb); /* priority */
410 } else
411 get_be16(pb); /* ID */
413 len = mp4_read_descr(fc, pb, &tag);
414 if (tag == MP4DecConfigDescrTag) {
415 int object_type_id = get_byte(pb);
416 get_byte(pb); /* stream type */
417 get_be24(pb); /* buffer size db */
418 get_be32(pb); /* max bitrate */
419 get_be32(pb); /* avg bitrate */
421 st->codec->codec_id= ff_codec_get_id(ff_mp4_obj_type, object_type_id);
422 dprintf(fc, "esds object type id %d\n", object_type_id);
423 len = mp4_read_descr(fc, pb, &tag);
424 if (tag == MP4DecSpecificDescrTag) {
425 dprintf(fc, "Specific MPEG4 header len=%d\n", len);
426 if((uint64_t)len > (1<<30))
427 return -1;
428 st->codec->extradata = av_mallocz(len + FF_INPUT_BUFFER_PADDING_SIZE);
429 if (!st->codec->extradata)
430 return AVERROR(ENOMEM);
431 get_buffer(pb, st->codec->extradata, len);
432 st->codec->extradata_size = len;
433 if (st->codec->codec_id == CODEC_ID_AAC) {
434 MPEG4AudioConfig cfg;
435 ff_mpeg4audio_get_config(&cfg, st->codec->extradata,
436 st->codec->extradata_size);
437 if (cfg.chan_config > 7)
438 return -1;
439 st->codec->channels = ff_mpeg4audio_channels[cfg.chan_config];
440 if (cfg.object_type == 29 && cfg.sampling_index < 3) // old mp3on4
441 st->codec->sample_rate = ff_mpa_freq_tab[cfg.sampling_index];
442 else
443 st->codec->sample_rate = cfg.sample_rate; // ext sample rate ?
444 dprintf(fc, "mp4a config channels %d obj %d ext obj %d "
445 "sample rate %d ext sample rate %d\n", st->codec->channels,
446 cfg.object_type, cfg.ext_object_type,
447 cfg.sample_rate, cfg.ext_sample_rate);
448 if (!(st->codec->codec_id = ff_codec_get_id(mp4_audio_types,
449 cfg.object_type)))
450 st->codec->codec_id = CODEC_ID_AAC;
454 return 0;
457 static int mov_read_esds(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
459 return ff_mov_read_esds(c->fc, pb, atom);
462 static int mov_read_pasp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
464 const int num = get_be32(pb);
465 const int den = get_be32(pb);
466 AVStream *st;
468 if (c->fc->nb_streams < 1)
469 return 0;
470 st = c->fc->streams[c->fc->nb_streams-1];
472 if (den != 0) {
473 if ((st->sample_aspect_ratio.den != 1 || st->sample_aspect_ratio.num) && // default
474 (den != st->sample_aspect_ratio.den || num != st->sample_aspect_ratio.num))
475 av_log(c->fc, AV_LOG_WARNING,
476 "sample aspect ratio already set to %d:%d, overriding by 'pasp' atom\n",
477 st->sample_aspect_ratio.num, st->sample_aspect_ratio.den);
478 st->sample_aspect_ratio.num = num;
479 st->sample_aspect_ratio.den = den;
481 return 0;
484 /* this atom contains actual media data */
485 static int mov_read_mdat(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
487 if(atom.size == 0) /* wrong one (MP4) */
488 return 0;
489 c->found_mdat=1;
490 return 0; /* now go for moov */
493 /* read major brand, minor version and compatible brands and store them as metadata */
494 static int mov_read_ftyp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
496 uint32_t minor_ver;
497 int comp_brand_size;
498 char minor_ver_str[11]; /* 32 bit integer -> 10 digits + null */
499 char* comp_brands_str;
500 uint8_t type[5] = {0};
502 get_buffer(pb, type, 4);
503 if (strcmp(type, "qt "))
504 c->isom = 1;
505 av_log(c->fc, AV_LOG_DEBUG, "ISO: File Type Major Brand: %.4s\n",(char *)&type);
506 av_metadata_set(&c->fc->metadata, "major_brand", type);
507 minor_ver = get_be32(pb); /* minor version */
508 snprintf(minor_ver_str, sizeof(minor_ver_str), "%d", minor_ver);
509 av_metadata_set(&c->fc->metadata, "minor_version", minor_ver_str);
511 comp_brand_size = atom.size - 8;
512 if (comp_brand_size < 0)
513 return -1;
514 comp_brands_str = av_malloc(comp_brand_size + 1); /* Add null terminator */
515 if (!comp_brands_str)
516 return AVERROR(ENOMEM);
517 get_buffer(pb, comp_brands_str, comp_brand_size);
518 comp_brands_str[comp_brand_size] = 0;
519 av_metadata_set(&c->fc->metadata, "compatible_brands", comp_brands_str);
520 av_freep(&comp_brands_str);
522 return 0;
525 /* this atom should contain all header atoms */
526 static int mov_read_moov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
528 if (mov_read_default(c, pb, atom) < 0)
529 return -1;
530 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
531 /* so we don't parse the whole file if over a network */
532 c->found_moov=1;
533 return 0; /* now go for mdat */
536 static int mov_read_moof(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
538 c->fragment.moof_offset = url_ftell(pb) - 8;
539 dprintf(c->fc, "moof offset %llx\n", c->fragment.moof_offset);
540 return mov_read_default(c, pb, atom);
543 static int mov_read_mdhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
545 AVStream *st;
546 MOVStreamContext *sc;
547 int version;
548 char language[4] = {0};
549 unsigned lang;
551 if (c->fc->nb_streams < 1)
552 return 0;
553 st = c->fc->streams[c->fc->nb_streams-1];
554 sc = st->priv_data;
556 version = get_byte(pb);
557 if (version > 1)
558 return -1; /* unsupported */
560 get_be24(pb); /* flags */
561 if (version == 1) {
562 get_be64(pb);
563 get_be64(pb);
564 } else {
565 get_be32(pb); /* creation time */
566 get_be32(pb); /* modification time */
569 sc->time_scale = get_be32(pb);
570 st->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
572 lang = get_be16(pb); /* language */
573 if (ff_mov_lang_to_iso639(lang, language))
574 av_metadata_set(&st->metadata, "language", language);
575 get_be16(pb); /* quality */
577 return 0;
580 static int mov_read_mvhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
582 int version = get_byte(pb); /* version */
583 get_be24(pb); /* flags */
585 if (version == 1) {
586 get_be64(pb);
587 get_be64(pb);
588 } else {
589 get_be32(pb); /* creation time */
590 get_be32(pb); /* modification time */
592 c->time_scale = get_be32(pb); /* time scale */
594 dprintf(c->fc, "time scale = %i\n", c->time_scale);
596 c->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
597 get_be32(pb); /* preferred scale */
599 get_be16(pb); /* preferred volume */
601 url_fskip(pb, 10); /* reserved */
603 url_fskip(pb, 36); /* display matrix */
605 get_be32(pb); /* preview time */
606 get_be32(pb); /* preview duration */
607 get_be32(pb); /* poster time */
608 get_be32(pb); /* selection time */
609 get_be32(pb); /* selection duration */
610 get_be32(pb); /* current time */
611 get_be32(pb); /* next track ID */
613 return 0;
616 static int mov_read_smi(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
618 AVStream *st;
620 if (c->fc->nb_streams < 1)
621 return 0;
622 st = c->fc->streams[c->fc->nb_streams-1];
624 if((uint64_t)atom.size > (1<<30))
625 return -1;
627 // currently SVQ3 decoder expect full STSD header - so let's fake it
628 // this should be fixed and just SMI header should be passed
629 av_free(st->codec->extradata);
630 st->codec->extradata = av_mallocz(atom.size + 0x5a + FF_INPUT_BUFFER_PADDING_SIZE);
631 if (!st->codec->extradata)
632 return AVERROR(ENOMEM);
633 st->codec->extradata_size = 0x5a + atom.size;
634 memcpy(st->codec->extradata, "SVQ3", 4); // fake
635 get_buffer(pb, st->codec->extradata + 0x5a, atom.size);
636 dprintf(c->fc, "Reading SMI %"PRId64" %s\n", atom.size, st->codec->extradata + 0x5a);
637 return 0;
640 static int mov_read_enda(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
642 AVStream *st;
643 int little_endian;
645 if (c->fc->nb_streams < 1)
646 return 0;
647 st = c->fc->streams[c->fc->nb_streams-1];
649 little_endian = get_be16(pb);
650 dprintf(c->fc, "enda %d\n", little_endian);
651 if (little_endian == 1) {
652 switch (st->codec->codec_id) {
653 case CODEC_ID_PCM_S24BE:
654 st->codec->codec_id = CODEC_ID_PCM_S24LE;
655 break;
656 case CODEC_ID_PCM_S32BE:
657 st->codec->codec_id = CODEC_ID_PCM_S32LE;
658 break;
659 case CODEC_ID_PCM_F32BE:
660 st->codec->codec_id = CODEC_ID_PCM_F32LE;
661 break;
662 case CODEC_ID_PCM_F64BE:
663 st->codec->codec_id = CODEC_ID_PCM_F64LE;
664 break;
665 default:
666 break;
669 return 0;
672 /* FIXME modify qdm2/svq3/h264 decoders to take full atom as extradata */
673 static int mov_read_extradata(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
675 AVStream *st;
676 uint64_t size;
677 uint8_t *buf;
679 if (c->fc->nb_streams < 1) // will happen with jp2 files
680 return 0;
681 st= c->fc->streams[c->fc->nb_streams-1];
682 size= (uint64_t)st->codec->extradata_size + atom.size + 8 + FF_INPUT_BUFFER_PADDING_SIZE;
683 if(size > INT_MAX || (uint64_t)atom.size > INT_MAX)
684 return -1;
685 buf= av_realloc(st->codec->extradata, size);
686 if(!buf)
687 return -1;
688 st->codec->extradata= buf;
689 buf+= st->codec->extradata_size;
690 st->codec->extradata_size= size - FF_INPUT_BUFFER_PADDING_SIZE;
691 AV_WB32( buf , atom.size + 8);
692 AV_WL32( buf + 4, atom.type);
693 get_buffer(pb, buf + 8, atom.size);
694 return 0;
697 static int mov_read_wave(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
699 AVStream *st;
701 if (c->fc->nb_streams < 1)
702 return 0;
703 st = c->fc->streams[c->fc->nb_streams-1];
705 if((uint64_t)atom.size > (1<<30))
706 return -1;
708 if (st->codec->codec_id == CODEC_ID_QDM2) {
709 // pass all frma atom to codec, needed at least for QDM2
710 av_free(st->codec->extradata);
711 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
712 if (!st->codec->extradata)
713 return AVERROR(ENOMEM);
714 st->codec->extradata_size = atom.size;
715 get_buffer(pb, st->codec->extradata, atom.size);
716 } else if (atom.size > 8) { /* to read frma, esds atoms */
717 if (mov_read_default(c, pb, atom) < 0)
718 return -1;
719 } else
720 url_fskip(pb, atom.size);
721 return 0;
725 * This function reads atom content and puts data in extradata without tag
726 * nor size unlike mov_read_extradata.
