You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

1489 lines
50KB

  1. /*
  2. * AVI demuxer
  3. * Copyright (c) 2001 Fabrice Bellard
  4. *
  5. * This file is part of FFmpeg.
  6. *
  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.
  11. *
  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.
  16. *
  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
  20. */
  21. #include "libavutil/intreadwrite.h"
  22. #include "libavutil/mathematics.h"
  23. #include "libavutil/bswap.h"
  24. #include "libavutil/opt.h"
  25. #include "libavutil/dict.h"
  26. #include "libavutil/avstring.h"
  27. #include "avformat.h"
  28. #include "internal.h"
  29. #include "avi.h"
  30. #include "dv.h"
  31. #include "riff.h"
  32. #undef NDEBUG
  33. #include <assert.h>
  34. typedef struct AVIStream {
  35. int64_t frame_offset; /* current frame (video) or byte (audio) counter
  36. (used to compute the pts) */
  37. int remaining;
  38. int packet_size;
  39. int scale;
  40. int rate;
  41. int sample_size; /* size of one sample (or packet) (in the rate/scale sense) in bytes */
  42. int64_t cum_len; /* temporary storage (used during seek) */
  43. int prefix; ///< normally 'd'<<8 + 'c' or 'w'<<8 + 'b'
  44. int prefix_count;
  45. uint32_t pal[256];
  46. int has_pal;
  47. int dshow_block_align; ///< block align variable used to emulate bugs in the MS dshow demuxer
  48. AVFormatContext *sub_ctx;
  49. AVPacket sub_pkt;
  50. uint8_t *sub_buffer;
  51. int64_t seek_pos;
  52. } AVIStream;
  53. typedef struct {
  54. const AVClass *class;
  55. int64_t riff_end;
  56. int64_t movi_end;
  57. int64_t fsize;
  58. int64_t movi_list;
  59. int64_t last_pkt_pos;
  60. int index_loaded;
  61. int is_odml;
  62. int non_interleaved;
  63. int stream_index;
  64. DVDemuxContext* dv_demux;
  65. int odml_depth;
  66. int use_odml;
  67. #define MAX_ODML_DEPTH 1000
  68. int64_t dts_max;
  69. } AVIContext;
  70. static const AVOption options[] = {
  71. { "use_odml", "use odml index", offsetof(AVIContext, use_odml), AV_OPT_TYPE_INT, {.dbl = 1}, -1, 1, AV_OPT_FLAG_DECODING_PARAM},
  72. { NULL },
  73. };
  74. static const AVClass demuxer_class = {
  75. "AVI demuxer",
  76. av_default_item_name,
  77. options,
  78. LIBAVUTIL_VERSION_INT,
  79. };
  80. static const char avi_headers[][8] = {
  81. { 'R', 'I', 'F', 'F', 'A', 'V', 'I', ' ' },
  82. { 'R', 'I', 'F', 'F', 'A', 'V', 'I', 'X' },
  83. { 'R', 'I', 'F', 'F', 'A', 'V', 'I', 0x19},
  84. { 'O', 'N', '2', ' ', 'O', 'N', '2', 'f' },
  85. { 'R', 'I', 'F', 'F', 'A', 'M', 'V', ' ' },
  86. { 0 }
  87. };
  88. static const AVMetadataConv avi_metadata_conv[] = {
  89. { "strn", "title" },
  90. { 0 },
  91. };
  92. static int avi_load_index(AVFormatContext *s);
  93. static int guess_ni_flag(AVFormatContext *s);
  94. #define print_tag(str, tag, size) \
  95. av_dlog(NULL, "%s: tag=%c%c%c%c size=0x%x\n", \
  96. str, tag & 0xff, \
  97. (tag >> 8) & 0xff, \
  98. (tag >> 16) & 0xff, \
  99. (tag >> 24) & 0xff, \
  100. size)
  101. static inline int get_duration(AVIStream *ast, int len){
  102. if(ast->sample_size){
  103. return len;
  104. }else if (ast->dshow_block_align){
  105. return (len + ast->dshow_block_align - 1)/ast->dshow_block_align;
  106. }else
  107. return 1;
  108. }
  109. static int get_riff(AVFormatContext *s, AVIOContext *pb)
  110. {
  111. AVIContext *avi = s->priv_data;
  112. char header[8];
  113. int i;
  114. /* check RIFF header */
  115. avio_read(pb, header, 4);
  116. avi->riff_end = avio_rl32(pb); /* RIFF chunk size */
  117. avi->riff_end += avio_tell(pb); /* RIFF chunk end */
  118. avio_read(pb, header+4, 4);
  119. for(i=0; avi_headers[i][0]; i++)
  120. if(!memcmp(header, avi_headers[i], 8))
  121. break;
  122. if(!avi_headers[i][0])
  123. return -1;
  124. if(header[7] == 0x19)
  125. av_log(s, AV_LOG_INFO, "This file has been generated by a totally broken muxer.\n");
  126. return 0;
  127. }
  128. static int read_braindead_odml_indx(AVFormatContext *s, int frame_num){
  129. AVIContext *avi = s->priv_data;
  130. AVIOContext *pb = s->pb;
  131. int longs_pre_entry= avio_rl16(pb);
  132. int index_sub_type = avio_r8(pb);
  133. int index_type = avio_r8(pb);
  134. int entries_in_use = avio_rl32(pb);
  135. int chunk_id = avio_rl32(pb);
  136. int64_t base = avio_rl64(pb);
  137. int stream_id= 10*((chunk_id&0xFF) - '0') + (((chunk_id>>8)&0xFF) - '0');
  138. AVStream *st;
  139. AVIStream *ast;
  140. int i;
  141. int64_t last_pos= -1;
  142. int64_t filesize= avi->fsize;
  143. av_dlog(s, "longs_pre_entry:%d index_type:%d entries_in_use:%d chunk_id:%X base:%16"PRIX64"\n",
  144. longs_pre_entry,index_type, entries_in_use, chunk_id, base);
  145. if(stream_id >= s->nb_streams || stream_id < 0)
  146. return -1;
  147. st= s->streams[stream_id];
  148. ast = st->priv_data;
  149. if(index_sub_type)
  150. return -1;
  151. avio_rl32(pb);
  152. if(index_type && longs_pre_entry != 2)
  153. return -1;
  154. if(index_type>1)
  155. return -1;
  156. if(filesize > 0 && base >= filesize){
  157. av_log(s, AV_LOG_ERROR, "ODML index invalid\n");
  158. if(base>>32 == (base & 0xFFFFFFFF) && (base & 0xFFFFFFFF) < filesize && filesize <= 0xFFFFFFFF)
  159. base &= 0xFFFFFFFF;
  160. else
  161. return -1;
  162. }
  163. for(i=0; i<entries_in_use; i++){
  164. if(index_type){
  165. int64_t pos= avio_rl32(pb) + base - 8;
  166. int len = avio_rl32(pb);
  167. int key= len >= 0;
  168. len &= 0x7FFFFFFF;
  169. #ifdef DEBUG_SEEK
  170. av_log(s, AV_LOG_ERROR, "pos:%"PRId64", len:%X\n", pos, len);
  171. #endif
  172. if(url_feof(pb))
  173. return -1;
  174. if(last_pos == pos || pos == base - 8)
  175. avi->non_interleaved= 1;
  176. if(last_pos != pos && (len || !ast->sample_size))
  177. av_add_index_entry(st, pos, ast->cum_len, len, 0, key ? AVINDEX_KEYFRAME : 0);
  178. ast->cum_len += get_duration(ast, len);
  179. last_pos= pos;
  180. }else{
  181. int64_t offset, pos;
  182. int duration;
  183. offset = avio_rl64(pb);
  184. avio_rl32(pb); /* size */
  185. duration = avio_rl32(pb);
  186. if(url_feof(pb))
  187. return -1;
  188. pos = avio_tell(pb);
  189. if(avi->odml_depth > MAX_ODML_DEPTH){
  190. av_log(s, AV_LOG_ERROR, "Too deeply nested ODML indexes\n");
