3 * Copyright (c) 2001 Fabrice Bellard.
5 * This library is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU Lesser General Public
7 * License as published by the Free Software Foundation; either
8 * version 2 of the License, or (at your option) any later version.
10 * This library is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
15 * You should have received a copy of the GNU Lesser General Public
16 * License along with this library; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
30 * First version by Francois Revol revol@free.fr
31 * Seek function by Gael Chardon gael.dev@4now.net
33 * Features and limitations:
34 * - reads most of the QT files I have (at least the structure),
35 * the exceptions are .mov with zlib compressed headers ('cmov' section). It shouldn't be hard to implement.
36 * FIXED, Francois Revol, 07/17/2002
37 * - ffmpeg has nearly none of the usual QuickTime codecs,
38 * although I succesfully dumped raw and mp3 audio tracks off .mov files.
39 * Sample QuickTime files with mp3 audio can be found at: http://www.3ivx.com/showcase.html
40 * - .mp4 parsing is still hazardous, although the format really is QuickTime with some minor changes
41 * (to make .mov parser crash maybe ?), despite what they say in the MPEG FAQ at
42 * http://mpeg.telecomitalialab.com/faq.htm
43 * - the code is quite ugly... maybe I won't do it recursive next time :-)
44 * - seek is not supported with files that contain edit list
46 * Funny I didn't know about http://sourceforge.net/projects/qt-ffmpeg/
47 * when coding this :) (it's a writer anyway)
49 * Reference documents:
50 * http://www.geocities.com/xhelmboyx/quicktime/formats/qtm-layout.txt
52 * http://developer.apple.com/documentation/QuickTime/QTFF/
53 * http://developer.apple.com/documentation/QuickTime/QTFF/qtff.pdf
54 * QuickTime is a trademark of Apple (AFAIK :))
63 #include "qtpalette.h"
66 /* Allows seeking (MOV_SPLIT_CHUNKS should also be defined) */
69 /* allows chunk splitting - should work now... */
70 /* in case you can't read a file, try commenting */
71 #define MOV_SPLIT_CHUNKS
73 /* Special handling for movies created with Minolta Dimaxe Xi*/
74 /* this fix should not interfere with other .mov files, but just in case*/
75 #define MOV_MINOLTA_FIX
77 /* some streams in QT (and in MP4 mostly) aren't either video nor audio */
78 /* so we first list them as this, then clean up the list of streams we give back, */
79 /* getting rid of these */
80 #define CODEC_TYPE_MOV_OTHER (enum CodecType) 2
82 static const CodecTag mov_video_tags[] = {
83 /* { CODEC_ID_, MKTAG('c', 'v', 'i', 'd') }, *//* Cinepak */
84 /* { CODEC_ID_H263, MKTAG('r', 'a', 'w', ' ') }, *//* Uncompressed RGB */
85 /* { CODEC_ID_H263, MKTAG('Y', 'u', 'v', '2') }, *//* Uncompressed YUV422 */
86 /* { CODEC_ID_RAWVIDEO, MKTAG('A', 'V', 'U', 'I') }, *//* YUV with alpha-channel (AVID Uncompressed) */
91 { CODEC_ID_MJPEG, MKTAG('j', 'p', 'e', 'g') }, /* PhotoJPEG */
92 { CODEC_ID_MPEG1VIDEO, MKTAG('m', 'p', 'e', 'g') }, /* MPEG */
93 { CODEC_ID_MJPEG, MKTAG('m', 'j', 'p', 'a') }, /* Motion-JPEG (format A) */
94 { CODEC_ID_MJPEG, MKTAG('m', 'j', 'p', 'b') }, /* Motion-JPEG (format B) */
95 { CODEC_ID_MJPEG, MKTAG('A', 'V', 'D', 'J') }, /* MJPEG with alpha-channel (AVID JFIF meridien compressed) */
96 /* { CODEC_ID_MJPEG, MKTAG('A', 'V', 'R', 'n') }, *//* MJPEG with alpha-channel (AVID ABVB/Truevision NuVista) */
97 /* { CODEC_ID_GIF, MKTAG('g', 'i', 'f', ' ') }, *//* embedded gif files as frames (usually one "click to play movie" frame) */
99 { CODEC_ID_SVQ1, MKTAG('S', 'V', 'Q', '1') }, /* Sorenson Video v1 */
100 { CODEC_ID_SVQ1, MKTAG('s', 'v', 'q', '1') }, /* Sorenson Video v1 */
101 { CODEC_ID_SVQ1, MKTAG('s', 'v', 'q', 'i') }, /* Sorenson Video v1 (from QT specs)*/
102 { CODEC_ID_SVQ3, MKTAG('S', 'V', 'Q', '3') }, /* Sorenson Video v3 */
103 { CODEC_ID_MPEG4, MKTAG('m', 'p', '4', 'v') },
104 { CODEC_ID_MPEG4, MKTAG('D', 'I', 'V', 'X') }, /* OpenDiVX *//* sample files at http://heroinewarrior.com/xmovie.php3 use this tag */
105 { CODEC_ID_MPEG4, MKTAG('X', 'V', 'I', 'D') },
106 { CODEC_ID_MPEG4, MKTAG('3', 'I', 'V', '2') }, /* experimental: 3IVX files before ivx D4 4.5.1 */
107 /* { CODEC_ID_, MKTAG('I', 'V', '5', '0') }, *//* Indeo 5.0 */
108 { CODEC_ID_H263, MKTAG('h', '2', '6', '3') }, /* H263 */
109 { CODEC_ID_H263, MKTAG('s', '2', '6', '3') }, /* H263 ?? works */
110 { CODEC_ID_DVVIDEO, MKTAG('d', 'v', 'c', ' ') }, /* DV NTSC */
111 { CODEC_ID_DVVIDEO, MKTAG('d', 'v', 'c', 'p') }, /* DV PAL */
112 /* { CODEC_ID_DVVIDEO, MKTAG('A', 'V', 'd', 'v') }, *//* AVID dv */
113 { CODEC_ID_VP3, MKTAG('V', 'P', '3', '1') }, /* On2 VP3 */
114 { CODEC_ID_RPZA, MKTAG('r', 'p', 'z', 'a') }, /* Apple Video (RPZA) */
115 { CODEC_ID_CINEPAK, MKTAG('c', 'v', 'i', 'd') }, /* Cinepak */
116 { CODEC_ID_8BPS, MKTAG('8', 'B', 'P', 'S') }, /* Planar RGB (8BPS) */
117 { CODEC_ID_SMC, MKTAG('s', 'm', 'c', ' ') }, /* Apple Graphics (SMC) */
118 { CODEC_ID_QTRLE, MKTAG('r', 'l', 'e', ' ') }, /* Apple Animation (RLE) */
119 { CODEC_ID_QDRAW, MKTAG('q', 'd', 'r', 'w') }, /* QuickDraw */
120 { CODEC_ID_H264, MKTAG('a', 'v', 'c', '1') }, /* AVC-1/H.264 */
121 { CODEC_ID_MPEG2VIDEO, MKTAG('h', 'd', 'v', '2') }, /* MPEG2 produced by Sony HD camera */
122 { CODEC_ID_NONE, 0 },
125 static const CodecTag mov_audio_tags[] = {
126 /* { CODEC_ID_PCM_S16BE, MKTAG('N', 'O', 'N', 'E') }, *//* uncompressed */
127 { CODEC_ID_PCM_S16BE, MKTAG('t', 'w', 'o', 's') }, /* 16 bits */
128 /* { CODEC_ID_PCM_S8, MKTAG('t', 'w', 'o', 's') },*/ /* 8 bits */
129 { CODEC_ID_PCM_U8, MKTAG('r', 'a', 'w', ' ') }, /* 8 bits unsigned */
130 { CODEC_ID_PCM_S16LE, MKTAG('s', 'o', 'w', 't') }, /* */
131 { CODEC_ID_PCM_MULAW, MKTAG('u', 'l', 'a', 'w') }, /* */
132 { CODEC_ID_PCM_ALAW, MKTAG('a', 'l', 'a', 'w') }, /* */
133 { CODEC_ID_ADPCM_IMA_QT, MKTAG('i', 'm', 'a', '4') }, /* IMA-4 ADPCM */
134 { CODEC_ID_MACE3, MKTAG('M', 'A', 'C', '3') }, /* Macintosh Audio Compression and Expansion 3:1 */
135 { CODEC_ID_MACE6, MKTAG('M', 'A', 'C', '6') }, /* Macintosh Audio Compression and Expansion 6:1 */
137 { CODEC_ID_MP2, MKTAG('.', 'm', 'p', '3') }, /* MPEG layer 3 */ /* sample files at http://www.3ivx.com/showcase.html use this tag */
138 { CODEC_ID_MP2, 0x6D730055 }, /* MPEG layer 3 */
139 { CODEC_ID_MP2, 0x5500736D }, /* MPEG layer 3 *//* XXX: check endianness */
140 /* { CODEC_ID_OGG_VORBIS, MKTAG('O', 'g', 'g', 'S') }, *//* sample files at http://heroinewarrior.com/xmovie.php3 use this tag */
142 { CODEC_ID_AAC, MKTAG('m', 'p', '4', 'a') }, /* MPEG-4 AAC */
143 /* The standard for mpeg4 audio is still not normalised AFAIK anyway */
144 { CODEC_ID_AMR_NB, MKTAG('s', 'a', 'm', 'r') }, /* AMR-NB 3gp */
145 { CODEC_ID_AMR_WB, MKTAG('s', 'a', 'w', 'b') }, /* AMR-WB 3gp */
146 { CODEC_ID_AC3, MKTAG('m', 's', 0x20, 0x00) }, /* Dolby AC-3 */
147 { CODEC_ID_ALAC,MKTAG('a', 'l', 'a', 'c') }, /* Apple Lossless */
148 { CODEC_ID_NONE, 0 },
151 /* the QuickTime file format is quite convoluted...
152 * it has lots of index tables, each indexing something in another one...