728 static int mov_read_glbl(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
730 AVStream *st;
732 if (c->fc->nb_streams < 1)
733 return 0;
734 st = c->fc->streams[c->fc->nb_streams-1];
736 if((uint64_t)atom.size > (1<<30))
737 return -1;
739 av_free(st->codec->extradata);
740 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
741 if (!st->codec->extradata)
742 return AVERROR(ENOMEM);
743 st->codec->extradata_size = atom.size;
744 get_buffer(pb, st->codec->extradata, atom.size);
745 return 0;
748 static int mov_read_stco(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
750 AVStream *st;
751 MOVStreamContext *sc;
752 unsigned int i, entries;
754 if (c->fc->nb_streams < 1)
755 return 0;
756 st = c->fc->streams[c->fc->nb_streams-1];
757 sc = st->priv_data;
759 get_byte(pb); /* version */
760 get_be24(pb); /* flags */
762 entries = get_be32(pb);
764 if(entries >= UINT_MAX/sizeof(int64_t))
765 return -1;
767 sc->chunk_offsets = av_malloc(entries * sizeof(int64_t));
768 if (!sc->chunk_offsets)
769 return AVERROR(ENOMEM);
770 sc->chunk_count = entries;
772 if (atom.type == MKTAG('s','t','c','o'))
773 for(i=0; i<entries; i++)
774 sc->chunk_offsets[i] = get_be32(pb);
775 else if (atom.type == MKTAG('c','o','6','4'))
776 for(i=0; i<entries; i++)
777 sc->chunk_offsets[i] = get_be64(pb);
778 else
779 return -1;
781 return 0;
785 * Compute codec id for 'lpcm' tag.
786 * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
788 enum CodecID ff_mov_get_lpcm_codec_id(int bps, int flags)
790 if (flags & 1) { // floating point
791 if (flags & 2) { // big endian
792 if (bps == 32) return CODEC_ID_PCM_F32BE;
793 else if (bps == 64) return CODEC_ID_PCM_F64BE;
794 } else {
795 if (bps == 32) return CODEC_ID_PCM_F32LE;
796 else if (bps == 64) return CODEC_ID_PCM_F64LE;
798 } else {
799 if (flags & 2) {
800 if (bps == 8)
801 // signed integer
802 if (flags & 4) return CODEC_ID_PCM_S8;
803 else return CODEC_ID_PCM_U8;
804 else if (bps == 16) return CODEC_ID_PCM_S16BE;
805 else if (bps == 24) return CODEC_ID_PCM_S24BE;
806 else if (bps == 32) return CODEC_ID_PCM_S32BE;
807 } else {
808 if (bps == 8)
809 if (flags & 4) return CODEC_ID_PCM_S8;
810 else return CODEC_ID_PCM_U8;
811 else if (bps == 16) return CODEC_ID_PCM_S16LE;
812 else if (bps == 24) return CODEC_ID_PCM_S24LE;
813 else if (bps == 32) return CODEC_ID_PCM_S32LE;
816 return CODEC_ID_NONE;
819 static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
821 AVStream *st;
822 MOVStreamContext *sc;
823 int j, entries, pseudo_stream_id;
825 if (c->fc->nb_streams < 1)
826 return 0;
827 st = c->fc->streams[c->fc->nb_streams-1];
828 sc = st->priv_data;
830 get_byte(pb); /* version */
831 get_be24(pb); /* flags */
833 entries = get_be32(pb);
835 for(pseudo_stream_id=0; pseudo_stream_id<entries; pseudo_stream_id++) {
836 //Parsing Sample description table
837 enum CodecID id;
838 int dref_id = 1;
839 MOVAtom a = { 0, 0, 0 };
840 int64_t start_pos = url_ftell(pb);
841 int size = get_be32(pb); /* size */
842 uint32_t format = get_le32(pb); /* data format */
844 if (size >= 16) {
845 get_be32(pb); /* reserved */
846 get_be16(pb); /* reserved */
847 dref_id = get_be16(pb);
850 if (st->codec->codec_tag &&
851 st->codec->codec_tag != format &&
852 (c->fc->video_codec_id ? ff_codec_get_id(codec_movvideo_tags, format) != c->fc->video_codec_id
853 : st->codec->codec_tag != MKTAG('j','p','e','g'))
855 /* Multiple fourcc, we skip JPEG. This is not correct, we should
856 * export it as a separate AVStream but this needs a few changes
857 * in the MOV demuxer, patch welcome. */
858 av_log(c->fc, AV_LOG_WARNING, "multiple fourcc not supported\n");
859 url_fskip(pb, size - (url_ftell(pb) - start_pos));
860 continue;
862 sc->pseudo_stream_id = st->codec->codec_tag ? -1 : pseudo_stream_id;
863 sc->dref_id= dref_id;
865 st->codec->codec_tag = format;
866 id = ff_codec_get_id(codec_movaudio_tags, format);
867 if (id<=0 && ((format&0xFFFF) == 'm'+('s'<<8) || (format&0xFFFF) == 'T'+('S'<<8)))
868 id = ff_codec_get_id(ff_codec_wav_tags, bswap_32(format)&0xFFFF);
870 if (st->codec->codec_type != CODEC_TYPE_VIDEO && id > 0) {
871 st->codec->codec_type = CODEC_TYPE_AUDIO;
872 } else if (st->codec->codec_type != CODEC_TYPE_AUDIO && /* do not overwrite codec type */
873 format && format != MKTAG('m','p','4','s')) { /* skip old asf mpeg4 tag */
874 id = ff_codec_get_id(codec_movvideo_tags, format);
875 if (id <= 0)
876 id = ff_codec_get_id(ff_codec_bmp_tags, format);
877 if (id > 0)
878 st->codec->codec_type = CODEC_TYPE_VIDEO;
879 else if(st->codec->codec_type == CODEC_TYPE_DATA){
880 id = ff_codec_get_id(ff_codec_movsubtitle_tags, format);
881 if(id > 0)
882 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
886 dprintf(c->fc, "size=%d 4CC= %c%c%c%c codec_type=%d\n", size,
887 (format >> 0) & 0xff, (format >> 8) & 0xff, (format >> 16) & 0xff,
888 (format >> 24) & 0xff, st->codec->codec_type);
890 if(st->codec->codec_type==CODEC_TYPE_VIDEO) {
891 uint8_t codec_name[32];
892 unsigned int color_depth;
893 int color_greyscale;
895 st->codec->codec_id = id;
896 get_be16(pb); /* version */
897 get_be16(pb); /* revision level */
898 get_be32(pb); /* vendor */
899 get_be32(pb); /* temporal quality */
900 get_be32(pb); /* spatial quality */
902 st->codec->width = get_be16(pb); /* width */
903 st->codec->height = get_be16(pb); /* height */
905 get_be32(pb); /* horiz resolution */
906 get_be32(pb); /* vert resolution */
907 get_be32(pb); /* data size, always 0 */
908 get_be16(pb); /* frames per samples */
910 get_buffer(pb, codec_name, 32); /* codec name, pascal string */
911 if (codec_name[0] <= 31) {
912 int i;
913 int pos = 0;
914 for (i = 0; i < codec_name[0] && pos < sizeof(st->codec->codec_name) - 3; i++) {
915 uint8_t tmp;
916 PUT_UTF8(codec_name[i+1], tmp, st->codec->codec_name[pos++] = tmp;)
918 st->codec->codec_name[pos] = 0;
921 st->codec->bits_per_coded_sample = get_be16(pb); /* depth */
922 st->codec->color_table_id = get_be16(pb); /* colortable id */
923 dprintf(c->fc, "depth %d, ctab id %d\n",
924 st->codec->bits_per_coded_sample, st->codec->color_table_id);
925 /* figure out the palette situation */
926 color_depth = st->codec->bits_per_coded_sample & 0x1F;
927 color_greyscale = st->codec->bits_per_coded_sample & 0x20;
929 /* if the depth is 2, 4, or 8 bpp, file is palettized */
930 if ((color_depth == 2) || (color_depth == 4) ||
931 (color_depth == 8)) {
932 /* for palette traversal */
933 unsigned int color_start, color_count, color_end;
934 unsigned char r, g, b;
936 st->codec->palctrl = av_malloc(sizeof(*st->codec->palctrl));
937 if (color_greyscale) {
938 int color_index, color_dec;
939 /* compute the greyscale palette */
940 st->codec->bits_per_coded_sample = color_depth;
941 color_count = 1 << color_depth;
942 color_index = 255;
943 color_dec = 256 / (color_count - 1);
944 for (j = 0; j < color_count; j++) {
945 r = g = b = color_index;
946 st->codec->palctrl->palette[j] =
