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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
29 typedef struct AVIStream {
30 int64_t frame_offset; /* current frame (video) or byte (audio) counter
31 (used to compute the pts) */
37 int sample_size; /* audio only data */
40 int64_t cum_len; /* temporary storage (used during seek) */
42 int prefix; ///< normally 'd'<<8 + 'c' or 'w'<<8 + 'b'
54 DVDemuxContext* dv_demux;
57 static int avi_load_index(AVFormatContext *s);
58 static int guess_ni_flag(AVFormatContext *s);
61 static void print_tag(const char *str, unsigned int tag, int size)
63 printf("%s: tag=%c%c%c%c size=0x%x\n",
72 static int get_riff(AVIContext *avi, ByteIOContext *pb)
75 /* check RIFF header */
78 if (tag != MKTAG('R', 'I', 'F', 'F'))
80 avi->riff_end = get_le32(pb); /* RIFF chunk size */
81 avi->riff_end += url_ftell(pb); /* RIFF chunk end */
83 if (tag != MKTAG('A', 'V', 'I', ' ') && tag != MKTAG('A', 'V', 'I', 'X'))
89 static int read_braindead_odml_indx(AVFormatContext *s, int frame_num){
90 ByteIOContext *pb = &s->pb;
91 int longs_pre_entry= get_le16(pb);
92 int index_sub_type = get_byte(pb);
93 int index_type = get_byte(pb);
94 int entries_in_use = get_le32(pb);
95 int chunk_id = get_le32(pb);
96 int64_t base = get_le64(pb);
97 int stream_id= 10*((chunk_id&0xFF) - '0') + (((chunk_id>>8)&0xFF) - '0');
102 // av_log(s, AV_LOG_ERROR, "longs_pre_entry:%d index_type:%d entries_in_use:%d chunk_id:%X base:%Ld\n",
103 // longs_pre_entry,index_type, entries_in_use, chunk_id, base);
105 if(stream_id > s->nb_streams || stream_id < 0)
107 st= s->streams[stream_id];
115 if(index_type && longs_pre_entry != 2)
120 for(i=0; i<entries_in_use; i++){
122 int64_t pos= get_le32(pb) + base - 8;
123 int len = get_le32(pb);
127 //av_log(s, AV_LOG_ERROR, "pos:%Ld, len:%X\n", pos, len);
128 av_add_index_entry(st, pos, ast->cum_len, len, 0, key ? AVINDEX_KEYFRAME : 0);
131 ast->cum_len += len / ast->sample_size;
135 int64_t offset= get_le64(pb);
136 int size = get_le32(pb);
137 int duration = get_le32(pb);
138 int64_t pos= url_ftell(pb);
140 url_fseek(pb, offset+8, SEEK_SET);
141 read_braindead_odml_indx(s, frame_num);
142 frame_num += duration;
144 url_fseek(pb, pos, SEEK_SET);
150 static int avi_read_header(AVFormatContext *s, AVFormatParameters *ap)
152 AVIContext *avi = s->priv_data;
153 ByteIOContext *pb = &s->pb;
154 uint32_t tag, tag1, handler;
155 int codec_type, stream_index, frame_period, bit_rate;
156 unsigned int size, nb_frames;
160 int xan_video = 0; /* hack to support Xan A/V */
162 avi->stream_index= -1;
164 if (get_riff(avi, pb) < 0)
177 print_tag("tag", tag, size);
181 case MKTAG('L', 'I', 'S', 'T'):
182 /* ignored, except when start of video packets */
185 print_tag("list", tag1, 0);
187 if (tag1 == MKTAG('m', 'o', 'v', 'i')) {
188 avi->movi_list = url_ftell(pb) - 4;
189 if(size) avi->movi_end = avi->movi_list + size;
190 else avi->movi_end = url_fsize(pb);
192 printf("movi end=%Lx\n", avi->movi_end);
197 case MKTAG('d', 'm', 'l', 'h'):
199 url_fskip(pb, size + (size & 1));
201 case MKTAG('a', 'v', 'i', 'h'):
203 /* using frame_period is bad idea */
204 frame_period = get_le32(pb);
205 bit_rate = get_le32(pb) * 8;
207 avi->non_interleaved |= get_le32(pb) & AVIF_MUSTUSEINDEX;
209 url_fskip(pb, 2 * 4);
213 st = av_new_stream(s, i);
217 ast = av_mallocz(sizeof(AVIStream));
222 url_fskip(pb, size - 7 * 4);
224 case MKTAG('s', 't', 'r', 'h'):
228 handler = get_le32(pb); /* codec tag */
230 print_tag("strh", tag1, -1);
232 if(tag1 == MKTAG('i', 'a', 'v', 's') || tag1 == MKTAG('i', 'v', 'a', 's')){
234 * After some consideration -- I don't think we
235 * have to support anything but DV in a type1 AVIs.