153 * Here we just use what is needed to read the chunks
156 typedef struct MOV_sample_to_chunk_tbl {
160 } MOV_sample_to_chunk_tbl;
165 int64_t size; /* total size (excluding the size and type fields) */
177 uint32_t flags; // 24bit
179 /* 0x03 ESDescrTag */
181 #define MP4ODescrTag 0x01
182 #define MP4IODescrTag 0x02
183 #define MP4ESDescrTag 0x03
184 #define MP4DecConfigDescrTag 0x04
185 #define MP4DecSpecificDescrTag 0x05
186 #define MP4SLConfigDescrTag 0x06
187 #define MP4ContentIdDescrTag 0x07
188 #define MP4SupplContentIdDescrTag 0x08
189 #define MP4IPIPtrDescrTag 0x09
190 #define MP4IPMPPtrDescrTag 0x0A
191 #define MP4IPMPDescrTag 0x0B
192 #define MP4RegistrationDescrTag 0x0D
193 #define MP4ESIDIncDescrTag 0x0E
194 #define MP4ESIDRefDescrTag 0x0F
195 #define MP4FileIODescrTag 0x10
196 #define MP4FileODescrTag 0x11
197 #define MP4ExtProfileLevelDescrTag 0x13
198 #define MP4ExtDescrTagsStart 0x80
199 #define MP4ExtDescrTagsEnd 0xFE
200 uint8_t stream_priority;
202 /* 0x04 DecConfigDescrTag */
203 uint8_t object_type_id;
205 /* XXX: really streamType is
206 * only 6bit, followed by:
210 uint32_t buffer_size_db; // 24
211 uint32_t max_bitrate;
212 uint32_t avg_bitrate;
214 /* 0x05 DecSpecificDescrTag */
215 uint8_t decoder_cfg_len;
216 uint8_t *decoder_cfg;
218 /* 0x06 SLConfigDescrTag */
219 uint8_t sl_config_len;
223 struct MOVParseTableEntry;
225 typedef struct Time2Sample{
230 typedef struct MOVStreamContext {
231 int ffindex; /* the ffmpeg stream id */
232 int is_ff_stream; /* Is this stream presented to ffmpeg ? i.e. is this an audio or video stream ? */
235 int64_t *chunk_offsets;
237 Time2Sample *stts_data;
239 Time2Sample *ctts_data;
240 int edit_count; /* number of 'edit' (elst atom) */
241 long sample_to_chunk_sz;
242 MOV_sample_to_chunk_tbl *sample_to_chunk;
243 long sample_to_chunk_index;
244 int sample_to_time_index;
245 long sample_to_time_sample;
246 uint64_t sample_to_time_time;
247 int sample_to_ctime_index;
248 int sample_to_ctime_sample;
256 long left_in_chunk; /* how many samples before next chunk */
260 typedef struct MOVContext {
261 int mp4; /* set to 1 as soon as we are sure that the file is an .mp4 file (even some header parsing depends on this) */
264 int duration; /* duration of the longest track */
265 int found_moov; /* when both 'moov' and 'mdat' sections has been found */
266 int found_mdat; /* we suppose we have enough data to read the file */
269 int ni; ///< non interleaved mode
271 /* some streams listed here aren't presented to the ffmpeg API, since they aren't either video nor audio
272 * but we need the info to be able to skip data from those streams in the 'mdat' section
274 MOVStreamContext *streams[MAX_STREAMS];
276 int64_t next_chunk_offset;
277 MOVStreamContext *partial; /* != 0 : there is still to read in the current chunk */
279 MOV_ctab_t **ctab; /* color tables */
280 const struct MOVParseTableEntry *parse_table; /* could be eventually used to change the table */
281 /* NOTE: for recursion save to/ restore from local variable! */
283 AVPaletteControl palette_control;
287 /* XXX: it's the first time I make a recursive parser I think... sorry if it's ugly :P */
289 /* those functions parse an atom */
291 1: found what I wanted, exit
292 0: continue to parse next atom
293 -1: error occured, exit
295 typedef int (*mov_parse_function)(MOVContext *ctx, ByteIOContext *pb, MOV_atom_t atom);
297 /* links atom IDs to parse functions */
298 typedef struct MOVParseTableEntry {
300 mov_parse_function func;
301 } MOVParseTableEntry;
305 * XXX: static sux, even more in a multithreaded environment...
306 * Avoid them. This is here just to help debugging.
308 static int debug_indent = 0;
309 void print_atom(const char *str, MOV_atom_t atom)
312 tag = (unsigned int) atom.type;
314 if(tag == 0) tag = MKTAG('N', 'U', 'L', 'L');
316 av_log(NULL, AV_LOG_DEBUG, "|");
317 av_log(NULL, AV_LOG_DEBUG, "parse:");
318 av_log(NULL, AV_LOG_DEBUG, " %s: tag=%c%c%c%c offset=0x%x size=0x%x\n",
323 (unsigned int)atom.offset,
324 (unsigned int)atom.size);
325 assert((unsigned int)atom.size < 0x7fffffff);// catching errors
328 #define print_atom(a,b)
332 static int mov_read_leaf(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
334 print_atom("leaf", atom);
337 url_fskip(pb, atom.size);
338 /* url_seek(pb, atom_offset+atom.size, SEEK_SET); */
342 static int mov_read_default(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
344 int64_t total_size = 0;
350 print_atom("default", atom);
354 a.offset = atom.offset;
357 atom.size = 0x7fffffffffffffffLL;
358 while(((total_size + 8) < atom.size) && !url_feof(pb) && !err) {
362 a.size = get_be32(pb);
363 a.type = get_le32(pb);
367 //av_log(NULL, AV_LOG_DEBUG, "type: %08x %.4s sz: %Lx %Lx %Lx\n", type, (char*)&type, size, atom.size, total_size);
368 if (a.size == 1) { /* 64 bit extended size */
369 a.size = get_be64(pb) - 8;
374 a.size = atom.size - total_size;
378 for (i = 0; c->parse_table[i].type != 0L
379 && c->parse_table[i].type != a.type; i++)
387 // av_log(NULL, AV_LOG_DEBUG, " i=%ld\n", i);
388 if (c->parse_table[i].type == 0) { /* skip leaf atoms data */
389 // url_seek(pb, atom.offset+atom.size, SEEK_SET);
391 print_atom("unknown", a);
393 url_fskip(pb, a.size);
396 //char b[5] = { type & 0xff, (type >> 8) & 0xff, (type >> 16) & 0xff, (type >> 24) & 0xff, 0 };
397 //print_atom(b, type, offset, size);
399 err = (c->parse_table[i].func)(c, pb, a);
403 total_size += a.size;
406 if (!err && total_size < atom.size && atom.size < 0x7ffff) {
407 //av_log(NULL, AV_LOG_DEBUG, "RESET %Ld %Ld err:%d\n", atom.size, total_size, err);
408 url_fskip(pb, atom.size - total_size);
417 static int mov_read_ctab(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
420 url_fskip(pb, atom.size); // for now
422 VERY VERY BROKEN, NEVER execute this, needs rewrite
425 c->ctab = av_realloc(c->ctab, ++c->ctab_size);
426 t = c->ctab[c->ctab_size];
427 t->seed = get_be32(pb);
428 t->flags = get_be16(pb);
429 t->size = get_be16(pb) + 1;
430 len = 2 * t->size * 4;
432 t->clrs = av_malloc(len); // 16bit A R G B
434 get_buffer(pb, t->clrs, len);
441 static int mov_read_hdlr(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
443 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
448 print_atom("hdlr", atom);
450 get_byte(pb); /* version */
451 get_byte(pb); get_byte(pb); get_byte(pb); /* flags */
454 ctype = get_le32(pb);
455 type = get_le32(pb); /* component subtype */
458 av_log(NULL, AV_LOG_DEBUG, "ctype= %c%c%c%c (0x%08lx)\n", *((char *)&ctype), ((char *)&ctype)[1], ((char *)&ctype)[2], ((char *)&ctype)[3], (long) ctype);
459 av_log(NULL, AV_LOG_DEBUG, "stype= %c%c%c%c\n", *((char *)&type), ((char *)&type)[1], ((char *)&type)[2], ((char *)&type)[3]);
462 /* XXX: yeah this is ugly... */
463 if(ctype == MKTAG('m', 'h', 'l', 'r')) { /* MOV */
464 if(type == MKTAG('v', 'i', 'd', 'e'))
466 else if(type == MKTAG('s', 'o', 'u', 'n'))
468 } else if(ctype == 0) { /* MP4 */
469 if(type == MKTAG('v', 'i', 'd', 'e'))
471 else if(type == MKTAG('s', 'o', 'u', 'n'))
473 else if(type == MKTAG('o', 'd', 's', 'm'))
475 else if(type == MKTAG('s', 'd', 's', 'm'))
477 } else puts("hdlr: meta");
480 if(ctype == MKTAG('m', 'h', 'l', 'r')) { /* MOV */
481 /* helps parsing the string hereafter... */
483 if(type == MKTAG('v', 'i', 'd', 'e'))
484 st->codec.codec_type = CODEC_TYPE_VIDEO;
485 else if(type == MKTAG('s', 'o', 'u', 'n'))
486 st->codec.codec_type = CODEC_TYPE_AUDIO;
487 } else if(ctype == 0) { /* MP4 */
488 /* helps parsing the string hereafter... */
490 if(type == MKTAG('v', 'i', 'd', 'e'))
491 st->codec.codec_type = CODEC_TYPE_VIDEO;
492 else if(type == MKTAG('s', 'o', 'u', 'n'))
493 st->codec.codec_type = CODEC_TYPE_AUDIO;
495 get_be32(pb); /* component manufacture */
496 get_be32(pb); /* component flags */
497 get_be32(pb); /* component flags mask */
500 return 0; /* nothing left to read */
501 /* XXX: MP4 uses a C string, not a pascal one */
506 while(get_byte(pb) && (++len < (atom.size - 24)));
508 /* .mov: PASCAL string */
514 buf = (uint8_t*) av_malloc(len+1);
516 get_buffer(pb, buf, len);
518 av_log(NULL, AV_LOG_DEBUG, "**buf='%s'\n", buf);
525 url_fskip(pb, atom.size - (url_ftell(pb) - atom.offset));
529 static int mov_mp4_read_descr_len(ByteIOContext *pb)
534 int c = get_byte(pb);
535 len = (len << 7) | (c & 0x7f);
542 static int mov_mp4_read_descr(ByteIOContext *pb, int *tag)
546 len = mov_mp4_read_descr_len(pb);
548 av_log(NULL, AV_LOG_DEBUG, "MPEG4 description: tag=0x%02x len=%d\n", *tag, len);
553 static inline unsigned int get_be24(ByteIOContext *s)
