3 * Copyright (c) 2001 Fabrice Bellard.
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
36 * First version by Francois Revol revol@free.fr
37 * Seek function by Gael Chardon gael.dev@4now.net
39 * Features and limitations:
40 * - reads most of the QT files I have (at least the structure),
41 * Sample QuickTime files with mp3 audio can be found at: http://www.3ivx.com/showcase.html
42 * - the code is quite ugly... maybe I won't do it recursive next time :-)
44 * Funny I didn't know about http://sourceforge.net/projects/qt-ffmpeg/
45 * when coding this :) (it's a writer anyway)
47 * Reference documents:
48 * http://www.geocities.com/xhelmboyx/quicktime/formats/qtm-layout.txt
50 * http://developer.apple.com/documentation/QuickTime/QTFF/
51 * http://developer.apple.com/documentation/QuickTime/QTFF/qtff.pdf
52 * QuickTime is a trademark of Apple (AFAIK :))
55 #include "qtpalette.h"
61 /* the QuickTime file format is quite convoluted...
62 * it has lots of index tables, each indexing something in another one...
63 * Here we just use what is needed to read the chunks
80 int64_t size; /* total size (excluding the size and type fields) */
83 struct MOVParseTableEntry;
87 uint64_t base_data_offset;
103 typedef struct MOVStreamContext {
105 int ffindex; /* the ffmpeg stream id */
107 unsigned int chunk_count;
108 int64_t *chunk_offsets;
109 unsigned int stts_count;
110 MOV_stts_t *stts_data;
111 unsigned int ctts_count;
112 MOV_stts_t *ctts_data;
113 unsigned int edit_count; /* number of 'edit' (elst atom) */
114 unsigned int sample_to_chunk_sz;
115 MOV_stsc_t *sample_to_chunk;
116 int sample_to_ctime_index;
117 int sample_to_ctime_sample;
118 unsigned int sample_size;
119 unsigned int sample_count;
121 unsigned int keyframe_count;
126 unsigned int bytes_per_frame;
127 unsigned int samples_per_frame;
128 int dv_audio_container;
129 int pseudo_stream_id;
130 int16_t audio_cid; ///< stsd audio compression id
131 unsigned drefs_count;
136 typedef struct MOVContext {
139 int64_t duration; /* duration of the longest track */
140 int found_moov; /* when both 'moov' and 'mdat' sections has been found */
141 int found_mdat; /* we suppose we have enough data to read the file */
142 AVPaletteControl palette_control;
143 DVDemuxContext *dv_demux;
144 AVFormatContext *dv_fctx;
145 int isom; /* 1 if file is ISO Media (mp4/3gp) */
146 MOVFragment fragment; ///< current fragment in moof atom
147 MOVTrackExt *trex_data;
152 /* XXX: it's the first time I make a recursive parser I think... sorry if it's ugly :P */
154 /* those functions parse an atom */
156 1: found what I wanted, exit
157 0: continue to parse next atom
158 <0: error occurred, exit
160 /* links atom IDs to parse functions */
161 typedef struct MOVParseTableEntry {
163 int (*parse)(MOVContext *ctx, ByteIOContext *pb, MOV_atom_t atom);
164 } MOVParseTableEntry;
166 static const MOVParseTableEntry mov_default_parse_table[];
168 static int mov_read_default(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
170 int64_t total_size = 0;
175 a.offset = atom.offset;
178 atom.size = INT64_MAX;
179 while(((total_size + 8) < atom.size) && !url_feof(pb) && !err) {
183 a.size = get_be32(pb);
184 a.type = get_le32(pb);
188 dprintf(c->fc, "type: %08x %.4s sz: %"PRIx64" %"PRIx64" %"PRIx64"\n",
189 a.type, (char*)&a.type, a.size, atom.size, total_size);
190 if (a.size == 1) { /* 64 bit extended size */
191 a.size = get_be64(pb) - 8;
196 a.size = atom.size - total_size;
203 a.size = FFMIN(a.size, atom.size - total_size);
205 for (i = 0; mov_default_parse_table[i].type != 0
206 && mov_default_parse_table[i].type != a.type; i++)
209 if (mov_default_parse_table[i].type == 0) { /* skip leaf atoms data */
210 url_fskip(pb, a.size);
212 offset_t start_pos = url_ftell(pb);
214 err = mov_default_parse_table[i].parse(c, pb, a);
215 if (url_is_streamed(pb) && c->found_moov && c->found_mdat)
217 left = a.size - url_ftell(pb) + start_pos;
218 if (left > 0) /* skip garbage at atom end */
223 total_size += a.size;
226 if (!err && total_size < atom.size && atom.size < 0x7ffff)
227 url_fskip(pb, atom.size - total_size);
232 static int mov_read_dref(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
234 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
235 MOVStreamContext *sc = st->priv_data;
238 get_be32(pb); // version + flags
239 entries = get_be32(pb);
240 if (entries >= UINT_MAX / sizeof(*sc->drefs))
242 sc->drefs_count = entries;
243 sc->drefs = av_mallocz(entries * sizeof(*sc->drefs));
245 for (i = 0; i < sc->drefs_count; i++) {
246 MOV_dref_t *dref = &sc->drefs[i];
247 uint32_t size = get_be32(pb);
248 offset_t next = url_ftell(pb) + size - 4;
250 dref->type = get_le32(pb);
251 get_be32(pb); // version + flags
252 dprintf(c->fc, "type %.4s size %d\n", (char*)&dref->type, size);
254 if (dref->type == MKTAG('a','l','i','s') && size > 150) {
255 /* macintosh alias record */
256 uint16_t volume_len, len;
262 volume_len = get_byte(pb);
263 volume_len = FFMIN(volume_len, 27);
264 get_buffer(pb, volume, 27);
265 volume[volume_len] = 0;
266 av_log(c->fc, AV_LOG_DEBUG, "volume %s, len %d\n", volume, volume_len);
270 for (type = 0; type != -1 && url_ftell(pb) < next; ) {
273 av_log(c->fc, AV_LOG_DEBUG, "type %d, len %d\n", type, len);
276 if (type == 2) { // absolute path
278 dref->path = av_mallocz(len+1);
280 return AVERROR(ENOMEM);
281 get_buffer(pb, dref->path, len);
282 if (len > volume_len && !strncmp(dref->path, volume, volume_len)) {
284 memmove(dref->path, dref->path+volume_len, len);
287 for (j = 0; j < len; j++)
288 if (dref->path[j] == ':')
290 av_log(c->fc, AV_LOG_DEBUG, "path %s\n", dref->path);
295 url_fseek(pb, next, SEEK_SET);
300 static int mov_read_hdlr(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
302 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
306 get_byte(pb); /* version */
307 get_be24(pb); /* flags */
310 ctype = get_le32(pb);
311 type = get_le32(pb); /* component subtype */
313 dprintf(c->fc, "ctype= %c%c%c%c (0x%08x)\n", *((char *)&ctype), ((char *)&ctype)[1],
314 ((char *)&ctype)[2], ((char *)&ctype)[3], (int) ctype);
315 dprintf(c->fc, "stype= %c%c%c%c\n",
316 *((char *)&type), ((char *)&type)[1], ((char *)&type)[2], ((char *)&type)[3]);
319 if(type == MKTAG('v', 'i', 'd', 'e'))
320 st->codec->codec_type = CODEC_TYPE_VIDEO;
321 else if(type == MKTAG('s', 'o', 'u', 'n'))
322 st->codec->codec_type = CODEC_TYPE_AUDIO;
323 else if(type == MKTAG('m', '1', 'a', ' '))
324 st->codec->codec_id = CODEC_ID_MP2;
325 else if(type == MKTAG('s', 'u', 'b', 'p')) {
326 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
328 get_be32(pb); /* component manufacture */
329 get_be32(pb); /* component flags */
330 get_be32(pb); /* component flags mask */
333 return 0; /* nothing left to read */