237 if (s->nb_streams != 1)
240 if (handler != MKTAG('d', 'v', 's', 'd') &&
241 handler != MKTAG('d', 'v', 'h', 'd') &&
242 handler != MKTAG('d', 'v', 's', 'l'))
245 ast = s->streams[0]->priv_data;
246 av_freep(&s->streams[0]->codec->extradata);
247 av_freep(&s->streams[0]);
249 avi->dv_demux = dv_init_demux(s);
252 s->streams[0]->priv_data = ast;
253 url_fskip(pb, 3 * 4);
254 ast->scale = get_le32(pb);
255 ast->rate = get_le32(pb);
256 stream_index = s->nb_streams - 1;
257 url_fskip(pb, size - 7*4);
261 if (stream_index >= s->nb_streams) {
262 url_fskip(pb, size - 8);
265 st = s->streams[stream_index];
267 st->codec->stream_codec_tag= handler;
269 get_le32(pb); /* flags */
270 get_le16(pb); /* priority */
271 get_le16(pb); /* language */
272 get_le32(pb); /* initial frame */
273 ast->scale = get_le32(pb);
274 ast->rate = get_le32(pb);
275 if(ast->scale && ast->rate){
276 }else if(frame_period){
278 ast->scale = frame_period;
283 av_set_pts_info(st, 64, ast->scale, ast->rate);
285 ast->start= get_le32(pb); /* start */
286 nb_frames = get_le32(pb);
289 st->duration = nb_frames;
290 get_le32(pb); /* buffer size */
291 get_le32(pb); /* quality */
292 ast->sample_size = get_le32(pb); /* sample ssize */
293 // av_log(NULL, AV_LOG_DEBUG, "%d %d %d %d\n", ast->rate, ast->scale, ast->start, ast->sample_size);
296 case MKTAG('v', 'i', 'd', 's'):
297 codec_type = CODEC_TYPE_VIDEO;
299 ast->sample_size = 0;
301 case MKTAG('a', 'u', 'd', 's'):
302 codec_type = CODEC_TYPE_AUDIO;
304 case MKTAG('t', 'x', 't', 's'):
306 codec_type = CODEC_TYPE_DATA; //CODEC_TYPE_SUB ? FIXME
308 case MKTAG('p', 'a', 'd', 's'):
309 codec_type = CODEC_TYPE_UNKNOWN;
313 av_log(s, AV_LOG_ERROR, "unknown stream type %X\n", tag1);
316 url_fskip(pb, size - 12 * 4);
318 case MKTAG('s', 't', 'r', 'f'):
320 if (stream_index >= s->nb_streams || avi->dv_demux) {
323 st = s->streams[stream_index];
325 case CODEC_TYPE_VIDEO:
326 get_le32(pb); /* size */
327 st->codec->width = get_le32(pb);
328 st->codec->height = get_le32(pb);
329 get_le16(pb); /* panes */
330 st->codec->bits_per_sample= get_le16(pb); /* depth */
332 get_le32(pb); /* ImageSize */
333 get_le32(pb); /* XPelsPerMeter */
334 get_le32(pb); /* YPelsPerMeter */
335 get_le32(pb); /* ClrUsed */
336 get_le32(pb); /* ClrImportant */
338 if(size > 10*4 && size<(1<<30)){
339 st->codec->extradata_size= size - 10*4;
340 st->codec->extradata= av_malloc(st->codec->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
341 get_buffer(pb, st->codec->extradata, st->codec->extradata_size);
344 if(st->codec->extradata_size & 1) //FIXME check if the encoder really did this correctly
347 /* Extract palette from extradata if bpp <= 8 */
348 /* This code assumes that extradata contains only palette */
