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 "pcm_reader.h"
  37 #include "aacenc.h"
  38 #include "m4af.h"
  39 #include "progress.h"
  40 #include "version.h"
  41 #include "metadata.h"
  42
  43 #define PROGNAME "fdkaac"
  44
  45 static volatile int g_interrupted = 0;
  46
  47 #if HAVE_SIGACTION
  48 static void signal_handler(int signum)
  49 {
  50     g_interrupted = 1;
  51 }
  52 static void handle_signals(void)
  53 {
  54     int i, sigs[] = { SIGINT, SIGHUP, SIGTERM };
  55     for (i = 0; i < sizeof(sigs)/sizeof(sigs[0]); ++i) {
  56         struct sigaction sa;
  57         memset(&sa, 0, sizeof sa);
  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 " -b, --bitrate <n>             Bitrate in bits per seconds (for CBR)\n"
 121 " -m, --bitrate-mode <n>        Bitrate configuration\n"
 122 "                                 0: CBR (default)\n"
 123 "                                 1-5: VBR\n"
 124 "                               (VBR mode is not officially supported, and\n"
 125 "                                works only on a certain combination of\n"
 126 "                                parameter settings, sample rate, and\n"
 127 "                                channel configuration)\n"
 128 " -w, --bandwidth <n>           Frequency bandwidth in Hz (AAC LC only)\n"
 129 " -a, --afterburner <n>         Afterburner\n"
 130 "                                 0: Off\n"
 131 "                                 1: On(default)\n"
 132 " -L, --lowdelay-sbr <-1|0|1>   Configure SBR activity on AAC ELD\n"
 133 "                                -1: Use ELD SBR auto configurator\n"
 134 "                                 0: Disable SBR on ELD (default)\n"
 135 "                                 1: Enable SBR on ELD\n"
 136 " -s, --sbr-ratio <0|1|2>       Controls activation of downsampled SBR\n"
 137 "                                 0: Use lib default (default)\n"
 138 "                                 1: downsampled SBR (default for ELD+SBR)\n"
 139 "                                 2: dual-rate SBR (default for HE-AAC)\n"
 140 " -f, --transport-format <n>    Transport format\n"
 141 "                                 0: RAW (default, muxed into M4A)\n"
 142 "                                 1: ADIF\n"
 143 "                                 2: ADTS\n"
 144 "                                 6: LATM MCP=1\n"
 145 "                                 7: LATM MCP=0\n"
 146 "                                10: LOAS/LATM (LATM within LOAS)\n"
 147 " -C, --adts-crc-check          Add CRC protection on ADTS header\n"
 148 " -h, --header-period <n>       StreamMuxConfig/PCE repetition period in\n"
 149 "                               transport layer\n"
 150 "\n"
 151 " -o <filename>                 Output filename\n"
 152 " -G, --gapless-mode <n>        Encoder delay signaling for gapless playback\n"
 153 "                                 0: iTunSMPB (default)\n"
 154 "                                 1: ISO standard (edts + sgpd)\n"
 155 "                                 2: Both\n"
 156 " --include-sbr-delay           Count SBR decoder delay in encoder delay\n"
 157 "                               This is not iTunes compatible, but is default\n"
 158 "                               behavior of FDK library.\n"
 159 " -I, --ignorelength            Ignore length of WAV header\n"
 160 " -S, --silent                  Don't print progress messages\n"
 161 " --moov-before-mdat            Place moov box before mdat box on m4a output\n"
 162 "\n"
 163 "Options for raw (headerless) input:\n"
 164 " -R, --raw                     Treat input as raw (by default WAV is\n"
 165 "                               assumed)\n"
 166 " --raw-channels <n>            Number of channels (default: 2)\n"
 167 " --raw-rate     <n>            Sample rate (default: 44100)\n"
 168 " --raw-format   <spec>         Sample format, default is \"S16L\".\n"
 169 "                               Spec is as follows:\n"
