src/main.c

   1 /*
   2  * Copyright (C) 2013 nu774
   3  * For conditions of distribution and use, see copyright notice in COPYING
   4  */
   5 #if HAVE_CONFIG_H
   6 #  include "config.h"
   7 #endif
   8 #if HAVE_STDINT_H
   9 #  include <stdint.h>
  10 #endif
  11 #if HAVE_INTTYPES_H
  12 #  include <inttypes.h>
  13 #elif defined(_MSC_VER)
  14 #  define SCNd64 "I64d"
  15 #endif
  16 #include <stdio.h>
  17 #include <stdlib.h>
  18 #include <string.h>
  19 #include <ctype.h>
  20 #include <locale.h>
  21 #include <errno.h>
  22 #include <sys/stat.h>
  23 #include <getopt.h>
  24 #if HAVE_UNISTD_H
  25 #include <unistd.h>
  26 #endif
  27 #if HAVE_SIGACTION
  28 #include <signal.h>
  29 #endif
  30 #ifdef _WIN32
  31 #include <io.h>
  32 #define WIN32_LEAN_AND_MEAN
  33 #include <windows.h>
  34 #endif
  35 #include "compat.h"
  36 #include "wav_reader.h"
  37 #include "caf_reader.h"
  38 #include "aacenc.h"
  39 #include "m4af.h"
  40 #include "progress.h"
  41 #include "version.h"
  42 #include "metadata.h"
  43
  44 #define PROGNAME "fdkaac"
  45
  46 static volatile int g_interrupted = 0;
  47
  48 #if HAVE_SIGACTION
  49 static void signal_handler(int signum)
  50 {
  51     g_interrupted = 1;
  52 }
  53 static void handle_signals(void)
  54 {
  55     int i, sigs[] = { SIGINT, SIGHUP, SIGTERM };
  56     for (i = 0; i < sizeof(sigs)/sizeof(sigs[0]); ++i) {
  57         struct sigaction sa = { 0 };
  58         sa.sa_handler = signal_handler;
  59         sa.sa_flags |= SA_RESTART;
  60         sigaction(sigs[i], &sa, 0);
  61     }
  62 }
  63 #elif defined(_WIN32)
  64 static BOOL WINAPI signal_handler(DWORD type)
  65 {
  66     g_interrupted = 1;
  67     return TRUE;
  68 }
  69
  70 static void handle_signals(void)
  71 {
  72     SetConsoleCtrlHandler(signal_handler, TRUE);
  73 }
  74 #else
  75 static void handle_signals(void)
  76 {
  77 }
  78 #endif
  79
  80 static
  81 int read_callback(void *cookie, void *data, uint32_t size)
  82 {
  83     size_t rc = fread(data, 1, size, (FILE*)cookie);
  84     return ferror((FILE*)cookie) ? -1 : (int)rc;
  85 }
  86
  87 static
  88 int write_callback(void *cookie, const void *data, uint32_t size)
  89 {
  90     size_t rc = fwrite(data, 1, size, (FILE*)cookie);
  91     return ferror((FILE*)cookie) ? -1 : (int)rc;
  92 }
  93
  94 static
  95 int seek_callback(void *cookie, int64_t off, int whence)
  96 {
  97     return fseeko((FILE*)cookie, off, whence);
  98 }
  99
 100 static
 101 int64_t tell_callback(void *cookie)
 102 {
 103     return ftello((FILE*)cookie);
 104 }
 105
 106 static
 107 void usage(void)
 108 {
 109     printf(
 110 PROGNAME " %s\n"
 111 "Usage: " PROGNAME " [options] input_file\n"
 112 "Options:\n"
 113 " -h, --help                    Print this help message\n"
 114 " -p, --profile <n>             Profile (audio object type)\n"
 115 "                                 2: MPEG-4 AAC LC (default)\n"
 116 "                                 5: MPEG-4 HE-AAC (SBR)\n"
 117 "                                29: MPEG-4 HE-AAC v2 (SBR+PS)\n"
 118 "                                23: MPEG-4 AAC LD\n"
 119 "                                39: MPEG-4 AAC ELD\n"
 120 "                               129: MPEG-2 AAC LC\n"
 121 "                               132: MPEG-2 HE-AAC (SBR)\n"
 122 "                               156: MPEG-2 HE-AAC v2 (SBR+PS)\n"
 123 " -b, --bitrate <n>             Bitrate in bits per seconds (for CBR)\n"
 124 " -m, --bitrate-mode <n>        Bitrate configuration\n"
 125 "                                 0: CBR (default)\n"
 126 "                                 1-5: VBR\n"
 127 "                               (VBR mode is not officially supported, and\n"
 128 "                                works only on a certain combination of\n"
 129 "                                parameter settings, sample rate, and\n"
 130 "                                channel configuration)\n"
 131 " -w, --bandwidth <n>           Frequency bandwidth in Hz (AAC LC only)\n"
 132 " -a, --afterburner <n>         Afterburner\n"
 133 "                                 0: Off\n"
 134 "                                 1: On(default)\n"
 135 " -L, --lowdelay-sbr <-1|0|1>   Configure SBR activity on AAC ELD\n"
 136 "                                -1: Use ELD SBR auto configurator\n"
 137 "                                 0: Disable SBR on ELD (default)\n"
 138 "                                 1: Enable SBR on ELD\n"
 139 " -s, --sbr-ratio <0|1|2>       Controls activation of downsampled SBR\n"