947 (r << 16) | (g << 8) | (b);
948 color_index -= color_dec;
949 if (color_index < 0)
950 color_index = 0;
952 } else if (st->codec->color_table_id) {
953 const uint8_t *color_table;
954 /* if flag bit 3 is set, use the default palette */
955 color_count = 1 << color_depth;
956 if (color_depth == 2)
957 color_table = ff_qt_default_palette_4;
958 else if (color_depth == 4)
959 color_table = ff_qt_default_palette_16;
960 else
961 color_table = ff_qt_default_palette_256;
963 for (j = 0; j < color_count; j++) {
964 r = color_table[j * 3 + 0];
965 g = color_table[j * 3 + 1];
966 b = color_table[j * 3 + 2];
967 st->codec->palctrl->palette[j] =
968 (r << 16) | (g << 8) | (b);
970 } else {
971 /* load the palette from the file */
972 color_start = get_be32(pb);
973 color_count = get_be16(pb);
974 color_end = get_be16(pb);
975 if ((color_start <= 255) &&
976 (color_end <= 255)) {
977 for (j = color_start; j <= color_end; j++) {
978 /* each R, G, or B component is 16 bits;
979 * only use the top 8 bits; skip alpha bytes
980 * up front */
981 get_byte(pb);
982 get_byte(pb);
983 r = get_byte(pb);
984 get_byte(pb);
985 g = get_byte(pb);
986 get_byte(pb);
987 b = get_byte(pb);
988 get_byte(pb);
989 st->codec->palctrl->palette[j] =
990 (r << 16) | (g << 8) | (b);
994 st->codec->palctrl->palette_changed = 1;
996 } else if(st->codec->codec_type==CODEC_TYPE_AUDIO) {
997 int bits_per_sample, flags;
998 uint16_t version = get_be16(pb);
1000 st->codec->codec_id = id;
1001 get_be16(pb); /* revision level */
1002 get_be32(pb); /* vendor */
1004 st->codec->channels = get_be16(pb); /* channel count */
1005 dprintf(c->fc, "audio channels %d\n", st->codec->channels);
1006 st->codec->bits_per_coded_sample = get_be16(pb); /* sample size */
1008 sc->audio_cid = get_be16(pb);
1009 get_be16(pb); /* packet size = 0 */
1011 st->codec->sample_rate = ((get_be32(pb) >> 16));
1013 //Read QT version 1 fields. In version 0 these do not exist.
1014 dprintf(c->fc, "version =%d, isom =%d\n",version,c->isom);
1015 if(!c->isom) {
1016 if(version==1) {
1017 sc->samples_per_frame = get_be32(pb);
1018 get_be32(pb); /* bytes per packet */
1019 sc->bytes_per_frame = get_be32(pb);
1020 get_be32(pb); /* bytes per sample */
1021 } else if(version==2) {
1022 get_be32(pb); /* sizeof struct only */
1023 st->codec->sample_rate = av_int2dbl(get_be64(pb)); /* float 64 */
1024 st->codec->channels = get_be32(pb);
1025 get_be32(pb); /* always 0x7F000000 */
1026 st->codec->bits_per_coded_sample = get_be32(pb); /* bits per channel if sound is uncompressed */
1027 flags = get_be32(pb); /* lcpm format specific flag */
1028 sc->bytes_per_frame = get_be32(pb); /* bytes per audio packet if constant */
1029 sc->samples_per_frame = get_be32(pb); /* lpcm frames per audio packet if constant */
1030 if (format == MKTAG('l','p','c','m'))
1031 st->codec->codec_id = ff_mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, flags);
1035 switch (st->codec->codec_id) {
1036 case CODEC_ID_PCM_S8:
1037 case CODEC_ID_PCM_U8:
1038 if (st->codec->bits_per_coded_sample == 16)
1039 st->codec->codec_id = CODEC_ID_PCM_S16BE;
1040 break;
1041 case CODEC_ID_PCM_S16LE:
1042 case CODEC_ID_PCM_S16BE:
1043 if (st->codec->bits_per_coded_sample == 8)
1044 st->codec->codec_id = CODEC_ID_PCM_S8;
1045 else if (st->codec->bits_per_coded_sample == 24)
1046 st->codec->codec_id =
1047 st->codec->codec_id == CODEC_ID_PCM_S16BE ?
1048 CODEC_ID_PCM_S24BE : CODEC_ID_PCM_S24LE;
1049 break;
1050 /* set values for old format before stsd version 1 appeared */
1051 case CODEC_ID_MACE3:
1052 sc->samples_per_frame = 6;
1053 sc->bytes_per_frame = 2*st->codec->channels;
1054 break;
1055 case CODEC_ID_MACE6:
1056 sc->samples_per_frame = 6;
1057 sc->bytes_per_frame = 1*st->codec->channels;
1058 break;
1059 case CODEC_ID_ADPCM_IMA_QT:
1060 sc->samples_per_frame = 64;
1061 sc->bytes_per_frame = 34*st->codec->channels;
1062 break;
1063 case CODEC_ID_GSM:
1064 sc->samples_per_frame = 160;
1065 sc->bytes_per_frame = 33;
1066 break;
1067 default:
1068 break;
1071 bits_per_sample = av_get_bits_per_sample(st->codec->codec_id);
1072 if (bits_per_sample) {
1073 st->codec->bits_per_coded_sample = bits_per_sample;
1074 sc->sample_size = (bits_per_sample >> 3) * st->codec->channels;
1076 } else if(st->codec->codec_type==CODEC_TYPE_SUBTITLE){
1077 // ttxt stsd contains display flags, justification, background
1078 // color, fonts, and default styles, so fake an atom to read it
1079 MOVAtom fake_atom = { .size = size - (url_ftell(pb) - start_pos) };
1080 if (format != AV_RL32("mp4s")) // mp4s contains a regular esds atom
1081 mov_read_glbl(c, pb, fake_atom);
1082 st->codec->codec_id= id;
1083 st->codec->width = sc->width;
1084 st->codec->height = sc->height;
1085 } else {
1086 /* other codec type, just skip (rtp, mp4s, tmcd ...) */
1087 url_fskip(pb, size - (url_ftell(pb) - start_pos));
1089 /* this will read extra atoms at the end (wave, alac, damr, avcC, SMI ...) */
1090 a.size = size - (url_ftell(pb) - start_pos);
1091 if (a.size > 8) {
1092 if (mov_read_default(c, pb, a) < 0)
1093 return -1;
1094 } else if (a.size > 0)
1095 url_fskip(pb, a.size);
1098 if(st->codec->codec_type==CODEC_TYPE_AUDIO && st->codec->sample_rate==0 && sc->time_scale>1)
1099 st->codec->sample_rate= sc->time_scale;
1101 /* special codec parameters handling */
1102 switch (st->codec->codec_id) {
1103 #if CONFIG_DV_DEMUXER
1104 case CODEC_ID_DVAUDIO:
1105 c->dv_fctx = avformat_alloc_context();
1106 c->dv_demux = dv_init_demux(c->dv_fctx);
1107 if (!c->dv_demux) {
1108 av_log(c->fc, AV_LOG_ERROR, "dv demux context init error\n");
1109 return -1;
1111 sc->dv_audio_container = 1;
1112 st->codec->codec_id = CODEC_ID_PCM_S16LE;
1113 break;
1114 #endif
1115 /* no ifdef since parameters are always those */
1116 case CODEC_ID_QCELP:
1117 // force sample rate for qcelp when not stored in mov
1118 if (st->codec->codec_tag != MKTAG('Q','c','l','p'))
1119 st->codec->sample_rate = 8000;
1120 st->codec->frame_size= 160;
1121 st->codec->channels= 1; /* really needed */
1122 break;
1123 case CODEC_ID_AMR_NB:
1124 case CODEC_ID_AMR_WB:
1125 st->codec->frame_size= sc->samples_per_frame;
1126 st->codec->channels= 1; /* really needed */
1127 /* force sample rate for amr, stsd in 3gp does not store sample rate */
1128 if (st->codec->codec_id == CODEC_ID_AMR_NB)
1129 st->codec->sample_rate = 8000;
1130 else if (st->codec->codec_id == CODEC_ID_AMR_WB)
1131 st->codec->sample_rate = 16000;
1132 break;
1133 case CODEC_ID_MP2:
1134 case CODEC_ID_MP3:
1135 st->codec->codec_type = CODEC_TYPE_AUDIO; /* force type after stsd for m1a hdlr */
1136 st->need_parsing = AVSTREAM_PARSE_FULL;
1137 break;
1138 case CODEC_ID_GSM:
1139 case CODEC_ID_ADPCM_MS:
1140 case CODEC_ID_ADPCM_IMA_WAV:
1141 st->codec->block_align = sc->bytes_per_frame;
1142 break;
1143 case CODEC_ID_ALAC:
1144 if (st->codec->extradata_size == 36) {
1145 st->codec->frame_size = AV_RB32(st->codec->extradata+12);