556 val = get_byte(s) << 16;
557 val |= get_byte(s) << 8;
562 static int mov_read_esds(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
564 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
565 MOVStreamContext *sc = (MOVStreamContext *)st->priv_data;
566 int64_t start_pos = url_ftell(pb);
569 print_atom("esds", atom);
571 /* Well, broken but suffisant for some MP4 streams */
572 get_be32(pb); /* version + flags */
573 len = mov_mp4_read_descr(pb, &tag);
574 if (tag == MP4ESDescrTag) {
575 get_be16(pb); /* ID */
576 get_byte(pb); /* priority */
578 get_be16(pb); /* ID */
580 len = mov_mp4_read_descr(pb, &tag);
581 if (tag == MP4DecConfigDescrTag) {
582 sc->esds.object_type_id = get_byte(pb);
583 sc->esds.stream_type = get_byte(pb);
584 sc->esds.buffer_size_db = get_be24(pb);
585 sc->esds.max_bitrate = get_be32(pb);
586 sc->esds.avg_bitrate = get_be32(pb);
588 len = mov_mp4_read_descr(pb, &tag);
589 //av_log(NULL, AV_LOG_DEBUG, "LEN %d TAG %d m:%d a:%d\n", len, tag, sc->esds.max_bitrate, sc->esds.avg_bitrate);
590 if (tag == MP4DecSpecificDescrTag) {
592 av_log(NULL, AV_LOG_DEBUG, "Specific MPEG4 header len=%d\n", len);
594 st->codec.extradata = (uint8_t*) av_mallocz(len + FF_INPUT_BUFFER_PADDING_SIZE);
595 if (st->codec.extradata) {
596 get_buffer(pb, st->codec.extradata, len);
597 st->codec.extradata_size = len;
601 /* in any case, skip garbage */
602 url_fskip(pb, atom.size - ((url_ftell(pb) - start_pos)));
606 /* this atom contains actual media data */
607 static int mov_read_mdat(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
609 print_atom("mdat", atom);
611 if(atom.size == 0) /* wrong one (MP4) */
614 c->mdat_offset = atom.offset;
615 c->mdat_size = atom.size;
617 return 1; /* found both, just go */
618 url_fskip(pb, atom.size);
619 return 0; /* now go for moov */
622 /* this atom should contain all header atoms */
623 static int mov_read_moov(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
627 print_atom("moov", atom);
629 err = mov_read_default(c, pb, atom);
630 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
631 /* so we don't parse the whole file if over a network */
634 return 1; /* found both, just go */
635 return 0; /* now go for mdat */
639 static int mov_read_mdhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
641 print_atom("mdhd", atom);
643 get_byte(pb); /* version */
645 get_byte(pb); get_byte(pb);
646 get_byte(pb); /* flags */
648 get_be32(pb); /* creation time */
649 get_be32(pb); /* modification time */
651 c->streams[c->fc->nb_streams-1]->time_scale = get_be32(pb);
654 av_log(NULL, AV_LOG_DEBUG, "track[%i].time_scale = %i\n", c->fc->nb_streams-1, c->streams[c->fc->nb_streams-1]->time_scale); /* time scale */
656 c->fc->streams[c->fc->nb_streams-1]->duration= get_be32(pb); /* duration */
658 get_be16(pb); /* language */
659 get_be16(pb); /* quality */
664 static int mov_read_mvhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
666 print_atom("mvhd", atom);
668 get_byte(pb); /* version */
669 get_byte(pb); get_byte(pb); get_byte(pb); /* flags */
671 get_be32(pb); /* creation time */
672 get_be32(pb); /* modification time */
673 c->time_scale = get_be32(pb); /* time scale */
675 av_log(NULL, AV_LOG_DEBUG, "time scale = %i\n", c->time_scale);
677 c->duration = get_be32(pb); /* duration */
678 get_be32(pb); /* preferred scale */
680 get_be16(pb); /* preferred volume */
682 url_fskip(pb, 10); /* reserved */
684 url_fskip(pb, 36); /* display matrix */
686 get_be32(pb); /* preview time */
687 get_be32(pb); /* preview duration */
688 get_be32(pb); /* poster time */
689 get_be32(pb); /* selection time */
690 get_be32(pb); /* selection duration */
691 get_be32(pb); /* current time */
692 get_be32(pb); /* next track ID */
697 static int mov_read_smi(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
699 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
701 if((uint64_t)atom.size > (1<<30))
704 // currently SVQ3 decoder expect full STSD header - so let's fake it
705 // this should be fixed and just SMI header should be passed
706 av_free(st->codec.extradata);
707 st->codec.extradata_size = 0x5a + atom.size;
708 st->codec.extradata = (uint8_t*) av_mallocz(st->codec.extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
710 if (st->codec.extradata) {
711 strcpy(st->codec.extradata, "SVQ3"); // fake
712 get_buffer(pb, st->codec.extradata + 0x5a, atom.size);
713 //av_log(NULL, AV_LOG_DEBUG, "Reading SMI %Ld %s\n", atom.size, (char*)st->codec.extradata + 0x5a);
715 url_fskip(pb, atom.size);
720 static int mov_read_avcC(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
722 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
724 if((uint64_t)atom.size > (1<<30))
727 av_free(st->codec.extradata);
729 st->codec.extradata_size = atom.size;
730 st->codec.extradata = (uint8_t*) av_mallocz(st->codec.extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
732 if (st->codec.extradata) {
733 get_buffer(pb, st->codec.extradata, atom.size);
735 url_fskip(pb, atom.size);
740 static int mov_read_stco(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
742 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
743 MOVStreamContext *sc = (MOVStreamContext *)st->priv_data;
744 unsigned int i, entries;
746 print_atom("stco", atom);
748 get_byte(pb); /* version */
749 get_byte(pb); get_byte(pb); get_byte(pb); /* flags */
751 entries = get_be32(pb);
753 if(entries >= UINT_MAX/sizeof(int64_t))
756 sc->chunk_count = entries;
757 sc->chunk_offsets = (int64_t*) av_malloc(entries * sizeof(int64_t));
758 if (!sc->chunk_offsets)
760 if (atom.type == MKTAG('s', 't', 'c', 'o')) {
761 for(i=0; i<entries; i++) {
762 sc->chunk_offsets[i] = get_be32(pb);
764 } else if (atom.type == MKTAG('c', 'o', '6', '4')) {
765 for(i=0; i<entries; i++) {
766 sc->chunk_offsets[i] = get_be64(pb);
771 for(i=0; i<c->fc->nb_streams; i++){
772 MOVStreamContext *sc2 = (MOVStreamContext *)c->fc->streams[i]->priv_data;
773 if(sc2 && sc2->chunk_offsets){
774 int64_t first= sc2->chunk_offsets[0];
775 int64_t last= sc2->chunk_offsets[sc2->chunk_count-1];
776 if(first >= sc->chunk_offsets[entries-1] || last <= sc->chunk_offsets[0])
782 for(i=0; i<entries; i++) {
783 av_log(NULL, AV_LOG_DEBUG, "chunk offset=0x%Lx\n", sc->chunk_offsets[i]);
790 static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
792 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
793 //MOVStreamContext *sc = (MOVStreamContext *)st->priv_data;
794 int entries, frames_per_sample;
796 uint8_t codec_name[32];
798 /* for palette traversal */
806 unsigned char *color_table;
808 unsigned char r, g, b;
810 print_atom("stsd", atom);
812 get_byte(pb); /* version */
813 get_byte(pb); get_byte(pb); get_byte(pb); /* flags */
815 entries = get_be32(pb);
817 while(entries--) { //Parsing Sample description table
819 int size = get_be32(pb); /* size */
820 format = get_le32(pb); /* data format */
822 get_be32(pb); /* reserved */
823 get_be16(pb); /* reserved */
824 get_be16(pb); /* index */
826 /* for MPEG4: set codec type by looking for it */
827 id = codec_get_id(mov_video_tags, format);
830 codec = avcodec_find_decoder(id);
832 st->codec.codec_type = codec->type;
835 av_log(NULL, AV_LOG_DEBUG, "size=%d 4CC= %c%c%c%c codec_type=%d\n",
837 (format >> 0) & 0xff,
838 (format >> 8) & 0xff,
839 (format >> 16) & 0xff,
840 (format >> 24) & 0xff,
841 st->codec.codec_type);
843 st->codec.codec_tag = format;
844 if(st->codec.codec_type==CODEC_TYPE_VIDEO) {
845 MOV_atom_t a = { 0, 0, 0 };
846 st->codec.codec_id = id;
847 get_be16(pb); /* version */
848 get_be16(pb); /* revision level */
849 get_be32(pb); /* vendor */
850 get_be32(pb); /* temporal quality */
851 get_be32(pb); /* spacial quality */
852 if(st->codec.codec_id == CODEC_ID_MPEG4){ //FIXME this is silly
856 st->codec.width = get_be16(pb); /* width */
857 st->codec.height = get_be16(pb); /* height */
859 get_be32(pb); /* horiz resolution */
860 get_be32(pb); /* vert resolution */
861 get_be32(pb); /* data size, always 0 */
862 frames_per_sample = get_be16(pb); /* frames per samples */
864 av_log(NULL, AV_LOG_DEBUG, "frames/samples = %d\n", frames_per_sample);
866 get_buffer(pb, codec_name, 32); /* codec name, pascal string (FIXME: true for mp4?) */
867 if (codec_name[0] <= 31) {
868 memcpy(st->codec.codec_name, &codec_name[1],codec_name[0]);
869 st->codec.codec_name[codec_name[0]] = 0;
872 st->codec.bits_per_sample = get_be16(pb); /* depth */
873 st->codec.color_table_id = get_be16(pb); /* colortable id */
875 /* These are set in mov_read_stts and might already be set!
