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