1146 st->codec->channels = AV_RB8 (st->codec->extradata+21);
1148 break;
1149 default:
1150 break;
1153 return 0;
1156 static int mov_read_stsc(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1158 AVStream *st;
1159 MOVStreamContext *sc;
1160 unsigned int i, entries;
1162 if (c->fc->nb_streams < 1)
1163 return 0;
1164 st = c->fc->streams[c->fc->nb_streams-1];
1165 sc = st->priv_data;
1167 get_byte(pb); /* version */
1168 get_be24(pb); /* flags */
1170 entries = get_be32(pb);
1172 dprintf(c->fc, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
1174 if(entries >= UINT_MAX / sizeof(*sc->stsc_data))
1175 return -1;
1176 sc->stsc_data = av_malloc(entries * sizeof(*sc->stsc_data));
1177 if (!sc->stsc_data)
1178 return AVERROR(ENOMEM);
1179 sc->stsc_count = entries;
1181 for(i=0; i<entries; i++) {
1182 sc->stsc_data[i].first = get_be32(pb);
1183 sc->stsc_data[i].count = get_be32(pb);
1184 sc->stsc_data[i].id = get_be32(pb);
1186 return 0;
1189 static int mov_read_stps(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1191 AVStream *st;
1192 MOVStreamContext *sc;
1193 unsigned i, entries;
1195 if (c->fc->nb_streams < 1)
1196 return 0;
1197 st = c->fc->streams[c->fc->nb_streams-1];
1198 sc = st->priv_data;
1200 get_be32(pb); // version + flags
1202 entries = get_be32(pb);
1203 if (entries >= UINT_MAX / sizeof(*sc->stps_data))
1204 return -1;
1205 sc->stps_data = av_malloc(entries * sizeof(*sc->stps_data));
1206 if (!sc->stps_data)
1207 return AVERROR(ENOMEM);
1208 sc->stps_count = entries;
1210 for (i = 0; i < entries; i++) {
1211 sc->stps_data[i] = get_be32(pb);
1212 //dprintf(c->fc, "stps %d\n", sc->stps_data[i]);
1215 return 0;
1218 static int mov_read_stss(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1220 AVStream *st;
1221 MOVStreamContext *sc;
1222 unsigned int i, entries;
1224 if (c->fc->nb_streams < 1)
1225 return 0;
1226 st = c->fc->streams[c->fc->nb_streams-1];
1227 sc = st->priv_data;
1229 get_byte(pb); /* version */
1230 get_be24(pb); /* flags */
1232 entries = get_be32(pb);
1234 dprintf(c->fc, "keyframe_count = %d\n", entries);
1236 if(entries >= UINT_MAX / sizeof(int))
1237 return -1;
1238 sc->keyframes = av_malloc(entries * sizeof(int));
1239 if (!sc->keyframes)
1240 return AVERROR(ENOMEM);
1241 sc->keyframe_count = entries;
1243 for(i=0; i<entries; i++) {
1244 sc->keyframes[i] = get_be32(pb);
1245 //dprintf(c->fc, "keyframes[]=%d\n", sc->keyframes[i]);
1247 return 0;
1250 static int mov_read_stsz(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1252 AVStream *st;
1253 MOVStreamContext *sc;
1254 unsigned int i, entries, sample_size, field_size, num_bytes;
1255 GetBitContext gb;
1256 unsigned char* buf;
1258 if (c->fc->nb_streams < 1)
1259 return 0;
1260 st = c->fc->streams[c->fc->nb_streams-1];
1261 sc = st->priv_data;
1263 get_byte(pb); /* version */
1264 get_be24(pb); /* flags */
1266 if (atom.type == MKTAG('s','t','s','z')) {
1267 sample_size = get_be32(pb);
1268 if (!sc->sample_size) /* do not overwrite value computed in stsd */
1269 sc->sample_size = sample_size;
1270 field_size = 32;
1271 } else {
1272 sample_size = 0;
1273 get_be24(pb); /* reserved */
1274 field_size = get_byte(pb);
1276 entries = get_be32(pb);
1278 dprintf(c->fc, "sample_size = %d sample_count = %d\n", sc->sample_size, entries);
1280 sc->sample_count = entries;
1281 if (sample_size)
1282 return 0;
1284 if (field_size != 4 && field_size != 8 && field_size != 16 && field_size != 32) {
1285 av_log(c->fc, AV_LOG_ERROR, "Invalid sample field size %d\n", field_size);
1286 return -1;
1289 if (entries >= UINT_MAX / sizeof(int) || entries >= (UINT_MAX - 4) / field_size)
1290 return -1;
1291 sc->sample_sizes = av_malloc(entries * sizeof(int));
1292 if (!sc->sample_sizes)
1293 return AVERROR(ENOMEM);
1295 num_bytes = (entries*field_size+4)>>3;
1297 buf = av_malloc(num_bytes+FF_INPUT_BUFFER_PADDING_SIZE);
1298 if (!buf) {
1299 av_freep(&sc->sample_sizes);
1300 return AVERROR(ENOMEM);
1303 if (get_buffer(pb, buf, num_bytes) < num_bytes) {
1304 av_freep(&sc->sample_sizes);
1305 av_free(buf);
1306 return -1;
1309 init_get_bits(&gb, buf, 8*num_bytes);
1311 for(i=0; i<entries; i++)
1312 sc->sample_sizes[i] = get_bits_long(&gb, field_size);
1314 av_free(buf);
1315 return 0;
1318 static int mov_read_stts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1320 AVStream *st;
1321 MOVStreamContext *sc;
1322 unsigned int i, entries;
1323 int64_t duration=0;
1324 int64_t total_sample_count=0;
1326 if (c->fc->nb_streams < 1)
1327 return 0;
1328 st = c->fc->streams[c->fc->nb_streams-1];
1329 sc = st->priv_data;
1331 get_byte(pb); /* version */
1332 get_be24(pb); /* flags */
1333 entries = get_be32(pb);
1335 dprintf(c->fc, "track[%i].stts.entries = %i\n", c->fc->nb_streams-1, entries);
1337 if(entries >= UINT_MAX / sizeof(*sc->stts_data))
1338 return -1;
1339 sc->stts_data = av_malloc(entries * sizeof(*sc->stts_data));
1340 if (!sc->stts_data)
1341 return AVERROR(ENOMEM);
1342 sc->stts_count = entries;
1344 for(i=0; i<entries; i++) {
1345 int sample_duration;
1346 int sample_count;
1348 sample_count=get_be32(pb);
1349 sample_duration = get_be32(pb);
1350 sc->stts_data[i].count= sample_count;
1351 sc->stts_data[i].duration= sample_duration;
1353 dprintf(c->fc, "sample_count=%d, sample_duration=%d\n",sample_count,sample_duration);
1355 duration+=(int64_t)sample_duration*sample_count;
1356 total_sample_count+=sample_count;
1359 st->nb_frames= total_sample_count;
1360 if(duration)
1361 st->duration= duration;
1362 return 0;
1365 static int mov_read_cslg(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1367 AVStream *st;
1368 MOVStreamContext *sc;
1370 if (c->fc->nb_streams < 1)
1371 return 0;
1372 st = c->fc->streams[c->fc->nb_streams-1];
1373 sc = st->priv_data;
1375 get_be32(pb); // version + flags
1377 sc->dts_shift = get_be32(pb);
1378 dprintf(c->fc, "dts shift %d\n", sc->dts_shift);
1380 get_be32(pb); // least dts to pts delta
1381 get_be32(pb); // greatest dts to pts delta
1382 get_be32(pb); // pts start
1383 get_be32(pb); // pts end
1385 return 0;
1388 static int mov_read_ctts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1390 AVStream *st;
1391 MOVStreamContext *sc;
1392 unsigned int i, entries;
1394 if (c->fc->nb_streams < 1)
1395 return 0;
1396 st = c->fc->streams[c->fc->nb_streams-1];
1397 sc = st->priv_data;
1399 get_byte(pb); /* version */
1400 get_be24(pb); /* flags */
1401 entries = get_be32(pb);
1403 dprintf(c->fc, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
1405 if(entries >= UINT_MAX / sizeof(*sc->ctts_data))
1406 return -1;
1407 sc->ctts_data = av_malloc(entries * sizeof(*sc->ctts_data));
1408 if (!sc->ctts_data)
1409 return AVERROR(ENOMEM);
1410 sc->ctts_count = entries;
1412 for(i=0; i<entries; i++) {
1413 int count =get_be32(pb);
1414 int duration =get_be32(pb);
1416 sc->ctts_data[i].count = count;
1417 sc->ctts_data[i].duration= duration;
1419 return 0;