  191. return -1;
  192. }
  193. if(avio_seek(pb, offset+8, SEEK_SET) < 0)
  194. return -1;
  195. avi->odml_depth++;
  196. read_braindead_odml_indx(s, frame_num);
  197. avi->odml_depth--;
  198. frame_num += duration;
  199. if(avio_seek(pb, pos, SEEK_SET) < 0) {
  200. av_log(s, AV_LOG_ERROR, "Failed to restore position after reading index");
  201. return -1;
  202. }
  203. }
  204. }
  205. avi->index_loaded=2;
  206. return 0;
  207. }
  208. static void clean_index(AVFormatContext *s){
  209. int i;
  210. int64_t j;
  211. for(i=0; i<s->nb_streams; i++){
  212. AVStream *st = s->streams[i];
  213. AVIStream *ast = st->priv_data;
  214. int n= st->nb_index_entries;
  215. int max= ast->sample_size;
  216. int64_t pos, size, ts;
  217. if(n != 1 || ast->sample_size==0)
  218. continue;
  219. while(max < 1024) max+=max;
  220. pos= st->index_entries[0].pos;
  221. size= st->index_entries[0].size;
  222. ts= st->index_entries[0].timestamp;
  223. for(j=0; j<size; j+=max){
  224. av_add_index_entry(st, pos+j, ts+j, FFMIN(max, size-j), 0, AVINDEX_KEYFRAME);
  225. }
  226. }
  227. }
  228. static int avi_read_tag(AVFormatContext *s, AVStream *st, uint32_t tag, uint32_t size)
  229. {
  230. AVIOContext *pb = s->pb;
  231. char key[5] = {0}, *value;
  232. size += (size & 1);
  233. if (size == UINT_MAX)
  234. return -1;
  235. value = av_malloc(size+1);
  236. if (!value)
  237. return -1;
  238. avio_read(pb, value, size);
  239. value[size]=0;
  240. AV_WL32(key, tag);
  241. return av_dict_set(st ? &st->metadata : &s->metadata, key, value,
  242. AV_DICT_DONT_STRDUP_VAL);
  243. }
  244. static const char months[12][4] = { "Jan", "Feb", "Mar", "Apr", "May", "Jun",
  245. "Jul", "Aug", "Sep", "Oct", "Nov", "Dec" };
  246. static void avi_metadata_creation_time(AVDictionary **metadata, char *date)
  247. {
  248. char month[4], time[9], buffer[64];
  249. int i, day, year;
  250. /* parse standard AVI date format (ie. "Mon Mar 10 15:04:43 2003") */
  251. if (sscanf(date, "%*3s%*[ ]%3s%*[ ]%2d%*[ ]%8s%*[ ]%4d",
  252. month, &day, time, &year) == 4) {
  253. for (i=0; i<12; i++)
  254. if (!av_strcasecmp(month, months[i])) {
  255. snprintf(buffer, sizeof(buffer), "%.4d-%.2d-%.2d %s",
  256. year, i+1, day, time);
  257. av_dict_set(metadata, "creation_time", buffer, 0);
  258. }
  259. } else if (date[4] == '/' && date[7] == '/') {
  260. date[4] = date[7] = '-';
  261. av_dict_set(metadata, "creation_time", date, 0);
  262. }
  263. }
  264. static void avi_read_nikon(AVFormatContext *s, uint64_t end)
  265. {
  266. while (avio_tell(s->pb) < end) {
  267. uint32_t tag = avio_rl32(s->pb);
  268. uint32_t size = avio_rl32(s->pb);
  269. switch (tag) {
  270. case MKTAG('n', 'c', 't', 'g'): { /* Nikon Tags */
  271. uint64_t tag_end = avio_tell(s->pb) + size;
  272. while (avio_tell(s->pb) < tag_end) {
  273. uint16_t tag = avio_rl16(s->pb);
  274. uint16_t size = avio_rl16(s->pb);
  275. const char *name = NULL;
  276. char buffer[64] = {0};
  277. size -= avio_read(s->pb, buffer,
  278. FFMIN(size, sizeof(buffer)-1));
  279. switch (tag) {
  280. case 0x03: name = "maker"; break;
  281. case 0x04: name = "model"; break;
  282. case 0x13: name = "creation_time";
  283. if (buffer[4] == ':' && buffer[7] == ':')
  284. buffer[4] = buffer[7] = '-';
  285. break;
  286. }
  287. if (name)
  288. av_dict_set(&s->metadata, name, buffer, 0);
  289. avio_skip(s->pb, size);
  290. }
  291. break;
  292. }
  293. default:
  294. avio_skip(s->pb, size);
  295. break;
  296. }
  297. }
  298. }
  299. static int avi_read_header(AVFormatContext *s, AVFormatParameters *ap)
  300. {
  301. AVIContext *avi = s->priv_data;
  302. AVIOContext *pb = s->pb;
  303. unsigned int tag, tag1, handler;
  304. int codec_type, stream_index, frame_period;
  305. unsigned int size;
  306. int i;
  307. AVStream *st;
  308. AVIStream *ast = NULL;
  309. int avih_width=0, avih_height=0;
  310. int amv_file_format=0;
  311. uint64_t list_end = 0;
  312. int ret;
  313. avi->stream_index= -1;
  314. if (get_riff(s, pb) < 0)
  315. return -1;
  316. av_log(avi, AV_LOG_DEBUG, "use odml:%d\n", avi->use_odml);
  317. avi->fsize = avio_size(pb);
  318. if(avi->fsize<=0 || avi->fsize < avi->riff_end)
  319. avi->fsize= avi->riff_end == 8 ? INT64_MAX : avi->riff_end;
  320. /* first list tag */
  321. stream_index = -1;
  322. codec_type = -1;
  323. frame_period = 0;
  324. for(;;) {
  325. if (url_feof(pb))
  326. goto fail;
  327. tag = avio_rl32(pb);
  328. size = avio_rl32(pb);
  329. print_tag("tag", tag, size);
  330. switch(tag) {
  331. case MKTAG('L', 'I', 'S', 'T'):
  332. list_end = avio_tell(pb) + size;
  333. /* Ignored, except at start of video packets. */
  334. tag1 = avio_rl32(pb);
  335. print_tag("list", tag1, 0);
  336. if (tag1 == MKTAG('m', 'o', 'v', 'i')) {
  337. avi->movi_list = avio_tell(pb) - 4;
  338. if(size) avi->movi_end = avi->movi_list + size + (size & 1);
  339. else avi->movi_end = avi->fsize;
  340. av_dlog(NULL, "movi end=%"PRIx64"\n", avi->movi_end);
  341. goto end_of_header;
  342. }
  343. else if (tag1 == MKTAG('I', 'N', 'F', 'O'))
  344. ff_read_riff_info(s, size - 4);
  345. else if (tag1 == MKTAG('n', 'c', 'd', 't'))
  346. avi_read_nikon(s, list_end);
  347. break;
  348. case MKTAG('I', 'D', 'I', 'T'): {
  349. unsigned char date[64] = {0};
  350. size += (size & 1);
  351. size -= avio_read(pb, date, FFMIN(size, sizeof(date)-1));
  352. avio_skip(pb, size);
  353. avi_metadata_creation_time(&s->metadata, date);
  354. break;
  355. }
  356. case MKTAG('d', 'm', 'l', 'h'):
  357. avi->is_odml = 1;
  358. avio_skip(pb, size + (size & 1));
  359. break;
  360. case MKTAG('a', 'm', 'v', 'h'):
  361. amv_file_format=1;
  362. case MKTAG('a', 'v', 'i', 'h'):
  363. /* AVI header */
  364. /* using frame_period is bad idea */
  365. frame_period = avio_rl32(pb);
  366. avio_rl32(pb); /* max. bytes per second */
  367. avio_rl32(pb);
  368. avi->non_interleaved |= avio_rl32(pb) & AVIF_MUSTUSEINDEX;
  369. avio_skip(pb, 2 * 4);
  370. avio_rl32(pb);
  371. avio_rl32(pb);
  372. avih_width=avio_rl32(pb);
  373. avih_height=avio_rl32(pb);
  374. avio_skip(pb, size - 10 * 4);
  375. break;
  376. case MKTAG('s', 't', 'r', 'h'):
  377. /* stream header */