335 url_fskip(pb, atom.size - (url_ftell(pb) - atom.offset));
339 static int mp4_read_descr_len(ByteIOContext *pb)
344 int c = get_byte(pb);
345 len = (len << 7) | (c & 0x7f);
352 static int mp4_read_descr(MOVContext *c, ByteIOContext *pb, int *tag)
356 len = mp4_read_descr_len(pb);
357 dprintf(c->fc, "MPEG4 description: tag=0x%02x len=%d\n", *tag, len);
361 #define MP4ESDescrTag 0x03
362 #define MP4DecConfigDescrTag 0x04
363 #define MP4DecSpecificDescrTag 0x05
365 static int mov_read_esds(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
367 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
370 get_be32(pb); /* version + flags */
371 len = mp4_read_descr(c, pb, &tag);
372 if (tag == MP4ESDescrTag) {
373 get_be16(pb); /* ID */
374 get_byte(pb); /* priority */
376 get_be16(pb); /* ID */
378 len = mp4_read_descr(c, pb, &tag);
379 if (tag == MP4DecConfigDescrTag) {
380 int object_type_id = get_byte(pb);
381 get_byte(pb); /* stream type */
382 get_be24(pb); /* buffer size db */
383 get_be32(pb); /* max bitrate */
384 get_be32(pb); /* avg bitrate */
386 st->codec->codec_id= codec_get_id(ff_mp4_obj_type, object_type_id);
387 dprintf(c->fc, "esds object type id %d\n", object_type_id);
388 len = mp4_read_descr(c, pb, &tag);
389 if (tag == MP4DecSpecificDescrTag) {
390 dprintf(c->fc, "Specific MPEG4 header len=%d\n", len);
391 if((uint64_t)len > (1<<30))
393 st->codec->extradata = av_mallocz(len + FF_INPUT_BUFFER_PADDING_SIZE);
394 if (!st->codec->extradata)
395 return AVERROR(ENOMEM);
396 get_buffer(pb, st->codec->extradata, len);
397 st->codec->extradata_size = len;
399 if ((*st->codec->extradata >> 3) == 29) {
400 st->codec->codec_id = CODEC_ID_MP3ON4;
407 /* this atom contains actual media data */
408 static int mov_read_mdat(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
410 if(atom.size == 0) /* wrong one (MP4) */
413 return 0; /* now go for moov */
416 static int mov_read_ftyp(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
418 uint32_t type = get_le32(pb);
420 if (type != MKTAG('q','t',' ',' '))
422 av_log(c->fc, AV_LOG_DEBUG, "ISO: File Type Major Brand: %.4s\n",(char *)&type);
423 get_be32(pb); /* minor version */
424 url_fskip(pb, atom.size - 8);
428 /* this atom should contain all header atoms */
429 static int mov_read_moov(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
431 if (mov_read_default(c, pb, atom) < 0)
433 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
434 /* so we don't parse the whole file if over a network */
436 return 0; /* now go for mdat */
439 static int mov_read_moof(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
441 c->fragment.moof_offset = url_ftell(pb) - 8;
442 dprintf(c->fc, "moof offset %llx\n", c->fragment.moof_offset);
443 return mov_read_default(c, pb, atom);
446 static int mov_read_mdhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
448 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
449 MOVStreamContext *sc = st->priv_data;
450 int version = get_byte(pb);
454 return 1; /* unsupported */
456 get_be24(pb); /* flags */
461 get_be32(pb); /* creation time */
462 get_be32(pb); /* modification time */
465 sc->time_scale = get_be32(pb);
466 st->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
468 lang = get_be16(pb); /* language */
469 ff_mov_lang_to_iso639(lang, st->language);
470 get_be16(pb); /* quality */
475 static int mov_read_mvhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
477 int version = get_byte(pb); /* version */
478 get_be24(pb); /* flags */
484 get_be32(pb); /* creation time */
485 get_be32(pb); /* modification time */
487 c->time_scale = get_be32(pb); /* time scale */
489 dprintf(c->fc, "time scale = %i\n", c->time_scale);
491 c->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
492 get_be32(pb); /* preferred scale */
494 get_be16(pb); /* preferred volume */
496 url_fskip(pb, 10); /* reserved */
498 url_fskip(pb, 36); /* display matrix */
500 get_be32(pb); /* preview time */
501 get_be32(pb); /* preview duration */
502 get_be32(pb); /* poster time */
503 get_be32(pb); /* selection time */
504 get_be32(pb); /* selection duration */
505 get_be32(pb); /* current time */
506 get_be32(pb); /* next track ID */
511 static int mov_read_smi(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
513 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
515 if((uint64_t)atom.size > (1<<30))
518 // currently SVQ3 decoder expect full STSD header - so let's fake it
519 // this should be fixed and just SMI header should be passed
520 av_free(st->codec->extradata);
521 st->codec->extradata = av_mallocz(atom.size + 0x5a + FF_INPUT_BUFFER_PADDING_SIZE);
522 if (!st->codec->extradata)
523 return AVERROR(ENOMEM);
524 st->codec->extradata_size = 0x5a + atom.size;
525 memcpy(st->codec->extradata, "SVQ3", 4); // fake
526 get_buffer(pb, st->codec->extradata + 0x5a, atom.size);
527 dprintf(c->fc, "Reading SMI %"PRId64" %s\n", atom.size, st->codec->extradata + 0x5a);
531 static int mov_read_enda(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
533 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
534 int little_endian = get_be16(pb);
537 switch (st->codec->codec_id) {
538 case CODEC_ID_PCM_S24BE:
539 st->codec->codec_id = CODEC_ID_PCM_S24LE;
541 case CODEC_ID_PCM_S32BE:
542 st->codec->codec_id = CODEC_ID_PCM_S32LE;
551 /* FIXME modify qdm2/svq3/h264 decoders to take full atom as extradata */
552 static int mov_read_extradata(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
554 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
555 uint64_t size= (uint64_t)st->codec->extradata_size + atom.size + 8 + FF_INPUT_BUFFER_PADDING_SIZE;
557 if(size > INT_MAX || (uint64_t)atom.size > INT_MAX)
559 buf= av_realloc(st->codec->extradata, size);
562 st->codec->extradata= buf;
563 buf+= st->codec->extradata_size;
564 st->codec->extradata_size= size - FF_INPUT_BUFFER_PADDING_SIZE;
565 AV_WB32( buf , atom.size + 8);
566 AV_WL32( buf + 4, atom.type);
567 get_buffer(pb, buf + 8, atom.size);
571 static int mov_read_wave(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
573 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
575 if((uint64_t)atom.size > (1<<30))
578 if (st->codec->codec_id == CODEC_ID_QDM2) {
579 // pass all frma atom to codec, needed at least for QDM2
580 av_free(st->codec->extradata);
581 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
582 if (!st->codec->extradata)
583 return AVERROR(ENOMEM);
584 st->codec->extradata_size = atom.size;
585 get_buffer(pb, st->codec->extradata, atom.size);
586 } else if (atom.size > 8) { /* to read frma, esds atoms */
587 if (mov_read_default(c, pb, atom) < 0)
590 url_fskip(pb, atom.size);
595 * This function reads atom content and puts data in extradata without tag
596 * nor size unlike mov_read_extradata.