1422 static void mov_build_index(MOVContext *mov, AVStream *st)
1424 MOVStreamContext *sc = st->priv_data;
1425 int64_t current_offset;
1426 int64_t current_dts = 0;
1427 unsigned int stts_index = 0;
1428 unsigned int stsc_index = 0;
1429 unsigned int stss_index = 0;
1430 unsigned int stps_index = 0;
1431 unsigned int i, j;
1433 /* adjust first dts according to edit list */
1434 if (sc->time_offset) {
1435 int rescaled = sc->time_offset < 0 ? av_rescale(sc->time_offset, sc->time_scale, mov->time_scale) : sc->time_offset;
1436 current_dts = -rescaled;
1437 if (sc->ctts_data && sc->ctts_data[0].duration / sc->stts_data[0].duration > 16) {
1438 /* more than 16 frames delay, dts are likely wrong
1439 this happens with files created by iMovie */
1440 sc->wrong_dts = 1;
1441 st->codec->has_b_frames = 1;
1445 /* only use old uncompressed audio chunk demuxing when stts specifies it */
1446 if (!(st->codec->codec_type == CODEC_TYPE_AUDIO &&
1447 sc->stts_count == 1 && sc->stts_data[0].duration == 1)) {
1448 unsigned int current_sample = 0;
1449 unsigned int stts_sample = 0;
1450 unsigned int sample_size;
1451 unsigned int distance = 0;
1452 int key_off = sc->keyframes && sc->keyframes[0] == 1;
1454 current_dts -= sc->dts_shift;
1456 st->nb_frames = sc->sample_count;
1457 for (i = 0; i < sc->chunk_count; i++) {
1458 current_offset = sc->chunk_offsets[i];
1459 if (stsc_index + 1 < sc->stsc_count &&
1460 i + 1 == sc->stsc_data[stsc_index + 1].first)
1461 stsc_index++;
1462 for (j = 0; j < sc->stsc_data[stsc_index].count; j++) {
1463 int keyframe = 0;
1464 if (current_sample >= sc->sample_count) {
1465 av_log(mov->fc, AV_LOG_ERROR, "wrong sample count\n");
1466 return;
1469 if (!sc->keyframe_count || current_sample+key_off == sc->keyframes[stss_index]) {
1470 keyframe = 1;
1471 if (stss_index + 1 < sc->keyframe_count)
1472 stss_index++;
1473 } else if (sc->stps_count && current_sample+key_off == sc->stps_data[stps_index]) {
1474 keyframe = 1;
1475 if (stps_index + 1 < sc->stps_count)
1476 stps_index++;
1478 if (keyframe)
1479 distance = 0;
1480 sample_size = sc->sample_size > 0 ? sc->sample_size : sc->sample_sizes[current_sample];
1481 if(sc->pseudo_stream_id == -1 ||
1482 sc->stsc_data[stsc_index].id - 1 == sc->pseudo_stream_id) {
1483 av_add_index_entry(st, current_offset, current_dts, sample_size, distance,
1484 keyframe ? AVINDEX_KEYFRAME : 0);
1485 dprintf(mov->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1486 "size %d, distance %d, keyframe %d\n", st->index, current_sample,
1487 current_offset, current_dts, sample_size, distance, keyframe);
1490 current_offset += sample_size;
1491 current_dts += sc->stts_data[stts_index].duration;
1492 distance++;
1493 stts_sample++;
1494 current_sample++;
1495 if (stts_index + 1 < sc->stts_count && stts_sample == sc->stts_data[stts_index].count) {
1496 stts_sample = 0;
1497 stts_index++;
1501 } else {
1502 for (i = 0; i < sc->chunk_count; i++) {
1503 unsigned chunk_samples;
1505 current_offset = sc->chunk_offsets[i];
1506 if (stsc_index + 1 < sc->stsc_count &&
1507 i + 1 == sc->stsc_data[stsc_index + 1].first)
1508 stsc_index++;
1509 chunk_samples = sc->stsc_data[stsc_index].count;
1511 if (sc->samples_per_frame && chunk_samples % sc->samples_per_frame) {
1512 av_log(mov->fc, AV_LOG_ERROR, "error unaligned chunk\n");
1513 return;
1516 while (chunk_samples > 0) {
1517 unsigned size, samples;
1519 if (sc->samples_per_frame >= 160) { // gsm
1520 samples = sc->samples_per_frame;
1521 size = sc->bytes_per_frame;
1522 } else {
1523 if (sc->samples_per_frame > 1) {
1524 samples = FFMIN((1024 / sc->samples_per_frame)*
1525 sc->samples_per_frame, chunk_samples);
1526 size = (samples / sc->samples_per_frame) * sc->bytes_per_frame;
1527 } else {
1528 samples = FFMIN(1024, chunk_samples);
1529 size = samples * sc->sample_size;
1533 av_add_index_entry(st, current_offset, current_dts, size, 0, AVINDEX_KEYFRAME);
1534 dprintf(mov->fc, "AVIndex stream %d, chunk %d, offset %"PRIx64", dts %"PRId64", "
1535 "size %d, duration %d\n", st->index, i, current_offset, current_dts,
1536 size, samples);
1538 current_offset += size;
1539 current_dts += samples;
1540 chunk_samples -= samples;
1546 static int mov_read_trak(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1548 AVStream *st;
1549 MOVStreamContext *sc;
1550 int ret;
1552 st = av_new_stream(c->fc, c->fc->nb_streams);
1553 if (!st) return AVERROR(ENOMEM);
1554 sc = av_mallocz(sizeof(MOVStreamContext));
1555 if (!sc) return AVERROR(ENOMEM);
1557 st->priv_data = sc;
1558 st->codec->codec_type = CODEC_TYPE_DATA;
1559 sc->ffindex = st->index;
1561 if ((ret = mov_read_default(c, pb, atom)) < 0)
1562 return ret;
1564 /* sanity checks */
1565 if (sc->chunk_count && (!sc->stts_count || !sc->stsc_count ||
1566 (!sc->sample_size && !sc->sample_count))) {
1567 av_log(c->fc, AV_LOG_ERROR, "stream %d, missing mandatory atoms, broken header\n",
1568 st->index);
1569 return 0;
1572 if (!sc->time_scale)
1573 sc->time_scale = c->time_scale;
1575 av_set_pts_info(st, 64, 1, sc->time_scale);
1577 if (st->codec->codec_type == CODEC_TYPE_AUDIO &&
1578 !st->codec->frame_size && sc->stts_count == 1) {
1579 st->codec->frame_size = av_rescale(sc->stts_data[0].duration,
1580 st->codec->sample_rate, sc->time_scale);
1581 dprintf(c->fc, "frame size %d\n", st->codec->frame_size);
1584 mov_build_index(c, st);
1586 if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
1587 if (url_fopen(&sc->pb, sc->drefs[sc->dref_id-1].path, URL_RDONLY) < 0)
1588 av_log(c->fc, AV_LOG_ERROR, "stream %d, error opening file %s: %s\n",
1589 st->index, sc->drefs[sc->dref_id-1].path, strerror(errno));
1590 } else
1591 sc->pb = c->fc->pb;
1593 switch (st->codec->codec_id) {
1594 #if CONFIG_H261_DECODER
1595 case CODEC_ID_H261:
1596 #endif
1597 #if CONFIG_H263_DECODER
1598 case CODEC_ID_H263:
1599 #endif
1600 #if CONFIG_H264_DECODER
1601 case CODEC_ID_H264:
1602 #endif
1603 #if CONFIG_MPEG4_DECODER
1604 case CODEC_ID_MPEG4:
1605 #endif
1606 st->codec->width = 0; /* let decoder init width/height */
1607 st->codec->height= 0;
1608 break;
1611 /* Do not need those anymore. */
1612 av_freep(&sc->chunk_offsets);
1613 av_freep(&sc->stsc_data);
1614 av_freep(&sc->sample_sizes);
1615 av_freep(&sc->keyframes);
1616 av_freep(&sc->stts_data);
1617 av_freep(&sc->stps_data);
1619 return 0;
1622 static int mov_read_ilst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1624 int ret;
1625 c->itunes_metadata = 1;
1626 ret = mov_read_default(c, pb, atom);
1627 c->itunes_metadata = 0;
1628 return ret;
1631 static int mov_read_meta(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1633 while (atom.size > 8) {
1634 uint32_t tag = get_le32(pb);
1635 atom.size -= 4;
1636 if (tag == MKTAG('h','d','l','r')) {
1637 url_fseek(pb, -8, SEEK_CUR);
1638 atom.size += 8;
1639 return mov_read_default(c, pb, atom);
1642 return 0;
1645 static int mov_read_tkhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1647 int i;
1648 int width;