  378. tag1 = avio_rl32(pb);
  379. handler = avio_rl32(pb); /* codec tag */
  380. if(tag1 == MKTAG('p', 'a', 'd', 's')){
  381. avio_skip(pb, size - 8);
  382. break;
  383. }else{
  384. stream_index++;
  385. st = avformat_new_stream(s, NULL);
  386. if (!st)
  387. goto fail;
  388. st->id = stream_index;
  389. ast = av_mallocz(sizeof(AVIStream));
  390. if (!ast)
  391. goto fail;
  392. st->priv_data = ast;
  393. }
  394. if(amv_file_format)
  395. tag1 = stream_index ? MKTAG('a','u','d','s') : MKTAG('v','i','d','s');
  396. print_tag("strh", tag1, -1);
  397. if(tag1 == MKTAG('i', 'a', 'v', 's') || tag1 == MKTAG('i', 'v', 'a', 's')){
  398. int64_t dv_dur;
  399. /*
  400. * After some consideration -- I don't think we
  401. * have to support anything but DV in type1 AVIs.
  402. */
  403. if (s->nb_streams != 1)
  404. goto fail;
  405. if (handler != MKTAG('d', 'v', 's', 'd') &&
  406. handler != MKTAG('d', 'v', 'h', 'd') &&
  407. handler != MKTAG('d', 'v', 's', 'l'))
  408. goto fail;
  409. ast = s->streams[0]->priv_data;
  410. av_freep(&s->streams[0]->codec->extradata);
  411. av_freep(&s->streams[0]->codec);
  412. av_freep(&s->streams[0]);
  413. s->nb_streams = 0;
  414. if (CONFIG_DV_DEMUXER) {
  415. avi->dv_demux = avpriv_dv_init_demux(s);
  416. if (!avi->dv_demux)
  417. goto fail;
  418. }
  419. s->streams[0]->priv_data = ast;
  420. avio_skip(pb, 3 * 4);
  421. ast->scale = avio_rl32(pb);
  422. ast->rate = avio_rl32(pb);
  423. avio_skip(pb, 4); /* start time */
  424. dv_dur = avio_rl32(pb);
  425. if (ast->scale > 0 && ast->rate > 0 && dv_dur > 0) {
  426. dv_dur *= AV_TIME_BASE;
  427. s->duration = av_rescale(dv_dur, ast->scale, ast->rate);
  428. }
  429. /*
  430. * else, leave duration alone; timing estimation in utils.c
  431. * will make a guess based on bitrate.
  432. */
  433. stream_index = s->nb_streams - 1;
  434. avio_skip(pb, size - 9*4);
  435. break;
  436. }
  437. assert(stream_index < s->nb_streams);
  438. st->codec->stream_codec_tag= handler;
  439. avio_rl32(pb); /* flags */
  440. avio_rl16(pb); /* priority */
  441. avio_rl16(pb); /* language */
  442. avio_rl32(pb); /* initial frame */
  443. ast->scale = avio_rl32(pb);
  444. ast->rate = avio_rl32(pb);
  445. if(!(ast->scale && ast->rate)){
  446. av_log(s, AV_LOG_WARNING, "scale/rate is %u/%u which is invalid. (This file has been generated by broken software.)\n", ast->scale, ast->rate);
  447. if(frame_period){
  448. ast->rate = 1000000;
  449. ast->scale = frame_period;
  450. }else{
  451. ast->rate = 25;
  452. ast->scale = 1;
  453. }
  454. }
  455. avpriv_set_pts_info(st, 64, ast->scale, ast->rate);
  456. ast->cum_len=avio_rl32(pb); /* start */
  457. st->nb_frames = avio_rl32(pb);
  458. st->start_time = 0;
  459. avio_rl32(pb); /* buffer size */
  460. avio_rl32(pb); /* quality */
  461. ast->sample_size = avio_rl32(pb); /* sample ssize */
  462. ast->cum_len *= FFMAX(1, ast->sample_size);
  463. // av_log(s, AV_LOG_DEBUG, "%d %d %d %d\n", ast->rate, ast->scale, ast->start, ast->sample_size);
  464. switch(tag1) {
  465. case MKTAG('v', 'i', 'd', 's'):
  466. codec_type = AVMEDIA_TYPE_VIDEO;
  467. ast->sample_size = 0;
  468. break;
  469. case MKTAG('a', 'u', 'd', 's'):
  470. codec_type = AVMEDIA_TYPE_AUDIO;
  471. break;
  472. case MKTAG('t', 'x', 't', 's'):
  473. codec_type = AVMEDIA_TYPE_SUBTITLE;
  474. break;
  475. case MKTAG('d', 'a', 't', 's'):
  476. codec_type = AVMEDIA_TYPE_DATA;
  477. break;
  478. default:
  479. av_log(s, AV_LOG_INFO, "unknown stream type %X\n", tag1);
  480. }
  481. if(ast->sample_size == 0)
  482. st->duration = st->nb_frames;
  483. ast->frame_offset= ast->cum_len;
  484. avio_skip(pb, size - 12 * 4);
  485. break;
  486. case MKTAG('s', 't', 'r', 'f'):
  487. /* stream header */
  488. if (!size)
  489. break;
  490. if (stream_index >= (unsigned)s->nb_streams || avi->dv_demux) {
  491. avio_skip(pb, size);
  492. } else {
  493. uint64_t cur_pos = avio_tell(pb);
  494. if (cur_pos < list_end)
  495. size = FFMIN(size, list_end - cur_pos);
  496. st = s->streams[stream_index];
  497. switch(codec_type) {
  498. case AVMEDIA_TYPE_VIDEO:
  499. if(amv_file_format){
  500. st->codec->width=avih_width;
  501. st->codec->height=avih_height;
  502. st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
  503. st->codec->codec_id = CODEC_ID_AMV;
  504. avio_skip(pb, size);
  505. break;
  506. }
  507. tag1 = ff_get_bmp_header(pb, st);
  508. if (tag1 == MKTAG('D', 'X', 'S', 'B') || tag1 == MKTAG('D','X','S','A')) {
  509. st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
  510. st->codec->codec_tag = tag1;
  511. st->codec->codec_id = CODEC_ID_XSUB;
  512. break;
  513. }
  514. if(size > 10*4 && size<(1<<30)){
  515. st->codec->extradata_size= size - 10*4;
  516. st->codec->extradata= av_malloc(st->codec->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
  517. if (!st->codec->extradata) {
  518. st->codec->extradata_size= 0;
  519. return AVERROR(ENOMEM);
  520. }
  521. avio_read(pb, st->codec->extradata, st->codec->extradata_size);
  522. }
  523. if(st->codec->extradata_size & 1) //FIXME check if the encoder really did this correctly
  524. avio_r8(pb);
  525. /* Extract palette from extradata if bpp <= 8. */
  526. /* This code assumes that extradata contains only palette. */
  527. /* This is true for all paletted codecs implemented in FFmpeg. */
  528. if (st->codec->extradata_size && (st->codec->bits_per_coded_sample <= 8)) {
  529. int pal_size = (1 << st->codec->bits_per_coded_sample) << 2;
  530. const uint8_t *pal_src;
  531. pal_size = FFMIN(pal_size, st->codec->extradata_size);
  532. pal_src = st->codec->extradata + st->codec->extradata_size - pal_size;
  533. #if HAVE_BIGENDIAN
  534. for (i = 0; i < pal_size/4; i++)
  535. ast->pal[i] = AV_RL32(pal_src+4*i);
  536. #else
  537. memcpy(ast->pal, pal_src, pal_size);
  538. #endif
  539. ast->has_pal = 1;
  540. }
  541. print_tag("video", tag1, 0);
  542. st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
  543. st->codec->codec_tag = tag1;
  544. st->codec->codec_id = ff_codec_get_id(ff_codec_bmp_tags, tag1);
  545. st->need_parsing = AVSTREAM_PARSE_HEADERS; // This is needed to get the pict type which is necessary for generating correct pts.