349 /* This is true for all paletted codecs implemented in ffmpeg */
350 if (st->codec->extradata_size && (st->codec->bits_per_sample <= 8)) {
351 st->codec->palctrl = av_mallocz(sizeof(AVPaletteControl));
352 #ifdef WORDS_BIGENDIAN
353 for (i = 0; i < FFMIN(st->codec->extradata_size, AVPALETTE_SIZE)/4; i++)
354 st->codec->palctrl->palette[i] = bswap_32(((uint32_t*)st->codec->extradata)[i]);
356 memcpy(st->codec->palctrl->palette, st->codec->extradata,
357 FFMIN(st->codec->extradata_size, AVPALETTE_SIZE));
359 st->codec->palctrl->palette_changed = 1;
363 print_tag("video", tag1, 0);
365 st->codec->codec_type = CODEC_TYPE_VIDEO;
366 st->codec->codec_tag = tag1;
367 st->codec->codec_id = codec_get_id(codec_bmp_tags, tag1);
368 if (st->codec->codec_id == CODEC_ID_XAN_WC4)
370 st->need_parsing = 2; //only parse headers dont do slower repacketization, this is needed to get the pict type which is needed for generating correct pts
371 // url_fskip(pb, size - 5 * 4);
373 case CODEC_TYPE_AUDIO:
374 get_wav_header(pb, st->codec, size);
375 if (size%2) /* 2-aligned (fix for Stargate SG-1 - 3x18 - Shades of Grey.avi) */
377 /* special case time: To support Xan DPCM, hardcode
378 * the format if Xxan is the video codec */
379 st->need_parsing = 1;
380 /* force parsing as several audio frames can be in
383 st->codec->codec_id = CODEC_ID_XAN_DPCM;
386 st->codec->codec_type = CODEC_TYPE_DATA;
387 st->codec->codec_id= CODEC_ID_NONE;
388 st->codec->codec_tag= 0;
394 case MKTAG('i', 'n', 'd', 'x'):
396 read_braindead_odml_indx(s, 0);
398 url_fseek(pb, i+size, SEEK_SET);
408 /* check stream number */
409 if (stream_index != s->nb_streams - 1) {
411 for(i=0;i<s->nb_streams;i++) {
412 av_freep(&s->streams[i]->codec->extradata);
413 av_freep(&s->streams[i]);
418 if(!avi->index_loaded)
420 avi->index_loaded = 1;
421 avi->non_interleaved |= guess_ni_flag(s);
426 static int avi_read_packet(AVFormatContext *s, AVPacket *pkt)
428 AVIContext *avi = s->priv_data;
429 ByteIOContext *pb = &s->pb;
435 size = dv_get_packet(avi->dv_demux, pkt);
440 if(avi->non_interleaved){
441 int best_stream_index = 0;
442 AVStream *best_st= NULL;
444 int64_t best_ts= INT64_MAX;
447 for(i=0; i<s->nb_streams; i++){
448 AVStream *st = s->streams[i];
449 AVIStream *ast = st->priv_data;
450 int64_t ts= ast->frame_offset;
453 ts /= ast->sample_size;
454 ts= av_rescale(ts, AV_TIME_BASE * (int64_t)st->time_base.num, st->time_base.den);
456 // av_log(NULL, AV_LOG_DEBUG, "%Ld %d/%d %Ld\n", ts, st->time_base.num, st->time_base.den, ast->frame_offset);
460 best_stream_index= i;
463 best_ast = best_st->priv_data;
464 best_ts= av_rescale(best_ts, best_st->time_base.den, AV_TIME_BASE * (int64_t)best_st->time_base.num); //FIXME a little ugly
465 if(best_ast->remaining)
