tools/ismindex.c

   1 /*
   2  * Copyright (c) 2012 Martin Storsjo
   3  *
   4  * This file is part of FFmpeg.
   5  *
   6  * FFmpeg is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (at your option) any later version.
  10  *
  11  * FFmpeg is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with FFmpeg; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19  */
  20
  21 /*
  22  * To create a simple file for smooth streaming:
  23  * ffmpeg <normal input/transcoding options> -movflags frag_keyframe foo.ismv
  24  * ismindex -n foo foo.ismv
  25  * This step creates foo.ism and foo.ismc that is required by IIS for
  26  * serving it.
  27  *
  28  * With -ismf, it also creates foo.ismf, which maps fragment names to
  29  * start-end offsets in the ismv, for use in your own streaming server.
  30  *
  31  * By adding -path-prefix path/, the produced foo.ism will refer to the
  32  * files foo.ismv as "path/foo.ismv" - the prefix for the generated ismc
  33  * file can be set with the -ismc-prefix option similarly.
  34  *
  35  * To pre-split files for serving as static files by a web server without
  36  * any extra server support, create the ismv file as above, and split it:
  37  * ismindex -split foo.ismv
  38  * This step creates a file Manifest and directories QualityLevel(...),
  39  * that can be read directly by a smooth streaming player.
  40  *
  41  * The -output dir option can be used to request that output files
  42  * (both .ism/.ismc, or Manifest/QualityLevels* when splitting)
  43  * should be written to this directory instead of in the current directory.
  44  * (The directory itself isn't created if it doesn't already exist.)
  45  */
  46
  47 #include <stdio.h>
  48 #include <string.h>
  49
  50 #include "libavformat/avformat.h"
  51 #include "libavformat/isom.h"
  52 #include "libavformat/os_support.h"
  53 #include "libavutil/intreadwrite.h"
  54 #include "libavutil/mathematics.h"
  55
  56 static int usage(const char *argv0, int ret)
  57 {
  58     fprintf(stderr, "%s [-split] [-ismf] [-n basename] [-path-prefix prefix] "
  59                     "[-ismc-prefix prefix] [-output dir] file1 [file2] ...\n", argv0);
  60     return ret;
  61 }
  62
  63 struct MoofOffset {
  64     int64_t time;
  65     int64_t offset;
  66     int64_t duration;
  67 };
  68
  69 struct Track {
  70     const char *name;
  71     int64_t duration;
  72     int bitrate;
  73     int track_id;
  74     int is_audio, is_video;
  75     int width, height;
  76     int chunks;
  77     int sample_rate, channels;
  78     uint8_t *codec_private;
  79     int codec_private_size;
  80     struct MoofOffset *offsets;
  81     int timescale;
  82     const char *fourcc;
  83     int blocksize;
  84     int tag;
  85 };
  86
  87 struct Tracks {
  88     int nb_tracks;
  89     int64_t duration;
  90     struct Track **tracks;
  91     int video_track, audio_track;
  92     int nb_video_tracks, nb_audio_tracks;
  93 };
  94
  95 static int expect_tag(int32_t got_tag, int32_t expected_tag) {
  96     if (got_tag != expected_tag) {
  97         char got_tag_str[4], expected_tag_str[4];
  98         AV_WB32(got_tag_str, got_tag);
  99         AV_WB32(expected_tag_str, expected_tag);
 100         fprintf(stderr, "wanted tag %.4s, got %.4s\n", expected_tag_str,
 101                 got_tag_str);
 102         return -1;
 103     }
 104     return 0;
 105 }
 106
 107 static int copy_tag(AVIOContext *in, AVIOContext *out, int32_t tag_name)
 108 {
 109     int32_t size, tag;
 110
 111     size = avio_rb32(in);
 112     tag  = avio_rb32(in);
 113     avio_wb32(out, size);
 114     avio_wb32(out, tag);
 115     if (expect_tag(tag, tag_name) != 0)
 116         return -1;
 117     size -= 8;
 118     while (size > 0) {
 119         char buf[1024];
 120         int len = FFMIN(sizeof(buf), size);
 121         int got;
 122         if ((got = avio_read(in, buf, len)) != len) {
 123             fprintf(stderr, "short read, wanted %d, got %d\n", len, got);
 124             break;
 125         }
 126         avio_write(out, buf, len);
 127         size -= len;
 128     }