  546. // Support "Resolution 1:1" for Avid AVI Codec
  547. if(tag1 == MKTAG('A', 'V', 'R', 'n') &&
  548. st->codec->extradata_size >= 31 &&
  549. !memcmp(&st->codec->extradata[28], "1:1", 3))
  550. st->codec->codec_id = CODEC_ID_RAWVIDEO;
  551. if(st->codec->codec_tag==0 && st->codec->height > 0 && st->codec->extradata_size < 1U<<30){
  552. st->codec->extradata_size+= 9;
  553. st->codec->extradata= av_realloc_f(st->codec->extradata, 1, st->codec->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
  554. if(st->codec->extradata)
  555. memcpy(st->codec->extradata + st->codec->extradata_size - 9, "BottomUp", 9);
  556. }
  557. st->codec->height= FFABS(st->codec->height);
  558. // avio_skip(pb, size - 5 * 4);
  559. break;
  560. case AVMEDIA_TYPE_AUDIO:
  561. ret = ff_get_wav_header(pb, st->codec, size);
  562. if (ret < 0)
  563. return ret;
  564. ast->dshow_block_align= st->codec->block_align;
  565. if(ast->sample_size && st->codec->block_align && ast->sample_size != st->codec->block_align){
  566. av_log(s, AV_LOG_WARNING, "sample size (%d) != block align (%d)\n", ast->sample_size, st->codec->block_align);
  567. ast->sample_size= st->codec->block_align;
  568. }
  569. if (size&1) /* 2-aligned (fix for Stargate SG-1 - 3x18 - Shades of Grey.avi) */
  570. avio_skip(pb, 1);
  571. /* Force parsing as several audio frames can be in
  572. * one packet and timestamps refer to packet start. */
  573. st->need_parsing = AVSTREAM_PARSE_TIMESTAMPS;
  574. /* ADTS header is in extradata, AAC without header must be
  575. * stored as exact frames. Parser not needed and it will
  576. * fail. */
  577. if (st->codec->codec_id == CODEC_ID_AAC && st->codec->extradata_size)
  578. st->need_parsing = AVSTREAM_PARSE_NONE;
  579. /* AVI files with Xan DPCM audio (wrongly) declare PCM
  580. * audio in the header but have Axan as stream_code_tag. */
  581. if (st->codec->stream_codec_tag == AV_RL32("Axan")){
  582. st->codec->codec_id = CODEC_ID_XAN_DPCM;
  583. st->codec->codec_tag = 0;
  584. ast->dshow_block_align = 0;
  585. }
  586. if (amv_file_format){
  587. st->codec->codec_id = CODEC_ID_ADPCM_IMA_AMV;
  588. ast->dshow_block_align = 0;
  589. }
  590. break;
  591. case AVMEDIA_TYPE_SUBTITLE:
  592. st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
  593. st->request_probe= 1;
  594. break;
  595. default:
  596. st->codec->codec_type = AVMEDIA_TYPE_DATA;
  597. st->codec->codec_id= CODEC_ID_NONE;
  598. st->codec->codec_tag= 0;
  599. avio_skip(pb, size);
  600. break;
  601. }
  602. }
  603. break;
  604. case MKTAG('i', 'n', 'd', 'x'):
  605. i= avio_tell(pb);
  606. if(pb->seekable && !(s->flags & AVFMT_FLAG_IGNIDX) && avi->use_odml &&
  607. read_braindead_odml_indx(s, 0) < 0 && s->error_recognition >= FF_ER_EXPLODE){
  608. goto fail; }
  609. avio_seek(pb, i+size, SEEK_SET);
  610. break;
  611. case MKTAG('v', 'p', 'r', 'p'):
  612. if(stream_index < (unsigned)s->nb_streams && size > 9*4){
  613. AVRational active, active_aspect;
  614. st = s->streams[stream_index];
  615. avio_rl32(pb);
  616. avio_rl32(pb);
  617. avio_rl32(pb);
  618. avio_rl32(pb);
  619. avio_rl32(pb);
  620. active_aspect.den= avio_rl16(pb);
  621. active_aspect.num= avio_rl16(pb);
  622. active.num = avio_rl32(pb);
  623. active.den = avio_rl32(pb);
  624. avio_rl32(pb); //nbFieldsPerFrame
  625. if(active_aspect.num && active_aspect.den && active.num && active.den){
  626. st->sample_aspect_ratio= av_div_q(active_aspect, active);
  627. //av_log(s, AV_LOG_ERROR, "vprp %d/%d %d/%d\n", active_aspect.num, active_aspect.den, active.num, active.den);
  628. }
  629. size -= 9*4;
  630. }
  631. avio_skip(pb, size);
  632. break;
  633. case MKTAG('s', 't', 'r', 'n'):
  634. if(s->nb_streams){
  635. avi_read_tag(s, s->streams[s->nb_streams-1], tag, size);
  636. break;
  637. }
  638. default:
  639. if(size > 1000000){
  640. av_log(s, AV_LOG_ERROR, "Something went wrong during header parsing, "
  641. "I will ignore it and try to continue anyway.\n");
  642. if (s->error_recognition >= FF_ER_EXPLODE) goto fail;
  643. avi->movi_list = avio_tell(pb) - 4;
  644. avi->movi_end = avi->fsize;
  645. goto end_of_header;
  646. }
  647. /* skip tag */
  648. size += (size & 1);
  649. avio_skip(pb, size);
  650. break;
  651. }
  652. }
  653. end_of_header:
  654. /* check stream number */
  655. if (stream_index != s->nb_streams - 1) {
  656. fail:
  657. return -1;
  658. }
  659. if(!avi->index_loaded && pb->seekable)
  660. avi_load_index(s);
  661. avi->index_loaded |= 1;
  662. avi->non_interleaved |= guess_ni_flag(s) | (s->flags & AVFMT_FLAG_SORT_DTS);
  663. for(i=0; i<s->nb_streams; i++){
  664. AVStream *st = s->streams[i];
  665. if(st->nb_index_entries)
  666. break;
  667. }
  668. // DV-in-AVI cannot be non-interleaved, if set this must be
  669. // a mis-detection.