466 i= av_index_search_timestamp(best_st, best_ts, AVSEEK_FLAG_ANY | AVSEEK_FLAG_BACKWARD);
468 i= av_index_search_timestamp(best_st, best_ts, AVSEEK_FLAG_ANY);
470 // av_log(NULL, AV_LOG_DEBUG, "%d\n", i);
472 int64_t pos= best_st->index_entries[i].pos;
473 pos += best_ast->packet_size - best_ast->remaining;
474 url_fseek(&s->pb, pos + 8, SEEK_SET);
475 // av_log(NULL, AV_LOG_DEBUG, "pos=%Ld\n", pos);
477 avi->stream_index= best_stream_index;
478 if(!best_ast->remaining)
479 best_ast->remaining= best_st->index_entries[i].size;
484 if(avi->stream_index >= 0){
485 AVStream *st= s->streams[ avi->stream_index ];
486 AVIStream *ast= st->priv_data;
489 if(ast->sample_size == 0)
491 else if(ast->sample_size < 32)
492 size= 64*ast->sample_size;
494 size= ast->sample_size;
496 if(size > ast->remaining)
497 size= ast->remaining;
498 av_get_packet(pb, pkt, size);
501 dstr = pkt->destruct;
502 size = dv_produce_packet(avi->dv_demux, pkt,
503 pkt->data, pkt->size);
504 pkt->destruct = dstr;
505 pkt->flags |= PKT_FLAG_KEY;
507 /* XXX: how to handle B frames in avi ? */
508 pkt->dts = ast->frame_offset;
509 // pkt->dts += ast->start;
511 pkt->dts /= ast->sample_size;
512 //av_log(NULL, AV_LOG_DEBUG, "dts:%Ld offset:%d %d/%d smpl_siz:%d base:%d st:%d size:%d\n", pkt->dts, ast->frame_offset, ast->scale, ast->rate, ast->sample_size, AV_TIME_BASE, n, size);
513 pkt->stream_index = avi->stream_index;
515 if (st->codec->codec_type == CODEC_TYPE_VIDEO) {
516 if(st->index_entries){
520 index= av_index_search_timestamp(st, pkt->dts, 0);
521 e= &st->index_entries[index];
523 if(index >= 0 && e->timestamp == ast->frame_offset){
524 if (e->flags & AVINDEX_KEYFRAME)
525 pkt->flags |= PKT_FLAG_KEY;
528 /* if no index, better to say that all frames
530 pkt->flags |= PKT_FLAG_KEY;
533 pkt->flags |= PKT_FLAG_KEY;
536 ast->frame_offset += pkt->size;
540 ast->remaining -= size;
542 avi->stream_index= -1;
553 memset(d, -1, sizeof(int)*8);
554 for(i=sync=url_ftell(pb); !url_feof(pb); i++) {
557 if (i >= avi->movi_end) {
559 url_fskip(pb, avi->riff_end - i);
560 avi->riff_end = avi->movi_end = url_fsize(pb);
569 size= d[4] + (d[5]<<8) + (d[6]<<16) + (d[7]<<24);
571 if( d[2] >= '0' && d[2] <= '9'
572 && d[3] >= '0' && d[3] <= '9'){
573 n= (d[2] - '0') * 10 + (d[3] - '0');
575 n= 100; //invalid stream id
577 //av_log(NULL, AV_LOG_DEBUG, "%X %X %X %X %X %X %X %X %lld %d %d\n", d[0], d[1], d[2], d[3], d[4], d[5], d[6], d[7], i, size, n);
578 if(i + size > avi->movi_end || d[0]<0)
582 if( (d[0] == 'i' && d[1] == 'x' && n < s->nb_streams)
584 ||(d[0] == 'J' && d[1] == 'U' && d[2] == 'N' && d[3] == 'K')){
586 //av_log(NULL, AV_LOG_DEBUG, "SKIP\n");
590 if( d[0] >= '0' && d[0] <= '9'
591 && d[1] >= '0' && d[1] <= '9'){
592 n= (d[0] - '0') * 10 + (d[1] - '0');