 170 "                                1st char: S(igned)|U(nsigned)|F(loat)\n"
 171 "                                2nd part: bits per channel\n"
 172 "                                Last char: L(ittle)|B(ig)\n"
 173 "                               Last char can be omitted, in which case L is\n"
 174 "                               assumed. Spec is case insensitive, therefore\n"
 175 "                               \"u16b\" is same as \"U16B\".\n"
 176 "\n"
 177 "Tagging options:\n"
 178 " --title <string>\n"
 179 " --artist <string>\n"
 180 " --album <string>\n"
 181 " --genre <string>\n"
 182 " --date <string>\n"
 183 " --composer <string>\n"
 184 " --grouping <string>\n"
 185 " --comment <string>\n"
 186 " --album-artist <string>\n"
 187 " --track <number[/total]>\n"
 188 " --disk <number[/total]>\n"
 189 " --tempo <n>\n"
 190 " --tag <fcc>:<value>          Set iTunes predefined tag with four char code.\n"
 191 " --tag-from-file <fcc>:<filename>\n"
 192 "                              Same as above, but value is read from file.\n"
 193 " --long-tag <name>:<value>    Set arbitrary tag as iTunes custom metadata.\n"
 194 " --tag-from-json <filename[?dot_notation]>\n"
 195 "                              Read tags from JSON. By default, tags are\n"
 196 "                              assumed to be direct children of the root\n"
 197 "                              object(dictionary).\n"
 198 "                              Optionally, position of the dictionary\n"
 199 "                              that contains tags can be specified with\n"
 200 "                              dotted notation.\n"
 201 "                              Example:\n"
 202 "                                --tag-from-json /path/to/json?format.tags\n"
 203     , fdkaac_version);
 204 }
 205
 206 typedef struct aacenc_param_ex_t {
 207     AACENC_PARAMS
 208
 209     char *input_filename;
 210     FILE *input_fp;
 211     char *output_filename;
 212     FILE *output_fp;
 213     unsigned gapless_mode;
 214     unsigned include_sbr_delay;
 215     unsigned ignore_length;
 216     int silent;
 217     int moov_before_mdat;
 218
 219     int is_raw;
 220     unsigned raw_channels;
 221     unsigned raw_rate;
 222     const char *raw_format;
 223
 224     aacenc_tag_store_t tags;
 225     aacenc_tag_store_t source_tags;
 226     aacenc_translate_generic_text_tag_ctx_t source_tag_ctx;
 227
 228     char *json_filename;
 229 } aacenc_param_ex_t;
 230
 231 static
 232 int parse_options(int argc, char **argv, aacenc_param_ex_t *params)
 233 {
 234     int ch;
 235     int n;
 236
 237 #define OPT_INCLUDE_SBR_DELAY    M4AF_FOURCC('s','d','l','y')
 238 #define OPT_MOOV_BEFORE_MDAT     M4AF_FOURCC('m','o','o','v')
 239 #define OPT_RAW_CHANNELS         M4AF_FOURCC('r','c','h','n')
 240 #define OPT_RAW_RATE             M4AF_FOURCC('r','r','a','t')
 241 #define OPT_RAW_FORMAT           M4AF_FOURCC('r','f','m','t')
 242 #define OPT_SHORT_TAG            M4AF_FOURCC('s','t','a','g')
 243 #define OPT_SHORT_TAG_FILE       M4AF_FOURCC('s','t','g','f')
 244 #define OPT_LONG_TAG             M4AF_FOURCC('l','t','a','g')
 245 #define OPT_TAG_FROM_JSON        M4AF_FOURCC('t','f','j','s')
 246
 247     static struct option long_options[] = {
 248         { "help",             no_argument,       0, 'h' },
 249         { "profile",          required_argument, 0, 'p' },
 250         { "bitrate",          required_argument, 0, 'b' },
 251         { "bitrate-mode",     required_argument, 0, 'm' },
 252         { "bandwidth",        required_argument, 0, 'w' },
 253         { "afterburner",      required_argument, 0, 'a' },
 254         { "lowdelay-sbr",     required_argument, 0, 'L' },
 255         { "sbr-ratio",        required_argument, 0, 's' },
 256         { "transport-format", required_argument, 0, 'f' },
 257         { "adts-crc-check",   no_argument,       0, 'C' },
 258         { "header-period",    required_argument, 0, 'P' },