 129     return 0;
 130 }
 131
 132 static int skip_tag(AVIOContext *in, int32_t tag_name)
 133 {
 134     int64_t pos = avio_tell(in);
 135     int32_t size, tag;
 136
 137     size = avio_rb32(in);
 138     tag  = avio_rb32(in);
 139     if (expect_tag(tag, tag_name) != 0)
 140         return -1;
 141     avio_seek(in, pos + size, SEEK_SET);
 142     return 0;
 143 }
 144
 145 static int write_fragment(const char *filename, AVIOContext *in)
 146 {
 147     AVIOContext *out = NULL;
 148     int ret;
 149
 150     if ((ret = avio_open2(&out, filename, AVIO_FLAG_WRITE, NULL, NULL)) < 0) {
 151         char errbuf[100];
 152         av_strerror(ret, errbuf, sizeof(errbuf));
 153         fprintf(stderr, "Unable to open %s: %s\n", filename, errbuf);
 154         return ret;
 155     }
 156     ret = copy_tag(in, out, MKBETAG('m', 'o', 'o', 'f'));
 157     if (!ret)
 158         ret = copy_tag(in, out, MKBETAG('m', 'd', 'a', 't'));
 159
 160     avio_flush(out);
 161     avio_close(out);
 162
 163     return ret;
 164 }
 165
 166 static int skip_fragment(AVIOContext *in)
 167 {
 168     int ret;
 169     ret = skip_tag(in, MKBETAG('m', 'o', 'o', 'f'));
 170     if (!ret)
 171         ret = skip_tag(in, MKBETAG('m', 'd', 'a', 't'));
 172     return ret;
 173 }
 174
 175 static int write_fragments(struct Tracks *tracks, int start_index,
 176                            AVIOContext *in, const char *basename,
 177                            int split, int ismf, const char* output_prefix)
 178 {
 179     char dirname[2048], filename[2048], idxname[2048];
 180     int i, j, ret = 0, fragment_ret;
 181     FILE* out = NULL;
 182
 183     if (ismf) {
 184         snprintf(idxname, sizeof(idxname), "%s%s.ismf", output_prefix, basename);
 185         out = fopen(idxname, "w");
 186         if (!out) {
 187             ret = AVERROR(errno);
 188             perror(idxname);
 189             goto fail;
 190         }
 191     }
 192     for (i = start_index; i < tracks->nb_tracks; i++) {
 193         struct Track *track = tracks->tracks[i];
 194         const char *type    = track->is_video ? "video" : "audio";
 195         snprintf(dirname, sizeof(dirname), "%sQualityLevels(%d)", output_prefix, track->bitrate);
 196         if (split) {
 197             if (mkdir(dirname, 0777) == -1 && errno != EEXIST) {
 198                 ret = AVERROR(errno);
 199                 perror(dirname);
 200                 goto fail;
 201             }
 202         }
 203         for (j = 0; j < track->chunks; j++) {
 204             snprintf(filename, sizeof(filename), "%s/Fragments(%s=%"PRId64")",
 205                      dirname, type, track->offsets[j].time);
 206             avio_seek(in, track->offsets[j].offset, SEEK_SET);
 207             if (ismf)
 208                 fprintf(out, "%s %"PRId64, filename, avio_tell(in));
 209             if (split)
 210                 fragment_ret = write_fragment(filename, in);
 211             else
 212                 fragment_ret = skip_fragment(in);
 213             if (ismf)
 214                 fprintf(out, " %"PRId64"\n", avio_tell(in));
 215             if (fragment_ret != 0) {
 216                 fprintf(stderr, "failed fragment %d in track %d (%s)\n", j,
 217                         track->track_id, track->name);
 218                 ret = fragment_ret;
 219             }
 220         }
 221     }
 222 fail:
 223     if (out)
 224         fclose(out);
 225     return ret;
 226 }
 227
 228 static int64_t read_trun_duration(AVIOContext *in, int default_duration,
 229                                   int64_t end)
 230 {
 231     int64_t dts = 0;
 232     int64_t pos;
 233     int flags, i;
 234     int entries;
 235     int64_t first_pts = 0;
 236     int64_t max_pts = 0;
 237     avio_r8(in); /* version */
 238     flags = avio_rb24(in);
 239     if (default_duration <= 0 && !(flags & MOV_TRUN_SAMPLE_DURATION)) {
 240         fprintf(stderr, "No sample duration in trun flags\n");
 241         return -1;
 242     }
 243     entries = avio_rb32(in);
 244
 245     if (flags & MOV_TRUN_DATA_OFFSET)        avio_rb32(in);
 246     if (flags & MOV_TRUN_FIRST_SAMPLE_FLAGS) avio_rb32(in);