594 n= 100; //invalid stream id
598 if(n < s->nb_streams){
604 if( (st->discard >= AVDISCARD_DEFAULT && size==0)
605 /*|| (st->discard >= AVDISCARD_NONKEY && !(pkt->flags & PKT_FLAG_KEY))*/ //FIXME needs a little reordering
606 || st->discard >= AVDISCARD_ALL){
607 if(ast->sample_size) ast->frame_offset += pkt->size;
608 else ast->frame_offset++;
613 if( ((ast->prefix_count<5 || sync+9 > i) && d[2]<128 && d[3]<128) ||
614 d[2]*256+d[3] == ast->prefix /*||
615 (d[2] == 'd' && d[3] == 'c') ||
616 (d[2] == 'w' && d[3] == 'b')*/) {
618 //av_log(NULL, AV_LOG_DEBUG, "OK\n");
619 if(d[2]*256+d[3] == ast->prefix)
622 ast->prefix= d[2]*256+d[3];
623 ast->prefix_count= 0;
626 avi->stream_index= n;
627 ast->packet_size= size + 8;
628 ast->remaining= size;
632 /* palette changed chunk */
633 if ( d[0] >= '0' && d[0] <= '9'
634 && d[1] >= '0' && d[1] <= '9'
635 && ((d[2] == 'p' && d[3] == 'c'))
636 && n < s->nb_streams && i + size <= avi->movi_end) {
639 int first, clr, flags, k, p;
643 first = get_byte(pb);
645 if(!clr) /* all 256 colors used */
647 flags = get_le16(pb);
649 for (k = first; k < clr + first; k++) {
655 st->codec->palctrl->palette[k] = b + (g << 8) + (r << 16);
657 st->codec->palctrl->palette_changed = 1;
666 /* XXX: we make the implicit supposition that the position are sorted
668 static int avi_read_idx1(AVFormatContext *s, int size)
670 AVIContext *avi = s->priv_data;
671 ByteIOContext *pb = &s->pb;
672 int nb_index_entries, i;
675 unsigned int index, tag, flags, pos, len;
676 unsigned last_pos= -1;
678 nb_index_entries = size / 16;
679 if (nb_index_entries <= 0)
682 /* read the entries and sort them in each stream component */
683 for(i = 0; i < nb_index_entries; i++) {
685 flags = get_le32(pb);
688 #if defined(DEBUG_SEEK)
689 av_log(NULL, AV_LOG_DEBUG, "%d: tag=0x%x flags=0x%x pos=0x%x len=%d/",
690 i, tag, flags, pos, len);
692 if(i==0 && pos > avi->movi_list)
693 avi->movi_list= 0; //FIXME better check
694 pos += avi->movi_list;
696 index = ((tag & 0xff) - '0') * 10;
697 index += ((tag >> 8) & 0xff) - '0';
698 if (index >= s->nb_streams)
700 st = s->streams[index];
703 #if defined(DEBUG_SEEK)
704 av_log(NULL, AV_LOG_DEBUG, "%d cum_len=%d\n", len, ast->cum_len);
707 avi->non_interleaved= 1;
709 av_add_index_entry(st, pos, ast->cum_len, len, 0, (flags&AVIIF_INDEX) ? AVINDEX_KEYFRAME : 0);
711 ast->cum_len += len / ast->sample_size;
719 static int guess_ni_flag(AVFormatContext *s){
721 int64_t last_start=0;
722 int64_t first_end= INT64_MAX;
724 for(i=0; i<s->nb_streams; i++){
725 AVStream *st = s->streams[i];
726 int n= st->nb_index_entries;
731 if(st->index_entries[0].pos > last_start)
732 last_start= st->index_entries[0].pos;
733 if(st->index_entries[n-1].pos < first_end)
734 first_end= st->index_entries[n-1].pos;