 259
 260         { "gapless-mode",     required_argument, 0, 'G' },
 261         { "include-sbr-delay", no_argument,      0, OPT_INCLUDE_SBR_DELAY  },
 262         { "ignorelength",     no_argument,       0, 'I' },
 263         { "silent",           no_argument,       0, 'S' },
 264         { "moov-before-mdat", no_argument,       0, OPT_MOOV_BEFORE_MDAT   },
 265
 266         { "raw",              no_argument,       0, 'R' },
 267         { "raw-channels",     required_argument, 0, OPT_RAW_CHANNELS       },
 268         { "raw-rate",         required_argument, 0, OPT_RAW_RATE           },
 269         { "raw-format",       required_argument, 0, OPT_RAW_FORMAT         },
 270
 271         { "title",            required_argument, 0, M4AF_TAG_TITLE         },
 272         { "artist",           required_argument, 0, M4AF_TAG_ARTIST        },
 273         { "album",            required_argument, 0, M4AF_TAG_ALBUM         },
 274         { "genre",            required_argument, 0, M4AF_TAG_GENRE         },
 275         { "date",             required_argument, 0, M4AF_TAG_DATE          },
 276         { "composer",         required_argument, 0, M4AF_TAG_COMPOSER      },
 277         { "grouping",         required_argument, 0, M4AF_TAG_GROUPING      },
 278         { "comment",          required_argument, 0, M4AF_TAG_COMMENT       },
 279         { "album-artist",     required_argument, 0, M4AF_TAG_ALBUM_ARTIST  },
 280         { "track",            required_argument, 0, M4AF_TAG_TRACK         },
 281         { "disk",             required_argument, 0, M4AF_TAG_DISK          },
 282         { "tempo",            required_argument, 0, M4AF_TAG_TEMPO         },
 283         { "tag",              required_argument, 0, OPT_SHORT_TAG          },
 284         { "tag-from-file",    required_argument, 0, OPT_SHORT_TAG_FILE     },
 285         { "long-tag",         required_argument, 0, OPT_LONG_TAG           },
 286         { "tag-from-json",    required_argument, 0, OPT_TAG_FROM_JSON      },
 287         { 0,                  0,                 0, 0                      },
 288     };
 289     params->afterburner = 1;
 290
 291     aacenc_getmainargs(&argc, &argv);
 292     while ((ch = getopt_long(argc, argv, "hp:b:m:w:a:Ls:f:CP:G:Io:SR",
 293                              long_options, 0)) != EOF) {
 294         switch (ch) {
 295         case 'h':
 296             return usage(), -1;
 297         case 'p':
 298             if (sscanf(optarg, "%u", &n) != 1) {
 299                 fprintf(stderr, "invalid arg for profile\n");
 300                 return -1;
 301             }
 302             params->profile = n;
 303             break;
 304         case 'b':
 305             if (sscanf(optarg, "%u", &n) != 1) {
 306                 fprintf(stderr, "invalid arg for bitrate\n");
 307                 return -1;
 308             }
 309             params->bitrate = n;
 310             break;
 311         case 'm':
 312             if (sscanf(optarg, "%u", &n) != 1 || n > 5) {
 313                 fprintf(stderr, "invalid arg for bitrate-mode\n");
 314                 return -1;
 315             }
 316             params->bitrate_mode = n;
 317             break;
 318         case 'w':
 319             if (sscanf(optarg, "%u", &n) != 1) {
 320                 fprintf(stderr, "invalid arg for bandwidth\n");
 321                 return -1;
 322             }
 323             params->bandwidth = n;
 324             break;
 325         case 'a':
 326             if (sscanf(optarg, "%u", &n) != 1 || n > 1) {
 327                 fprintf(stderr, "invalid arg for afterburner\n");
 328                 return -1;
 329             }
 330             params->afterburner = n;
 331             break;
 332         case 'L':
 333             if (sscanf(optarg, "%d", &n) != 1 || n < -1 || n > 1) {
 334                 fprintf(stderr, "invalid arg for lowdelay-sbr\n");
 335                 return -1;