  670. if(avi->dv_demux)
  671. avi->non_interleaved=0;
  672. if(i==s->nb_streams && avi->non_interleaved) {
  673. av_log(s, AV_LOG_WARNING, "non-interleaved AVI without index, switching to interleaved\n");
  674. avi->non_interleaved=0;
  675. }
  676. if(avi->non_interleaved) {
  677. av_log(s, AV_LOG_INFO, "non-interleaved AVI\n");
  678. clean_index(s);
  679. }
  680. ff_metadata_conv_ctx(s, NULL, avi_metadata_conv);
  681. ff_metadata_conv_ctx(s, NULL, ff_riff_info_conv);
  682. return 0;
  683. }
  684. static int read_gab2_sub(AVStream *st, AVPacket *pkt) {
  685. if (!strcmp(pkt->data, "GAB2") && AV_RL16(pkt->data+5) == 2) {
  686. uint8_t desc[256];
  687. int score = AVPROBE_SCORE_MAX / 2, ret;
  688. AVIStream *ast = st->priv_data;
  689. AVInputFormat *sub_demuxer;
  690. AVRational time_base;
  691. AVIOContext *pb = avio_alloc_context( pkt->data + 7,
  692. pkt->size - 7,
  693. 0, NULL, NULL, NULL, NULL);
  694. AVProbeData pd;
  695. unsigned int desc_len = avio_rl32(pb);
  696. if (desc_len > pb->buf_end - pb->buf_ptr)
  697. goto error;
  698. ret = avio_get_str16le(pb, desc_len, desc, sizeof(desc));
  699. avio_skip(pb, desc_len - ret);
  700. if (*desc)
  701. av_dict_set(&st->metadata, "title", desc, 0);
  702. avio_rl16(pb); /* flags? */
  703. avio_rl32(pb); /* data size */
  704. pd = (AVProbeData) { .buf = pb->buf_ptr, .buf_size = pb->buf_end - pb->buf_ptr };
  705. if (!(sub_demuxer = av_probe_input_format2(&pd, 1, &score)))
  706. goto error;
  707. if (!(ast->sub_ctx = avformat_alloc_context()))
  708. goto error;
  709. ast->sub_ctx->pb = pb;
  710. if (!avformat_open_input(&ast->sub_ctx, "", sub_demuxer, NULL)) {
  711. av_read_packet(ast->sub_ctx, &ast->sub_pkt);
  712. *st->codec = *ast->sub_ctx->streams[0]->codec;
  713. ast->sub_ctx->streams[0]->codec->extradata = NULL;
  714. time_base = ast->sub_ctx->streams[0]->time_base;
  715. avpriv_set_pts_info(st, 64, time_base.num, time_base.den);
  716. }
  717. ast->sub_buffer = pkt->data;
  718. memset(pkt, 0, sizeof(*pkt));
  719. return 1;
  720. error:
  721. av_freep(&pb);
  722. }
  723. return 0;
  724. }
  725. static AVStream *get_subtitle_pkt(AVFormatContext *s, AVStream *next_st,
  726. AVPacket *pkt)
  727. {
  728. AVIStream *ast, *next_ast = next_st->priv_data;
  729. int64_t ts, next_ts, ts_min = INT64_MAX;
  730. AVStream *st, *sub_st = NULL;
  731. int i;
  732. next_ts = av_rescale_q(next_ast->frame_offset, next_st->time_base,
  733. AV_TIME_BASE_Q);
  734. for (i=0; i<s->nb_streams; i++) {
  735. st = s->streams[i];
  736. ast = st->priv_data;
  737. if (st->discard < AVDISCARD_ALL && ast && ast->sub_pkt.data) {
  738. ts = av_rescale_q(ast->sub_pkt.dts, st->time_base, AV_TIME_BASE_Q);
  739. if (ts <= next_ts && ts < ts_min) {
  740. ts_min = ts;
  741. sub_st = st;
  742. }
  743. }
  744. }
  745. if (sub_st) {
  746. ast = sub_st->priv_data;
  747. *pkt = ast->sub_pkt;
  748. pkt->stream_index = sub_st->index;
  749. if (av_read_packet(ast->sub_ctx, &ast->sub_pkt) < 0)
  750. ast->sub_pkt.data = NULL;
  751. }
  752. return sub_st;
  753. }
  754. static int get_stream_idx(int *d){
  755. if( d[0] >= '0' && d[0] <= '9'
  756. && d[1] >= '0' && d[1] <= '9'){
  757. return (d[0] - '0') * 10 + (d[1] - '0');
  758. }else{
  759. return 100; //invalid stream ID
  760. }
  761. }
  762. static int avi_sync(AVFormatContext *s, int exit_early)
  763. {
  764. AVIContext *avi = s->priv_data;
  765. AVIOContext *pb = s->pb;
  766. int n;
  767. unsigned int d[8];
  768. unsigned int size;
  769. int64_t i, sync;
  770. start_sync:
  771. memset(d, -1, sizeof(d));
  772. for(i=sync=avio_tell(pb); !url_feof(pb); i++) {
  773. int j;
  774. for(j=0; j<7; j++)
  775. d[j]= d[j+1];
  776. d[7]= avio_r8(pb);
  777. size= d[4] + (d[5]<<8) + (d[6]<<16) + (d[7]<<24);
  778. n= get_stream_idx(d+2);
  779. //av_log(s, AV_LOG_DEBUG, "%X %X %X %X %X %X %X %X %"PRId64" %d %d\n", d[0], d[1], d[2], d[3], d[4], d[5], d[6], d[7], i, size, n);
  780. if(i + (uint64_t)size > avi->fsize || d[0] > 127)
  781. continue;
  782. //parse ix##
  783. if( (d[0] == 'i' && d[1] == 'x' && n < s->nb_streams)
  784. //parse JUNK
  785. ||(d[0] == 'J' && d[1] == 'U' && d[2] == 'N' && d[3] == 'K')
  786. ||(d[0] == 'i' && d[1] == 'd' && d[2] == 'x' && d[3] == '1')){
  787. avio_skip(pb, size);
  788. //av_log(s, AV_LOG_DEBUG, "SKIP\n");
  789. goto start_sync;
  790. }
  791. //parse stray LIST
  792. if(d[0] == 'L' && d[1] == 'I' && d[2] == 'S' && d[3] == 'T'){
  793. avio_skip(pb, 4);
  794. goto start_sync;
  795. }
  796. n= get_stream_idx(d);
  797. if(!((i-avi->last_pkt_pos)&1) && get_stream_idx(d+1) < s->nb_streams)
  798. continue;
  799. //detect ##ix chunk and skip
  800. if(d[2] == 'i' && d[3] == 'x' && n < s->nb_streams){
  801. avio_skip(pb, size);
  802. goto start_sync;
  803. }
  804. //parse ##dc/##wb
  805. if(n < s->nb_streams){
  806. AVStream *st;
  807. AVIStream *ast;
  808. st = s->streams[n];
  809. ast = st->priv_data;
  810. if(s->nb_streams>=2){
  811. AVStream *st1 = s->streams[1];
  812. AVIStream *ast1= st1->priv_data;
  813. //workaround for broken small-file-bug402.avi
  814. if( d[2] == 'w' && d[3] == 'b'
  815. && n==0
  816. && st ->codec->codec_type == AVMEDIA_TYPE_VIDEO
  817. && st1->codec->codec_type == AVMEDIA_TYPE_AUDIO
  818. && ast->prefix == 'd'*256+'c'
  819. && (d[2]*256+d[3] == ast1->prefix || !ast1->prefix_count)
  820. ){
  821. n=1;
  822. st = st1;
  823. ast = ast1;
  824. av_log(s, AV_LOG_WARNING, "Invalid stream + prefix combination, assuming audio.\n");