598 static int mov_read_glbl(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
600 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
602 if((uint64_t)atom.size > (1<<30))
605 av_free(st->codec->extradata);
606 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
607 if (!st->codec->extradata)
608 return AVERROR(ENOMEM);
609 st->codec->extradata_size = atom.size;
610 get_buffer(pb, st->codec->extradata, atom.size);
614 static int mov_read_stco(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
616 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
617 MOVStreamContext *sc = st->priv_data;
618 unsigned int i, entries;
620 get_byte(pb); /* version */
621 get_be24(pb); /* flags */
623 entries = get_be32(pb);
625 if(entries >= UINT_MAX/sizeof(int64_t))
628 sc->chunk_count = entries;
629 sc->chunk_offsets = av_malloc(entries * sizeof(int64_t));
630 if (!sc->chunk_offsets)
632 if (atom.type == MKTAG('s', 't', 'c', 'o')) {
633 for(i=0; i<entries; i++) {
634 sc->chunk_offsets[i] = get_be32(pb);
636 } else if (atom.type == MKTAG('c', 'o', '6', '4')) {
637 for(i=0; i<entries; i++) {
638 sc->chunk_offsets[i] = get_be64(pb);
646 static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
648 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
649 MOVStreamContext *sc = st->priv_data;
650 int entries, frames_per_sample;
652 uint8_t codec_name[32];
654 /* for palette traversal */
655 unsigned int color_depth;
656 unsigned int color_start;
657 unsigned int color_count;
658 unsigned int color_end;
662 const uint8_t *color_table;
663 int j, pseudo_stream_id;
664 unsigned char r, g, b;
666 get_byte(pb); /* version */
667 get_be24(pb); /* flags */
669 entries = get_be32(pb);
671 for(pseudo_stream_id=0; pseudo_stream_id<entries; pseudo_stream_id++) {
672 //Parsing Sample description table
675 MOV_atom_t a = { 0, 0, 0 };
676 offset_t start_pos = url_ftell(pb);
677 int size = get_be32(pb); /* size */
678 format = get_le32(pb); /* data format */
680 get_be32(pb); /* reserved */
681 get_be16(pb); /* reserved */
682 dref_id = get_be16(pb);
684 if (st->codec->codec_tag &&
685 (c->fc->video_codec_id ? codec_get_id(codec_movvideo_tags, format) != c->fc->video_codec_id
686 : st->codec->codec_tag != MKTAG('j', 'p', 'e', 'g'))
688 /* Multiple fourcc, we skip JPEG. This is not correct, we should
689 * export it as a separate AVStream but this needs a few changes
690 * in the MOV demuxer, patch welcome. */
691 url_fskip(pb, size - (url_ftell(pb) - start_pos));
694 sc->pseudo_stream_id= pseudo_stream_id;
695 sc->dref_id= dref_id;
697 st->codec->codec_tag = format;
698 id = codec_get_id(codec_movaudio_tags, format);
699 if (id<=0 && (format&0xFFFF) == 'm' + ('s'<<8))
700 id = codec_get_id(codec_wav_tags, bswap_32(format)&0xFFFF);
702 if (st->codec->codec_type != CODEC_TYPE_VIDEO && id > 0) {
703 st->codec->codec_type = CODEC_TYPE_AUDIO;
704 } else if (st->codec->codec_type != CODEC_TYPE_AUDIO && /* do not overwrite codec type */
705 format && format != MKTAG('m', 'p', '4', 's')) { /* skip old asf mpeg4 tag */
706 id = codec_get_id(codec_movvideo_tags, format);
708 id = codec_get_id(codec_bmp_tags, format);
710 st->codec->codec_type = CODEC_TYPE_VIDEO;
711 else if(st->codec->codec_type == CODEC_TYPE_DATA){
712 id = codec_get_id(ff_codec_movsubtitle_tags, format);
714 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
718 dprintf(c->fc, "size=%d 4CC= %c%c%c%c codec_type=%d\n", size,
719 (format >> 0) & 0xff, (format >> 8) & 0xff, (format >> 16) & 0xff,
720 (format >> 24) & 0xff, st->codec->codec_type);
722 if(st->codec->codec_type==CODEC_TYPE_VIDEO) {
723 st->codec->codec_id = id;
724 get_be16(pb); /* version */
725 get_be16(pb); /* revision level */
726 get_be32(pb); /* vendor */
727 get_be32(pb); /* temporal quality */
728 get_be32(pb); /* spatial quality */
730 st->codec->width = get_be16(pb); /* width */
731 st->codec->height = get_be16(pb); /* height */
733 get_be32(pb); /* horiz resolution */
734 get_be32(pb); /* vert resolution */
735 get_be32(pb); /* data size, always 0 */
736 frames_per_sample = get_be16(pb); /* frames per samples */
738 dprintf(c->fc, "frames/samples = %d\n", frames_per_sample);
740 get_buffer(pb, codec_name, 32); /* codec name, pascal string (FIXME: true for mp4?) */
741 if (codec_name[0] <= 31) {
742 memcpy(st->codec->codec_name, &codec_name[1],codec_name[0]);
743 st->codec->codec_name[codec_name[0]] = 0;
746 st->codec->bits_per_sample = get_be16(pb); /* depth */
747 st->codec->color_table_id = get_be16(pb); /* colortable id */
748 dprintf(c->fc, "depth %d, ctab id %d\n",
749 st->codec->bits_per_sample, st->codec->color_table_id);
750 /* figure out the palette situation */
751 color_depth = st->codec->bits_per_sample & 0x1F;
752 color_greyscale = st->codec->bits_per_sample & 0x20;
754 /* if the depth is 2, 4, or 8 bpp, file is palettized */
755 if ((color_depth == 2) || (color_depth == 4) ||
756 (color_depth == 8)) {
757 if (color_greyscale) {
758 /* compute the greyscale palette */
759 st->codec->bits_per_sample = color_depth;
760 color_count = 1 << color_depth;
762 color_dec = 256 / (color_count - 1);
763 for (j = 0; j < color_count; j++) {
764 r = g = b = color_index;
765 c->palette_control.palette[j] =
766 (r << 16) | (g << 8) | (b);
767 color_index -= color_dec;
771 } else if (st->codec->color_table_id) {
772 /* if flag bit 3 is set, use the default palette */
773 color_count = 1 << color_depth;
774 if (color_depth == 2)
775 color_table = ff_qt_default_palette_4;
776 else if (color_depth == 4)
777 color_table = ff_qt_default_palette_16;
779 color_table = ff_qt_default_palette_256;
781 for (j = 0; j < color_count; j++) {
782 r = color_table[j * 4 + 0];
783 g = color_table[j * 4 + 1];
784 b = color_table[j * 4 + 2];
785 c->palette_control.palette[j] =
786 (r << 16) | (g << 8) | (b);
789 /* load the palette from the file */
790 color_start = get_be32(pb);
791 color_count = get_be16(pb);
792 color_end = get_be16(pb);
793 if ((color_start <= 255) &&
794 (color_end <= 255)) {
795 for (j = color_start; j <= color_end; j++) {
796 /* each R, G, or B component is 16 bits;
797 * only use the top 8 bits; skip alpha bytes
807 c->palette_control.palette[j] =
808 (r << 16) | (g << 8) | (b);
812 st->codec->palctrl = &c->palette_control;
813 st->codec->palctrl->palette_changed = 1;
815 st->codec->palctrl = NULL;
816 } else if(st->codec->codec_type==CODEC_TYPE_AUDIO) {
818 uint16_t version = get_be16(pb);
820 st->codec->codec_id = id;
821 get_be16(pb); /* revision level */
822 get_be32(pb); /* vendor */
824 st->codec->channels = get_be16(pb); /* channel count */
825 dprintf(c->fc, "audio channels %d\n", st->codec->channels);
826 st->codec->bits_per_sample = get_be16(pb); /* sample size */
828 sc->audio_cid = get_be16(pb);
829 get_be16(pb); /* packet size = 0 */
831 st->codec->sample_rate = ((get_be32(pb) >> 16));
833 switch (st->codec->codec_id) {
834 case CODEC_ID_PCM_S8:
835 case CODEC_ID_PCM_U8:
836 if (st->codec->bits_per_sample == 16)
837 st->codec->codec_id = CODEC_ID_PCM_S16BE;
839 case CODEC_ID_PCM_S16LE:
840 case CODEC_ID_PCM_S16BE:
841 if (st->codec->bits_per_sample == 8)
842 st->codec->codec_id = CODEC_ID_PCM_S8;
843 else if (st->codec->bits_per_sample == 24)
844 st->codec->codec_id = CODEC_ID_PCM_S24BE;
846 /* set values for old format before stsd version 1 appeared */
848 sc->samples_per_frame = 6;
849 sc->bytes_per_frame = 2*st->codec->channels;
852 sc->samples_per_frame = 6;
853 sc->bytes_per_frame = 1*st->codec->channels;
855 case CODEC_ID_ADPCM_IMA_QT:
856 sc->samples_per_frame = 64;
857 sc->bytes_per_frame = 34*st->codec->channels;
863 //Read QT version 1 fields. In version 0 these do not exist.