1649 int height;
1650 int64_t disp_transform[2];
1651 int display_matrix[3][2];
1652 AVStream *st;
1653 MOVStreamContext *sc;
1654 int version;
1656 if (c->fc->nb_streams < 1)
1657 return 0;
1658 st = c->fc->streams[c->fc->nb_streams-1];
1659 sc = st->priv_data;
1661 version = get_byte(pb);
1662 get_be24(pb); /* flags */
1664 MOV_TRACK_ENABLED 0x0001
1665 MOV_TRACK_IN_MOVIE 0x0002
1666 MOV_TRACK_IN_PREVIEW 0x0004
1667 MOV_TRACK_IN_POSTER 0x0008
1670 if (version == 1) {
1671 get_be64(pb);
1672 get_be64(pb);
1673 } else {
1674 get_be32(pb); /* creation time */
1675 get_be32(pb); /* modification time */
1677 st->id = (int)get_be32(pb); /* track id (NOT 0 !)*/
1678 get_be32(pb); /* reserved */
1680 /* highlevel (considering edits) duration in movie timebase */
1681 (version == 1) ? get_be64(pb) : get_be32(pb);
1682 get_be32(pb); /* reserved */
1683 get_be32(pb); /* reserved */
1685 get_be16(pb); /* layer */
1686 get_be16(pb); /* alternate group */
1687 get_be16(pb); /* volume */
1688 get_be16(pb); /* reserved */
1690 //read in the display matrix (outlined in ISO 14496-12, Section 6.2.2)
1691 // they're kept in fixed point format through all calculations
1692 // ignore u,v,z b/c we don't need the scale factor to calc aspect ratio
1693 for (i = 0; i < 3; i++) {
1694 display_matrix[i][0] = get_be32(pb); // 16.16 fixed point
1695 display_matrix[i][1] = get_be32(pb); // 16.16 fixed point
1696 get_be32(pb); // 2.30 fixed point (not used)
1699 width = get_be32(pb); // 16.16 fixed point track width
1700 height = get_be32(pb); // 16.16 fixed point track height
1701 sc->width = width >> 16;
1702 sc->height = height >> 16;
1704 // transform the display width/height according to the matrix
1705 // skip this if the display matrix is the default identity matrix
1706 // or if it is rotating the picture, ex iPhone 3GS
1707 // to keep the same scale, use [width height 1<<16]
1708 if (width && height &&
1709 ((display_matrix[0][0] != 65536 ||
1710 display_matrix[1][1] != 65536) &&
1711 !display_matrix[0][1] &&
1712 !display_matrix[1][0] &&
1713 !display_matrix[2][0] && !display_matrix[2][1])) {
1714 for (i = 0; i < 2; i++)
1715 disp_transform[i] =
1716 (int64_t) width * display_matrix[0][i] +
1717 (int64_t) height * display_matrix[1][i] +
1718 ((int64_t) display_matrix[2][i] << 16);
1720 //sample aspect ratio is new width/height divided by old width/height
1721 st->sample_aspect_ratio = av_d2q(
1722 ((double) disp_transform[0] * height) /
1723 ((double) disp_transform[1] * width), INT_MAX);
1725 return 0;
1728 static int mov_read_tfhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1730 MOVFragment *frag = &c->fragment;
1731 MOVTrackExt *trex = NULL;
1732 int flags, track_id, i;
1734 get_byte(pb); /* version */
1735 flags = get_be24(pb);
1737 track_id = get_be32(pb);
1738 if (!track_id)
1739 return -1;
1740 frag->track_id = track_id;
1741 for (i = 0; i < c->trex_count; i++)
1742 if (c->trex_data[i].track_id == frag->track_id) {
1743 trex = &c->trex_data[i];
1744 break;
1746 if (!trex) {
1747 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding trex\n");
1748 return -1;
1751 if (flags & 0x01) frag->base_data_offset = get_be64(pb);
1752 else frag->base_data_offset = frag->moof_offset;
1753 if (flags & 0x02) frag->stsd_id = get_be32(pb);
1754 else frag->stsd_id = trex->stsd_id;
1756 frag->duration = flags & 0x08 ? get_be32(pb) : trex->duration;
1757 frag->size = flags & 0x10 ? get_be32(pb) : trex->size;
1758 frag->flags = flags & 0x20 ? get_be32(pb) : trex->flags;
1759 dprintf(c->fc, "frag flags 0x%x\n", frag->flags);
1760 return 0;
1763 static int mov_read_trex(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1765 MOVTrackExt *trex;
1767 if ((uint64_t)c->trex_count+1 >= UINT_MAX / sizeof(*c->trex_data))
1768 return -1;
1769 trex = av_realloc(c->trex_data, (c->trex_count+1)*sizeof(*c->trex_data));
1770 if (!trex)
1771 return AVERROR(ENOMEM);
1772 c->trex_data = trex;
1773 trex = &c->trex_data[c->trex_count++];
1774 get_byte(pb); /* version */
1775 get_be24(pb); /* flags */
1776 trex->track_id = get_be32(pb);
1777 trex->stsd_id = get_be32(pb);
1778 trex->duration = get_be32(pb);
1779 trex->size = get_be32(pb);
1780 trex->flags = get_be32(pb);
1781 return 0;
1784 static int mov_read_trun(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1786 MOVFragment *frag = &c->fragment;
1787 AVStream *st = NULL;
1788 MOVStreamContext *sc;
1789 uint64_t offset;
1790 int64_t dts;
1791 int data_offset = 0;
1792 unsigned entries, first_sample_flags = frag->flags;
1793 int flags, distance, i;
1795 for (i = 0; i < c->fc->nb_streams; i++) {
1796 if (c->fc->streams[i]->id == frag->track_id) {
1797 st = c->fc->streams[i];
1798 break;
1801 if (!st) {
1802 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding track id %d\n", frag->track_id);
1803 return -1;
1805 sc = st->priv_data;
1806 if (sc->pseudo_stream_id+1 != frag->stsd_id)
1807 return 0;
1808 get_byte(pb); /* version */
1809 flags = get_be24(pb);
1810 entries = get_be32(pb);
1811 dprintf(c->fc, "flags 0x%x entries %d\n", flags, entries);
1812 if (flags & 0x001) data_offset = get_be32(pb);
1813 if (flags & 0x004) first_sample_flags = get_be32(pb);
1814 if (flags & 0x800) {
1815 MOVStts *ctts_data;
1816 if ((uint64_t)entries+sc->ctts_count >= UINT_MAX/sizeof(*sc->ctts_data))
1817 return -1;
1818 ctts_data = av_realloc(sc->ctts_data,
1819 (entries+sc->ctts_count)*sizeof(*sc->ctts_data));
1820 if (!ctts_data)
1821 return AVERROR(ENOMEM);
1822 sc->ctts_data = ctts_data;
1824 dts = st->duration;
1825 offset = frag->base_data_offset + data_offset;
1826 distance = 0;
1827 dprintf(c->fc, "first sample flags 0x%x\n", first_sample_flags);
1828 for (i = 0; i < entries; i++) {
1829 unsigned sample_size = frag->size;
1830 int sample_flags = i ? frag->flags : first_sample_flags;
1831 unsigned sample_duration = frag->duration;
1832 int keyframe;
1834 if (flags & 0x100) sample_duration = get_be32(pb);
1835 if (flags & 0x200) sample_size = get_be32(pb);
1836 if (flags & 0x400) sample_flags = get_be32(pb);
1837 if (flags & 0x800) {
1838 sc->ctts_data[sc->ctts_count].count = 1;
1839 sc->ctts_data[sc->ctts_count].duration = get_be32(pb);
1840 sc->ctts_count++;
1842 if ((keyframe = st->codec->codec_type == CODEC_TYPE_AUDIO ||
1843 (flags & 0x004 && !i && !sample_flags) || sample_flags & 0x2000000))
1844 distance = 0;
1845 av_add_index_entry(st, offset, dts, sample_size, distance,
1846 keyframe ? AVINDEX_KEYFRAME : 0);
1847 dprintf(c->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1848 "size %d, distance %d, keyframe %d\n", st->index, sc->sample_count+i,
1849 offset, dts, sample_size, distance, keyframe);
1850 distance++;
1851 dts += sample_duration;
1852 offset += sample_size;
1854 frag->moof_offset = offset;
1855 st->duration = dts;
1856 return 0;
1859 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
1860 /* like the files created with Adobe Premiere 5.0, for samples see */