  825. }
  826. }
  827. if( (st->discard >= AVDISCARD_DEFAULT && size==0)
  828. /*|| (st->discard >= AVDISCARD_NONKEY && !(pkt->flags & AV_PKT_FLAG_KEY))*/ //FIXME needs a little reordering
  829. || st->discard >= AVDISCARD_ALL){
  830. if (!exit_early) {
  831. ast->frame_offset += get_duration(ast, size);
  832. }
  833. avio_skip(pb, size);
  834. goto start_sync;
  835. }
  836. if (d[2] == 'p' && d[3] == 'c' && size<=4*256+4) {
  837. int k = avio_r8(pb);
  838. int last = (k + avio_r8(pb) - 1) & 0xFF;
  839. avio_rl16(pb); //flags
  840. for (; k <= last; k++)
  841. ast->pal[k] = avio_rb32(pb)>>8;// b + (g << 8) + (r << 16);
  842. ast->has_pal= 1;
  843. goto start_sync;
  844. } else if( ((ast->prefix_count<5 || sync+9 > i) && d[2]<128 && d[3]<128) ||
  845. d[2]*256+d[3] == ast->prefix /*||
  846. (d[2] == 'd' && d[3] == 'c') ||
  847. (d[2] == 'w' && d[3] == 'b')*/) {
  848. if (exit_early)
  849. return 0;
  850. //av_log(s, AV_LOG_DEBUG, "OK\n");
  851. if(d[2]*256+d[3] == ast->prefix)
  852. ast->prefix_count++;
  853. else{
  854. ast->prefix= d[2]*256+d[3];
  855. ast->prefix_count= 0;
  856. }
  857. avi->stream_index= n;
  858. ast->packet_size= size + 8;
  859. ast->remaining= size;
  860. if(size || !ast->sample_size){
  861. uint64_t pos= avio_tell(pb) - 8;
  862. if(!st->index_entries || !st->nb_index_entries || st->index_entries[st->nb_index_entries - 1].pos < pos){
  863. av_add_index_entry(st, pos, ast->frame_offset, size, 0, AVINDEX_KEYFRAME);
  864. }
  865. }
  866. return 0;
  867. }
  868. }
  869. }
  870. if(pb->error)
  871. return pb->error;
  872. return AVERROR_EOF;
  873. }
  874. static int avi_read_packet(AVFormatContext *s, AVPacket *pkt)
  875. {
  876. AVIContext *avi = s->priv_data;
  877. AVIOContext *pb = s->pb;
  878. int err;
  879. void* dstr;
  880. if (CONFIG_DV_DEMUXER && avi->dv_demux) {
  881. int size = avpriv_dv_get_packet(avi->dv_demux, pkt);
  882. if (size >= 0)
  883. return size;
  884. }
  885. if(avi->non_interleaved){
  886. int best_stream_index = 0;
  887. AVStream *best_st= NULL;
  888. AVIStream *best_ast;
  889. int64_t best_ts= INT64_MAX;
  890. int i;
  891. for(i=0; i<s->nb_streams; i++){
  892. AVStream *st = s->streams[i];
  893. AVIStream *ast = st->priv_data;
  894. int64_t ts= ast->frame_offset;
  895. int64_t last_ts;
  896. if(!st->nb_index_entries)
  897. continue;
  898. last_ts = st->index_entries[st->nb_index_entries - 1].timestamp;
  899. if(!ast->remaining && ts > last_ts)
  900. continue;
  901. ts = av_rescale_q(ts, st->time_base, (AVRational){FFMAX(1, ast->sample_size), AV_TIME_BASE});
  902. // av_log(s, AV_LOG_DEBUG, "%"PRId64" %d/%d %"PRId64"\n", ts, st->time_base.num, st->time_base.den, ast->frame_offset);
  903. if(ts < best_ts){
  904. best_ts= ts;
  905. best_st= st;
  906. best_stream_index= i;
  907. }
  908. }
  909. if(!best_st)
  910. return AVERROR_EOF;
  911. best_ast = best_st->priv_data;
  912. best_ts = av_rescale_q(best_ts, (AVRational){FFMAX(1, best_ast->sample_size), AV_TIME_BASE}, best_st->time_base);
  913. if(best_ast->remaining)
  914. i= av_index_search_timestamp(best_st, best_ts, AVSEEK_FLAG_ANY | AVSEEK_FLAG_BACKWARD);
  915. else{
  916. i= av_index_search_timestamp(best_st, best_ts, AVSEEK_FLAG_ANY);
  917. if(i>=0)
  918. best_ast->frame_offset= best_st->index_entries[i].timestamp;
  919. }
  920. // av_log(s, AV_LOG_DEBUG, "%d\n", i);
  921. if(i>=0){
  922. int64_t pos= best_st->index_entries[i].pos;
  923. pos += best_ast->packet_size - best_ast->remaining;
  924. if(avio_seek(s->pb, pos + 8, SEEK_SET) < 0)
  925. return AVERROR_EOF;
  926. // av_log(s, AV_LOG_DEBUG, "pos=%"PRId64"\n", pos);
  927. assert(best_ast->remaining <= best_ast->packet_size);
  928. avi->stream_index= best_stream_index;
  929. if(!best_ast->remaining)
  930. best_ast->packet_size=
  931. best_ast->remaining= best_st->index_entries[i].size;
  932. }
  933. else
  934. return AVERROR_EOF;
  935. }
  936. resync:
  937. if(avi->stream_index >= 0){
  938. AVStream *st= s->streams[ avi->stream_index ];
  939. AVIStream *ast= st->priv_data;
  940. int size, err;
  941. if(get_subtitle_pkt(s, st, pkt))
  942. return 0;
  943. if(ast->sample_size <= 1) // minorityreport.AVI block_align=1024 sample_size=1 IMA-ADPCM
  944. size= INT_MAX;
  945. else if(ast->sample_size < 32)
  946. // arbitrary multiplier to avoid tiny packets for raw PCM data
  947. size= 1024*ast->sample_size;
  948. else
  949. size= ast->sample_size;
  950. if(size > ast->remaining)
  951. size= ast->remaining;
  952. avi->last_pkt_pos= avio_tell(pb);
  953. err= av_get_packet(pb, pkt, size);
  954. if(err<0)
  955. return err;
  956. if(ast->has_pal && pkt->data && pkt->size<(unsigned)INT_MAX/2){
  957. uint8_t *pal;
  958. pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
  959. if(!pal){
  960. av_log(s, AV_LOG_ERROR, "Failed to allocate data for palette\n");
  961. }else{
  962. memcpy(pal, ast->pal, AVPALETTE_SIZE);
  963. ast->has_pal = 0;
  964. }
  965. }
  966. if (CONFIG_DV_DEMUXER && avi->dv_demux) {
  967. dstr = pkt->destruct;
  968. size = avpriv_dv_produce_packet(avi->dv_demux, pkt,
  969. pkt->data, pkt->size, pkt->pos);
  970. pkt->destruct = dstr;
  971. pkt->flags |= AV_PKT_FLAG_KEY;
  972. if (size < 0)
  973. av_free_packet(pkt);
  974. } else if (st->codec->codec_type == AVMEDIA_TYPE_SUBTITLE
  975. && !st->codec->codec_tag && read_gab2_sub(st, pkt)) {
  976. ast->frame_offset++;
  977. avi->stream_index = -1;
  978. ast->remaining = 0;