 247
 248     pos = avio_tell(in);
 249     for (i = 0; i < entries && pos < end; i++) {
 250         int sample_duration = default_duration;
 251         int64_t pts = dts;
 252         if (flags & MOV_TRUN_SAMPLE_DURATION) sample_duration = avio_rb32(in);
 253         if (flags & MOV_TRUN_SAMPLE_SIZE)     avio_rb32(in);
 254         if (flags & MOV_TRUN_SAMPLE_FLAGS)    avio_rb32(in);
 255         if (flags & MOV_TRUN_SAMPLE_CTS)      pts += avio_rb32(in);
 256         if (sample_duration < 0) {
 257             fprintf(stderr, "Negative sample duration %d\n", sample_duration);
 258             return -1;
 259         }
 260         if (i == 0)
 261             first_pts = pts;
 262         max_pts = FFMAX(max_pts, pts + sample_duration);
 263         dts += sample_duration;
 264         pos = avio_tell(in);
 265     }
 266
 267     return max_pts - first_pts;
 268 }
 269
 270 static int64_t read_moof_duration(AVIOContext *in, int64_t offset)
 271 {
 272     int64_t ret = -1;
 273     int32_t moof_size, size, tag;
 274     int64_t pos = 0;
 275     int default_duration = 0;
 276
 277     avio_seek(in, offset, SEEK_SET);
 278     moof_size = avio_rb32(in);
 279     tag  = avio_rb32(in);
 280     if (expect_tag(tag, MKBETAG('m', 'o', 'o', 'f')) != 0)
 281         goto fail;
 282     while (pos < offset + moof_size) {
 283         pos = avio_tell(in);
 284         size = avio_rb32(in);
 285         tag  = avio_rb32(in);
 286         if (tag == MKBETAG('t', 'r', 'a', 'f')) {
 287             int64_t traf_pos = pos;
 288             int64_t traf_size = size;
 289             while (pos < traf_pos + traf_size) {
 290                 pos = avio_tell(in);
 291                 size = avio_rb32(in);
 292                 tag  = avio_rb32(in);
 293                 if (tag == MKBETAG('t', 'f', 'h', 'd')) {
 294                     int flags = 0;
 295                     avio_r8(in); /* version */
 296                     flags = avio_rb24(in);
 297                     avio_rb32(in); /* track_id */
 298                     if (flags & MOV_TFHD_BASE_DATA_OFFSET)
 299                         avio_rb64(in);
 300                     if (flags & MOV_TFHD_STSD_ID)
 301                         avio_rb32(in);
 302                     if (flags & MOV_TFHD_DEFAULT_DURATION)
 303                         default_duration = avio_rb32(in);
 304                 }
 305                 if (tag == MKBETAG('t', 'r', 'u', 'n')) {
 306                     return read_trun_duration(in, default_duration,
 307                                               pos + size);
 308                 }
 309                 avio_seek(in, pos + size, SEEK_SET);
 310             }
 311             fprintf(stderr, "Couldn't find trun\n");
 312             goto fail;
 313         }
 314         avio_seek(in, pos + size, SEEK_SET);
 315     }
 316     fprintf(stderr, "Couldn't find traf\n");
 317
 318 fail:
 319     return ret;
 320 }
 321
 322 static int read_tfra(struct Tracks *tracks, int start_index, AVIOContext *f)
 323 {
 324     int ret = AVERROR_EOF, track_id;
 325     int version, fieldlength, i, j;
 326     int64_t pos   = avio_tell(f);
 327     uint32_t size = avio_rb32(f);
 328     struct Track *track = NULL;
 329
 330     if (avio_rb32(f) != MKBETAG('t', 'f', 'r', 'a'))
 331         goto fail;
 332     version = avio_r8(f);
 333     avio_rb24(f);
 334     track_id = avio_rb32(f); /* track id */
 335     for (i = start_index; i < tracks->nb_tracks && !track; i++)
 336         if (tracks->tracks[i]->track_id == track_id)
 337             track = tracks->tracks[i];
 338     if (!track) {
 339         /* Ok, continue parsing the next atom */
 340         ret = 0;
 341         goto fail;
 342     }
 343     fieldlength = avio_rb32(f);
 344     track->chunks  = avio_rb32(f);
 345     track->offsets = av_calloc(track->chunks, sizeof(*track->offsets));
 346     if (!track->offsets) {
 347         track->chunks = 0;
 348         ret = AVERROR(ENOMEM);
 349         goto fail;
 350     }
 351     // The duration here is always the difference between consecutive
 352     // start times.