 336             }
 337             params->lowdelay_sbr = n;
 338             break;
 339         case 's':
 340             if (sscanf(optarg, "%u", &n) != 1 || n > 2) {
 341                 fprintf(stderr, "invalid arg for sbr-ratio\n");
 342                 return -1;
 343             }
 344             params->sbr_ratio = n;
 345             break;
 346         case 'f':
 347             if (sscanf(optarg, "%u", &n) != 1) {
 348                 fprintf(stderr, "invalid arg for transport-format\n");
 349                 return -1;
 350             }
 351             params->transport_format = n;
 352             break;
 353         case 'C':
 354             params->adts_crc_check = 1;
 355             break;
 356         case 'P':
 357             if (sscanf(optarg, "%u", &n) != 1) {
 358                 fprintf(stderr, "invalid arg for header-period\n");
 359                 return -1;
 360             }
 361             params->header_period = n;
 362             break;
 363         case 'o':
 364             params->output_filename = optarg;
 365             break;
 366         case 'G':
 367             if (sscanf(optarg, "%u", &n) != 1 || n > 2) {
 368                 fprintf(stderr, "invalid arg for gapless-mode\n");
 369                 return -1;
 370             }
 371             params->gapless_mode = n;
 372             break;
 373         case OPT_INCLUDE_SBR_DELAY:
 374             params->include_sbr_delay = 1;
 375             break;
 376         case 'I':
 377             params->ignore_length = 1;
 378             break;
 379         case 'S':
 380             params->silent = 1;
 381             break;
 382         case OPT_MOOV_BEFORE_MDAT:
 383             params->moov_before_mdat = 1;
 384             break;
 385         case 'R':
 386             params->is_raw = 1;
 387             break;
 388         case OPT_RAW_CHANNELS:
 389             if (sscanf(optarg, "%u", &n) != 1) {
 390                 fprintf(stderr, "invalid arg for raw-channels\n");
 391                 return -1;
 392             }
 393             params->raw_channels = n;
 394             break;
 395         case OPT_RAW_RATE:
 396             if (sscanf(optarg, "%u", &n) != 1) {
 397                 fprintf(stderr, "invalid arg for raw-rate\n");
 398                 return -1;
 399             }
 400             params->raw_rate = n;
 401             break;
 402         case OPT_RAW_FORMAT:
 403             params->raw_format = optarg;
 404             break;
 405         case M4AF_TAG_TITLE:
 406         case M4AF_TAG_ARTIST:
 407         case M4AF_TAG_ALBUM:
 408         case M4AF_TAG_GENRE:
 409         case M4AF_TAG_DATE:
 410         case M4AF_TAG_COMPOSER:
 411         case M4AF_TAG_GROUPING:
 412         case M4AF_TAG_COMMENT:
 413         case M4AF_TAG_ALBUM_ARTIST:
 414         case M4AF_TAG_TRACK:
 415         case M4AF_TAG_DISK:
 416         case M4AF_TAG_TEMPO:
 417             aacenc_add_tag_to_store(&params->tags, ch, 0, optarg,
 418                                     strlen(optarg), 0);
 419             break;
 420         case OPT_SHORT_TAG:
 421         case OPT_SHORT_TAG_FILE:
 422         case OPT_LONG_TAG:
 423             {
 424                 char *val;
 425                 size_t klen;
 426                 unsigned fcc = M4AF_FOURCC('-','-','-','-');
 427
 428                 if ((val = strchr(optarg, ':')) == 0) {
 429                     fprintf(stderr, "invalid arg for tag\n");
 430                     return -1;
 431                 }
 432                 *val++ = '\0';
 433                 if (ch == OPT_SHORT_TAG || ch == OPT_SHORT_TAG_FILE) {
 434                     /*
 435                      * take care of U+00A9(COPYRIGHT SIGN).
 436                      * 1) if length of fcc is 3, we prepend '\xa9'.
 437                      * 2) U+00A9 becomes "\xc2\xa9" in UTF-8. Therefore
 438                      *    we remove first '\xc2'.