 140 "                                 0: Use lib default (default)\n"
 141 "                                 1: downsampled SBR (default for ELD+SBR)\n"
 142 "                                 2: dual-rate SBR (default for HE-AAC)\n"
 143 " -f, --transport-format <n>    Transport format\n"
 144 "                                 0: RAW (default, muxed into M4A)\n"
 145 "                                 1: ADIF\n"
 146 "                                 2: ADTS\n"
 147 "                                 6: LATM MCP=1\n"
 148 "                                 7: LATM MCP=0\n"
 149 "                                10: LOAS/LATM (LATM within LOAS)\n"
 150 " -C, --adts-crc-check          Add CRC protection on ADTS header\n"
 151 " -h, --header-period <n>       StreamMuxConfig/PCE repetition period in\n"
 152 "                               transport layer\n"
 153 "\n"
 154 " -o <filename>                 Output filename\n"
 155 " -G, --gapless-mode <n>        Encoder delay signaling for gapless playback\n"
 156 "                                 0: iTunSMPB (default)\n"
 157 "                                 1: ISO standard (edts + sgpd)\n"
 158 "                                 2: Both\n"
 159 " --include-sbr-delay           Count SBR decoder delay in encoder delay\n"
 160 "                               This is not iTunes compatible, but is default\n"
 161 "                               behavior of FDK library.\n"
 162 " -I, --ignorelength            Ignore length of WAV header\n"
 163 " -S, --silent                  Don't print progress messages\n"
 164 " --moov-before-mdat            Place moov box before mdat box on m4a output\n"
 165 "\n"
 166 "Options for raw (headerless) input:\n"
 167 " -R, --raw                     Treat input as raw (by default WAV is\n"
 168 "                               assumed)\n"
 169 " --raw-channels <n>            Number of channels (default: 2)\n"
 170 " --raw-rate     <n>            Sample rate (default: 44100)\n"
 171 " --raw-format   <spec>         Sample format, default is \"S16L\".\n"
 172 "                               Spec is as follows:\n"
 173 "                                1st char: S(igned)|U(nsigned)|F(loat)\n"
 174 "                                2nd part: bits per channel\n"
 175 "                                Last char: L(ittle)|B(ig)\n"
 176 "                               Last char can be omitted, in which case L is\n"
 177 "                               assumed. Spec is case insensitive, therefore\n"
 178 "                               \"u16b\" is same as \"U16B\".\n"
 179 "\n"
 180 "Tagging options:\n"
 181 " --title <string>\n"
 182 " --artist <string>\n"
 183 " --album <string>\n"
 184 " --genre <string>\n"
 185 " --date <string>\n"
 186 " --composer <string>\n"
 187 " --grouping <string>\n"
 188 " --comment <string>\n"
 189 " --album-artist <string>\n"
 190 " --track <number[/total]>\n"
 191 " --disk <number[/total]>\n"
 192 " --tempo <n>\n"
 193 " --tag <fcc>:<value>          Set iTunes predefined tag with four char code.\n"
 194 " --tag-from-file <fcc>:<filename>\n"
 195 "                              Same as above, but value is read from file.\n"
 196 " --long-tag <name>:<value>    Set arbitrary tag as iTunes custom metadata.\n"
 197 " --tag-from-json <filename[?dot_notation]>\n"
 198 "                              Read tags from JSON. By default, tags are\n"
 199 "                              assumed to be direct children of the root\n"
 200 "                              object(dictionary).\n"
 201 "                              Optionally, position of the dictionary\n"
 202 "                              that contains tags can be specified with\n"
 203 "                              dotted notation.\n"
 204 "                              Example:\n"
 205 "                                --tag-from-json /path/to/json?format.tags\n"
 206     , fdkaac_version);
 207 }
 208
 209 typedef struct aacenc_param_ex_t {
 210     AACENC_PARAMS
 211
 212     char *input_filename;
 213     FILE *input_fp;
 214     char *output_filename;
 215     FILE *output_fp;
 216     unsigned gapless_mode;
 217     unsigned include_sbr_delay;
 218     unsigned ignore_length;
 219     int silent;
 220     int moov_before_mdat;
 221
 222     int is_raw;
 223     unsigned raw_channels;
 224     unsigned raw_rate;
 225     const char *raw_format;
 226
 227     aacenc_tag_store_t tags;
 228     aacenc_tag_store_t source_tags;
 229     aacenc_translate_generic_text_tag_ctx_t source_tag_ctx;
 230
 231     char *json_filename;
 232 } aacenc_param_ex_t;
 233
 234 static