 353     for (i = 0; i < track->chunks; i++) {
 354         if (version == 1) {
 355             track->offsets[i].time   = avio_rb64(f);
 356             track->offsets[i].offset = avio_rb64(f);
 357         } else {
 358             track->offsets[i].time   = avio_rb32(f);
 359             track->offsets[i].offset = avio_rb32(f);
 360         }
 361         for (j = 0; j < ((fieldlength >> 4) & 3) + 1; j++)
 362             avio_r8(f);
 363         for (j = 0; j < ((fieldlength >> 2) & 3) + 1; j++)
 364             avio_r8(f);
 365         for (j = 0; j < ((fieldlength >> 0) & 3) + 1; j++)
 366             avio_r8(f);
 367         if (i > 0)
 368             track->offsets[i - 1].duration = track->offsets[i].time -
 369                                              track->offsets[i - 1].time;
 370     }
 371     if (track->chunks > 0) {
 372         track->offsets[track->chunks - 1].duration = track->offsets[0].time +
 373                                                      track->duration -
 374                                                      track->offsets[track->chunks - 1].time;
 375     }
 376     // Now try to read the actual durations from the trun sample data.
 377     for (i = 0; i < track->chunks; i++) {
 378         int64_t duration = read_moof_duration(f, track->offsets[i].offset);
 379         if (duration > 0 && llabs(duration - track->offsets[i].duration) > 3) {
 380             // 3 allows for integer duration to drift a few units,
 381             // e.g., for 1/3 durations
 382             track->offsets[i].duration = duration;
 383         }
 384     }
 385     if (track->chunks > 0) {
 386         if (track->offsets[track->chunks - 1].duration <= 0) {
 387             fprintf(stderr, "Calculated last chunk duration for track %d "
 388                     "was non-positive (%"PRId64"), probably due to missing "
 389                     "fragments ", track->track_id,
 390                     track->offsets[track->chunks - 1].duration);
 391             if (track->chunks > 1) {
 392                 track->offsets[track->chunks - 1].duration =
 393                     track->offsets[track->chunks - 2].duration;
 394             } else {
 395                 track->offsets[track->chunks - 1].duration = 1;
 396             }
 397             fprintf(stderr, "corrected to %"PRId64"\n",
 398                     track->offsets[track->chunks - 1].duration);
 399             track->duration = track->offsets[track->chunks - 1].time +
 400                               track->offsets[track->chunks - 1].duration -
 401                               track->offsets[0].time;
 402             fprintf(stderr, "Track duration corrected to %"PRId64"\n",
 403                     track->duration);
 404         }
 405     }
 406     ret = 0;
 407
 408 fail:
 409     avio_seek(f, pos + size, SEEK_SET);
 410     return ret;
 411 }
 412
 413 static int read_mfra(struct Tracks *tracks, int start_index,
 414                      const char *file, int split, int ismf,
 415                      const char *basename, const char* output_prefix)
 416 {
 417     int err = 0;
 418     const char* err_str = "";
 419     AVIOContext *f = NULL;
 420     int32_t mfra_size;
 421
 422     if ((err = avio_open2(&f, file, AVIO_FLAG_READ, NULL, NULL)) < 0)
 423         goto fail;
 424     avio_seek(f, avio_size(f) - 4, SEEK_SET);
 425     mfra_size = avio_rb32(f);
 426     avio_seek(f, -mfra_size, SEEK_CUR);
 427     if (avio_rb32(f) != mfra_size) {
 428         err = AVERROR_INVALIDDATA;
 429         err_str = "mfra size mismatch";
 430         goto fail;
 431     }
 432     if (avio_rb32(f) != MKBETAG('m', 'f', 'r', 'a')) {
 433         err = AVERROR_INVALIDDATA;
 434         err_str = "mfra tag mismatch";
 435         goto fail;
 436     }
 437     while (!read_tfra(tracks, start_index, f)) {
 438         /* Empty */
 439     }
 440
 441     if (split || ismf)
 442         err = write_fragments(tracks, start_index, f, basename, split, ismf,
 443                               output_prefix);
 444     err_str = "error in write_fragments";
 445
 446 fail:
 447     if (f)