 439                      */
 440                     if (optarg[0] == '\xc2')
 441                         ++optarg;
 442                     if ((klen = strlen(optarg))== 3)
 443                         fcc = 0xa9;
 444                     else if (klen != 4) {
 445                         fprintf(stderr, "invalid arg for tag\n");
 446                         return -1;
 447                     }
 448                     for (; *optarg; ++optarg)
 449                         fcc = ((fcc << 8) | (*optarg & 0xff));
 450                 }
 451                 aacenc_add_tag_to_store(&params->tags, fcc, optarg,
 452                                         val, strlen(val),
 453                                         ch == OPT_SHORT_TAG_FILE);
 454             }
 455             break;
 456         case OPT_TAG_FROM_JSON:
 457             params->json_filename = optarg;
 458             break;
 459         default:
 460             return usage(), -1;
 461         }
 462     }
 463     if (argc == optind)
 464         return usage(), -1;
 465
 466     if (!params->bitrate && !params->bitrate_mode) {
 467         fprintf(stderr, "bitrate or bitrate-mode is mandatory\n");
 468         return -1;
 469     }
 470     if (params->output_filename && !strcmp(params->output_filename, "-") &&
 471         !params->transport_format) {
 472         fprintf(stderr, "stdout streaming is not available on M4A output\n");
 473         return -1;
 474     }
 475     if (params->bitrate && params->bitrate < 10000)
 476         params->bitrate *= 1000;
 477
 478     if (params->is_raw) {
 479         if (!params->raw_channels)
 480             params->raw_channels = 2;
 481         if (!params->raw_rate)
 482             params->raw_rate = 44100;
 483         if (!params->raw_format)
 484             params->raw_format = "S16L";
 485     }
 486     params->input_filename = argv[optind];
 487     return 0;
 488 };
 489
 490 static
 491 int write_sample(FILE *ofp, m4af_ctx_t *m4af, aacenc_frame_t *frame)
 492 {
 493     if (!m4af) {
 494         fwrite(frame->data, 1, frame->size, ofp);
 495         if (ferror(ofp)) {
 496             fprintf(stderr, "ERROR: fwrite(): %s\n", strerror(errno));
 497             return -1;
 498         }
 499     } else if (m4af_write_sample(m4af, 0, frame->data, frame->size, 0) < 0) {
 500         fprintf(stderr, "ERROR: failed to write m4a sample\n");
 501         return -1;
 502     }
 503     return 0;
 504 }
 505
 506 static
 507 int encode(aacenc_param_ex_t *params, pcm_reader_t *reader,
 508            HANDLE_AACENCODER encoder, uint32_t frame_length,
 509            m4af_ctx_t *m4af)
 510 {
 511     int16_t *ibuf = 0, *ip;
 512     aacenc_frame_t obuf[2] = {{ 0 }}, *obp;
 513     unsigned flip = 0;
 514     int nread = 1;
 515     int rc = -1;
 516     int remaining, consumed;
 517     int frames_written = 0, encoded = 0;
 518     aacenc_progress_t progress = { 0 };
 519     const pcm_sample_description_t *fmt = pcm_get_format(reader);
 520
 521     ibuf = malloc(frame_length * fmt->bytes_per_frame);
 522     aacenc_progress_init(&progress, pcm_get_length(reader), fmt->sample_rate);
 523
 524     for (;;) {
 525         /*
 526          * Since we delay the write, we cannot just exit loop when interrupted.
 527          * Instead, we regard it as EOF.
 528          */
 529         if (g_interrupted)
 530             nread = 0;
 531         if (nread > 0) {
 532             if ((nread = pcm_read_frames(reader, ibuf, frame_length)) < 0) {
 533                 fprintf(stderr, "ERROR: read failed\n");
 534                 goto END;
 535             }
 536             if (!params->silent)
 537                 aacenc_progress_update(&progress, pcm_get_position(reader),
 538                                        fmt->sample_rate * 2);
 539         }
 540         ip = ibuf;
 541         remaining = nread;
 542         do {
 543             obp = &obuf[flip];
 544             consumed = aac_encode_frame(encoder, fmt, ip, remaining, obp);
 545             if (consumed < 0) goto END;
 546             if (consumed == 0 && obp->size == 0) goto DONE;
 547             if (obp->size == 0) break;
 548
 549             remaining -= consumed;
 550             ip += consumed * fmt->channels_per_frame;
 551             flip ^= 1;
 552             /*
 553              * As we pad 1 frame at beginning and ending by our extrapolator,
 554              * we want to drop them.
 555              * We delay output by 1 frame by double buffering, and discard
 556              * second frame and final frame from the encoder.
 557              * Since sbr_header is included in the first frame (in case of
 558              * SBR), we cannot discard first frame. So we pick second instead.
 559              */
 560             ++encoded;
 561             if (encoded == 1 || encoded == 3)
 562                 continue;
 563
 564             if (write_sample(params->output_fp, m4af, &obuf[flip]) < 0)
 565                 goto END;
 566             ++frames_written;
 567         } while (remaining > 0);
 568         /*
 569          * When interrupted, we haven't pulled out last extrapolated frames
 570          * from the reader. Therefore, we have to write the final outcome.