1861 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
1862 static int mov_read_wide(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1864 int err;
1866 if (atom.size < 8)
1867 return 0; /* continue */
1868 if (get_be32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
1869 url_fskip(pb, atom.size - 4);
1870 return 0;
1872 atom.type = get_le32(pb);
1873 atom.offset += 8;
1874 atom.size -= 8;
1875 if (atom.type != MKTAG('m','d','a','t')) {
1876 url_fskip(pb, atom.size);
1877 return 0;
1879 err = mov_read_mdat(c, pb, atom);
1880 return err;
1883 static int mov_read_cmov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1885 #if CONFIG_ZLIB
1886 ByteIOContext ctx;
1887 uint8_t *cmov_data;
1888 uint8_t *moov_data; /* uncompressed data */
1889 long cmov_len, moov_len;
1890 int ret = -1;
1892 get_be32(pb); /* dcom atom */
1893 if (get_le32(pb) != MKTAG('d','c','o','m'))
1894 return -1;
1895 if (get_le32(pb) != MKTAG('z','l','i','b')) {
1896 av_log(c->fc, AV_LOG_ERROR, "unknown compression for cmov atom !");
1897 return -1;
1899 get_be32(pb); /* cmvd atom */
1900 if (get_le32(pb) != MKTAG('c','m','v','d'))
1901 return -1;
1902 moov_len = get_be32(pb); /* uncompressed size */
1903 cmov_len = atom.size - 6 * 4;
1905 cmov_data = av_malloc(cmov_len);
1906 if (!cmov_data)
1907 return AVERROR(ENOMEM);
1908 moov_data = av_malloc(moov_len);
1909 if (!moov_data) {
1910 av_free(cmov_data);
1911 return AVERROR(ENOMEM);
1913 get_buffer(pb, cmov_data, cmov_len);
1914 if(uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
1915 goto free_and_return;
1916 if(init_put_byte(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL, NULL) != 0)
1917 goto free_and_return;
1918 atom.type = MKTAG('m','o','o','v');
1919 atom.offset = 0;
1920 atom.size = moov_len;
1921 #ifdef DEBUG
1922 // { int fd = open("/tmp/uncompheader.mov", O_WRONLY | O_CREAT); write(fd, moov_data, moov_len); close(fd); }
1923 #endif
1924 ret = mov_read_default(c, &ctx, atom);
1925 free_and_return:
1926 av_free(moov_data);
1927 av_free(cmov_data);
1928 return ret;
1929 #else
1930 av_log(c->fc, AV_LOG_ERROR, "this file requires zlib support compiled in\n");
1931 return -1;
1932 #endif
1935 /* edit list atom */
1936 static int mov_read_elst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1938 MOVStreamContext *sc;
1939 int i, edit_count;
1941 if (c->fc->nb_streams < 1)
1942 return 0;
1943 sc = c->fc->streams[c->fc->nb_streams-1]->priv_data;
1945 get_byte(pb); /* version */
1946 get_be24(pb); /* flags */
1947 edit_count = get_be32(pb); /* entries */
1949 if((uint64_t)edit_count*12+8 > atom.size)
1950 return -1;
1952 for(i=0; i<edit_count; i++){
1953 int time;
1954 int duration = get_be32(pb); /* Track duration */
1955 time = get_be32(pb); /* Media time */
1956 get_be32(pb); /* Media rate */
1957 if (i == 0 && time >= -1) {
1958 sc->time_offset = time != -1 ? time : -duration;
1962 if(edit_count > 1)
1963 av_log(c->fc, AV_LOG_WARNING, "multiple edit list entries, "
1964 "a/v desync might occur, patch welcome\n");
1966 dprintf(c->fc, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, edit_count);
1967 return 0;
1970 static const MOVParseTableEntry mov_default_parse_table[] = {
1971 { MKTAG('a','v','s','s'), mov_read_extradata },
1972 { MKTAG('c','o','6','4'), mov_read_stco },
1973 { MKTAG('c','s','l','g'), mov_read_cslg },
1974 { MKTAG('c','t','t','s'), mov_read_ctts }, /* composition time to sample */
1975 { MKTAG('d','i','n','f'), mov_read_default },
1976 { MKTAG('d','r','e','f'), mov_read_dref },
1977 { MKTAG('e','d','t','s'), mov_read_default },
1978 { MKTAG('e','l','s','t'), mov_read_elst },
1979 { MKTAG('e','n','d','a'), mov_read_enda },
1980 { MKTAG('f','i','e','l'), mov_read_extradata },
1981 { MKTAG('f','t','y','p'), mov_read_ftyp },
1982 { MKTAG('g','l','b','l'), mov_read_glbl },
1983 { MKTAG('h','d','l','r'), mov_read_hdlr },
1984 { MKTAG('i','l','s','t'), mov_read_ilst },
1985 { MKTAG('j','p','2','h'), mov_read_extradata },
1986 { MKTAG('m','d','a','t'), mov_read_mdat },
1987 { MKTAG('m','d','h','d'), mov_read_mdhd },
1988 { MKTAG('m','d','i','a'), mov_read_default },
1989 { MKTAG('m','e','t','a'), mov_read_meta },
1990 { MKTAG('m','i','n','f'), mov_read_default },
1991 { MKTAG('m','o','o','f'), mov_read_moof },
1992 { MKTAG('m','o','o','v'), mov_read_moov },
1993 { MKTAG('m','v','e','x'), mov_read_default },
1994 { MKTAG('m','v','h','d'), mov_read_mvhd },
1995 { MKTAG('S','M','I',' '), mov_read_smi }, /* Sorenson extension ??? */
1996 { MKTAG('a','l','a','c'), mov_read_extradata }, /* alac specific atom */
1997 { MKTAG('a','v','c','C'), mov_read_glbl },
1998 { MKTAG('p','a','s','p'), mov_read_pasp },
1999 { MKTAG('s','t','b','l'), mov_read_default },
2000 { MKTAG('s','t','c','o'), mov_read_stco },
2001 { MKTAG('s','t','p','s'), mov_read_stps },
2002 { MKTAG('s','t','s','c'), mov_read_stsc },
2003 { MKTAG('s','t','s','d'), mov_read_stsd }, /* sample description */
2004 { MKTAG('s','t','s','s'), mov_read_stss }, /* sync sample */
2005 { MKTAG('s','t','s','z'), mov_read_stsz }, /* sample size */
2006 { MKTAG('s','t','t','s'), mov_read_stts },
2007 { MKTAG('s','t','z','2'), mov_read_stsz }, /* compact sample size */
2008 { MKTAG('t','k','h','d'), mov_read_tkhd }, /* track header */
2009 { MKTAG('t','f','h','d'), mov_read_tfhd }, /* track fragment header */
2010 { MKTAG('t','r','a','k'), mov_read_trak },
2011 { MKTAG('t','r','a','f'), mov_read_default },
2012 { MKTAG('t','r','e','x'), mov_read_trex },
2013 { MKTAG('t','r','u','n'), mov_read_trun },
2014 { MKTAG('u','d','t','a'), mov_read_default },
2015 { MKTAG('w','a','v','e'), mov_read_wave },
2016 { MKTAG('e','s','d','s'), mov_read_esds },
2017 { MKTAG('w','i','d','e'), mov_read_wide }, /* place holder */
2018 { MKTAG('c','m','o','v'), mov_read_cmov },
2019 { 0, NULL }
2022 static int mov_probe(AVProbeData *p)
2024 unsigned int offset;
2025 uint32_t tag;
2026 int score = 0;
2028 /* check file header */
2029 offset = 0;
2030 for(;;) {
2031 /* ignore invalid offset */
2032 if ((offset + 8) > (unsigned int)p->buf_size)
2033 return score;
2034 tag = AV_RL32(p->buf + offset + 4);
2035 switch(tag) {
2036 /* check for obvious tags */
2037 case MKTAG('j','P',' ',' '): /* jpeg 2000 signature */
2038 case MKTAG('m','o','o','v'):
2039 case MKTAG('m','d','a','t'):
2040 case MKTAG('p','n','o','t'): /* detect movs with preview pics like ew.mov and april.mov */
2041 case MKTAG('u','d','t','a'): /* Packet Video PVAuthor adds this and a lot of more junk */
2042 case MKTAG('f','t','y','p'):
2043 return AVPROBE_SCORE_MAX;
2044 /* those are more common words, so rate then a bit less */
2045 case MKTAG('e','d','i','w'): /* xdcam files have reverted first tags */
2046 case MKTAG('w','i','d','e'):
2047 case MKTAG('f','r','e','e'):
2048 case MKTAG('j','u','n','k'):
2049 case MKTAG('p','i','c','t'):
2050 return AVPROBE_SCORE_MAX - 5;
2051 case MKTAG(0x82,0x82,0x7f,0x7d):
2052 case MKTAG('s','k','i','p'):