 235 int parse_options(int argc, char **argv, aacenc_param_ex_t *params)
 236 {
 237     int ch;
 238     int n;
 239
 240 #define OPT_INCLUDE_SBR_DELAY    M4AF_FOURCC('s','d','l','y')
 241 #define OPT_MOOV_BEFORE_MDAT     M4AF_FOURCC('m','o','o','v')
 242 #define OPT_RAW_CHANNELS         M4AF_FOURCC('r','c','h','n')
 243 #define OPT_RAW_RATE             M4AF_FOURCC('r','r','a','t')
 244 #define OPT_RAW_FORMAT           M4AF_FOURCC('r','f','m','t')
 245 #define OPT_SHORT_TAG            M4AF_FOURCC('s','t','a','g')
 246 #define OPT_SHORT_TAG_FILE       M4AF_FOURCC('s','t','g','f')
 247 #define OPT_LONG_TAG             M4AF_FOURCC('l','t','a','g')
 248 #define OPT_TAG_FROM_JSON        M4AF_FOURCC('t','f','j','s')
 249
 250     static struct option long_options[] = {
 251         { "help",             no_argument,       0, 'h' },
 252         { "profile",          required_argument, 0, 'p' },
 253         { "bitrate",          required_argument, 0, 'b' },
 254         { "bitrate-mode",     required_argument, 0, 'm' },
 255         { "bandwidth",        required_argument, 0, 'w' },
 256         { "afterburner",      required_argument, 0, 'a' },
 257         { "lowdelay-sbr",     required_argument, 0, 'L' },
 258         { "sbr-ratio",        required_argument, 0, 's' },
 259         { "transport-format", required_argument, 0, 'f' },
 260         { "adts-crc-check",   no_argument,       0, 'C' },
 261         { "header-period",    required_argument, 0, 'P' },
 262
 263         { "gapless-mode",     required_argument, 0, 'G' },
 264         { "include-sbr-delay", no_argument,      0, OPT_INCLUDE_SBR_DELAY  },
 265         { "ignorelength",     no_argument,       0, 'I' },
 266         { "silent",           no_argument,       0, 'S' },
 267         { "moov-before-mdat", no_argument,       0, OPT_MOOV_BEFORE_MDAT   },
 268
 269         { "raw",              no_argument,       0, 'R' },
 270         { "raw-channels",     required_argument, 0, OPT_RAW_CHANNELS       },
 271         { "raw-rate",         required_argument, 0, OPT_RAW_RATE           },
 272         { "raw-format",       required_argument, 0, OPT_RAW_FORMAT         },
 273
 274         { "title",            required_argument, 0, M4AF_TAG_TITLE         },
 275         { "artist",           required_argument, 0, M4AF_TAG_ARTIST        },
 276         { "album",            required_argument, 0, M4AF_TAG_ALBUM         },
 277         { "genre",            required_argument, 0, M4AF_TAG_GENRE         },
 278         { "date",             required_argument, 0, M4AF_TAG_DATE          },
 279         { "composer",         required_argument, 0, M4AF_TAG_COMPOSER      },
 280         { "grouping",         required_argument, 0, M4AF_TAG_GROUPING      },
 281         { "comment",          required_argument, 0, M4AF_TAG_COMMENT       },
 282         { "album-artist",     required_argument, 0, M4AF_TAG_ALBUM_ARTIST  },
 283         { "track",            required_argument, 0, M4AF_TAG_TRACK         },
 284         { "disk",             required_argument, 0, M4AF_TAG_DISK          },
 285         { "tempo",            required_argument, 0, M4AF_TAG_TEMPO         },
 286         { "tag",              required_argument, 0, OPT_SHORT_TAG          },
 287         { "tag-from-file",    required_argument, 0, OPT_SHORT_TAG_FILE     },
 288         { "long-tag",         required_argument, 0, OPT_LONG_TAG           },
 289         { "tag-from-json",    required_argument, 0, OPT_TAG_FROM_JSON      },
 290         { 0,                  0,                 0, 0                      },
 291     };
 292     params->afterburner = 1;
 293
 294     aacenc_getmainargs(&argc, &argv);
 295     while ((ch = getopt_long(argc, argv, "hp:b:m:w:a:Ls:f:CP:G:Io:SR",
 296                              long_options, 0)) != EOF) {
 297         switch (ch) {
 298         case 'h':
 299             return usage(), -1;
 300         case 'p':
 301             if (sscanf(optarg, "%u", &n) != 1) {
 302                 fprintf(stderr, "invalid arg for profile\n");
 303                 return -1;
 304             }
 305             params->profile = n;
 306             break;
 307         case 'b':
 308             if (sscanf(optarg, "%u", &n) != 1) {
 309                 fprintf(stderr, "invalid arg for bitrate\n");
 310                 return -1;
 311             }
 312             params->bitrate = n;
 313             break;
 314         case 'm':
 315             if (sscanf(optarg, "%u", &n) != 1 || n > 5) {
 316                 fprintf(stderr, "invalid arg for bitrate-mode\n");