 571          */
 572         if (g_interrupted) {
 573             if (write_sample(params->output_fp, m4af, &obp[flip^1]) < 0)
 574                 goto END;
 575             ++frames_written;
 576         }
 577     }
 578 DONE:
 579     if (!params->silent)
 580         aacenc_progress_finish(&progress, pcm_get_position(reader));
 581     rc = frames_written;
 582 END:
 583     if (ibuf) free(ibuf);
 584     if (obuf[0].data) free(obuf[0].data);
 585     if (obuf[1].data) free(obuf[1].data);
 586     return rc;
 587 }
 588
 589 static
 590 void put_tool_tag(m4af_ctx_t *m4af, const aacenc_param_ex_t *params,
 591                   HANDLE_AACENCODER encoder)
 592 {
 593     char tool_info[256];
 594     char *p = tool_info;
 595     LIB_INFO lib_info;
 596
 597     p += sprintf(p, PROGNAME " %s, ", fdkaac_version);
 598     aacenc_get_lib_info(&lib_info);
 599     p += sprintf(p, "libfdk-aac %s, ", lib_info.versionStr);
 600     if (params->bitrate_mode)
 601         sprintf(p, "VBR mode %d", params->bitrate_mode);
 602     else
 603         sprintf(p, "CBR %dkbps",
 604                 aacEncoder_GetParam(encoder, AACENC_BITRATE) / 1000);
 605
 606     m4af_add_itmf_string_tag(m4af, M4AF_TAG_TOOL, tool_info);
 607 }
 608
 609 static
 610 int finalize_m4a(m4af_ctx_t *m4af, const aacenc_param_ex_t *params,
 611                  HANDLE_AACENCODER encoder)
 612 {
 613     unsigned i;
 614     aacenc_tag_entry_t *tag;
 615
 616     tag = params->source_tags.tag_table;
 617     for (i = 0; i < params->source_tags.tag_count; ++i, ++tag)
 618         aacenc_write_tag_entry(m4af, tag);
 619
 620     if (params->json_filename)
 621         aacenc_write_tags_from_json(m4af, params->json_filename);
 622
 623     tag = params->tags.tag_table;
 624     for (i = 0; i < params->tags.tag_count; ++i, ++tag)
 625         aacenc_write_tag_entry(m4af, tag);
 626
 627     put_tool_tag(m4af, params, encoder);
 628
 629     if (m4af_finalize(m4af, params->moov_before_mdat) < 0) {
 630         fprintf(stderr, "ERROR: failed to finalize m4a\n");
 631         return -1;
 632     }
 633     return 0;
 634 }
 635
 636 static
 637 char *generate_output_filename(const char *filename, const char *ext)
 638 {
 639     char *p = 0;
 640     size_t ext_len = strlen(ext);
 641
 642     if (strcmp(filename, "-") == 0) {
 643         p = malloc(ext_len + 6);
 644         sprintf(p, "stdin%s", ext);
 645     } else {
 646         const char *base = aacenc_basename(filename);
 647         size_t ilen = strlen(base);
 648         const char *ext_org = strrchr(base, '.');
 649         if (ext_org) ilen = ext_org - base;
 650         p = malloc(ilen + ext_len + 1);
 651         sprintf(p, "%.*s%s", (int)ilen, base, ext);
 652     }
 653     return p;
 654 }
 655
 656 static
 657 int parse_raw_spec(const char *spec, pcm_sample_description_t *desc)
 658 {
 659     unsigned bits;
 660     unsigned char c_type, c_endian = 'L';
 661     int type;
 662
 663     if (sscanf(spec, "%c%u%c", &c_type, &bits, &c_endian) < 2)
 664         return -1;
 665     c_type = toupper(c_type);
 666     c_endian = toupper(c_endian);
 667
 668     if (c_type == 'S')
 669         type = 1;
 670     else if (c_type == 'U')
 671         type = 2;
 672     else if (c_type == 'F')
 673         type = 4;
 674     else
 675         return -1;
 676
 677     if (c_endian == 'B')
 678         type |= 8;
 679     else if (c_endian != 'L')
 680         return -1;
 681
 682     if (c_type == 'F' && bits != 32 && bits != 64)
 683         return -1;
 684     if (c_type != 'F' && (bits < 8 || bits > 32))
 685         return -1;
 686
 687     desc->sample_type = type;
 688     desc->bits_per_channel = bits;
 689     return 0;
 690 }
 691
 692 static pcm_io_vtbl_t pcm_io_vtbl = {