876 st->codec.time_base.den = 25;
877 st->codec.time_base.num = 1;
879 size -= (16+8*4+2+32+2*2);
885 a.size = get_be32(pb);
886 a.type = get_le32(pb);
889 av_log(NULL, AV_LOG_DEBUG, "VIDEO: atom_type=%c%c%c%c atom.size=%Ld size_left=%d\n",
890 (a.type >> 0) & 0xff,
891 (a.type >> 8) & 0xff,
892 (a.type >> 16) & 0xff,
893 (a.type >> 24) & 0xff,
896 start_pos = url_ftell(pb);
899 case MKTAG('e', 's', 'd', 's'):
902 /* Well, broken but suffisant for some MP4 streams */
903 get_be32(pb); /* version + flags */
904 len = mov_mp4_read_descr(pb, &tag);
907 get_be16(pb); /* ID */
908 get_byte(pb); /* priority */
909 len = mov_mp4_read_descr(pb, &tag);
912 /* MP4DecConfigDescrTag */
913 get_byte(pb); /* objectTypeId */
914 get_be32(pb); /* streamType + buffer size */
915 get_be32(pb); /* max bit rate */
916 get_be32(pb); /* avg bit rate */
917 len = mov_mp4_read_descr(pb, &tag);
920 /* MP4DecSpecificDescrTag */
922 av_log(NULL, AV_LOG_DEBUG, "Specific MPEG4 header len=%d\n", len);
924 sc->header_data = av_mallocz(len);
925 if (sc->header_data) {
926 get_buffer(pb, sc->header_data, len);
927 sc->header_len = len;
930 /* in any case, skip garbage */
937 av_log(NULL, AV_LOG_DEBUG, "ATOMENEWSIZE %Ld %d\n", atom.size, url_ftell(pb) - start_pos);
939 url_fskip(pb, (atom.size - 8) -
940 ((url_ftell(pb) - start_pos)));
941 size -= atom.size - 8;
945 /* unknown extension */
950 /* figure out the palette situation */
951 color_depth = st->codec.bits_per_sample & 0x1F;
952 color_greyscale = st->codec.bits_per_sample & 0x20;
954 /* if the depth is 2, 4, or 8 bpp, file is palettized */
955 if ((color_depth == 2) || (color_depth == 4) ||
956 (color_depth == 8)) {
958 if (color_greyscale) {
960 /* compute the greyscale palette */
961 color_count = 1 << color_depth;
963 color_dec = 256 / (color_count - 1);
964 for (j = 0; j < color_count; j++) {
965 r = g = b = color_index;
966 c->palette_control.palette[j] =
967 (r << 16) | (g << 8) | (b);
968 color_index -= color_dec;
973 } else if (st->codec.color_table_id & 0x08) {
975 /* if flag bit 3 is set, use the default palette */
976 color_count = 1 << color_depth;
977 if (color_depth == 2)
978 color_table = ff_qt_default_palette_4;
979 else if (color_depth == 4)
980 color_table = ff_qt_default_palette_16;
982 color_table = ff_qt_default_palette_256;
984 for (j = 0; j < color_count; j++) {
985 r = color_table[j * 4 + 0];
986 g = color_table[j * 4 + 1];
987 b = color_table[j * 4 + 2];
988 c->palette_control.palette[j] =
989 (r << 16) | (g << 8) | (b);
994 /* load the palette from the file */
995 color_start = get_be32(pb);
996 color_count = get_be16(pb);
997 color_end = get_be16(pb);
998 for (j = color_start; j <= color_end; j++) {
999 /* each R, G, or B component is 16 bits;
1000 * only use the top 8 bits; skip alpha bytes
1010 c->palette_control.palette[j] =
1011 (r << 16) | (g << 8) | (b);
1015 st->codec.palctrl = &c->palette_control;
1016 st->codec.palctrl->palette_changed = 1;
1018 st->codec.palctrl = NULL;
1021 mov_read_default(c, pb, a);
1024 st->codec.codec_id = codec_get_id(mov_audio_tags, format);
1025 if(st->codec.codec_id==CODEC_ID_AMR_NB || st->codec.codec_id==CODEC_ID_AMR_WB) //from TS26.244
1028 av_log(NULL, AV_LOG_DEBUG, "AMR-NB or AMR-WB audio identified!!\n");
1030 get_be32(pb);get_be32(pb); //Reserved_8
1031 get_be16(pb);//Reserved_2
1032 get_be16(pb);//Reserved_2
1033 get_be32(pb);//Reserved_4
1034 get_be16(pb);//TimeScale
1035 get_be16(pb);//Reserved_2
1037 //AMRSpecificBox.(10 bytes)
1039 get_be32(pb); //size
1040 get_be32(pb); //type=='damr'
1041 get_be32(pb); //vendor
1042 get_byte(pb); //decoder version
1043 get_be16(pb); //mode_set
1044 get_byte(pb); //mode_change_period
1045 get_byte(pb); //frames_per_sample
1047 st->duration = AV_NOPTS_VALUE;//Not possible to get from this info, must count number of AMR frames
1048 if(st->codec.codec_id==CODEC_ID_AMR_NB)
1050 st->codec.sample_rate=8000;
1051 st->codec.channels=1;
1055 st->codec.sample_rate=16000;
1056 st->codec.channels=1;
1058 st->codec.bits_per_sample=16;
1059 st->codec.bit_rate=0; /*It is not possible to tell this before we have
1060 an audio frame and even then every frame can be different*/
1062 else if( st->codec.codec_tag == MKTAG( 'm', 'p', '4', 's' ))
1064 //This is some stuff for the hint track, lets ignore it!
1065 //Do some mp4 auto detect.
1068 url_fskip(pb, size); /* The mp4s atom also contians a esds atom that we can skip*/
1070 else if( st->codec.codec_tag == MKTAG( 'm', 'p', '4', 'a' ))
1075 /* Handle mp4 audio tag */
1076 mp4_version=get_be16(pb);/*version*/
1077 get_be16(pb); /*revesion*/
1079 st->codec.channels = get_be16(pb); /* channels */
1080 st->codec.bits_per_sample = get_be16(pb); /* bits per sample */
1082 st->codec.sample_rate = get_be16(pb); /* sample rate, not always correct */
1089 a.size=size-(16+20+16);
1092 a.size=size-(16+20);
1094 a.offset=url_ftell(pb);
1096 mov_read_default(c, pb, a);
1098 /* Get correct sample rate from extradata */
1099 if(st->codec.extradata_size) {
1100 const int samplerate_table[] = {
1101 96000, 88200, 64000, 48000, 44100, 32000,
1102 24000, 22050, 16000, 12000, 11025, 8000,
1105 unsigned char *px = st->codec.extradata;
1106 // 5 bits objectTypeIndex, 4 bits sampleRateIndex, 4 bits channels
1107 int samplerate_index = ((px[0] & 7) << 1) + ((px[1] >> 7) & 1);
1108 st->codec.sample_rate = samplerate_table[samplerate_index];
1109 st->codec.channels = (px[1] >> 3) & 15;
1112 else if( st->codec.codec_tag == MKTAG( 'a', 'l', 'a', 'c' ))
1114 /* Handle alac audio tag + special extradata */
1115 get_be32(pb); /* version */
1117 st->codec.channels = get_be16(pb); /* channels */
1118 st->codec.bits_per_sample = get_be16(pb); /* bits per sample */
1120 st->codec.sample_rate = get_be16(pb);
1123 /* fetch the 36-byte extradata needed for alac decoding */
1124 st->codec.extradata_size = 36;
1125 st->codec.extradata = (uint8_t*)
1126 av_mallocz(st->codec.extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1127 get_buffer(pb, st->codec.extradata, st->codec.extradata_size);
1129 else if(size>=(16+20))
1130 {//16 bytes read, reading atleast 20 more
1133 av_log(NULL, AV_LOG_DEBUG, "audio size=0x%X\n",size);
1135 version = get_be16(pb); /* version */
1136 get_be16(pb); /* revision level */
1137 get_be32(pb); /* vendor */
1139 st->codec.channels = get_be16(pb); /* channel count */
1140 st->codec.bits_per_sample = get_be16(pb); /* sample size */
1142 /* handle specific s8 codec */
1143 get_be16(pb); /* compression id = 0*/
1144 get_be16(pb); /* packet size = 0 */
1146 st->codec.sample_rate = ((get_be32(pb) >> 16));
1147 //av_log(NULL, AV_LOG_DEBUG, "CODECID %d %d %.4s\n", st->codec.codec_id, CODEC_ID_PCM_S16BE, (char*)&format);
1149 switch (st->codec.codec_id) {
1150 case CODEC_ID_PCM_S16BE:
1151 if (st->codec.bits_per_sample == 8)
1152 st->codec.codec_id = CODEC_ID_PCM_S8;
1154 case CODEC_ID_PCM_U8:
1155 st->codec.bit_rate = st->codec.sample_rate * 8;
1161 //Read QT version 1 fields. In version 0 theese dont exist
1163 av_log(NULL, AV_LOG_DEBUG, "version =%d mp4=%d\n",version,c->mp4);
1164 av_log(NULL, AV_LOG_DEBUG, "size-(16+20+16)=%d\n",size-(16+20+16));
1166 if((version==1) && size>=(16+20+16))
1168 get_be32(pb); /* samples per packet */
1169 get_be32(pb); /* bytes per packet */
1170 get_be32(pb); /* bytes per frame */
1171 get_be32(pb); /* bytes per sample */
1174 //Optional, additional atom-based fields
1175 MOV_atom_t a = { format, url_ftell(pb), size - (16 + 20 + 16 + 8) };
1177 av_log(NULL, AV_LOG_DEBUG, "offest=0x%X, sizeleft=%d=0x%x,format=%c%c%c%c\n",(int)url_ftell(pb),size - (16 + 20 + 16 ),size - (16 + 20 + 16 ),
1178 (format >> 0) & 0xff,
1179 (format >> 8) & 0xff,
1180 (format >> 16) & 0xff,
1181 (format >> 24) & 0xff);
1183 mov_read_default(c, pb, a);
1188 //We should be down to 0 bytes here, but lets make sure.
1192 av_log(NULL, AV_LOG_DEBUG, "skipping 0x%X bytes\n",size-(16+20));
1194 url_fskip(pb, size);
1200 //Unknown size, but lets do our best and skip the rest.