 317                 return -1;
 318             }
 319             params->bitrate_mode = n;
 320             break;
 321         case 'w':
 322             if (sscanf(optarg, "%u", &n) != 1) {
 323                 fprintf(stderr, "invalid arg for bandwidth\n");
 324                 return -1;
 325             }
 326             params->bandwidth = n;
 327             break;
 328         case 'a':
 329             if (sscanf(optarg, "%u", &n) != 1 || n > 1) {
 330                 fprintf(stderr, "invalid arg for afterburner\n");
 331                 return -1;
 332             }
 333             params->afterburner = n;
 334             break;
 335         case 'L':
 336             if (sscanf(optarg, "%d", &n) != 1 || n < -1 || n > 1) {
 337                 fprintf(stderr, "invalid arg for lowdelay-sbr\n");
 338                 return -1;
 339             }
 340             params->lowdelay_sbr = n;
 341             break;
 342         case 's':
 343             if (sscanf(optarg, "%u", &n) != 1 || n > 2) {
 344                 fprintf(stderr, "invalid arg for sbr-ratio\n");
 345                 return -1;
 346             }
 347             params->sbr_ratio = n;
 348             break;
 349         case 'f':
 350             if (sscanf(optarg, "%u", &n) != 1) {
 351                 fprintf(stderr, "invalid arg for transport-format\n");
 352                 return -1;
 353             }
 354             params->transport_format = n;
 355             break;
 356         case 'C':
 357             params->adts_crc_check = 1;
 358             break;
 359         case 'P':
 360             if (sscanf(optarg, "%u", &n) != 1) {
 361                 fprintf(stderr, "invalid arg for header-period\n");
 362                 return -1;
 363             }
 364             params->header_period = n;
 365             break;
 366         case 'o':
 367             params->output_filename = optarg;
 368             break;
 369         case 'G':
 370             if (sscanf(optarg, "%u", &n) != 1 || n > 2) {
 371                 fprintf(stderr, "invalid arg for gapless-mode\n");
 372                 return -1;
 373             }
 374             params->gapless_mode = n;
 375             break;
 376         case OPT_INCLUDE_SBR_DELAY:
 377             params->include_sbr_delay = 1;
 378             break;
 379         case 'I':
 380             params->ignore_length = 1;
 381             break;
 382         case 'S':
 383             params->silent = 1;
 384             break;
 385         case OPT_MOOV_BEFORE_MDAT:
 386             params->moov_before_mdat = 1;
 387             break;
 388         case 'R':
 389             params->is_raw = 1;
 390             break;
 391         case OPT_RAW_CHANNELS:
 392             if (sscanf(optarg, "%u", &n) != 1) {
 393                 fprintf(stderr, "invalid arg for raw-channels\n");
 394                 return -1;
 395             }
 396             params->raw_channels = n;
 397             break;
 398         case OPT_RAW_RATE:
 399             if (sscanf(optarg, "%u", &n) != 1) {
 400                 fprintf(stderr, "invalid arg for raw-rate\n");
 401                 return -1;
 402             }
 403             params->raw_rate = n;
 404             break;
 405         case OPT_RAW_FORMAT:
 406             params->raw_format = optarg;
 407             break;
 408         case M4AF_TAG_TITLE:
 409         case M4AF_TAG_ARTIST:
 410         case M4AF_TAG_ALBUM:
 411         case M4AF_TAG_GENRE:
 412         case M4AF_TAG_DATE:
 413         case M4AF_TAG_COMPOSER:
 414         case M4AF_TAG_GROUPING:
 415         case M4AF_TAG_COMMENT:
 416         case M4AF_TAG_ALBUM_ARTIST:
 417         case M4AF_TAG_TRACK:
 418         case M4AF_TAG_DISK:
 419         case M4AF_TAG_TEMPO:
 420             aacenc_add_tag_to_store(&params->tags, ch, 0, optarg,
 421                                     strlen(optarg), 0);
 422             break;
 423         case OPT_SHORT_TAG:
 424         case OPT_SHORT_TAG_FILE:
 425         case OPT_LONG_TAG:
 426             {
 427                 char *val;
 428                 size_t klen;
 429                 unsigned fcc = M4AF_FOURCC('-','-','-','-');
 430
 431                 if ((val = strchr(optarg, ':')) == 0) {
 432                     fprintf(stderr, "invalid arg for tag\n");
 433                     return -1;
 434                 }
 435                 *val++ = '\0';
 436                 if (ch == OPT_SHORT_TAG || ch == OPT_SHORT_TAG_FILE) {
 437                     /*
 438                      * take care of U+00A9(COPYRIGHT SIGN).