 448         avio_close(f);
 449     if (err)
 450         fprintf(stderr, "Unable to read the MFRA atom in %s (%s)\n", file, err_str);
 451     return err;
 452 }
 453
 454 static int get_private_data(struct Track *track, AVCodecParameters *codecpar)
 455 {
 456     track->codec_private_size = 0;
 457     track->codec_private      = av_mallocz(codecpar->extradata_size);
 458     if (!track->codec_private)
 459         return AVERROR(ENOMEM);
 460     track->codec_private_size = codecpar->extradata_size;
 461     memcpy(track->codec_private, codecpar->extradata, codecpar->extradata_size);
 462     return 0;
 463 }
 464
 465 static int get_video_private_data(struct Track *track, AVCodecParameters *codecpar)
 466 {
 467     AVIOContext *io = NULL;
 468     uint16_t sps_size, pps_size;
 469     int err;
 470
 471     if (codecpar->codec_id == AV_CODEC_ID_VC1)
 472         return get_private_data(track, codecpar);
 473
 474     if ((err = avio_open_dyn_buf(&io)) < 0)
 475         goto fail;
 476     err = AVERROR(EINVAL);
 477     if (codecpar->extradata_size < 11 || codecpar->extradata[0] != 1)
 478         goto fail;
 479     sps_size = AV_RB16(&codecpar->extradata[6]);
 480     if (11 + sps_size > codecpar->extradata_size)
 481         goto fail;
 482     avio_wb32(io, 0x00000001);
 483     avio_write(io, &codecpar->extradata[8], sps_size);
 484     pps_size = AV_RB16(&codecpar->extradata[9 + sps_size]);
 485     if (11 + sps_size + pps_size > codecpar->extradata_size)
 486         goto fail;
 487     avio_wb32(io, 0x00000001);
 488     avio_write(io, &codecpar->extradata[11 + sps_size], pps_size);
 489     err = 0;
 490
 491 fail:
 492     track->codec_private_size = avio_close_dyn_buf(io, &track->codec_private);
 493     return err;
 494 }
 495
 496 static int handle_file(struct Tracks *tracks, const char *file, int split,
 497                        int ismf, const char *basename,
 498                        const char* output_prefix)
 499 {
 500     AVFormatContext *ctx = NULL;
 501     int err = 0, i, orig_tracks = tracks->nb_tracks;
 502     char errbuf[50], *ptr;
 503     struct Track *track;
 504
 505     err = avformat_open_input(&ctx, file, NULL, NULL);
 506     if (err < 0) {
 507         av_strerror(err, errbuf, sizeof(errbuf));
 508         fprintf(stderr, "Unable to open %s: %s\n", file, errbuf);
 509         return 1;
 510     }
 511
 512     err = avformat_find_stream_info(ctx, NULL);
 513     if (err < 0) {
 514         av_strerror(err, errbuf, sizeof(errbuf));
 515         fprintf(stderr, "Unable to identify %s: %s\n", file, errbuf);
 516         goto fail;
 517     }
 518
 519     if (ctx->nb_streams < 1) {
 520         fprintf(stderr, "No streams found in %s\n", file);
 521         goto fail;
 522     }
 523
 524     for (i = 0; i < ctx->nb_streams; i++) {
 525         struct Track **temp;
 526         AVStream *st = ctx->streams[i];
 527
 528         if (st->codecpar->bit_rate == 0) {
 529             fprintf(stderr, "Skipping track %d in %s as it has zero bitrate\n",
 530                     st->id, file);
 531             continue;
 532         }
 533
 534         track = av_mallocz(sizeof(*track));
 535         if (!track) {
 536             err = AVERROR(ENOMEM);
 537             goto fail;
 538         }
 539         temp = av_realloc_array(tracks->tracks,
 540                                 tracks->nb_tracks + 1,
 541                                 sizeof(*tracks->tracks));
 542         if (!temp) {
 543             av_free(track);
 544             err = AVERROR(ENOMEM);
 545             goto fail;
 546         }
 547         tracks->tracks = temp;
 548         tracks->tracks[tracks->nb_tracks] = track;
 549
 550         track->name = file;
 551         if ((ptr = strrchr(file, '/')))
 552             track->name = ptr + 1;
 553
 554         track->bitrate   = st->codecpar->bit_rate;
 555         track->track_id  = st->id;
 556         track->timescale = st->time_base.den;
 557         track->duration  = st->duration;
 558         track->is_audio  = st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO;
 559         track->is_video  = st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO;