2053 case MKTAG('u','u','i','d'):
2054 case MKTAG('p','r','f','l'):
2055 offset = AV_RB32(p->buf+offset) + offset;
2056 /* if we only find those cause probedata is too small at least rate them */
2057 score = AVPROBE_SCORE_MAX - 50;
2058 break;
2059 default:
2060 /* unrecognized tag */
2061 return score;
2064 return score;
2067 static int mov_read_header(AVFormatContext *s, AVFormatParameters *ap)
2069 MOVContext *mov = s->priv_data;
2070 ByteIOContext *pb = s->pb;
2071 int err;
2072 MOVAtom atom = { 0, 0, 0 };
2074 mov->fc = s;
2075 /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
2076 if(!url_is_streamed(pb))
2077 atom.size = url_fsize(pb);
2078 else
2079 atom.size = INT64_MAX;
2081 /* check MOV header */
2082 if ((err = mov_read_default(mov, pb, atom)) < 0) {
2083 av_log(s, AV_LOG_ERROR, "error reading header: %d\n", err);
2084 return err;
2086 if (!mov->found_moov) {
2087 av_log(s, AV_LOG_ERROR, "moov atom not found\n");
2088 return -1;
2090 dprintf(mov->fc, "on_parse_exit_offset=%lld\n", url_ftell(pb));
2092 return 0;
2095 static AVIndexEntry *mov_find_next_sample(AVFormatContext *s, AVStream **st)
2097 AVIndexEntry *sample = NULL;
2098 int64_t best_dts = INT64_MAX;
2099 int i;
2100 for (i = 0; i < s->nb_streams; i++) {
2101 AVStream *avst = s->streams[i];
2102 MOVStreamContext *msc = avst->priv_data;
2103 if (msc->pb && msc->current_sample < avst->nb_index_entries) {
2104 AVIndexEntry *current_sample = &avst->index_entries[msc->current_sample];
2105 int64_t dts = av_rescale(current_sample->timestamp, AV_TIME_BASE, msc->time_scale);
2106 dprintf(s, "stream %d, sample %d, dts %"PRId64"\n", i, msc->current_sample, dts);
2107 if (!sample || (url_is_streamed(s->pb) && current_sample->pos < sample->pos) ||
2108 (!url_is_streamed(s->pb) &&
2109 ((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
2110 ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
2111 (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
2112 sample = current_sample;
2113 best_dts = dts;
2114 *st = avst;
2118 return sample;
2121 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
2123 MOVContext *mov = s->priv_data;
2124 MOVStreamContext *sc;
2125 AVIndexEntry *sample;
2126 AVStream *st = NULL;
2127 int ret;
2128 retry:
2129 sample = mov_find_next_sample(s, &st);
2130 if (!sample) {
2131 mov->found_mdat = 0;
2132 if (!url_is_streamed(s->pb) ||
2133 mov_read_default(mov, s->pb, (MOVAtom){ 0, 0, INT64_MAX }) < 0 ||
2134 url_feof(s->pb))
2135 return AVERROR_EOF;
2136 dprintf(s, "read fragments, offset 0x%llx\n", url_ftell(s->pb));
2137 goto retry;
2139 sc = st->priv_data;
2140 /* must be done just before reading, to avoid infinite loop on sample */
2141 sc->current_sample++;
2143 if (st->discard != AVDISCARD_ALL) {
2144 if (url_fseek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
2145 av_log(mov->fc, AV_LOG_ERROR, "stream %d, offset 0x%"PRIx64": partial file\n",
2146 sc->ffindex, sample->pos);
2147 return -1;
2149 ret = av_get_packet(sc->pb, pkt, sample->size);
2150 if (ret < 0)
2151 return ret;
2152 #if CONFIG_DV_DEMUXER
2153 if (mov->dv_demux && sc->dv_audio_container) {
2154 dv_produce_packet(mov->dv_demux, pkt, pkt->data, pkt->size);
2155 av_free(pkt->data);
2156 pkt->size = 0;
2157 ret = dv_get_packet(mov->dv_demux, pkt);
2158 if (ret < 0)
2159 return ret;
2161 #endif
2164 pkt->stream_index = sc->ffindex;
2165 pkt->dts = sample->timestamp;
2166 if (sc->ctts_data) {
2167 pkt->pts = pkt->dts + sc->dts_shift + sc->ctts_data[sc->ctts_index].duration;
2168 /* update ctts context */
2169 sc->ctts_sample++;
2170 if (sc->ctts_index < sc->ctts_count &&
2171 sc->ctts_data[sc->ctts_index].count == sc->ctts_sample) {
2172 sc->ctts_index++;
2173 sc->ctts_sample = 0;
2175 if (sc->wrong_dts)
2176 pkt->dts = AV_NOPTS_VALUE;
2177 } else {
2178 int64_t next_dts = (sc->current_sample < st->nb_index_entries) ?
2179 st->index_entries[sc->current_sample].timestamp : st->duration;
2180 pkt->duration = next_dts - pkt->dts;
2181 pkt->pts = pkt->dts;
2183 if (st->discard == AVDISCARD_ALL)
2184 goto retry;
2185 pkt->flags |= sample->flags & AVINDEX_KEYFRAME ? PKT_FLAG_KEY : 0;
2186 pkt->pos = sample->pos;
2187 dprintf(s, "stream %d, pts %"PRId64", dts %"PRId64", pos 0x%"PRIx64", duration %d\n",
2188 pkt->stream_index, pkt->pts, pkt->dts, pkt->pos, pkt->duration);
2189 return 0;
2192 static int mov_seek_stream(AVStream *st, int64_t timestamp, int flags)
2194 MOVStreamContext *sc = st->priv_data;
2195 int sample, time_sample;
2196 int i;
2198 sample = av_index_search_timestamp(st, timestamp, flags);
2199 dprintf(st->codec, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
2200 if (sample < 0) /* not sure what to do */
2201 return -1;
2202 sc->current_sample = sample;
2203 dprintf(st->codec, "stream %d, found sample %d\n", st->index, sc->current_sample);
2204 /* adjust ctts index */
2205 if (sc->ctts_data) {
2206 time_sample = 0;
2207 for (i = 0; i < sc->ctts_count; i++) {
2208 int next = time_sample + sc->ctts_data[i].count;
2209 if (next > sc->current_sample) {
2210 sc->ctts_index = i;
2211 sc->ctts_sample = sc->current_sample - time_sample;
2212 break;
2214 time_sample = next;
2217 return sample;
2220 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
2222 AVStream *st;
2223 int64_t seek_timestamp, timestamp;
2224 int sample;
2225 int i;
2227 if (stream_index >= s->nb_streams)
2228 return -1;
2229 if (sample_time < 0)
2230 sample_time = 0;
2232 st = s->streams[stream_index];
2233 sample = mov_seek_stream(st, sample_time, flags);
2234 if (sample < 0)
2235 return -1;
2237 /* adjust seek timestamp to found sample timestamp */
2238 seek_timestamp = st->index_entries[sample].timestamp;
2240 for (i = 0; i < s->nb_streams; i++) {
2241 st = s->streams[i];
2242 if (stream_index == i)
2243 continue;
2245 timestamp = av_rescale_q(seek_timestamp, s->streams[stream_index]->time_base, st->time_base);
2246 mov_seek_stream(st, timestamp, flags);
2248 return 0;
2251 static int mov_read_close(AVFormatContext *s)
2253 MOVContext *mov = s->priv_data;
2254 int i, j;
2256 for (i = 0; i < s->nb_streams; i++) {
2257 AVStream *st = s->streams[i];
2258 MOVStreamContext *sc = st->priv_data;
2260 av_freep(&sc->ctts_data);
2261 for (j = 0; j < sc->drefs_count; j++)
2262 av_freep(&sc->drefs[j].path);
2263 av_freep(&sc->drefs);
2264 if (sc->pb && sc->pb != s->pb)
2265 url_fclose(sc->pb);
2267 av_freep(&st->codec->palctrl);
2270 if (mov->dv_demux) {
2271 for(i = 0; i < mov->dv_fctx->nb_streams; i++) {
2272 av_freep(&mov->dv_fctx->streams[i]->codec);
2273 av_freep(&mov->dv_fctx->streams[i]);
2275 av_freep(&mov->dv_fctx);
2276 av_freep(&mov->dv_demux);
2279 av_freep(&mov->trex_data);
2281 return 0;
2284 AVInputFormat mov_demuxer = {
2285 "mov,mp4,m4a,3gp,3g2,mj2",
2286 NULL_IF_CONFIG_SMALL("QuickTime/MPEG-4/Motion JPEG 2000 format"),
2287 sizeof(MOVContext),
2288 mov_probe,
2289 mov_read_header,
2290 mov_read_packet,
2291 mov_read_close,
2292 mov_read_seek,