 693     read_callback, seek_callback, tell_callback
 694 };
 695 static pcm_io_vtbl_t pcm_io_vtbl_noseek = { read_callback, 0, tell_callback };
 696
 697 static
 698 pcm_reader_t *open_input(aacenc_param_ex_t *params)
 699 {
 700     pcm_io_context_t io = { 0 };
 701     pcm_reader_t *reader = 0;
 702
 703     if ((params->input_fp = aacenc_fopen(params->input_filename, "rb")) == 0) {
 704         aacenc_fprintf(stderr, "ERROR: %s: %s\n", params->input_filename,
 705                        strerror(errno));
 706         goto FAIL;
 707     }
 708     io.cookie = params->input_fp;
 709     if (aacenc_seekable(params->input_fp))
 710         io.vtbl = &pcm_io_vtbl;
 711     else
 712         io.vtbl = &pcm_io_vtbl_noseek;
 713
 714     if (params->is_raw) {
 715         int bytes_per_channel;
 716         pcm_sample_description_t desc = { 0 };
 717         if (parse_raw_spec(params->raw_format, &desc) < 0) {
 718             fprintf(stderr, "ERROR: invalid raw-format spec\n");
 719             goto FAIL;
 720         }
 721         desc.sample_rate = params->raw_rate;
 722         desc.channels_per_frame = params->raw_channels;
 723         bytes_per_channel = (desc.bits_per_channel + 7) / 8;
 724         desc.bytes_per_frame = params->raw_channels * bytes_per_channel;
 725         if ((reader = raw_open(&io, &desc)) == 0) {
 726             fprintf(stderr, "ERROR: failed to open raw input\n");
 727             goto FAIL;
 728         }
 729     } else {
 730         int c;
 731         ungetc(c = getc(params->input_fp), params->input_fp);
 732
 733         switch (c) {
 734         case 'R':
 735             if ((reader = wav_open(&io, params->ignore_length)) == 0) {
 736                 fprintf(stderr, "ERROR: broken / unsupported input file\n");
 737                 goto FAIL;
 738             }
 739             break;
 740         case 'c':
 741             params->source_tag_ctx.add = aacenc_add_tag_entry_to_store;
 742             params->source_tag_ctx.add_ctx = &params->source_tags;
 743             if ((reader = caf_open(&io,
 744                                    aacenc_translate_generic_text_tag,
 745                                    &params->source_tag_ctx)) == 0) {
 746                 fprintf(stderr, "ERROR: broken / unsupported input file\n");
 747                 goto FAIL;
 748             }
 749             break;
 750         default:
 751             fprintf(stderr, "ERROR: unsupported input file\n");
 752             goto FAIL;
 753         }
 754     }
 755     reader = pcm_open_native_converter(reader);
 756     if (reader && PCM_IS_FLOAT(pcm_get_format(reader)))
 757         reader = limiter_open(reader);
 758     if (reader && (reader = pcm_open_sint16_converter(reader)) != 0)
 759         reader = extrapolater_open(reader);
 760     return reader;
 761 FAIL:
 762     return 0;
 763 }
 764
 765 int main(int argc, char **argv)
 766 {
 767     static m4af_io_callbacks_t m4af_io = {
 768         read_callback, write_callback, seek_callback, tell_callback
 769     };
 770     aacenc_param_ex_t params = { 0 };
 771
 772     int result = 2;
 773     char *output_filename = 0;
 774     pcm_reader_t *reader = 0;
 775     HANDLE_AACENCODER encoder = 0;
 776     AACENC_InfoStruct aacinfo = { 0 };
 777     m4af_ctx_t *m4af = 0;
 778     const pcm_sample_description_t *sample_format;
 779     int frame_count = 0;
 780     int sbr_mode = 0;
 781     unsigned scale_shift = 0;
 782
 783     setlocale(LC_CTYPE, "");
 784     setbuf(stderr, 0);
 785
 786     if (parse_options(argc, argv, &params) < 0)
 787         return 1;
 788
 789     if ((reader = open_input(&params)) == 0)
 790         goto END;
 791
 792     sample_format = pcm_get_format(reader);