864 dprintf(c->fc, "version =%d, isom =%d\n",version,c->isom);
867 sc->samples_per_frame = get_be32(pb);
868 get_be32(pb); /* bytes per packet */
869 sc->bytes_per_frame = get_be32(pb);
870 get_be32(pb); /* bytes per sample */
871 } else if(version==2) {
872 get_be32(pb); /* sizeof struct only */
873 st->codec->sample_rate = av_int2dbl(get_be64(pb)); /* float 64 */
874 st->codec->channels = get_be32(pb);
875 get_be32(pb); /* always 0x7F000000 */
876 get_be32(pb); /* bits per channel if sound is uncompressed */
877 get_be32(pb); /* lcpm format specific flag */
878 get_be32(pb); /* bytes per audio packet if constant */
879 get_be32(pb); /* lpcm frames per audio packet if constant */
883 bits_per_sample = av_get_bits_per_sample(st->codec->codec_id);
884 if (bits_per_sample) {
885 st->codec->bits_per_sample = bits_per_sample;
886 sc->sample_size = (bits_per_sample >> 3) * st->codec->channels;
888 } else if(st->codec->codec_type==CODEC_TYPE_SUBTITLE){
889 st->codec->codec_id= id;
891 /* other codec type, just skip (rtp, mp4s, tmcd ...) */
892 url_fskip(pb, size - (url_ftell(pb) - start_pos));
894 /* this will read extra atoms at the end (wave, alac, damr, avcC, SMI ...) */
895 a.size = size - (url_ftell(pb) - start_pos);
897 if (mov_read_default(c, pb, a) < 0)
899 } else if (a.size > 0)
900 url_fskip(pb, a.size);
903 if(st->codec->codec_type==CODEC_TYPE_AUDIO && st->codec->sample_rate==0 && sc->time_scale>1)
904 st->codec->sample_rate= sc->time_scale;
906 /* special codec parameters handling */
907 switch (st->codec->codec_id) {
908 #ifdef CONFIG_DV_DEMUXER
909 case CODEC_ID_DVAUDIO:
910 c->dv_fctx = av_alloc_format_context();
911 c->dv_demux = dv_init_demux(c->dv_fctx);
913 av_log(c->fc, AV_LOG_ERROR, "dv demux context init error\n");
916 sc->dv_audio_container = 1;
917 st->codec->codec_id = CODEC_ID_PCM_S16LE;
920 /* no ifdef since parameters are always those */
921 case CODEC_ID_AMR_WB:
922 st->codec->sample_rate= 16000;
923 st->codec->channels= 1; /* really needed */
925 case CODEC_ID_AMR_NB:
926 st->codec->sample_rate= 8000;
927 st->codec->channels= 1; /* really needed */
931 st->codec->codec_type = CODEC_TYPE_AUDIO; /* force type after stsd for m1a hdlr */
932 st->need_parsing = AVSTREAM_PARSE_FULL;
935 case CODEC_ID_ADPCM_MS:
936 case CODEC_ID_ADPCM_IMA_WAV:
937 st->codec->block_align = sc->bytes_per_frame;
940 if (st->codec->extradata_size == 36)
941 st->codec->frame_size = AV_RB32((st->codec->extradata+12));
950 static int mov_read_stsc(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
952 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
953 MOVStreamContext *sc = st->priv_data;
954 unsigned int i, entries;
956 get_byte(pb); /* version */
957 get_be24(pb); /* flags */
959 entries = get_be32(pb);
961 if(entries >= UINT_MAX / sizeof(MOV_stsc_t))
964 dprintf(c->fc, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
966 sc->sample_to_chunk_sz = entries;
967 sc->sample_to_chunk = av_malloc(entries * sizeof(MOV_stsc_t));
968 if (!sc->sample_to_chunk)
970 for(i=0; i<entries; i++) {
971 sc->sample_to_chunk[i].first = get_be32(pb);
972 sc->sample_to_chunk[i].count = get_be32(pb);
973 sc->sample_to_chunk[i].id = get_be32(pb);
978 static int mov_read_stss(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
980 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
981 MOVStreamContext *sc = st->priv_data;
982 unsigned int i, entries;
984 get_byte(pb); /* version */
985 get_be24(pb); /* flags */
987 entries = get_be32(pb);
989 if(entries >= UINT_MAX / sizeof(int))
992 sc->keyframe_count = entries;
994 dprintf(c->fc, "keyframe_count = %d\n", sc->keyframe_count);
996 sc->keyframes = av_malloc(entries * sizeof(int));
999 for(i=0; i<entries; i++) {
1000 sc->keyframes[i] = get_be32(pb);
1001 //dprintf(c->fc, "keyframes[]=%d\n", sc->keyframes[i]);
1006 static int mov_read_stsz(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1008 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1009 MOVStreamContext *sc = st->priv_data;
1010 unsigned int i, entries, sample_size;
1012 get_byte(pb); /* version */
1013 get_be24(pb); /* flags */
1015 sample_size = get_be32(pb);
1016 if (!sc->sample_size) /* do not overwrite value computed in stsd */
1017 sc->sample_size = sample_size;
1018 entries = get_be32(pb);
1019 if(entries >= UINT_MAX / sizeof(int))
1022 sc->sample_count = entries;
1026 dprintf(c->fc, "sample_size = %d sample_count = %d\n", sc->sample_size, sc->sample_count);
1028 sc->sample_sizes = av_malloc(entries * sizeof(int));
1029 if (!sc->sample_sizes)
1031 for(i=0; i<entries; i++)
1032 sc->sample_sizes[i] = get_be32(pb);
1036 static int mov_read_stts(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1038 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1039 MOVStreamContext *sc = st->priv_data;
1040 unsigned int i, entries;
1042 int64_t total_sample_count=0;
1044 get_byte(pb); /* version */
1045 get_be24(pb); /* flags */
1046 entries = get_be32(pb);
1047 if(entries >= UINT_MAX / sizeof(MOV_stts_t))
1050 sc->stts_count = entries;
1051 sc->stts_data = av_malloc(entries * sizeof(MOV_stts_t));
1054 dprintf(c->fc, "track[%i].stts.entries = %i\n", c->fc->nb_streams-1, entries);
1058 for(i=0; i<entries; i++) {
1059 int sample_duration;
1062 sample_count=get_be32(pb);
1063 sample_duration = get_be32(pb);
1064 sc->stts_data[i].count= sample_count;
1065 sc->stts_data[i].duration= sample_duration;
1067 sc->time_rate= ff_gcd(sc->time_rate, sample_duration);
1069 dprintf(c->fc, "sample_count=%d, sample_duration=%d\n",sample_count,sample_duration);
1071 duration+=(int64_t)sample_duration*sample_count;
1072 total_sample_count+=sample_count;
1075 st->nb_frames= total_sample_count;
1077 st->duration= duration;
1081 static int mov_read_ctts(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1083 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1084 MOVStreamContext *sc = st->priv_data;
1085 unsigned int i, entries;