 560
 561         if (!track->is_audio && !track->is_video) {
 562             fprintf(stderr,
 563                     "Track %d in %s is neither video nor audio, skipping\n",
 564                     track->track_id, file);
 565             av_freep(&tracks->tracks[tracks->nb_tracks]);
 566             continue;
 567         }
 568
 569         tracks->duration = FFMAX(tracks->duration,
 570                                  av_rescale_rnd(track->duration, AV_TIME_BASE,
 571                                                 track->timescale, AV_ROUND_UP));
 572
 573         if (track->is_audio) {
 574             if (tracks->audio_track < 0)
 575                 tracks->audio_track = tracks->nb_tracks;
 576             tracks->nb_audio_tracks++;
 577             track->channels    = st->codecpar->ch_layout.nb_channels;
 578             track->sample_rate = st->codecpar->sample_rate;
 579             if (st->codecpar->codec_id == AV_CODEC_ID_AAC) {
 580                 track->fourcc    = "AACL";
 581                 track->tag       = 255;
 582                 track->blocksize = 4;
 583             } else if (st->codecpar->codec_id == AV_CODEC_ID_WMAPRO) {
 584                 track->fourcc    = "WMAP";
 585                 track->tag       = st->codecpar->codec_tag;
 586                 track->blocksize = st->codecpar->block_align;
 587             }
 588             get_private_data(track, st->codecpar);
 589         }
 590         if (track->is_video) {
 591             if (tracks->video_track < 0)
 592                 tracks->video_track = tracks->nb_tracks;
 593             tracks->nb_video_tracks++;
 594             track->width  = st->codecpar->width;
 595             track->height = st->codecpar->height;
 596             if (st->codecpar->codec_id == AV_CODEC_ID_H264)
 597                 track->fourcc = "H264";
 598             else if (st->codecpar->codec_id == AV_CODEC_ID_VC1)
 599                 track->fourcc = "WVC1";
 600             get_video_private_data(track, st->codecpar);
 601         }
 602
 603         tracks->nb_tracks++;
 604     }
 605
 606     avformat_close_input(&ctx);
 607
 608     err = read_mfra(tracks, orig_tracks, file, split, ismf, basename,
 609                     output_prefix);
 610
 611 fail:
 612     if (ctx)
 613         avformat_close_input(&ctx);
 614     return err;
 615 }
 616
 617 static void output_server_manifest(struct Tracks *tracks, const char *basename,
 618                                    const char *output_prefix,
 619                                    const char *path_prefix,
 620                                    const char *ismc_prefix)
 621 {
 622     char filename[1000];
 623     FILE *out;
 624     int i;
 625
 626     snprintf(filename, sizeof(filename), "%s%s.ism", output_prefix, basename);
 627     out = fopen(filename, "w");
 628     if (!out) {
 629         perror(filename);
 630         return;
 631     }
 632     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
 633     fprintf(out, "<smil xmlns=\"http://www.w3.org/2001/SMIL20/Language\">\n");
 634     fprintf(out, "\t<head>\n");
 635     fprintf(out, "\t\t<meta name=\"clientManifestRelativePath\" "
 636                  "content=\"%s%s.ismc\" />\n", ismc_prefix, basename);
 637     fprintf(out, "\t</head>\n");
 638     fprintf(out, "\t<body>\n");
 639     fprintf(out, "\t\t<switch>\n");
 640     for (i = 0; i < tracks->nb_tracks; i++) {
 641         struct Track *track = tracks->tracks[i];
 642         const char *type    = track->is_video ? "video" : "audio";
 643         fprintf(out, "\t\t\t<%s src=\"%s%s\" systemBitrate=\"%d\">\n",
 644                 type, path_prefix, track->name, track->bitrate);
 645         fprintf(out, "\t\t\t\t<param name=\"trackID\" value=\"%d\" "
 646                      "valueType=\"data\" />\n", track->track_id);
 647         fprintf(out, "\t\t\t</%s>\n", type);
 648     }
 649     fprintf(out, "\t\t</switch>\n");
 650     fprintf(out, "\t</body>\n");
 651     fprintf(out, "</smil>\n");
 652     fclose(out);
 653 }
 654
 655 static void print_track_chunks(FILE *out, struct Tracks *tracks, int main,