1202 av_log(NULL, AV_LOG_DEBUG, "Strange size, skipping 0x%X bytes\n",size);
1204 url_fskip(pb, size);
1212 static int mov_read_stsc(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1214 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1215 MOVStreamContext *sc = (MOVStreamContext *)st->priv_data;
1216 unsigned int i, entries;
1218 print_atom("stsc", atom);
1220 get_byte(pb); /* version */
1221 get_byte(pb); get_byte(pb); get_byte(pb); /* flags */
1223 entries = get_be32(pb);
1225 if(entries >= UINT_MAX / sizeof(MOV_sample_to_chunk_tbl))
1229 av_log(NULL, AV_LOG_DEBUG, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
1231 sc->sample_to_chunk_sz = entries;
1232 sc->sample_to_chunk = (MOV_sample_to_chunk_tbl*) av_malloc(entries * sizeof(MOV_sample_to_chunk_tbl));
1233 if (!sc->sample_to_chunk)
1235 for(i=0; i<entries; i++) {
1236 sc->sample_to_chunk[i].first = get_be32(pb);
1237 sc->sample_to_chunk[i].count = get_be32(pb);
1238 sc->sample_to_chunk[i].id = get_be32(pb);
1240 /* av_log(NULL, AV_LOG_DEBUG, "sample_to_chunk first=%ld count=%ld, id=%ld\n", sc->sample_to_chunk[i].first, sc->sample_to_chunk[i].count, sc->sample_to_chunk[i].id); */
1246 static int mov_read_stss(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1248 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1249 MOVStreamContext *sc = (MOVStreamContext *)st->priv_data;
1250 unsigned int i, entries;
1252 print_atom("stss", atom);
1254 get_byte(pb); /* version */
1255 get_byte(pb); get_byte(pb); get_byte(pb); /* flags */
1257 entries = get_be32(pb);
1259 if(entries >= UINT_MAX / sizeof(long))
1262 sc->keyframe_count = entries;
1264 av_log(NULL, AV_LOG_DEBUG, "keyframe_count = %ld\n", sc->keyframe_count);
1266 sc->keyframes = (long*) av_malloc(entries * sizeof(long));
1269 for(i=0; i<entries; i++) {
1270 sc->keyframes[i] = get_be32(pb);
1272 /* av_log(NULL, AV_LOG_DEBUG, "keyframes[]=%ld\n", sc->keyframes[i]); */
1278 static int mov_read_stsz(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1280 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1281 MOVStreamContext *sc = (MOVStreamContext *)st->priv_data;
1282 unsigned int i, entries;
1284 print_atom("stsz", atom);
1286 get_byte(pb); /* version */
1287 get_byte(pb); get_byte(pb); get_byte(pb); /* flags */
1289 sc->sample_size = get_be32(pb);
1290 entries = get_be32(pb);
1291 if(entries >= UINT_MAX / sizeof(long))
1294 sc->sample_count = entries;
1296 av_log(NULL, AV_LOG_DEBUG, "sample_size = %ld sample_count = %ld\n", sc->sample_size, sc->sample_count);
1299 return 0; /* there isn't any table following */
1300 sc->sample_sizes = (long*) av_malloc(entries * sizeof(long));
1301 if (!sc->sample_sizes)
1303 for(i=0; i<entries; i++) {
1304 sc->sample_sizes[i] = get_be32(pb);
1306 av_log(NULL, AV_LOG_DEBUG, "sample_sizes[]=%ld\n", sc->sample_sizes[i]);
1312 static int mov_read_stts(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1314 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1315 //MOVStreamContext *sc = (MOVStreamContext *)st->priv_data;
1316 unsigned int i, entries;
1318 int64_t total_sample_count=0;
1320 print_atom("stts", atom);
1322 get_byte(pb); /* version */
1323 get_byte(pb); get_byte(pb); get_byte(pb); /* flags */
1324 entries = get_be32(pb);
1325 if(entries >= UINT_MAX / sizeof(Time2Sample))
1328 c->streams[c->fc->nb_streams-1]->stts_count = entries;
1329 c->streams[c->fc->nb_streams-1]->stts_data = av_malloc(entries * sizeof(Time2Sample));
1332 av_log(NULL, AV_LOG_DEBUG, "track[%i].stts.entries = %i\n", c->fc->nb_streams-1, entries);
1334 for(i=0; i<entries; i++) {
1335 int sample_duration;
1338 sample_count=get_be32(pb);
1339 sample_duration = get_be32(pb);
1340 c->streams[c->fc->nb_streams - 1]->stts_data[i].count= sample_count;
1341 c->streams[c->fc->nb_streams - 1]->stts_data[i].duration= sample_duration;
1344 av_log(NULL, AV_LOG_DEBUG, "sample_count=%d, sample_duration=%d\n",sample_count,sample_duration);
1346 duration+=sample_duration*sample_count;
1347 total_sample_count+=sample_count;
1350 av_set_pts_info(st, 64, 1, c->streams[c->fc->nb_streams-1]->time_scale);
1351 // st->codec.time_base.num = 1;
1352 // st->codec.time_base.den = c->streams[c->fc->nb_streams-1]->time_scale;
1353 st->nb_frames= total_sample_count;
1355 st->duration= duration;
1359 static int mov_read_ctts(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1361 // AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1362 //MOVStreamContext *sc = (MOVStreamContext *)st->priv_data;
1363 unsigned int i, entries;
1365 print_atom("ctts", atom);
1367 get_byte(pb); /* version */
1368 get_byte(pb); get_byte(pb); get_byte(pb); /* flags */
1369 entries = get_be32(pb);
1370 if(entries >= UINT_MAX / sizeof(Time2Sample))
1373 c->streams[c->fc->nb_streams-1]->ctts_count = entries;
1374 c->streams[c->fc->nb_streams-1]->ctts_data = av_malloc(entries * sizeof(Time2Sample));
1377 av_log(NULL, AV_LOG_DEBUG, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
1379 for(i=0; i<entries; i++) {
1380 c->streams[c->fc->nb_streams - 1]->ctts_data[i].count= get_be32(pb);
1381 c->streams[c->fc->nb_streams - 1]->ctts_data[i].duration= get_be32(pb);
1386 static int mov_read_trak(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1389 MOVStreamContext *sc;
1391 print_atom("trak", atom);
1393 st = av_new_stream(c->fc, c->fc->nb_streams);
1395 sc = (MOVStreamContext*) av_mallocz(sizeof(MOVStreamContext));
1401 sc->sample_to_chunk_index = -1;
1403 st->codec.codec_type = CODEC_TYPE_MOV_OTHER;
1404 st->start_time = 0; /* XXX: check */
1405 c->streams[c->fc->nb_streams-1] = sc;
1407 return mov_read_default(c, pb, atom);
1410 static int mov_read_tkhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1414 print_atom("tkhd", atom);
1416 st = c->fc->streams[c->fc->nb_streams-1];
1418 get_byte(pb); /* version */
1420 get_byte(pb); get_byte(pb);
1421 get_byte(pb); /* flags */
1423 MOV_TRACK_ENABLED 0x0001
1424 MOV_TRACK_IN_MOVIE 0x0002
1425 MOV_TRACK_IN_PREVIEW 0x0004
1426 MOV_TRACK_IN_POSTER 0x0008
1429 get_be32(pb); /* creation time */
1430 get_be32(pb); /* modification time */
1431 st->id = (int)get_be32(pb); /* track id (NOT 0 !)*/
1432 get_be32(pb); /* reserved */
1433 st->start_time = 0; /* check */
1434 get_be32(pb); /* highlevel (considering edits) duration in movie timebase */
1435 get_be32(pb); /* reserved */
1436 get_be32(pb); /* reserved */
1438 get_be16(pb); /* layer */
1439 get_be16(pb); /* alternate group */
1440 get_be16(pb); /* volume */
1441 get_be16(pb); /* reserved */
1443 url_fskip(pb, 36); /* display matrix */
1445 /* those are fixed-point */
1446 /*st->codec.width =*/ get_be32(pb) >> 16; /* track width */
1447 /*st->codec.height =*/ get_be32(pb) >> 16; /* track height */
1452 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
1453 /* like the files created with Adobe Premiere 5.0, for samples see */
1454 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
1455 static int mov_read_wide(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1460 print_atom("wide", atom);
1464 return 0; /* continue */
1465 if (get_be32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
1466 url_fskip(pb, atom.size - 4);
1469 atom.type = get_le32(pb);
1472 if (atom.type != MKTAG('m', 'd', 'a', 't')) {
1473 url_fskip(pb, atom.size);
1476 err = mov_read_mdat(c, pb, atom);
1485 static int null_read_packet(void *opaque, uint8_t *buf, int buf_size)
1490 static int mov_read_cmov(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1494 uint8_t *moov_data; /* uncompressed data */
1495 long cmov_len, moov_len;
1498 print_atom("cmov", atom);
1500 get_be32(pb); /* dcom atom */
1501 if (get_le32(pb) != MKTAG( 'd', 'c', 'o', 'm' ))
1503 if (get_le32(pb) != MKTAG( 'z', 'l', 'i', 'b' )) {
1504 av_log(NULL, AV_LOG_DEBUG, "unknown compression for cmov atom !");
1507 get_be32(pb); /* cmvd atom */
1508 if (get_le32(pb) != MKTAG( 'c', 'm', 'v', 'd' ))
1510 moov_len = get_be32(pb); /* uncompressed size */
1511 cmov_len = atom.size - 6 * 4;
1513 cmov_data = (uint8_t *) av_malloc(cmov_len);
1516 moov_data = (uint8_t *) av_malloc(moov_len);
1521 get_buffer(pb, cmov_data, cmov_len);
1522 if(uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
1524 if(init_put_byte(&ctx, moov_data, moov_len, 0, NULL, null_read_packet, NULL, NULL) != 0)
1526 ctx.buf_end = ctx.buffer + moov_len;
1527 atom.type = MKTAG( 'm', 'o', 'o', 'v' );
1529 atom.size = moov_len;
1531 { int fd = open("/tmp/uncompheader.mov", O_WRONLY | O_CREAT); write(fd, moov_data, moov_len); close(fd); }
1533 ret = mov_read_default(c, &ctx, atom);
1541 /* edit list atom */
1542 static int mov_read_elst(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1545 print_atom("elst", atom);