 793
 794     sbr_mode = aacenc_is_sbr_active((aacenc_param_t*)&params);
 795     if (sbr_mode && !aacenc_is_sbr_ratio_available()) {
 796         fprintf(stderr, "WARNING: Only dual-rate SBR is available "
 797                         "for this version\n");
 798         params.sbr_ratio = 2;
 799     }
 800     scale_shift = aacenc_is_dual_rate_sbr((aacenc_param_t*)&params);
 801     params.sbr_signaling = 0;
 802     if (sbr_mode) {
 803         if (params.transport_format == TT_MP4_LOAS || !scale_shift)
 804             params.sbr_signaling = 2;
 805         if (params.transport_format == TT_MP4_RAW &&
 806             aacenc_is_explicit_bw_compatible_sbr_signaling_available())
 807             params.sbr_signaling = 1;
 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         unsigned framelen = aacinfo.frameLength;
 829         scale = sample_format->sample_rate >> scale_shift;
 830         if ((m4af = m4af_create(M4AF_CODEC_MP4A, scale, &m4af_io,
 831                                 params.output_fp)) < 0)
 832             goto END;
 833         m4af_set_num_channels(m4af, 0, sample_format->channels_per_frame);
 834         m4af_set_fixed_frame_duration(m4af, 0, framelen >> scale_shift);
 835         if (aacenc_is_explicit_bw_compatible_sbr_signaling_available())
 836             m4af_set_decoder_specific_info(m4af, 0,
 837                                            aacinfo.confBuf, aacinfo.confSize);
 838         else {
 839             uint8_t mp4asc[32];
 840             uint32_t ascsize = sizeof(mp4asc);
 841             aacenc_mp4asc((aacenc_param_t*)&params, aacinfo.confBuf,
 842                           aacinfo.confSize, mp4asc, &ascsize);
 843             m4af_set_decoder_specific_info(m4af, 0, mp4asc, ascsize);
 844         }
 845         m4af_set_vbr_mode(m4af, 0, params.bitrate_mode);
 846         m4af_set_priming_mode(m4af, params.gapless_mode + 1);
 847         m4af_begin_write(m4af);
 848     }
 849     if (scale_shift && (aacinfo.encoderDelay & 1)) {
 850         /*
 851          * Since odd delay cannot be exactly expressed in downsampled scale,
 852          * we push one zero frame to the encoder here, to make delay even
 853          */
 854         int16_t zero[8] = { 0 };
 855         aacenc_frame_t frame = { 0 };
 856         aac_encode_frame(encoder, sample_format, zero, 1, &frame);
 857         free(frame.data);
 858     }
 859     frame_count = encode(&params, reader, encoder, aacinfo.frameLength, m4af);
 860     if (frame_count < 0)
 861         goto END;
 862     if (m4af) {
 863         uint32_t delay = aacinfo.encoderDelay;
 864         uint32_t padding;
 865         int64_t frames_read = pcm_get_position(reader);
 866
 867         if (sbr_mode && params.profile != AOT_ER_AAC_ELD &&
 868             !params.include_sbr_delay)
 869             delay -= 481 << scale_shift;
 870         if (scale_shift && (delay & 1))
 871             ++delay;
 872         padding = frame_count * aacinfo.frameLength - frames_read - delay;
 873         m4af_set_priming(m4af, 0, delay >> scale_shift, padding >> scale_shift);
 874         if (finalize_m4a(m4af, &params, encoder) < 0)
 875             goto END;
 876     }
 877     result = 0;
 878 END:
 879     if (reader) pcm_teardown(&reader);
 880     if (params.input_fp) fclose(params.input_fp);
 881     if (m4af) m4af_teardown(&m4af);
 882     if (params.output_fp) fclose(params.output_fp);
 883     if (encoder) aacEncClose(&encoder);
 884     if (output_filename) free(output_filename);
 885     if (params.tags.tag_table)
 886         aacenc_free_tag_store(&params.tags);
 887     if (params.source_tags.tag_table)
 888         aacenc_free_tag_store(&params.source_tags);
 889
 890     return result;
 891 }