  979. goto resync;
  980. } else {
  981. /* XXX: How to handle B-frames in AVI? */
  982. pkt->dts = ast->frame_offset;
  983. // pkt->dts += ast->start;
  984. if(ast->sample_size)
  985. pkt->dts /= ast->sample_size;
  986. //av_log(s, AV_LOG_DEBUG, "dts:%"PRId64" offset:%"PRId64" %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, avi->stream_index, size);
  987. pkt->stream_index = avi->stream_index;
  988. if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
  989. AVIndexEntry *e;
  990. int index;
  991. assert(st->index_entries);
  992. index= av_index_search_timestamp(st, ast->frame_offset, 0);
  993. e= &st->index_entries[index];
  994. if(index >= 0 && e->timestamp == ast->frame_offset){
  995. if (index == st->nb_index_entries-1){
  996. int key=1;
  997. int i;
  998. uint32_t state=-1;
  999. for(i=0; i<FFMIN(size,256); i++){
  1000. if(st->codec->codec_id == CODEC_ID_MPEG4){
  1001. if(state == 0x1B6){
  1002. key= !(pkt->data[i]&0xC0);
  1003. break;
  1004. }
  1005. }else
  1006. break;
  1007. state= (state<<8) + pkt->data[i];
  1008. }
  1009. if(!key)
  1010. e->flags &= ~AVINDEX_KEYFRAME;
  1011. }
  1012. if (e->flags & AVINDEX_KEYFRAME)
  1013. pkt->flags |= AV_PKT_FLAG_KEY;
  1014. }
  1015. } else {
  1016. pkt->flags |= AV_PKT_FLAG_KEY;
  1017. }
  1018. ast->frame_offset += get_duration(ast, pkt->size);
  1019. }
  1020. ast->remaining -= size;
  1021. if(!ast->remaining){
  1022. avi->stream_index= -1;
  1023. ast->packet_size= 0;
  1024. }
  1025. if(!avi->non_interleaved && pkt->pos >= 0 && ast->seek_pos > pkt->pos){
  1026. av_free_packet(pkt);
  1027. goto resync;
  1028. }
  1029. ast->seek_pos= 0;
  1030. if(!avi->non_interleaved && st->nb_index_entries>1 && avi->index_loaded>1){
  1031. int64_t dts= av_rescale_q(pkt->dts, st->time_base, AV_TIME_BASE_Q);
  1032. if(avi->dts_max - dts > 2*AV_TIME_BASE){
  1033. avi->non_interleaved= 1;
  1034. av_log(s, AV_LOG_INFO, "Switching to NI mode, due to poor interleaving\n");
  1035. }else if(avi->dts_max < dts)
  1036. avi->dts_max = dts;
  1037. }
  1038. return size;
  1039. }
  1040. if ((err = avi_sync(s, 0)) < 0)
  1041. return err;
  1042. goto resync;
  1043. }
  1044. /* XXX: We make the implicit supposition that the positions are sorted
  1045. for each stream. */
  1046. static int avi_read_idx1(AVFormatContext *s, int size)
  1047. {
  1048. AVIContext *avi = s->priv_data;
  1049. AVIOContext *pb = s->pb;
  1050. int nb_index_entries, i;
  1051. AVStream *st;
  1052. AVIStream *ast;
  1053. unsigned int index, tag, flags, pos, len, first_packet = 1;
  1054. unsigned last_pos= -1;
  1055. int64_t idx1_pos, first_packet_pos = 0, data_offset = 0;
  1056. nb_index_entries = size / 16;
  1057. if (nb_index_entries <= 0)
  1058. return -1;
  1059. idx1_pos = avio_tell(pb);
  1060. avio_seek(pb, avi->movi_list+4, SEEK_SET);
  1061. if (avi_sync(s, 1) == 0) {
  1062. first_packet_pos = avio_tell(pb) - 8;
  1063. }
  1064. avi->stream_index = -1;
  1065. avio_seek(pb, idx1_pos, SEEK_SET);
  1066. /* Read the entries and sort them in each stream component. */
  1067. for(i = 0; i < nb_index_entries; i++) {
  1068. tag = avio_rl32(pb);
  1069. flags = avio_rl32(pb);
  1070. pos = avio_rl32(pb);
  1071. len = avio_rl32(pb);
  1072. av_dlog(s, "%d: tag=0x%x flags=0x%x pos=0x%x len=%d/",
  1073. i, tag, flags, pos, len);
  1074. index = ((tag & 0xff) - '0') * 10;
  1075. index += ((tag >> 8) & 0xff) - '0';
  1076. if (index >= s->nb_streams)
  1077. continue;
  1078. st = s->streams[index];
  1079. ast = st->priv_data;
  1080. if(first_packet && first_packet_pos && len) {
  1081. data_offset = first_packet_pos - pos;
  1082. first_packet = 0;
  1083. }
  1084. pos += data_offset;
  1085. av_dlog(s, "%d cum_len=%"PRId64"\n", len, ast->cum_len);
  1086. if(url_feof(pb))
  1087. return -1;
  1088. if(last_pos == pos)
  1089. avi->non_interleaved= 1;
  1090. else if(len || !ast->sample_size)
  1091. av_add_index_entry(st, pos, ast->cum_len, len, 0, (flags&AVIIF_INDEX) ? AVINDEX_KEYFRAME : 0);
  1092. ast->cum_len += get_duration(ast, len);
  1093. last_pos= pos;
  1094. }
  1095. return 0;
  1096. }
  1097. static int guess_ni_flag(AVFormatContext *s){
  1098. int i;
  1099. int64_t last_start=0;
  1100. int64_t first_end= INT64_MAX;
  1101. int64_t oldpos= avio_tell(s->pb);
  1102. for(i=0; i<s->nb_streams; i++){
  1103. AVStream *st = s->streams[i];
  1104. int n= st->nb_index_entries;
  1105. unsigned int size;
  1106. if(n <= 0)
  1107. continue;
  1108. if(n >= 2){
  1109. int64_t pos= st->index_entries[0].pos;
  1110. avio_seek(s->pb, pos + 4, SEEK_SET);
  1111. size= avio_rl32(s->pb);
  1112. if(pos + size > st->index_entries[1].pos)
  1113. last_start= INT64_MAX;
  1114. }
  1115. if(st->index_entries[0].pos > last_start)
  1116. last_start= st->index_entries[0].pos;
  1117. if(st->index_entries[n-1].pos < first_end)
  1118. first_end= st->index_entries[n-1].pos;
  1119. }
  1120. avio_seek(s->pb, oldpos, SEEK_SET);
  1121. return last_start > first_end;
  1122. }
  1123. static int avi_load_index(AVFormatContext *s)
  1124. {
  1125. AVIContext *avi = s->priv_data;
  1126. AVIOContext *pb = s->pb;
  1127. uint32_t tag, size;
  1128. int64_t pos= avio_tell(pb);
  1129. int ret = -1;
  1130. if (avio_seek(pb, avi->movi_end, SEEK_SET) < 0)
  1131. goto the_end; // maybe truncated file
  1132. av_dlog(s, "movi_end=0x%"PRIx64"\n", avi->movi_end);
  1133. for(;;) {
  1134. if (url_feof(pb))
  1135. break;
  1136. tag = avio_rl32(pb);
  1137. size = avio_rl32(pb);