1547 get_byte(pb); /* version */
1548 get_byte(pb); get_byte(pb); get_byte(pb); /* flags */
1549 edit_count= c->streams[c->fc->nb_streams-1]->edit_count = get_be32(pb); /* entries */
1551 for(i=0; i<edit_count; i++){
1552 get_be32(pb); /* Track duration */
1553 get_be32(pb); /* Media time */
1554 get_be32(pb); /* Media rate */
1557 av_log(NULL, AV_LOG_DEBUG, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, c->streams[c->fc->nb_streams-1]->edit_count);
1562 static const MOVParseTableEntry mov_default_parse_table[] = {
1564 { MKTAG( 'c', 'o', '6', '4' ), mov_read_stco },
1565 { MKTAG( 'c', 'p', 'r', 't' ), mov_read_default },
1566 { MKTAG( 'c', 'r', 'h', 'd' ), mov_read_default },
1567 { MKTAG( 'c', 't', 't', 's' ), mov_read_ctts }, /* composition time to sample */
1568 { MKTAG( 'd', 'i', 'n', 'f' ), mov_read_default }, /* data information */
1569 { MKTAG( 'd', 'p', 'n', 'd' ), mov_read_leaf },
1570 { MKTAG( 'd', 'r', 'e', 'f' ), mov_read_leaf },
1571 { MKTAG( 'e', 'd', 't', 's' ), mov_read_default },
1572 { MKTAG( 'e', 'l', 's', 't' ), mov_read_elst },
1573 { MKTAG( 'f', 'r', 'e', 'e' ), mov_read_leaf },
1574 { MKTAG( 'h', 'd', 'l', 'r' ), mov_read_hdlr },
1575 { MKTAG( 'h', 'i', 'n', 't' ), mov_read_leaf },
1576 { MKTAG( 'h', 'm', 'h', 'd' ), mov_read_leaf },
1577 { MKTAG( 'i', 'o', 'd', 's' ), mov_read_leaf },
1578 { MKTAG( 'm', 'd', 'a', 't' ), mov_read_mdat },
1579 { MKTAG( 'm', 'd', 'h', 'd' ), mov_read_mdhd },
1580 { MKTAG( 'm', 'd', 'i', 'a' ), mov_read_default },
1581 { MKTAG( 'm', 'i', 'n', 'f' ), mov_read_default },
1582 { MKTAG( 'm', 'o', 'o', 'v' ), mov_read_moov },
1583 { MKTAG( 'm', 'p', '4', 'a' ), mov_read_default },
1584 { MKTAG( 'm', 'p', '4', 's' ), mov_read_default },
1585 { MKTAG( 'm', 'p', '4', 'v' ), mov_read_default },
1586 { MKTAG( 'm', 'p', 'o', 'd' ), mov_read_leaf },
1587 { MKTAG( 'm', 'v', 'h', 'd' ), mov_read_mvhd },
1588 { MKTAG( 'n', 'm', 'h', 'd' ), mov_read_leaf },
1589 { MKTAG( 'o', 'd', 'h', 'd' ), mov_read_default },
1590 { MKTAG( 's', 'd', 'h', 'd' ), mov_read_default },
1591 { MKTAG( 's', 'k', 'i', 'p' ), mov_read_leaf },
1592 { MKTAG( 's', 'm', 'h', 'd' ), mov_read_leaf }, /* sound media info header */
1593 { MKTAG( 'S', 'M', 'I', ' ' ), mov_read_smi }, /* Sorenson extension ??? */
1594 { MKTAG( 'a', 'v', 'c', 'C' ), mov_read_avcC },
1595 { MKTAG( 's', 't', 'b', 'l' ), mov_read_default },
1596 { MKTAG( 's', 't', 'c', 'o' ), mov_read_stco },
1597 { MKTAG( 's', 't', 'd', 'p' ), mov_read_default },
1598 { MKTAG( 's', 't', 's', 'c' ), mov_read_stsc },
1599 { MKTAG( 's', 't', 's', 'd' ), mov_read_stsd }, /* sample description */
1600 { MKTAG( 's', 't', 's', 'h' ), mov_read_default },
1601 { MKTAG( 's', 't', 's', 's' ), mov_read_stss }, /* sync sample */
1602 { MKTAG( 's', 't', 's', 'z' ), mov_read_stsz }, /* sample size */
1603 { MKTAG( 's', 't', 't', 's' ), mov_read_stts },
1604 { MKTAG( 't', 'k', 'h', 'd' ), mov_read_tkhd }, /* track header */
1605 { MKTAG( 't', 'r', 'a', 'k' ), mov_read_trak },
1606 { MKTAG( 't', 'r', 'e', 'f' ), mov_read_default }, /* not really */
1607 { MKTAG( 'u', 'd', 't', 'a' ), mov_read_leaf },
1608 { MKTAG( 'u', 'r', 'l', ' ' ), mov_read_leaf },
1609 { MKTAG( 'u', 'r', 'n', ' ' ), mov_read_leaf },
1610 { MKTAG( 'u', 'u', 'i', 'd' ), mov_read_leaf },
1611 { MKTAG( 'v', 'm', 'h', 'd' ), mov_read_leaf }, /* video media info header */
1612 { MKTAG( 'w', 'a', 'v', 'e' ), mov_read_default },
1614 { MKTAG( 'M', 'D', 'E', 'S' ), mov_read_leaf },
1616 { MKTAG( 'c', 'h', 'a', 'p' ), mov_read_leaf },
1617 { MKTAG( 'c', 'l', 'i', 'p' ), mov_read_default },
1618 { MKTAG( 'c', 'r', 'g', 'n' ), mov_read_leaf },
1619 { MKTAG( 'c', 't', 'a', 'b' ), mov_read_ctab },
1620 { MKTAG( 'e', 's', 'd', 's' ), mov_read_esds },
1621 { MKTAG( 'k', 'm', 'a', 't' ), mov_read_leaf },
1622 { MKTAG( 'm', 'a', 't', 't' ), mov_read_default },
1623 { MKTAG( 'r', 'd', 'r', 'f' ), mov_read_leaf },
1624 { MKTAG( 'r', 'm', 'd', 'a' ), mov_read_default },
1625 { MKTAG( 'r', 'm', 'd', 'r' ), mov_read_leaf },
1626 { MKTAG( 'r', 'm', 'r', 'a' ), mov_read_default },
1627 { MKTAG( 's', 'c', 'p', 't' ), mov_read_leaf },
1628 { MKTAG( 's', 's', 'r', 'c' ), mov_read_leaf },
1629 { MKTAG( 's', 'y', 'n', 'c' ), mov_read_leaf },
1630 { MKTAG( 't', 'c', 'm', 'd' ), mov_read_leaf },
1631 { MKTAG( 'w', 'i', 'd', 'e' ), mov_read_wide }, /* place holder */
1632 //{ MKTAG( 'r', 'm', 'q', 'u' ), mov_read_leaf },
1634 { MKTAG( 'c', 'm', 'o', 'v' ), mov_read_cmov },
1636 { MKTAG( 'c', 'm', 'o', 'v' ), mov_read_leaf },
1638 { 0L, mov_read_leaf }
1641 static void mov_free_stream_context(MOVStreamContext *sc)
1644 av_freep(&sc->chunk_offsets);
1645 av_freep(&sc->sample_to_chunk);
1646 av_freep(&sc->sample_sizes);
1647 av_freep(&sc->keyframes);
1648 av_freep(&sc->stts_data);
1649 av_freep(&sc->ctts_data);
1654 static inline uint32_t mov_to_tag(uint8_t *buf)
1656 return MKTAG(buf[0], buf[1], buf[2], buf[3]);
1659 static inline uint32_t to_be32(uint8_t *buf)
1661 return (buf[0] << 24) | (buf[1] << 16) | (buf[2] << 8) | buf[3];
1664 /* XXX: is it sufficient ? */
1665 static int mov_probe(AVProbeData *p)
1667 unsigned int offset;
1671 /* check file header */
1672 if (p->buf_size <= 12)
1676 /* ignore invalid offset */
1677 if ((offset + 8) > (unsigned int)p->buf_size)
1679 tag = mov_to_tag(p->buf + offset + 4);
1681 /* check for obvious tags */
1682 case MKTAG( 'm', 'o', 'o', 'v' ):
1683 case MKTAG( 'm', 'd', 'a', 't' ):
1684 case MKTAG( 'p', 'n', 'o', 't' ): /* detect movs with preview pics like ew.mov and april.mov */
1685 case MKTAG( 'u', 'd', 't', 'a' ): /* Packet Video PVAuthor adds this and a lot of more junk */
1686 return AVPROBE_SCORE_MAX;
1687 /* those are more common words, so rate then a bit less */
1688 case MKTAG( 'w', 'i', 'd', 'e' ):
1689 case MKTAG( 'f', 'r', 'e', 'e' ):
1690 case MKTAG( 'j', 'u', 'n', 'k' ):
1691 case MKTAG( 'p', 'i', 'c', 't' ):
1692 return AVPROBE_SCORE_MAX - 5;
1693 case MKTAG( 'f', 't', 'y', 'p' ):
1694 case MKTAG( 's', 'k', 'i', 'p' ):
1695 case MKTAG( 'u', 'u', 'i', 'd' ):
1696 offset = to_be32(p->buf+offset) + offset;
1697 /* if we only find those cause probedata is too small at least rate them */
1698 score = AVPROBE_SCORE_MAX - 50;
1701 /* unrecognized tag */
1708 static int mov_read_header(AVFormatContext *s, AVFormatParameters *ap)
1710 MOVContext *mov = (MOVContext *) s->priv_data;
1711 ByteIOContext *pb = &s->pb;
1713 MOV_atom_t atom = { 0, 0, 0 };
1716 mov->parse_table = mov_default_parse_table;
1718 /* XXX: I think we should auto detect */
1719 if(s->iformat->name[1] == 'p')
1722 if(!url_is_streamed(pb)) /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
1723 atom.size = url_fsize(pb);
1725 atom.size = 0x7FFFFFFFFFFFFFFFLL;
1728 av_log(NULL, AV_LOG_DEBUG, "filesz=%Ld\n", atom.size);
1731 /* check MOV header */
1732 err = mov_read_default(mov, pb, atom);
1733 if (err<0 || (!mov->found_moov && !mov->found_mdat)) {
1734 av_log(s, AV_LOG_ERROR, "mov: header not found !!! (err:%d, moov:%d, mdat:%d) pos:%lld\n",
1735 err, mov->found_moov, mov->found_mdat, url_ftell(pb));
1739 av_log(NULL, AV_LOG_DEBUG, "on_parse_exit_offset=%d\n", (int) url_ftell(pb));
1741 /* some cleanup : make sure we are on the mdat atom */
1742 if(!url_is_streamed(pb) && (url_ftell(pb) != mov->mdat_offset))
1743 url_fseek(pb, mov->mdat_offset, SEEK_SET);
1745 mov->next_chunk_offset = mov->mdat_offset; /* initialise reading */
1748 av_log(NULL, AV_LOG_DEBUG, "mdat_reset_offset=%d\n", (int) url_ftell(pb));
1752 av_log(NULL, AV_LOG_DEBUG, "streams= %d\n", s->nb_streams);
1754 mov->total_streams = nb = s->nb_streams;
1757 for(i=0; i<s->nb_streams;) {
1758 if(s->streams[i]->codec.codec_type == CODEC_TYPE_MOV_OTHER) {/* not audio, not video, delete */
1759 av_free(s->streams[i]);
1760 for(j=i+1; j<s->nb_streams; j++)
1761 s->streams[j-1] = s->streams[j];
1766 for(i=0; i<s->nb_streams;i++) {
1767 MOVStreamContext *sc;
1768 sc = (MOVStreamContext *)s->streams[i]->priv_data;
1770 sc->is_ff_stream = 1;
1774 av_log(NULL, AV_LOG_DEBUG, "real streams= %d\n", s->nb_streams);
1779 /* Yes, this is ugly... I didn't write the specs of QT :p */
1780 /* XXX:remove useless commented code sometime */
1781 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
1783 MOVContext *mov = (MOVContext *) s->priv_data;
1784 MOVStreamContext *sc;
1785 int64_t offset = INT64_MAX;
1786 int64_t best_dts = INT64_MAX;
1792 #ifdef MOV_SPLIT_CHUNKS
1795 idx = sc->sample_to_chunk_index;
1797 if (idx < 0) return 0;
1799 fprintf(stderr, "sc[ffid %d]->sample_size = %d\n", sc->ffindex, sc->sample_size);
1801 //size = sc->sample_sizes[sc->current_sample];
1802 // that ain't working...