 439                      * 1) if length of fcc is 3, we prepend '\xa9'.
 440                      * 2) U+00A9 becomes "\xc2\xa9" in UTF-8. Therefore
 441                      *    we remove first '\xc2'.
 442                      */
 443                     if (optarg[0] == '\xc2')
 444                         ++optarg;
 445                     if ((klen = strlen(optarg))== 3)
 446                         fcc = 0xa9;
 447                     else if (klen != 4) {
 448                         fprintf(stderr, "invalid arg for tag\n");
 449                         return -1;
 450                     }
 451                     for (; *optarg; ++optarg)
 452                         fcc = ((fcc << 8) | (*optarg & 0xff));
 453                 }
 454                 aacenc_add_tag_to_store(&params->tags, fcc, optarg,
 455                                         val, strlen(val),
 456                                         ch == OPT_SHORT_TAG_FILE);
 457             }
 458             break;
 459         case OPT_TAG_FROM_JSON:
 460             params->json_filename = optarg;
 461             break;
 462         default:
 463             return usage(), -1;
 464         }
 465     }
 466     if (argc == optind)
 467         return usage(), -1;
 468
 469     if (!params->bitrate && !params->bitrate_mode) {
 470         fprintf(stderr, "bitrate or bitrate-mode is mandatory\n");
 471         return -1;
 472     }
 473     if (params->output_filename && !strcmp(params->output_filename, "-") &&
 474         !params->transport_format) {
 475         fprintf(stderr, "stdout streaming is not available on M4A output\n");
 476         return -1;
 477     }
 478     if (params->bitrate && params->bitrate < 10000)
 479         params->bitrate *= 1000;
 480
 481     if (params->is_raw) {
 482         if (!params->raw_channels)
 483             params->raw_channels = 2;
 484         if (!params->raw_rate)
 485             params->raw_rate = 44100;
 486         if (!params->raw_format)
 487             params->raw_format = "S16L";
 488     }
 489     params->input_filename = argv[optind];
 490     return 0;
 491 };
 492
 493 static
 494 int write_sample(FILE *ofp, m4af_ctx_t *m4af, aacenc_frame_t *frame)
 495 {
 496     if (!m4af) {
 497         fwrite(frame->data, 1, frame->size, ofp);
 498         if (ferror(ofp)) {
 499             fprintf(stderr, "ERROR: fwrite(): %s\n", strerror(errno));
 500             return -1;
 501         }
 502     } else if (m4af_write_sample(m4af, 0, frame->data, frame->size, 0) < 0) {
 503         fprintf(stderr, "ERROR: failed to write m4a sample\n");
 504         return -1;
 505     }
 506     return 0;
 507 }
 508
 509 static
 510 int encode(aacenc_param_ex_t *params, pcm_reader_t *reader,
 511            HANDLE_AACENCODER encoder, uint32_t frame_length,
 512            m4af_ctx_t *m4af)
 513 {
 514     int16_t *ibuf = 0, *ip;
 515     aacenc_frame_t obuf[2] = {{ 0 }}, *obp;
 516     unsigned flip = 0;
 517     int nread = 1;
 518     int rc = -1;
 519     int remaining, consumed;
 520     int frames_written = 0, encoded = 0;
 521     aacenc_progress_t progress = { 0 };
 522     const pcm_sample_description_t *fmt = pcm_get_format(reader);
 523
 524     ibuf = malloc(frame_length * fmt->bytes_per_frame);
 525     aacenc_progress_init(&progress, pcm_get_length(reader), fmt->sample_rate);
 526
 527     for (;;) {
 528         /*
 529          * Since we delay the write, we cannot just exit loop when interrupted.
 530          * Instead, we regard it as EOF.