1087 get_byte(pb); /* version */
1088 get_be24(pb); /* flags */
1089 entries = get_be32(pb);
1090 if(entries >= UINT_MAX / sizeof(MOV_stts_t))
1093 sc->ctts_count = entries;
1094 sc->ctts_data = av_malloc(entries * sizeof(MOV_stts_t));
1097 dprintf(c->fc, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
1099 for(i=0; i<entries; i++) {
1100 int count =get_be32(pb);
1101 int duration =get_be32(pb);
1104 av_log(c->fc, AV_LOG_ERROR, "negative ctts, ignoring\n");
1106 url_fskip(pb, 8 * (entries - i - 1));
1109 sc->ctts_data[i].count = count;
1110 sc->ctts_data[i].duration= duration;
1112 sc->time_rate= ff_gcd(sc->time_rate, duration);
1117 static void mov_build_index(MOVContext *mov, AVStream *st)
1119 MOVStreamContext *sc = st->priv_data;
1120 offset_t current_offset;
1121 int64_t current_dts = 0;
1122 unsigned int stts_index = 0;
1123 unsigned int stsc_index = 0;
1124 unsigned int stss_index = 0;
1127 /* only use old uncompressed audio chunk demuxing when stts specifies it */
1128 if (!(st->codec->codec_type == CODEC_TYPE_AUDIO &&
1129 sc->stts_count == 1 && sc->stts_data[0].duration == 1)) {
1130 unsigned int current_sample = 0;
1131 unsigned int stts_sample = 0;
1132 unsigned int keyframe, sample_size;
1133 unsigned int distance = 0;
1134 int key_off = sc->keyframes && sc->keyframes[0] == 1;
1136 st->nb_frames = sc->sample_count;
1137 for (i = 0; i < sc->chunk_count; i++) {
1138 current_offset = sc->chunk_offsets[i];
1139 if (stsc_index + 1 < sc->sample_to_chunk_sz &&
1140 i + 1 == sc->sample_to_chunk[stsc_index + 1].first)
1142 for (j = 0; j < sc->sample_to_chunk[stsc_index].count; j++) {
1143 if (current_sample >= sc->sample_count) {
1144 av_log(mov->fc, AV_LOG_ERROR, "wrong sample count\n");
1147 keyframe = !sc->keyframe_count || current_sample+key_off == sc->keyframes[stss_index];
1150 if (stss_index + 1 < sc->keyframe_count)
1153 sample_size = sc->sample_size > 0 ? sc->sample_size : sc->sample_sizes[current_sample];
1154 dprintf(mov->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1155 "size %d, distance %d, keyframe %d\n", st->index, current_sample,
1156 current_offset, current_dts, sample_size, distance, keyframe);
1157 if(sc->sample_to_chunk[stsc_index].id - 1 == sc->pseudo_stream_id)
1158 av_add_index_entry(st, current_offset, current_dts, sample_size, distance,
1159 keyframe ? AVINDEX_KEYFRAME : 0);
1160 current_offset += sample_size;
1161 assert(sc->stts_data[stts_index].duration % sc->time_rate == 0);
1162 current_dts += sc->stts_data[stts_index].duration / sc->time_rate;
1166 if (stts_index + 1 < sc->stts_count && stts_sample == sc->stts_data[stts_index].count) {
1172 } else { /* read whole chunk */
1173 unsigned int chunk_samples, chunk_size, chunk_duration;
1174 unsigned int frames = 1;
1175 for (i = 0; i < sc->chunk_count; i++) {
1176 current_offset = sc->chunk_offsets[i];
1177 if (stsc_index + 1 < sc->sample_to_chunk_sz &&
1178 i + 1 == sc->sample_to_chunk[stsc_index + 1].first)
1180 chunk_samples = sc->sample_to_chunk[stsc_index].count;
1181 /* get chunk size, beware of alaw/ulaw/mace */
1182 if (sc->samples_per_frame > 0 &&
1183 (chunk_samples * sc->bytes_per_frame % sc->samples_per_frame == 0)) {
1184 if (sc->samples_per_frame < 1024)
1185 chunk_size = chunk_samples * sc->bytes_per_frame / sc->samples_per_frame;
1187 chunk_size = sc->bytes_per_frame;
1188 frames = chunk_samples / sc->samples_per_frame;
1189 chunk_samples = sc->samples_per_frame;
1192 chunk_size = chunk_samples * sc->sample_size;
1193 for (j = 0; j < frames; j++) {
1194 av_add_index_entry(st, current_offset, current_dts, chunk_size, 0, AVINDEX_KEYFRAME);
1195 /* get chunk duration */
1197 while (chunk_samples > 0) {
1198 if (chunk_samples < sc->stts_data[stts_index].count) {
1199 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1200 sc->stts_data[stts_index].count -= chunk_samples;
1203 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1204 chunk_samples -= sc->stts_data[stts_index].count;
1205 if (stts_index + 1 < sc->stts_count)
1209 current_offset += sc->bytes_per_frame;
1210 dprintf(mov->fc, "AVIndex stream %d, chunk %d, offset %"PRIx64", dts %"PRId64", "
1211 "size %d, duration %d\n", st->index, i, current_offset, current_dts,
1212 chunk_size, chunk_duration);
1213 assert(chunk_duration % sc->time_rate == 0);
1214 current_dts += chunk_duration / sc->time_rate;
1219 /* adjust sample count to avindex entries */
1220 sc->sample_count = st->nb_index_entries;
1223 static int mov_read_trak(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1226 MOVStreamContext *sc;
1229 st = av_new_stream(c->fc, c->fc->nb_streams);
1230 if (!st) return AVERROR(ENOMEM);
1231 sc = av_mallocz(sizeof(MOVStreamContext));
1232 if (!sc) return AVERROR(ENOMEM);
1235 st->codec->codec_type = CODEC_TYPE_DATA;
1236 st->start_time = 0; /* XXX: check */
1238 if ((ret = mov_read_default(c, pb, atom)) < 0)
1242 if(sc->chunk_count && (!sc->stts_count || !sc->sample_to_chunk_sz ||
1243 (!sc->sample_size && !sc->sample_count))){
1244 av_log(c->fc, AV_LOG_ERROR, "stream %d, missing mandatory atoms, broken header\n",
1246 sc->sample_count = 0; //ignore track
1252 sc->time_scale= c->time_scale;
1253 av_set_pts_info(st, 64, sc->time_rate, sc->time_scale);
1255 if (st->codec->codec_type == CODEC_TYPE_AUDIO && sc->stts_count == 1)
1256 st->codec->frame_size = av_rescale(sc->time_rate, st->codec->sample_rate, sc->time_scale);
1258 if(st->duration != AV_NOPTS_VALUE){
1259 assert(st->duration % sc->time_rate == 0);
1260 st->duration /= sc->time_rate;
1262 sc->ffindex = st->index;
1263 mov_build_index(c, st);
1265 if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
1266 if (url_fopen(&sc->pb, sc->drefs[sc->dref_id-1].path, URL_RDONLY) < 0)
1267 av_log(c->fc, AV_LOG_ERROR, "stream %d, error opening file %s: %s\n",
1268 st->index, sc->drefs[sc->dref_id-1].path, strerror(errno));
1272 switch (st->codec->codec_id) {
1273 #ifdef CONFIG_H261_DECODER
1276 #ifdef CONFIG_H263_DECODER