1803 //size = (sc->sample_size)?sc->sample_size:sc->sample_sizes[sc->current_sample];
1804 size = (sc->sample_size > 1)?sc->sample_size:sc->sample_sizes[sc->current_sample];
1806 sc->current_sample++;
1807 sc->left_in_chunk--;
1809 if (sc->left_in_chunk <= 0)
1811 offset = mov->next_chunk_offset;
1812 /* extract the sample */
1820 if(offset == INT64_MAX)
1821 best_dts= INT64_MAX;
1822 for(i=0; i<mov->total_streams; i++) {
1823 MOVStreamContext *msc = mov->streams[i];
1825 if ((msc->next_chunk < msc->chunk_count) && msc->next_chunk >= 0){
1826 if (msc->sample_to_time_index < msc->stts_count && mov->ni) {
1828 int index= msc->sample_to_time_index;
1829 int sample= msc->sample_to_time_sample;
1830 int time= msc->sample_to_time_time;
1831 int duration = msc->stts_data[index].duration;
1832 int count = msc->stts_data[index].count;
1833 if (sample + count < msc->current_sample) {
1835 time += count*duration;
1837 duration = msc->stts_data[index].duration;
1839 dts = time + (msc->current_sample-1 - sample) * (int64_t)duration;
1840 dts = av_rescale(dts, AV_TIME_BASE, msc->time_scale);
1841 // av_log(NULL, AV_LOG_DEBUG, "%d %Ld %Ld %Ld \n", i, dts, best_dts, offset);
1845 offset = msc->chunk_offsets[msc->next_chunk];
1848 //av_log(NULL, AV_LOG_DEBUG, "MOCHUNK %ld %d %p pos:%Ld\n", mov->streams[i]->next_chunk, mov->total_streams, mov->streams[i], url_ftell(&s->pb));
1849 if ((msc->chunk_offsets[msc->next_chunk] < offset)) {
1851 offset = msc->chunk_offsets[msc->next_chunk];
1852 //av_log(NULL, AV_LOG_DEBUG, "SELETED %Ld i:%d\n", offset, i);
1857 if (!sc || offset==INT64_MAX)
1862 if(mov->next_chunk_offset < offset) { /* some meta data */
1863 url_fskip(&s->pb, (offset - mov->next_chunk_offset));
1864 mov->next_chunk_offset = offset;
1867 //av_log(NULL, AV_LOG_DEBUG, "chunk: [%i] %lli -> %lli\n", st_id, mov->next_chunk_offset, offset);
1868 if(!sc->is_ff_stream || (s->streams[sc->ffindex]->discard >= AVDISCARD_ALL)) {
1869 url_fskip(&s->pb, (offset - mov->next_chunk_offset));
1870 mov->next_chunk_offset = offset;
1875 /* now get the chunk size... */
1877 for(i=0; i<mov->total_streams; i++) {
1878 MOVStreamContext *msc = mov->streams[i];
1879 if ((msc->next_chunk < msc->chunk_count)
1880 && msc->chunk_offsets[msc->next_chunk] - offset < size
1881 && msc->chunk_offsets[msc->next_chunk] > offset)
1882 size = msc->chunk_offsets[msc->next_chunk] - offset;
1885 #ifdef MOV_MINOLTA_FIX
1886 //Make sure that size is according to sample_size (Needed by .mov files
1887 //created on a Minolta Dimage Xi where audio chunks contains waste data in the end)
1888 //Maybe we should really not only check sc->sample_size, but also sc->sample_sizes
1889 //but I have no such movies
1890 if (sc->sample_size > 0) {
1892 for(i=0; i<(sc->sample_to_chunk_sz); i++) {
1893 if( (sc->sample_to_chunk[i].first)<=(sc->next_chunk) )
1895 // I can't figure out why for PCM audio sample_size is always 1
1896 // (it should actually be channels*bits_per_second/8) but it is.
1897 AVCodecContext* cod = &s->streams[sc->ffindex]->codec;
1898 if (sc->sample_size == 1 && (cod->codec_id == CODEC_ID_PCM_S16BE || cod->codec_id == CODEC_ID_PCM_S16LE))
1899 foundsize=(sc->sample_to_chunk[i].count*cod->channels*cod->bits_per_sample)/8;
1901 foundsize=sc->sample_to_chunk[i].count*sc->sample_size;
1904 av_log(NULL, AV_LOG_DEBUG, "sample_to_chunk first=%ld count=%ld, id=%ld\n", sc->sample_to_chunk[i].first, sc->sample_to_chunk[i].count, sc->sample_to_chunk[i].id);
1907 if( (foundsize>0) && (foundsize<size) )
1910 /*av_log(NULL, AV_LOG_DEBUG, "this size should actually be %d\n",foundsize);*/
1915 #endif //MOV_MINOLTA_FIX
1917 idx = sc->sample_to_chunk_index;
1918 if (idx + 1 < sc->sample_to_chunk_sz && sc->next_chunk >= sc->sample_to_chunk[idx + 1].first)
1920 sc->sample_to_chunk_index = idx;
1921 #ifdef MOV_SPLIT_CHUNKS
1922 /* split chunks into samples */
1923 if (sc->sample_size == 0 || sc->sample_size > 100) {
1924 if (idx >= 0 && sc->sample_to_chunk[idx].count != 1) {
1926 /* we'll have to get those samples before next chunk */
1927 sc->left_in_chunk = sc->sample_to_chunk[idx].count - 1;
1928 size = (sc->sample_size > 1)?sc->sample_size:sc->sample_sizes[sc->current_sample];
1931 sc->current_sample++;
1932 }else if(idx + 1 < sc->sample_to_chunk_sz){
1933 sc->current_sample += sc->sample_size * sc->sample_to_chunk[idx].count;
1939 av_log(NULL, AV_LOG_DEBUG, "chunk: %lli -> %lli (%i)\n", offset, offset + size, size);
1941 if(size == 0x0FFFFFFF)
1942 size = mov->mdat_size + mov->mdat_offset - offset;
1947 url_fseek(&s->pb, offset, SEEK_SET);
1949 av_get_packet(&s->pb, pkt, size);
1950 pkt->stream_index = sc->ffindex;
1952 // If the keyframes table exists, mark any samples that are in the table as key frames.
1953 // If no table exists, treat very sample as a key frame.