 531          */
 532         if (g_interrupted)
 533             nread = 0;
 534         if (nread > 0) {
 535             if ((nread = pcm_read_frames(reader, ibuf, frame_length)) < 0) {
 536                 fprintf(stderr, "ERROR: read failed\n");
 537                 goto END;
 538             }
 539             if (!params->silent)
 540                 aacenc_progress_update(&progress, pcm_get_position(reader),
 541                                        fmt->sample_rate * 2);
 542         }
 543         ip = ibuf;
 544         remaining = nread;
 545         do {
 546             obp = &obuf[flip];
 547             consumed = aac_encode_frame(encoder, fmt, ip, remaining, obp);
 548             if (consumed < 0) goto END;
 549             if (consumed == 0 && obp->size == 0) goto DONE;
 550             if (obp->size == 0) break;
 551
 552             remaining -= consumed;
 553             ip += consumed * fmt->channels_per_frame;
 554             flip ^= 1;
 555             /*
 556              * As we pad 1 frame at beginning and ending by our extrapolator,
 557              * we want to drop them.
 558              * We delay output by 1 frame by double buffering, and discard
 559              * second frame and final frame from the encoder.
 560              * Since sbr_header is included in the first frame (in case of
 561              * SBR), we cannot discard first frame. So we pick second instead.
 562              */
 563             ++encoded;
 564             if (encoded == 1 || encoded == 3)
 565                 continue;
 566             obp = &obuf[flip];
 567             if (write_sample(params->output_fp, m4af, obp) < 0)
 568                 goto END;
 569             ++frames_written;
 570         } while (remaining > 0);
 571     }
 572 DONE:
 573     if (!params->silent)
 574         aacenc_progress_finish(&progress, pcm_get_position(reader));
 575     rc = frames_written;
 576 END:
 577     if (ibuf) free(ibuf);
 578     if (obuf[0].data) free(obuf[0].data);
 579     if (obuf[1].data) free(obuf[1].data);
 580     return rc;
 581 }
 582
 583 static
 584 void put_tool_tag(m4af_ctx_t *m4af, const aacenc_param_ex_t *params,
 585                   HANDLE_AACENCODER encoder)
 586 {
 587     char tool_info[256];
 588     char *p = tool_info;
 589     LIB_INFO lib_info;
 590
 591     p += sprintf(p, PROGNAME " %s, ", fdkaac_version);
 592     aacenc_get_lib_info(&lib_info);
 593     p += sprintf(p, "libfdk-aac %s, ", lib_info.versionStr);
 594     if (params->bitrate_mode)
 595         sprintf(p, "VBR mode %d", params->bitrate_mode);
 596     else
 597         sprintf(p, "CBR %dkbps",
 598                 aacEncoder_GetParam(encoder, AACENC_BITRATE) / 1000);
 599
 600     m4af_add_itmf_string_tag(m4af, M4AF_TAG_TOOL, tool_info);
 601 }
 602
 603 static
 604 int finalize_m4a(m4af_ctx_t *m4af, const aacenc_param_ex_t *params,
 605                  HANDLE_AACENCODER encoder)
 606 {
 607     unsigned i;
 608     aacenc_tag_entry_t *tag;
 609
 610     tag = params->source_tags.tag_table;
 611     for (i = 0; i < params->source_tags.tag_count; ++i, ++tag)
 612         aacenc_write_tag_entry(m4af, tag);
 613
 614     if (params->json_filename)
 615         aacenc_write_tags_from_json(m4af, params->json_filename);
 616
 617     tag = params->tags.tag_table;
 618     for (i = 0; i < params->tags.tag_count; ++i, ++tag)
 619         aacenc_write_tag_entry(m4af, tag);
 620
 621     put_tool_tag(m4af, params, encoder);
 622
 623     if (m4af_finalize(m4af, params->moov_before_mdat) < 0) {
 624         fprintf(stderr, "ERROR: failed to finalize m4a\n");
 625         return -1;
 626     }
 627     return 0;
 628 }
 629
 630 static
 631 char *generate_output_filename(const char *filename, const char *ext)
 632 {
 633     char *p = 0;
 634     size_t ext_len = strlen(ext);
 635
 636     if (strcmp(filename, "-") == 0) {
 637         p = malloc(ext_len + 6);
 638         sprintf(p, "stdin%s", ext);
 639     } else {
 640         const char *base = aacenc_basename(filename);
 641         size_t ilen = strlen(base);
 642         const char *ext_org = strrchr(base, '.');
 643         if (ext_org) ilen = ext_org - base;
 644         p = malloc(ilen + ext_len + 1);
 645         sprintf(p, "%.*s%s", (int)ilen, base, ext);
 646     }
 647     return p;
 648 }
 649
 650 static
 651 int parse_raw_spec(const char *spec, pcm_sample_description_t *desc)
 652 {
 653     unsigned bits;
 654     unsigned char c_type, c_endian = 'L';
 655     int type;
 656
 657     if (sscanf(spec, "%c%u%c", &c_type, &bits, &c_endian) < 2)