1279 #ifdef CONFIG_MPEG4_DECODER
1280 case CODEC_ID_MPEG4:
1282 st->codec->width= 0; /* let decoder init width/height */
1283 st->codec->height= 0;
1285 #ifdef CONFIG_LIBFAAD
1288 #ifdef CONFIG_VORBIS_DECODER
1289 case CODEC_ID_VORBIS:
1291 case CODEC_ID_MP3ON4:
1292 st->codec->sample_rate= 0; /* let decoder init parameters properly */
1296 /* Do not need those anymore. */
1297 av_freep(&sc->chunk_offsets);
1298 av_freep(&sc->sample_to_chunk);
1299 av_freep(&sc->sample_sizes);
1300 av_freep(&sc->keyframes);
1301 av_freep(&sc->stts_data);
1306 static void mov_parse_udta_string(ByteIOContext *pb, char *str, int size)
1308 uint16_t str_size = get_be16(pb); /* string length */;
1310 get_be16(pb); /* skip language */
1311 get_buffer(pb, str, FFMIN(size, str_size));
1314 static int mov_read_udta(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1316 uint64_t end = url_ftell(pb) + atom.size;
1318 while (url_ftell(pb) + 8 < end) {
1319 uint32_t tag_size = get_be32(pb);
1320 uint32_t tag = get_le32(pb);
1321 uint64_t next = url_ftell(pb) + tag_size - 8;
1323 if (next > end) // stop if tag_size is wrong
1327 case MKTAG(0xa9,'n','a','m'):
1328 mov_parse_udta_string(pb, c->fc->title, sizeof(c->fc->title));
1330 case MKTAG(0xa9,'w','r','t'):
1331 mov_parse_udta_string(pb, c->fc->author, sizeof(c->fc->author));
1333 case MKTAG(0xa9,'c','p','y'):
1334 mov_parse_udta_string(pb, c->fc->copyright, sizeof(c->fc->copyright));
1336 case MKTAG(0xa9,'i','n','f'):
1337 mov_parse_udta_string(pb, c->fc->comment, sizeof(c->fc->comment));
1343 url_fseek(pb, next, SEEK_SET);
1349 static int mov_read_tkhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1351 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1352 int version = get_byte(pb);
1354 get_be24(pb); /* flags */
1356 MOV_TRACK_ENABLED 0x0001
1357 MOV_TRACK_IN_MOVIE 0x0002
1358 MOV_TRACK_IN_PREVIEW 0x0004
1359 MOV_TRACK_IN_POSTER 0x0008
1366 get_be32(pb); /* creation time */
1367 get_be32(pb); /* modification time */
1369 st->id = (int)get_be32(pb); /* track id (NOT 0 !)*/
1370 get_be32(pb); /* reserved */
1371 st->start_time = 0; /* check */
1372 /* highlevel (considering edits) duration in movie timebase */
1373 (version == 1) ? get_be64(pb) : get_be32(pb);
1374 get_be32(pb); /* reserved */
1375 get_be32(pb); /* reserved */
1377 get_be16(pb); /* layer */
1378 get_be16(pb); /* alternate group */
1379 get_be16(pb); /* volume */
1380 get_be16(pb); /* reserved */
1382 url_fskip(pb, 36); /* display matrix */
1384 /* those are fixed-point */
1385 get_be32(pb); /* track width */
1386 get_be32(pb); /* track height */
1391 static int mov_read_tfhd(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1393 MOVFragment *frag = &c->fragment;
1394 MOVTrackExt *trex = NULL;
1395 int flags, track_id, i;
1397 get_byte(pb); /* version */
1398 flags = get_be24(pb);
1400 track_id = get_be32(pb);
1401 if (!track_id || track_id > c->fc->nb_streams)
1403 frag->track_id = track_id;
1404 for (i = 0; i < c->trex_count; i++)
1405 if (c->trex_data[i].track_id == frag->track_id) {
1406 trex = &c->trex_data[i];
1410 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding trex\n");
1414 if (flags & 0x01) frag->base_data_offset = get_be64(pb);
1415 else frag->base_data_offset = frag->moof_offset;
1416 if (flags & 0x02) frag->stsd_id = get_be32(pb);
1417 else frag->stsd_id = trex->stsd_id;
1419 frag->duration = flags & 0x08 ? get_be32(pb) : trex->duration;
1420 frag->size = flags & 0x10 ? get_be32(pb) : trex->size;
1421 frag->flags = flags & 0x20 ? get_be32(pb) : trex->flags;
1422 dprintf(c->fc, "frag flags 0x%x\n", frag->flags);
1426 static int mov_read_trex(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1430 if ((uint64_t)c->trex_count+1 >= UINT_MAX / sizeof(*c->trex_data))
1432 c->trex_data = av_realloc(c->trex_data, (c->trex_count+1)*sizeof(*c->trex_data));
1434 return AVERROR(ENOMEM);
1435 trex = &c->trex_data[c->trex_count++];
1436 get_byte(pb); /* version */
1437 get_be24(pb); /* flags */
1438 trex->track_id = get_be32(pb);
1439 trex->stsd_id = get_be32(pb);
1440 trex->duration = get_be32(pb);
1441 trex->size = get_be32(pb);
1442 trex->flags = get_be32(pb);
1446 static int mov_read_trun(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1448 MOVFragment *frag = &c->fragment;
1450 MOVStreamContext *sc;
1453 int data_offset = 0;
1454 unsigned entries, first_sample_flags = frag->flags;
1455 int flags, distance, i;
1457 if (!frag->track_id || frag->track_id > c->fc->nb_streams)
1459 st = c->fc->streams[frag->track_id-1];
1461 if (sc->pseudo_stream_id+1 != frag->stsd_id)
1463 get_byte(pb); /* version */
1464 flags = get_be24(pb);
1465 entries = get_be32(pb);
1466 dprintf(c->fc, "flags 0x%x entries %d\n", flags, entries);
1467 if (flags & 0x001) data_offset = get_be32(pb);
1468 if (flags & 0x004) first_sample_flags = get_be32(pb);
1469 if (flags & 0x800) {
1470 if ((uint64_t)entries+sc->ctts_count >= UINT_MAX/sizeof(*sc->ctts_data))
1472 sc->ctts_data = av_realloc(sc->ctts_data,
1473 (entries+sc->ctts_count)*sizeof(*sc->ctts_data));
1475 return AVERROR(ENOMEM);
1478 offset = frag->base_data_offset + data_offset;
1480 dprintf(c->fc, "first sample flags 0x%x\n", first_sample_flags);
1481 for (i = 0; i < entries; i++) {
1482 unsigned sample_size = frag->size;
1483 int sample_flags = i ? frag->flags : first_sample_flags;
1484 unsigned sample_duration = frag->duration;
1487 if (flags & 0x100) sample_duration = get_be32(pb);
1488 if (flags & 0x200) sample_size = get_be32(pb);
1489 if (flags & 0x400) sample_flags = get_be32(pb);
1490 if (flags & 0x800) {
1491 sc->ctts_data[sc->ctts_count].count = 1;
1492 sc->ctts_data[sc->ctts_count].duration = get_be32(pb);
1495 if ((keyframe = st->codec->codec_type == CODEC_TYPE_AUDIO ||
1496 (flags & 0x004 && !i && !sample_flags) || sample_flags & 0x2000000))
1498 av_add_index_entry(st, offset, dts, sample_size, distance,
1499 keyframe ? AVINDEX_KEYFRAME : 0);