1954 if (sc->keyframes) {
1956 b = sc->keyframe_count - 1;
1959 m = (a + b + 1) >> 1;
1960 if (sc->keyframes[m] > sc->current_sample) {
1967 if (sc->keyframes[a] == sc->current_sample)
1968 pkt->flags |= PKT_FLAG_KEY;
1971 pkt->flags |= PKT_FLAG_KEY;
1975 av_log(NULL, AV_LOG_DEBUG, "Packet (%d, %ld) ", pkt->stream_index, pkt->size);
1977 av_log(NULL, AV_LOG_DEBUG, "%02x ", pkt->data[i]);
1979 av_log(NULL, AV_LOG_DEBUG, "%c ", (pkt->data[i]) & 0x7F);
1980 av_log(NULL, AV_LOG_DEBUG, "\n");
1984 mov->next_chunk_offset = offset + size;
1986 /* find the corresponding dts */
1987 if (sc && sc->sample_to_time_index < sc->stts_count && pkt) {
1990 unsigned int duration = sc->stts_data[sc->sample_to_time_index].duration;
1991 count = sc->stts_data[sc->sample_to_time_index].count;
1992 if ((sc->sample_to_time_sample + count) < sc->current_sample) {
1993 sc->sample_to_time_sample += count;
1994 sc->sample_to_time_time += count*duration;
1995 sc->sample_to_time_index ++;
1996 duration = sc->stts_data[sc->sample_to_time_index].duration;
1998 dts = sc->sample_to_time_time + (sc->current_sample-1 - sc->sample_to_time_sample) * (int64_t)duration;
1999 /* find the corresponding pts */
2000 if (sc->sample_to_ctime_index < sc->ctts_count) {
2001 int duration = sc->ctts_data[sc->sample_to_ctime_index].duration;
2002 int count = sc->ctts_data[sc->sample_to_ctime_index].count;
2004 if ((sc->sample_to_ctime_sample + count) < sc->current_sample) {
2005 sc->sample_to_ctime_sample += count;
2006 sc->sample_to_ctime_index ++;
2007 duration = sc->ctts_data[sc->sample_to_ctime_index].duration;
2009 pts = dts + duration;
2015 av_log(NULL, AV_LOG_DEBUG, "stream #%d smp #%ld dts = %lld pts = %lld (smp:%ld time:%lld idx:%d ent:%d count:%d dur:%d)\n"
2016 , pkt->stream_index, sc->current_sample-1, pkt->dts, pkt->pts
2017 , sc->sample_to_time_sample
2018 , sc->sample_to_time_time
2019 , sc->sample_to_time_index
2029 #if defined(MOV_SPLIT_CHUNKS) && defined(MOV_SEEK)
2031 * Seek method based on the one described in the Appendix C of QTFileFormat.pdf
2033 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
2035 MOVContext* mov = (MOVContext *) s->priv_data;
2036 MOVStreamContext* sc;
2039 int32_t seek_sample, sample;
2043 int32_t left_in_chunk;
2044 int64_t chunk_file_offset;
2045 int64_t sample_file_offset;
2046 int32_t first_chunk_sample;
2047 int32_t sample_to_chunk_idx;
2048 int sample_to_time_index;
2049 long sample_to_time_sample = 0;
2050 uint64_t sample_to_time_time = 0;
2053 // Find the corresponding mov stream
2054 for (mov_idx = 0; mov_idx < mov->total_streams; mov_idx++)
2055 if (mov->streams[mov_idx]->ffindex == stream_index)
2057 if (mov_idx == mov->total_streams) {
2058 av_log(s, AV_LOG_ERROR, "mov: requested stream was not found in mov streams (idx=%i)\n", stream_index);
2061 sc = mov->streams[mov_idx];
2063 // Step 1. Find the edit that contains the requested time (elst)
2064 if (sc->edit_count) {
2065 // FIXME should handle edit list
2066 av_log(s, AV_LOG_ERROR, "mov: does not handle seeking in files that contain edit list (c:%d)\n", sc->edit_count);
2070 // Step 2. Find the corresponding sample using the Time-to-sample atom (stts) */
2072 av_log(s, AV_LOG_DEBUG, "Searching for time %li in stream #%i (time_scale=%i)\n", (long)timestamp, mov_idx, sc->time_scale);
2074 start_time = 0; // FIXME use elst atom
2075 sample = 1; // sample are 0 based in table
2077 av_log(s, AV_LOG_DEBUG, "Searching for sample_time %li \n", (long)sample_time);
2079 for (i = 0; i < sc->stts_count; i++) {
2080 count = sc->stts_data[i].count;
2081 duration = sc->stts_data[i].duration;
2082 //av_log(s, AV_LOG_DEBUG, "> sample_time %lli \n", (long)sample_time);
2083 //av_log(s, AV_LOG_DEBUG, "> count=%i duration=%i\n", count, duration);
2084 if ((start_time + count*duration) > sample_time) {
2085 sample_to_time_time = start_time;
2086 sample_to_time_index = i;
2087 sample_to_time_sample = sample;
2088 sample += (sample_time - start_time) / duration;
2092 start_time += count * duration;
2094 sample_to_time_time = start_time;
2095 sample_to_time_index = i;
2096 /* NOTE: despite what qt doc say, the dt value (Display Time in qt vocabulary) computed with the stts atom
2097 is a decoding time stamp (dts) not a presentation time stamp. And as usual dts != pts for stream with b frames */
2100 av_log(s, AV_LOG_DEBUG, "Found time %li at sample #%u\n", (long)sample_time, sample);
2102 if (sample > sc->sample_count) {
2103 av_log(s, AV_LOG_ERROR, "mov: sample pos is too high, unable to seek (req. sample=%i, sample count=%ld)\n", sample, sc->sample_count);
2107 // Step 3. Find the prior sync. sample using the Sync sample atom (stss)
2108 if (sc->keyframes) {
2110 b = sc->keyframe_count - 1;
2112 m = (a + b + 1) >> 1;
2113 if (sc->keyframes[m] > sample) {
2119 // av_log(s, AV_LOG_DEBUG, "a=%i (%i) b=%i (%i) m=%i (%i) stream #%i\n", a, sc->keyframes[a], b, sc->keyframes[b], m, sc->keyframes[m], mov_idx);
2122 // for low latency prob: always use the previous keyframe, just uncomment the next line
2124 seek_sample = sc->keyframes[a];
2127 seek_sample = sample; // else all samples are key frames
2129 av_log(s, AV_LOG_DEBUG, "Found nearest keyframe at sample #%i \n", seek_sample);
2132 // Step 4. Find the chunk of the sample using the Sample-to-chunk-atom (stsc)
2133 for (first_chunk_sample = 1, i = 0; i < (sc->sample_to_chunk_sz - 1); i++) {
2134 b = (sc->sample_to_chunk[i + 1].first - sc->sample_to_chunk[i].first) * sc->sample_to_chunk[i].count;
2135 if (seek_sample >= first_chunk_sample && seek_sample < (first_chunk_sample + b))
2137 first_chunk_sample += b;
2139 chunk = sc->sample_to_chunk[i].first + (seek_sample - first_chunk_sample) / sc->sample_to_chunk[i].count;
2140 left_in_chunk = sc->sample_to_chunk[i].count - (seek_sample - first_chunk_sample) % sc->sample_to_chunk[i].count;
2141 first_chunk_sample += ((seek_sample - first_chunk_sample) / sc->sample_to_chunk[i].count) * sc->sample_to_chunk[i].count;
2142 sample_to_chunk_idx = i;
2144 av_log(s, AV_LOG_DEBUG, "Sample was found in chunk #%i at sample offset %i (idx %i)\n", chunk, seek_sample - first_chunk_sample, sample_to_chunk_idx);
2147 // Step 5. Find the offset of the chunk using the chunk offset atom
2148 if (!sc->chunk_offsets) {
2149 av_log(s, AV_LOG_ERROR, "mov: no chunk offset atom, unable to seek\n");
2152 if (chunk > sc->chunk_count) {
2153 av_log(s, AV_LOG_ERROR, "mov: chunk offset atom too short, unable to seek (req. chunk=%i, chunk count=%li)\n", chunk, sc->chunk_count);
2156 chunk_file_offset = sc->chunk_offsets[chunk - 1];
2158 av_log(s, AV_LOG_DEBUG, "Chunk file offset is #%llu \n", chunk_file_offset);
2161 // Step 6. Find the byte offset within the chunk using the sample size atom
2162 sample_file_offset = chunk_file_offset;
2163 if (sc->sample_size)
2164 sample_file_offset += (seek_sample - first_chunk_sample) * sc->sample_size;
2166 for (i = 0; i < (seek_sample - first_chunk_sample); i++) {
2167 sample_file_offset += sc->sample_sizes[first_chunk_sample + i - 1];
2171 av_log(s, AV_LOG_DEBUG, "Sample file offset is #%llu \n", sample_file_offset);
2174 // Step 6. Update the parser
2176 mov->next_chunk_offset = sample_file_offset;
2177 // Update current stream state
2178 sc->current_sample = seek_sample - 1; // zero based
2179 sc->left_in_chunk = left_in_chunk;
2180 sc->next_chunk = chunk; // +1 -1 (zero based)
2181 sc->sample_to_chunk_index = sample_to_chunk_idx;
2183 // Update other streams
2184 for (i = 0; i<mov->total_streams; i++) {
2185 MOVStreamContext *msc;
2186 if (i == mov_idx) continue;
2187 // Find the nearest 'next' chunk
2188 msc = mov->streams[i];
2190 b = msc->chunk_count - 1;
2192 m = (a + b + 1) >> 1;
2193 if (msc->chunk_offsets[m] > chunk_file_offset) {
2199 /* av_log(s, AV_LOG_DEBUG, "a=%i (%li) b=%i (%li) m=%i (%li) stream #%i\n"
2200 , a, (long)msc->chunk_offsets[a], b, (long)msc->chunk_offsets[b], m, (long)msc->chunk_offsets[m], i); */
2203 msc->next_chunk = a;
2204 if (msc->chunk_offsets[a] < chunk_file_offset && a < (msc->chunk_count-1))
2207 av_log(s, AV_LOG_DEBUG, "Nearest next chunk for stream #%i is #%i @%lli\n", i, msc->next_chunk+1, msc->chunk_offsets[msc->next_chunk]);
2209 // Compute sample count and index in the sample_to_chunk table (what a pity)
2210 msc->sample_to_chunk_index = 0;
2211 msc->current_sample = 0;
2212 for(; msc->sample_to_chunk_index < (msc->sample_to_chunk_sz - 1)
2213 && msc->sample_to_chunk[msc->sample_to_chunk_index + 1].first <= (1 + msc->next_chunk); msc->sample_to_chunk_index++) {
2214 msc->current_sample += (msc->sample_to_chunk[msc->sample_to_chunk_index + 1].first - msc->sample_to_chunk[msc->sample_to_chunk_index].first) \
2215 * msc->sample_to_chunk[msc->sample_to_chunk_index].count;
2217 msc->current_sample += (msc->next_chunk - (msc->sample_to_chunk[msc->sample_to_chunk_index].first - 1)) * sc->sample_to_chunk[msc->sample_to_chunk_index].count;
2218 msc->left_in_chunk = msc->sample_to_chunk[msc->sample_to_chunk_index].count - 1;
2219 // Find corresponding position in stts (used later to compute dts)
2222 for (msc->sample_to_time_index = 0; msc->sample_to_time_index < msc->stts_count; msc->sample_to_time_index++) {
2223 count = msc->stts_data[msc->sample_to_time_index].count;
2224 duration = msc->stts_data[msc->sample_to_time_index].duration;
2225 if ((sample + count - 1) > msc->current_sample) {
2226 msc->sample_to_time_time = start_time;
2227 msc->sample_to_time_sample = sample;
2231 start_time += count * duration;
2234 for (msc->sample_to_ctime_index = 0; msc->sample_to_ctime_index < msc->ctts_count; msc->sample_to_ctime_index++) {
2235 count = msc->ctts_data[msc->sample_to_ctime_index].count;
2236 duration = msc->ctts_data[msc->sample_to_ctime_index].duration;
2237 if ((sample + count - 1) > msc->current_sample) {
2238 msc->sample_to_ctime_sample = sample;
2244 av_log(s, AV_LOG_DEBUG, "Next Sample for stream #%i is #%i @%i\n", i, msc->current_sample + 1, msc->sample_to_chunk_index + 1);
2251 static int mov_read_close(AVFormatContext *s)
2254 MOVContext *mov = (MOVContext *) s->priv_data;
2255 for(i=0; i<mov->total_streams; i++)
2256 mov_free_stream_context(mov->streams[i]);
2257 /* free color tabs */
2258 for(i=0; i<mov->ctab_size; i++)
2259 av_freep(&mov->ctab[i]);
2260 av_freep(&mov->ctab);
2264 static AVInputFormat mov_iformat = {
2265 "mov,mp4,m4a,3gp,3g2",
2266 "QuickTime/MPEG4 format",
2272 #if defined(MOV_SPLIT_CHUNKS) && defined(MOV_SEEK)
2279 av_register_input_format(&mov_iformat);