 656                                const char *type)
 657 {
 658     int i, j;
 659     int64_t pos = 0;
 660     struct Track *track = tracks->tracks[main];
 661     int should_print_time_mismatch = 1;
 662
 663     for (i = 0; i < track->chunks; i++) {
 664         for (j = main + 1; j < tracks->nb_tracks; j++) {
 665             if (tracks->tracks[j]->is_audio == track->is_audio) {
 666                 if (track->offsets[i].duration != tracks->tracks[j]->offsets[i].duration) {
 667                     fprintf(stderr, "Mismatched duration of %s chunk %d in %s (%d) and %s (%d)\n",
 668                             type, i, track->name, main, tracks->tracks[j]->name, j);
 669                     should_print_time_mismatch = 1;
 670                 }
 671                 if (track->offsets[i].time != tracks->tracks[j]->offsets[i].time) {
 672                     if (should_print_time_mismatch)
 673                         fprintf(stderr, "Mismatched (start) time of %s chunk %d in %s (%d) and %s (%d)\n",
 674                                 type, i, track->name, main, tracks->tracks[j]->name, j);
 675                     should_print_time_mismatch = 0;
 676                 }
 677             }
 678         }
 679         fprintf(out, "\t\t<c n=\"%d\" d=\"%"PRId64"\" ",
 680                 i, track->offsets[i].duration);
 681         if (pos != track->offsets[i].time) {
 682             fprintf(out, "t=\"%"PRId64"\" ", track->offsets[i].time);
 683             pos = track->offsets[i].time;
 684         }
 685         pos += track->offsets[i].duration;
 686         fprintf(out, "/>\n");
 687     }
 688 }
 689
 690 static void output_client_manifest(struct Tracks *tracks, const char *basename,
 691                                    const char *output_prefix, int split)
 692 {
 693     char filename[1000];
 694     FILE *out;
 695     int i, j;
 696
 697     if (split)
 698         snprintf(filename, sizeof(filename), "%sManifest", output_prefix);
 699     else
 700         snprintf(filename, sizeof(filename), "%s%s.ismc", output_prefix, basename);
 701     out = fopen(filename, "w");
 702     if (!out) {
 703         perror(filename);
 704         return;
 705     }
 706     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
 707     fprintf(out, "<SmoothStreamingMedia MajorVersion=\"2\" MinorVersion=\"0\" "
 708                  "Duration=\"%"PRId64 "\">\n", tracks->duration * 10);
 709     if (tracks->video_track >= 0) {
 710         struct Track *track = tracks->tracks[tracks->video_track];
 711         struct Track *first_track = track;
 712         int index = 0;
 713         fprintf(out,
 714                 "\t<StreamIndex Type=\"video\" QualityLevels=\"%d\" "
 715                 "Chunks=\"%d\" "
 716                 "Url=\"QualityLevels({bitrate})/Fragments(video={start time})\">\n",
 717                 tracks->nb_video_tracks, track->chunks);
 718         for (i = 0; i < tracks->nb_tracks; i++) {
 719             track = tracks->tracks[i];
 720             if (!track->is_video)
 721                 continue;
 722             fprintf(out,
 723                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
 724                     "FourCC=\"%s\" MaxWidth=\"%d\" MaxHeight=\"%d\" "
 725                     "CodecPrivateData=\"",
 726                     index, track->bitrate, track->fourcc, track->width, track->height);
 727             for (j = 0; j < track->codec_private_size; j++)
 728                 fprintf(out, "%02X", track->codec_private[j]);
 729             fprintf(out, "\" />\n");
 730             index++;
 731             if (track->chunks != first_track->chunks)
 732                 fprintf(stderr, "Mismatched number of video chunks in %s (id: %d, chunks %d) and %s (id: %d, chunks %d)\n",
 733                         track->name, track->track_id, track->chunks, first_track->name, first_track->track_id, first_track->chunks);
 734         }
 735         print_track_chunks(out, tracks, tracks->video_track, "video");
 736         fprintf(out, "\t</StreamIndex>\n");
 737     }
 738     if (tracks->audio_track >= 0) {