1500 dprintf(c->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1501 "size %d, distance %d, keyframe %d\n", st->index, sc->sample_count+i,
1502 offset, dts, sample_size, distance, keyframe);
1504 assert(sample_duration % sc->time_rate == 0);
1505 dts += sample_duration / sc->time_rate;
1506 offset += sample_size;
1508 frag->moof_offset = offset;
1509 sc->sample_count = st->nb_index_entries;
1514 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
1515 /* like the files created with Adobe Premiere 5.0, for samples see */
1516 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
1517 static int mov_read_wide(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1522 return 0; /* continue */
1523 if (get_be32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
1524 url_fskip(pb, atom.size - 4);
1527 atom.type = get_le32(pb);
1530 if (atom.type != MKTAG('m', 'd', 'a', 't')) {
1531 url_fskip(pb, atom.size);
1534 err = mov_read_mdat(c, pb, atom);
1538 static int mov_read_cmov(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1543 uint8_t *moov_data; /* uncompressed data */
1544 long cmov_len, moov_len;
1547 get_be32(pb); /* dcom atom */
1548 if (get_le32(pb) != MKTAG( 'd', 'c', 'o', 'm' ))
1550 if (get_le32(pb) != MKTAG( 'z', 'l', 'i', 'b' )) {
1551 av_log(NULL, AV_LOG_ERROR, "unknown compression for cmov atom !");
1554 get_be32(pb); /* cmvd atom */
1555 if (get_le32(pb) != MKTAG( 'c', 'm', 'v', 'd' ))
1557 moov_len = get_be32(pb); /* uncompressed size */
1558 cmov_len = atom.size - 6 * 4;
1560 cmov_data = av_malloc(cmov_len);
1563 moov_data = av_malloc(moov_len);
1568 get_buffer(pb, cmov_data, cmov_len);
1569 if(uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
1571 if(init_put_byte(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL, NULL) != 0)
1573 atom.type = MKTAG( 'm', 'o', 'o', 'v' );
1575 atom.size = moov_len;
1577 // { int fd = open("/tmp/uncompheader.mov", O_WRONLY | O_CREAT); write(fd, moov_data, moov_len); close(fd); }
1579 ret = mov_read_default(c, &ctx, atom);
1584 av_log(c->fc, AV_LOG_ERROR, "this file requires zlib support compiled in\n");
1589 /* edit list atom */
1590 static int mov_read_elst(MOVContext *c, ByteIOContext *pb, MOV_atom_t atom)
1592 MOVStreamContext *sc = c->fc->streams[c->fc->nb_streams-1]->priv_data;
1595 get_byte(pb); /* version */
1596 get_be24(pb); /* flags */
1597 edit_count= sc->edit_count = get_be32(pb); /* entries */
1599 for(i=0; i<edit_count; i++){
1601 get_be32(pb); /* Track duration */
1602 time = get_be32(pb); /* Media time */
1603 get_be32(pb); /* Media rate */
1605 av_log(c->fc, AV_LOG_WARNING, "edit list not starting at 0, "
1606 "a/v desync might occur, patch welcome\n");
1608 dprintf(c->fc, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, sc->edit_count);
1612 static const MOVParseTableEntry mov_default_parse_table[] = {
1613 { MKTAG( 'c', 'o', '6', '4' ), mov_read_stco },
1614 { MKTAG( 'c', 't', 't', 's' ), mov_read_ctts }, /* composition time to sample */
1615 { MKTAG( 'd', 'i', 'n', 'f' ), mov_read_default },
1616 { MKTAG( 'd', 'r', 'e', 'f' ), mov_read_dref },
1617 { MKTAG( 'e', 'd', 't', 's' ), mov_read_default },
1618 { MKTAG( 'e', 'l', 's', 't' ), mov_read_elst },
1619 { MKTAG( 'e', 'n', 'd', 'a' ), mov_read_enda },
1620 { MKTAG( 'f', 'i', 'e', 'l' ), mov_read_extradata },
1621 { MKTAG( 'f', 't', 'y', 'p' ), mov_read_ftyp },
1622 { MKTAG( 'g', 'l', 'b', 'l' ), mov_read_glbl },
1623 { MKTAG( 'h', 'd', 'l', 'r' ), mov_read_hdlr },
1624 { MKTAG( 'j', 'p', '2', 'h' ), mov_read_extradata },
1625 { MKTAG( 'm', 'd', 'a', 't' ), mov_read_mdat },
1626 { MKTAG( 'm', 'd', 'h', 'd' ), mov_read_mdhd },
1627 { MKTAG( 'm', 'd', 'i', 'a' ), mov_read_default },
1628 { MKTAG( 'm', 'i', 'n', 'f' ), mov_read_default },
1629 { MKTAG( 'm', 'o', 'o', 'f' ), mov_read_moof },
1630 { MKTAG( 'm', 'o', 'o', 'v' ), mov_read_moov },
1631 { MKTAG( 'm', 'v', 'e', 'x' ), mov_read_default },
1632 { MKTAG( 'm', 'v', 'h', 'd' ), mov_read_mvhd },
1633 { MKTAG( 'S', 'M', 'I', ' ' ), mov_read_smi }, /* Sorenson extension ??? */
1634 { MKTAG( 'a', 'l', 'a', 'c' ), mov_read_extradata }, /* alac specific atom */
1635 { MKTAG( 'a', 'v', 'c', 'C' ), mov_read_glbl },
1636 { MKTAG( 's', 't', 'b', 'l' ), mov_read_default },
1637 { MKTAG( 's', 't', 'c', 'o' ), mov_read_stco },
1638 { MKTAG( 's', 't', 's', 'c' ), mov_read_stsc },
1639 { MKTAG( 's', 't', 's', 'd' ), mov_read_stsd }, /* sample description */
1640 { MKTAG( 's', 't', 's', 's' ), mov_read_stss }, /* sync sample */
1641 { MKTAG( 's', 't', 's', 'z' ), mov_read_stsz }, /* sample size */
1642 { MKTAG( 's', 't', 't', 's' ), mov_read_stts },
1643 { MKTAG( 't', 'k', 'h', 'd' ), mov_read_tkhd }, /* track header */
1644 { MKTAG( 't', 'f', 'h', 'd' ), mov_read_tfhd }, /* track fragment header */
1645 { MKTAG( 't', 'r', 'a', 'k' ), mov_read_trak },
1646 { MKTAG( 't', 'r', 'a', 'f' ), mov_read_default },
1647 { MKTAG( 't', 'r', 'e', 'x' ), mov_read_trex },
1648 { MKTAG( 't', 'r', 'u', 'n' ), mov_read_trun },
1649 { MKTAG( 'u', 'd', 't', 'a' ), mov_read_udta },
1650 { MKTAG( 'w', 'a', 'v', 'e' ), mov_read_wave },
1651 { MKTAG( 'e', 's', 'd', 's' ), mov_read_esds },
1652 { MKTAG( 'w', 'i', 'd', 'e' ), mov_read_wide }, /* place holder */
1653 { MKTAG( 'c', 'm', 'o', 'v' ), mov_read_cmov },
1657 static int mov_probe(AVProbeData *p)
1659 unsigned int offset;
1663 /* check file header */
1666 /* ignore invalid offset */
1667 if ((offset + 8) > (unsigned int)p->buf_size)
1669 tag = AV_RL32(p->buf + offset + 4);
1671 /* check for obvious tags */
1672 case MKTAG( 'j', 'P', ' ', ' ' ): /* jpeg 2000 signature */
1673 case MKTAG( 'm', 'o', 'o', 'v' ):
1674 case MKTAG( 'm', 'd', 'a', 't' ):
1675 case MKTAG( 'p', 'n', 'o', 't' ): /* detect movs with preview pics like ew.mov and april.mov */
1676 case MKTAG( 'u', 'd', 't', 'a' ): /* Packet Video PVAuthor adds this and a lot of more junk */
1677 return AVPROBE_SCORE_MAX;
1678 /* those are more common words, so rate then a bit less */
