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