 658         return -1;
 659     c_type = toupper(c_type);
 660     c_endian = toupper(c_endian);
 661
 662     if (c_type == 'S')
 663         type = 1;
 664     else if (c_type == 'U')
 665         type = 2;
 666     else if (c_type == 'F')
 667         type = 4;
 668     else
 669         return -1;
 670
 671     if (c_endian == 'B')
 672         type |= 8;
 673     else if (c_endian != 'L')
 674         return -1;
 675
 676     if (c_type == 'F' && bits != 32 && bits != 64)
 677         return -1;
 678     if (c_type != 'F' && (bits < 8 || bits > 32))
 679         return -1;
 680
 681     desc->sample_type = type;
 682     desc->bits_per_channel = bits;
 683     return 0;
 684 }
 685
 686 static pcm_io_vtbl_t pcm_io_vtbl = {
 687     read_callback, seek_callback, tell_callback
 688 };
 689 static pcm_io_vtbl_t pcm_io_vtbl_noseek = { read_callback, 0, tell_callback };
 690
 691 static
 692 pcm_reader_t *open_input(aacenc_param_ex_t *params)
 693 {
 694     pcm_io_context_t io = { 0 };
 695     pcm_reader_t *reader = 0;
 696     struct stat stb = { 0 };
 697
 698     if ((params->input_fp = aacenc_fopen(params->input_filename, "rb")) == 0) {
 699         aacenc_fprintf(stderr, "ERROR: %s: %s\n", params->input_filename,
 700                        strerror(errno));
 701         goto END;
 702     }
 703     io.cookie = params->input_fp;
 704     if (fstat(fileno(params->input_fp), &stb) == 0
 705             && (stb.st_mode & S_IFMT) == S_IFREG)
 706         io.vtbl = &pcm_io_vtbl;
 707     else
 708         io.vtbl = &pcm_io_vtbl_noseek;
 709
 710     if (params->is_raw) {
 711         int bytes_per_channel;
 712         pcm_sample_description_t desc = { 0 };
 713         if (parse_raw_spec(params->raw_format, &desc) < 0) {
 714             fprintf(stderr, "ERROR: invalid raw-format spec\n");
 715             goto END;
 716         }
 717         desc.sample_rate = params->raw_rate;
 718         desc.channels_per_frame = params->raw_channels;
 719         bytes_per_channel = (desc.bits_per_channel + 7) / 8;
 720         desc.bytes_per_frame = params->raw_channels * bytes_per_channel;
 721         if ((reader = raw_open(&io, &desc)) == 0) {
 722             fprintf(stderr, "ERROR: failed to open raw input\n");
 723             goto END;
 724         }
 725     } else {
 726         int c;
 727         ungetc(c = getc(params->input_fp), params->input_fp);
 728
 729         switch (c) {
 730         case 'R':
 731             if ((reader = wav_open(&io, params->ignore_length)) == 0) {
 732                 fprintf(stderr, "ERROR: broken / unsupported input file\n");
 733                 goto END;
 734             }
 735             break;
 736         case 'c':
 737             params->source_tag_ctx.add = aacenc_add_tag_entry_to_store;
 738             params->source_tag_ctx.add_ctx = &params->source_tags;
 739             if ((reader = caf_open(&io,
 740                                    aacenc_translate_generic_text_tag,
 741                                    &params->source_tag_ctx)) == 0) {
 742                 fprintf(stderr, "ERROR: broken / unsupported input file\n");
 743                 goto END;
 744             }
 745             break;
 746         default:
 747             fprintf(stderr, "ERROR: unsupported input file\n");
 748             goto END;
 749         }
 750     }
 751     if ((reader = pcm_open_sint16_converter(reader)) != 0)
 752         reader = extrapolater_open(reader);
 753     return reader;
 754 END:
 755     return 0;
 756 }
 757
 758 int main(int argc, char **argv)
 759 {
 760     static m4af_io_callbacks_t m4af_io = {
 761         read_callback, write_callback, seek_callback, tell_callback
 762     };
 763     aacenc_param_ex_t params = { 0 };
 764
 765     int result = 2;
 766     char *output_filename = 0;
 767     pcm_reader_t *reader = 0;
 768     HANDLE_AACENCODER encoder = 0;
 769     AACENC_InfoStruct aacinfo = { 0 };
 770     LIB_INFO lib_info = { 0 };
 771     m4af_ctx_t *m4af = 0;
 772     const pcm_sample_description_t *sample_format;
 773     int downsampled_timescale = 0;
 774     int frame_count = 0;
 775     int sbr_mode = 0;
 776     unsigned scale_shift = 0;
 777
 778     setlocale(LC_CTYPE, "");
 779     setbuf(stderr, 0);
 780
 781     if (parse_options(argc, argv, &params) < 0)
 782         return 1;
 783
 784     if ((reader = open_input(&params)) == 0)
 785         goto END;
 786
 787     sample_format = pcm_get_format(reader);
 788
 789     /*
 790      * We use explicit/hierarchical signaling for LOAS.