1679 case MKTAG( 'e', 'd', 'i', 'w' ): /* xdcam files have reverted first tags */
1680 case MKTAG( 'w', 'i', 'd', 'e' ):
1681 case MKTAG( 'f', 'r', 'e', 'e' ):
1682 case MKTAG( 'j', 'u', 'n', 'k' ):
1683 case MKTAG( 'p', 'i', 'c', 't' ):
1684 return AVPROBE_SCORE_MAX - 5;
1685 case MKTAG(0x82,0x82,0x7f,0x7d ):
1686 case MKTAG( 'f', 't', 'y', 'p' ):
1687 case MKTAG( 's', 'k', 'i', 'p' ):
1688 case MKTAG( 'u', 'u', 'i', 'd' ):
1689 case MKTAG( 'p', 'r', 'f', 'l' ):
1690 offset = AV_RB32(p->buf+offset) + offset;
1691 /* if we only find those cause probedata is too small at least rate them */
1692 score = AVPROBE_SCORE_MAX - 50;
1695 /* unrecognized tag */
1702 static int mov_read_header(AVFormatContext *s, AVFormatParameters *ap)
1704 MOVContext *mov = s->priv_data;
1705 ByteIOContext *pb = s->pb;
1707 MOV_atom_t atom = { 0, 0, 0 };
1710 /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
1711 if(!url_is_streamed(pb))
1712 atom.size = url_fsize(pb);
1714 atom.size = INT64_MAX;
1716 /* check MOV header */
1717 if ((err = mov_read_default(mov, pb, atom)) < 0) {
1718 av_log(s, AV_LOG_ERROR, "error reading header: %d\n", err);
1721 if (!mov->found_moov) {
1722 av_log(s, AV_LOG_ERROR, "moov atom not found\n");
1725 dprintf(mov->fc, "on_parse_exit_offset=%lld\n", url_ftell(pb));
1730 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
1732 MOVContext *mov = s->priv_data;
1733 MOVStreamContext *sc = 0;
1734 AVIndexEntry *sample = 0;
1735 int64_t best_dts = INT64_MAX;
1738 for (i = 0; i < s->nb_streams; i++) {
1739 AVStream *st = s->streams[i];
1740 MOVStreamContext *msc = st->priv_data;
1741 if (st->discard != AVDISCARD_ALL && msc->pb && msc->current_sample < msc->sample_count) {
1742 AVIndexEntry *current_sample = &st->index_entries[msc->current_sample];
1743 int64_t dts = av_rescale(current_sample->timestamp * (int64_t)msc->time_rate,
1744 AV_TIME_BASE, msc->time_scale);
1745 dprintf(s, "stream %d, sample %d, dts %"PRId64"\n", i, msc->current_sample, dts);
1746 if (!sample || (url_is_streamed(s->pb) && current_sample->pos < sample->pos) ||
1747 (!url_is_streamed(s->pb) &&
1748 ((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
1749 ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
1750 (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
1751 sample = current_sample;
1758 mov->found_mdat = 0;
1759 if (!url_is_streamed(s->pb) ||
1760 mov_read_default(mov, s->pb, (MOV_atom_t){ 0, 0, INT64_MAX }) < 0 ||
1763 dprintf(s, "read fragments, offset 0x%llx\n", url_ftell(s->pb));
1766 /* must be done just before reading, to avoid infinite loop on sample */
1767 sc->current_sample++;
1768 if (url_fseek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
1769 av_log(mov->fc, AV_LOG_ERROR, "stream %d, offset 0x%"PRIx64": partial file\n",
1770 sc->ffindex, sample->pos);
1773 av_get_packet(sc->pb, pkt, sample->size);
1774 #ifdef CONFIG_DV_DEMUXER
1775 if (mov->dv_demux && sc->dv_audio_container) {
1776 dv_produce_packet(mov->dv_demux, pkt, pkt->data, pkt->size);
1779 if (dv_get_packet(mov->dv_demux, pkt) < 0)
1783 pkt->stream_index = sc->ffindex;
1784 pkt->dts = sample->timestamp;
1785 if (sc->ctts_data) {
1786 assert(sc->ctts_data[sc->sample_to_ctime_index].duration % sc->time_rate == 0);
1787 pkt->pts = pkt->dts + sc->ctts_data[sc->sample_to_ctime_index].duration / sc->time_rate;
1788 /* update ctts context */
1789 sc->sample_to_ctime_sample++;
1790 if (sc->sample_to_ctime_index < sc->ctts_count &&
1791 sc->ctts_data[sc->sample_to_ctime_index].count == sc->sample_to_ctime_sample) {
1792 sc->sample_to_ctime_index++;
1793 sc->sample_to_ctime_sample = 0;
1796 pkt->pts = pkt->dts;
1798 pkt->flags |= sample->flags & AVINDEX_KEYFRAME ? PKT_FLAG_KEY : 0;
1799 pkt->pos = sample->pos;
1800 dprintf(s, "stream %d, pts %"PRId64", dts %"PRId64", pos 0x%"PRIx64", duration %d\n",
1801 pkt->stream_index, pkt->pts, pkt->dts, pkt->pos, pkt->duration);
1805 static int mov_seek_stream(AVStream *st, int64_t timestamp, int flags)
1807 MOVStreamContext *sc = st->priv_data;
1808 int sample, time_sample;
1811 sample = av_index_search_timestamp(st, timestamp, flags);
1812 dprintf(st->codec, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
1813 if (sample < 0) /* not sure what to do */
1815 sc->current_sample = sample;
1816 dprintf(st->codec, "stream %d, found sample %d\n", st->index, sc->current_sample);
1817 /* adjust ctts index */
1818 if (sc->ctts_data) {
1820 for (i = 0; i < sc->ctts_count; i++) {
1821 int next = time_sample + sc->ctts_data[i].count;
1822 if (next > sc->current_sample) {
1823 sc->sample_to_ctime_index = i;
1824 sc->sample_to_ctime_sample = sc->current_sample - time_sample;
1833 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
1836 int64_t seek_timestamp, timestamp;
1840 if (stream_index >= s->nb_streams)
1843 st = s->streams[stream_index];
1844 sample = mov_seek_stream(st, sample_time, flags);
1848 /* adjust seek timestamp to found sample timestamp */
1849 seek_timestamp = st->index_entries[sample].timestamp;
1851 for (i = 0; i < s->nb_streams; i++) {
1853 if (stream_index == i || st->discard == AVDISCARD_ALL)
1856 timestamp = av_rescale_q(seek_timestamp, s->streams[stream_index]->time_base, st->time_base);
1857 mov_seek_stream(st, timestamp, flags);
1862 static int mov_read_close(AVFormatContext *s)
1865 MOVContext *mov = s->priv_data;
1866 for(i=0; i<s->nb_streams; i++) {
1867 MOVStreamContext *sc = s->streams[i]->priv_data;
1868 av_freep(&sc->ctts_data);
1869 for (j=0; j<sc->drefs_count; j++)
1870 av_freep(&sc->drefs[j].path);
1871 av_freep(&sc->drefs);
1872 if (sc->pb && sc->pb != s->pb)
1876 for(i=0; i<mov->dv_fctx->nb_streams; i++){
1877 av_freep(&mov->dv_fctx->streams[i]->codec);
1878 av_freep(&mov->dv_fctx->streams[i]);
1880 av_freep(&mov->dv_fctx);
1881 av_freep(&mov->dv_demux);
1883 av_freep(&mov->trex_data);
1887 AVInputFormat mov_demuxer = {
1888 "mov,mp4,m4a,3gp,3g2,mj2",
1889 "QuickTime/MPEG4/Motion JPEG 2000 format",