  1138. av_dlog(s, "tag=%c%c%c%c size=0x%x\n",
  1139. tag & 0xff,
  1140. (tag >> 8) & 0xff,
  1141. (tag >> 16) & 0xff,
  1142. (tag >> 24) & 0xff,
  1143. size);
  1144. if (tag == MKTAG('i', 'd', 'x', '1') &&
  1145. avi_read_idx1(s, size) >= 0) {
  1146. avi->index_loaded=2;
  1147. ret = 0;
  1148. break;
  1149. }
  1150. size += (size & 1);
  1151. if (avio_skip(pb, size) < 0)
  1152. break; // something is wrong here
  1153. }
  1154. the_end:
  1155. avio_seek(pb, pos, SEEK_SET);
  1156. return ret;
  1157. }
  1158. static void seek_subtitle(AVStream *st, AVStream *st2, int64_t timestamp)
  1159. {
  1160. AVIStream *ast2 = st2->priv_data;
  1161. int64_t ts2 = av_rescale_q(timestamp, st->time_base, st2->time_base);
  1162. av_free_packet(&ast2->sub_pkt);
  1163. if (avformat_seek_file(ast2->sub_ctx, 0, INT64_MIN, ts2, ts2, 0) >= 0 ||
  1164. avformat_seek_file(ast2->sub_ctx, 0, ts2, ts2, INT64_MAX, 0) >= 0)
  1165. av_read_packet(ast2->sub_ctx, &ast2->sub_pkt);
  1166. }
  1167. static int avi_read_seek(AVFormatContext *s, int stream_index, int64_t timestamp, int flags)
  1168. {
  1169. AVIContext *avi = s->priv_data;
  1170. AVStream *st;
  1171. int i, index;
  1172. int64_t pos, pos_min;
  1173. AVIStream *ast;
  1174. if (!avi->index_loaded) {
  1175. /* we only load the index on demand */
  1176. avi_load_index(s);
  1177. avi->index_loaded |= 1;
  1178. }
  1179. assert(stream_index>= 0);
  1180. st = s->streams[stream_index];
  1181. ast= st->priv_data;
  1182. index= av_index_search_timestamp(st, timestamp * FFMAX(ast->sample_size, 1), flags);
  1183. if(index<0)
  1184. return -1;
  1185. /* find the position */
  1186. pos = st->index_entries[index].pos;
  1187. timestamp = st->index_entries[index].timestamp / FFMAX(ast->sample_size, 1);
  1188. // av_log(s, AV_LOG_DEBUG, "XX %"PRId64" %d %"PRId64"\n", timestamp, index, st->index_entries[index].timestamp);
  1189. if (CONFIG_DV_DEMUXER && avi->dv_demux) {
  1190. /* One and only one real stream for DV in AVI, and it has video */
  1191. /* offsets. Calling with other stream indexes should have failed */
  1192. /* the av_index_search_timestamp call above. */
  1193. assert(stream_index == 0);
  1194. if(avio_seek(s->pb, pos, SEEK_SET) < 0)
  1195. return -1;
  1196. /* Feed the DV video stream version of the timestamp to the */
  1197. /* DV demux so it can synthesize correct timestamps. */
  1198. dv_offset_reset(avi->dv_demux, timestamp);
  1199. avi->stream_index= -1;
  1200. return 0;
  1201. }
  1202. pos_min= pos;
  1203. for(i = 0; i < s->nb_streams; i++) {
  1204. AVStream *st2 = s->streams[i];
  1205. AVIStream *ast2 = st2->priv_data;
  1206. ast2->packet_size=
  1207. ast2->remaining= 0;
  1208. if (ast2->sub_ctx) {
  1209. seek_subtitle(st, st2, timestamp);
  1210. continue;
  1211. }
  1212. if (st2->nb_index_entries <= 0)
  1213. continue;
  1214. // assert(st2->codec->block_align);
  1215. assert((int64_t)st2->time_base.num*ast2->rate == (int64_t)st2->time_base.den*ast2->scale);
  1216. index = av_index_search_timestamp(
  1217. st2,
  1218. av_rescale_q(timestamp, st->time_base, st2->time_base) * FFMAX(ast2->sample_size, 1),
  1219. flags | AVSEEK_FLAG_BACKWARD | (st2->codec->codec_type != AVMEDIA_TYPE_VIDEO ? AVSEEK_FLAG_ANY : 0));
  1220. if(index<0)
  1221. index=0;
  1222. ast2->seek_pos= st2->index_entries[index].pos;
  1223. pos_min= FFMIN(pos_min,ast2->seek_pos);
  1224. }
  1225. for(i = 0; i < s->nb_streams; i++) {
  1226. AVStream *st2 = s->streams[i];
  1227. AVIStream *ast2 = st2->priv_data;
  1228. if (ast2->sub_ctx || st2->nb_index_entries <= 0)
  1229. continue;
  1230. index = av_index_search_timestamp(
  1231. st2,
  1232. av_rescale_q(timestamp, st->time_base, st2->time_base) * FFMAX(ast2->sample_size, 1),
  1233. flags | AVSEEK_FLAG_BACKWARD | (st2->codec->codec_type != AVMEDIA_TYPE_VIDEO ? AVSEEK_FLAG_ANY : 0));
  1234. if(index<0)
  1235. index=0;
  1236. while(!avi->non_interleaved && index>0 && st2->index_entries[index-1].pos >= pos_min)
  1237. index--;
  1238. ast2->frame_offset = st2->index_entries[index].timestamp;
  1239. }
  1240. /* do the seek */
  1241. if (avio_seek(s->pb, pos_min, SEEK_SET) < 0)
  1242. return -1;
  1243. avi->stream_index= -1;
  1244. avi->dts_max= INT_MIN;
  1245. return 0;
  1246. }
  1247. static int avi_read_close(AVFormatContext *s)
  1248. {
  1249. int i;
  1250. AVIContext *avi = s->priv_data;
  1251. for(i=0;i<s->nb_streams;i++) {
  1252. AVStream *st = s->streams[i];
  1253. AVIStream *ast = st->priv_data;
  1254. if (ast) {
  1255. if (ast->sub_ctx) {
  1256. av_freep(&ast->sub_ctx->pb);
  1257. av_close_input_file(ast->sub_ctx);
  1258. }
  1259. av_free(ast->sub_buffer);
  1260. av_free_packet(&ast->sub_pkt);
  1261. }
  1262. }
  1263. av_free(avi->dv_demux);
  1264. return 0;
  1265. }
  1266. static int avi_probe(AVProbeData *p)
  1267. {
  1268. int i;
  1269. /* check file header */
  1270. for(i=0; avi_headers[i][0]; i++)
  1271. if(!memcmp(p->buf , avi_headers[i] , 4) &&
  1272. !memcmp(p->buf+8, avi_headers[i]+4, 4))
  1273. return AVPROBE_SCORE_MAX;
  1274. return 0;
  1275. }
  1276. AVInputFormat ff_avi_demuxer = {
  1277. .name = "avi",
  1278. .long_name = NULL_IF_CONFIG_SMALL("AVI format"),
  1279. .priv_data_size = sizeof(AVIContext),
  1280. .read_probe = avi_probe,
  1281. .read_header = avi_read_header,
  1282. .read_packet = avi_read_packet,
  1283. .read_close = avi_read_close,
  1284. .read_seek = avi_read_seek,
  1285. .priv_class = &demuxer_class,
  1286. };