 791      * Other than that, we request implicit signaling to FDK library, then
 792      * append explicit/backward-compatible signaling to ASC in case of MP4FF.
 793      *
 794      * Explicit/backward-compatible signaling of SBR is the most recommended
 795      * way in MPEG4 part3 spec, and seems the only way supported by iTunes.
 796      * Since FDK library does not support it, we have to do it on our side.
 797      */
 798     sbr_mode = aacenc_is_sbr_active((aacenc_param_t*)&params);
 799     if (sbr_mode && !aacenc_is_sbr_ratio_available()) {
 800         fprintf(stderr, "WARNING: Only dual-rate SBR is available "
 801                         "for this version\n");
 802         params.sbr_ratio = 2;
 803     }
 804     scale_shift = aacenc_is_dual_rate_sbr((aacenc_param_t*)&params);
 805     params.sbr_signaling = (params.transport_format == TT_MP4_LOAS) ? 2 : 0;
 806     if (sbr_mode && !scale_shift)
 807         params.sbr_signaling = 2;
 808
 809     if (aacenc_init(&encoder, (aacenc_param_t*)&params, sample_format,
 810                     &aacinfo) < 0)
 811         goto END;
 812
 813     if (!params.output_filename) {
 814         const char *ext = params.transport_format ? ".aac" : ".m4a";
 815         output_filename = generate_output_filename(params.input_filename, ext);
 816         params.output_filename = output_filename;
 817     }
 818
 819     if ((params.output_fp = aacenc_fopen(params.output_filename, "wb+")) == 0) {
 820         aacenc_fprintf(stderr, "ERROR: %s: %s\n", params.output_filename,
 821                        strerror(errno));
 822         goto END;
 823     }
 824     handle_signals();
 825
 826     if (!params.transport_format) {
 827         uint32_t scale;
 828         uint8_t mp4asc[32];
 829         uint32_t ascsize = sizeof(mp4asc);
 830         unsigned framelen = aacinfo.frameLength;
 831         scale = sample_format->sample_rate >> scale_shift;
 832         if ((m4af = m4af_create(M4AF_CODEC_MP4A, scale, &m4af_io,
 833                                 params.output_fp)) < 0)
 834             goto END;
 835         aacenc_mp4asc((aacenc_param_t*)&params, aacinfo.confBuf,
 836                       aacinfo.confSize, mp4asc, &ascsize);
 837         m4af_set_decoder_specific_info(m4af, 0, mp4asc, ascsize);
 838         m4af_set_fixed_frame_duration(m4af, 0,
 839                                       framelen >> scale_shift);
 840         m4af_set_vbr_mode(m4af, 0, params.bitrate_mode);
 841         m4af_set_priming_mode(m4af, params.gapless_mode + 1);
 842         m4af_begin_write(m4af);
 843     }
 844     if (scale_shift && (aacinfo.encoderDelay & 1)) {
 845         /*
 846          * Since odd delay cannot be exactly expressed in downsampled scale,
 847          * we push one zero frame to the encoder here, to make delay even
 848          */
 849         int16_t zero[8] = { 0 };
 850         aacenc_frame_t frame = { 0 };
 851         aac_encode_frame(encoder, sample_format, zero, 1, &frame);
 852         free(frame.data);
 853     }
 854     frame_count = encode(&params, reader, encoder, aacinfo.frameLength, m4af);
 855     if (frame_count < 0)
 856         goto END;
 857     if (m4af) {
 858         uint32_t delay = aacinfo.encoderDelay;
 859         uint32_t padding;
 860         int64_t frames_read = pcm_get_position(reader);
 861
 862         if (sbr_mode && params.profile != AOT_ER_AAC_ELD &&
 863             !params.include_sbr_delay)
 864             delay -= 481 << scale_shift;
 865         if (scale_shift && (delay & 1))
 866             ++delay;
 867         padding = frame_count * aacinfo.frameLength - frames_read - delay;
 868         m4af_set_priming(m4af, 0, delay >> scale_shift, padding >> scale_shift);
 869         if (finalize_m4a(m4af, &params, encoder) < 0)
 870             goto END;
 871     }
 872     result = 0;
 873 END:
 874     if (reader) pcm_teardown(&reader);
 875     if (params.input_fp) fclose(params.input_fp);
 876     if (m4af) m4af_teardown(&m4af);
 877     if (params.output_fp) fclose(params.output_fp);
 878     if (encoder) aacEncClose(&encoder);
 879     if (output_filename) free(output_filename);
 880     if (params.tags.tag_table)
 881         aacenc_free_tag_store(&params.tags);
 882     if (params.source_tags.tag_table)
 883         aacenc_free_tag_store(&params.source_tags);
 884
 885     return result;
 886 }