736 return last_start > first_end;
739 static int avi_load_index(AVFormatContext *s)
741 AVIContext *avi = s->priv_data;
742 ByteIOContext *pb = &s->pb;
744 offset_t pos= url_ftell(pb);
746 url_fseek(pb, avi->movi_end, SEEK_SET);
748 printf("movi_end=0x%llx\n", avi->movi_end);
756 printf("tag=%c%c%c%c size=0x%x\n",
764 case MKTAG('i', 'd', 'x', '1'):
765 if (avi_read_idx1(s, size) < 0)
778 url_fseek(pb, pos, SEEK_SET);
782 static int avi_read_seek(AVFormatContext *s, int stream_index, int64_t timestamp, int flags)
784 AVIContext *avi = s->priv_data;
789 if (!avi->index_loaded) {
790 /* we only load the index on demand */
792 avi->index_loaded = 1;
794 assert(stream_index>= 0);
796 st = s->streams[stream_index];
797 index= av_index_search_timestamp(st, timestamp, flags);
801 /* find the position */
802 pos = st->index_entries[index].pos;
803 timestamp = st->index_entries[index].timestamp;
805 // av_log(NULL, AV_LOG_DEBUG, "XX %Ld %d %Ld\n", timestamp, index, st->index_entries[index].timestamp);
807 for(i = 0; i < s->nb_streams; i++) {
808 AVStream *st2 = s->streams[i];
809 AVIStream *ast2 = st2->priv_data;
814 if (st2->nb_index_entries <= 0)
817 // assert(st2->codec.block_align);
818 assert(st2->time_base.den == ast2->rate);
819 assert(st2->time_base.num == ast2->scale);
820 index = av_index_search_timestamp(
822 av_rescale(timestamp, st2->time_base.den*(int64_t)st->time_base.num, st->time_base.den * (int64_t)st2->time_base.num),
823 flags | AVSEEK_FLAG_BACKWARD);
827 if(!avi->non_interleaved){
828 while(index>0 && st2->index_entries[index].pos > pos)
830 while(index+1 < st2->nb_index_entries && st2->index_entries[index].pos < pos)
834 // av_log(NULL, AV_LOG_DEBUG, "%Ld %d %Ld\n", timestamp, index, st2->index_entries[index].timestamp);
835 /* extract the current frame number */
836 ast2->frame_offset = st2->index_entries[index].timestamp;
837 if(ast2->sample_size)
838 ast2->frame_offset *=ast2->sample_size;
842 dv_flush_audio_packets(avi->dv_demux);
844 url_fseek(&s->pb, pos, SEEK_SET);
845 avi->stream_index= -1;
849 static int avi_read_close(AVFormatContext *s)
852 AVIContext *avi = s->priv_data;
854 for(i=0;i<s->nb_streams;i++) {
855 AVStream *st = s->streams[i];
856 AVIStream *ast = st->priv_data;
858 av_free(st->codec->extradata);
859 av_free(st->codec->palctrl);
863 av_free(avi->dv_demux);
868 static int avi_probe(AVProbeData *p)
870 /* check file header */
871 if (p->buf_size <= 32)
873 if (p->buf[0] == 'R' && p->buf[1] == 'I' &&
874 p->buf[2] == 'F' && p->buf[3] == 'F' &&
875 p->buf[8] == 'A' && p->buf[9] == 'V' &&
876 p->buf[10] == 'I' && p->buf[11] == ' ')
877 return AVPROBE_SCORE_MAX;
882 static AVInputFormat avi_iformat = {
893 int avidec_init(void)
895 av_register_input_format(&avi_iformat);