 739         struct Track *track = tracks->tracks[tracks->audio_track];
 740         struct Track *first_track = track;
 741         int index = 0;
 742         fprintf(out,
 743                 "\t<StreamIndex Type=\"audio\" QualityLevels=\"%d\" "
 744                 "Chunks=\"%d\" "
 745                 "Url=\"QualityLevels({bitrate})/Fragments(audio={start time})\">\n",
 746                 tracks->nb_audio_tracks, track->chunks);
 747         for (i = 0; i < tracks->nb_tracks; i++) {
 748             track = tracks->tracks[i];
 749             if (!track->is_audio)
 750                 continue;
 751             fprintf(out,
 752                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
 753                     "FourCC=\"%s\" SamplingRate=\"%d\" Channels=\"%d\" "
 754                     "BitsPerSample=\"16\" PacketSize=\"%d\" "
 755                     "AudioTag=\"%d\" CodecPrivateData=\"",
 756                     index, track->bitrate, track->fourcc, track->sample_rate,
 757                     track->channels, track->blocksize, track->tag);
 758             for (j = 0; j < track->codec_private_size; j++)
 759                 fprintf(out, "%02X", track->codec_private[j]);
 760             fprintf(out, "\" />\n");
 761             index++;
 762             if (track->chunks != first_track->chunks)
 763                 fprintf(stderr, "Mismatched number of audio chunks in %s and %s\n",
 764                         track->name, first_track->name);
 765         }
 766         print_track_chunks(out, tracks, tracks->audio_track, "audio");
 767         fprintf(out, "\t</StreamIndex>\n");
 768     }
 769     fprintf(out, "</SmoothStreamingMedia>\n");
 770     fclose(out);
 771 }
 772
 773 static void clean_tracks(struct Tracks *tracks)
 774 {
 775     int i;
 776     for (i = 0; i < tracks->nb_tracks; i++) {
 777         av_freep(&tracks->tracks[i]->codec_private);
 778         av_freep(&tracks->tracks[i]->offsets);
 779         av_freep(&tracks->tracks[i]);
 780     }
 781     av_freep(&tracks->tracks);
 782     tracks->nb_tracks = 0;
 783 }
 784
 785 int main(int argc, char **argv)
 786 {
 787     const char *basename = NULL;
 788     const char *path_prefix = "", *ismc_prefix = "";
 789     const char *output_prefix = "";
 790     char output_prefix_buf[2048];
 791     int split = 0, ismf = 0, i;
 792     struct Tracks tracks = { 0, .video_track = -1, .audio_track = -1 };
 793
 794     for (i = 1; i < argc; i++) {
 795         if (!strcmp(argv[i], "-n")) {
 796             basename = argv[i + 1];
 797             i++;
 798         } else if (!strcmp(argv[i], "-path-prefix")) {
 799             path_prefix = argv[i + 1];
 800             i++;
 801         } else if (!strcmp(argv[i], "-ismc-prefix")) {
 802             ismc_prefix = argv[i + 1];
 803             i++;
 804         } else if (!strcmp(argv[i], "-output")) {
 805             output_prefix = argv[i + 1];
 806             i++;
 807             if (output_prefix[strlen(output_prefix) - 1] != '/') {
 808                 snprintf(output_prefix_buf, sizeof(output_prefix_buf),
 809                          "%s/", output_prefix);
 810                 output_prefix = output_prefix_buf;
 811             }
 812         } else if (!strcmp(argv[i], "-split")) {
 813             split = 1;
 814         } else if (!strcmp(argv[i], "-ismf")) {
 815             ismf = 1;
 816         } else if (argv[i][0] == '-') {
 817             return usage(argv[0], 1);
 818         } else {
 819             if (!basename)
 820                 ismf = 0;
 821             if (handle_file(&tracks, argv[i], split, ismf,
 822                             basename, output_prefix))
 823                 return 1;
 824         }
 825     }
 826     if (!tracks.nb_tracks || (!basename && !split))
 827         return usage(argv[0], 1);
 828
 829     if (!split)
 830         output_server_manifest(&tracks, basename, output_prefix,
 831                                path_prefix, ismc_prefix);
 832     output_client_manifest(&tracks, basename, output_prefix, split);
 833
 834     clean_tracks(&tracks);
 835
 836     return 0;
 837 }