apps/talk.c

   1 /***************************************************************************
   2  *             __________               __   ___.
   3  *   Open      \______   \ ____   ____ |  | _\_ |__   _______  ___
   4  *   Source     |       _//  _ \_/ ___\|  |/ /| __ \ /  _ \  \/  /
   5  *   Jukebox    |    |   (  <_> )  \___|    < | \_\ (  <_> > <  <
   6  *   Firmware   |____|_  /\____/ \___  >__|_ \|___  /\____/__/\_ \
   7  *                     \/            \/     \/    \/            \/
   8  * $Id$
   9  *
  10  * Copyright (C) 2004 Jörg Hohensohn
  11  *
  12  * This module collects the Talkbox and voice UI functions.
  13  * (Talkbox reads directory names from mp3 clips called thumbnails,
  14  *  the voice UI lets menus and screens "talk" from a voicefile in memory.
  15  *
  16  * This program is free software; you can redistribute it and/or
  17  * modify it under the terms of the GNU General Public License
  18  * as published by the Free Software Foundation; either version 2
  19  * of the License, or (at your option) any later version.
  20  *
  21  * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
  22  * KIND, either express or implied.
  23  *
  24  ****************************************************************************/
  25
  26 #include <stdio.h>
  27 #include <stddef.h>
  28 #include <string.h>
  29 #include "file.h"
  30 #include "buffer.h"
  31 #include "system.h"
  32 #include "kernel.h"
  33 #include "settings.h"
  34 #include "settings_list.h"
  35 #include "mp3_playback.h"
  36 #include "audio.h"
  37 #include "lang.h"
  38 #include "talk.h"
  39 #include "metadata.h"
  40 #include "logf.h"
  41 #include "bitswap.h"
  42 #include "structec.h"
  43 #if CONFIG_CODEC == SWCODEC
  44 #include "playback.h"
  45 #endif
  46 #include "debug.h"
  47
  48
  49 /* Memory layout varies between targets because the
  50    Archos (MASCODEC) devices cannot mix voice and audio playback
  51
  52              MASCODEC  | MASCODEC  | SWCODEC
  53              (playing) | (stopped) |
  54     audiobuf-----------+-----------+------------
  55               audio    | voice     | thumbnail
  56                        |-----------|------------
  57                        | thumbnail | voice
  58                        |           |------------
  59                        |           | filebuf
  60                        |           |------------
  61                        |           | audio
  62   audiobufend----------+-----------+------------
  63
  64   SWCODEC allocates dedicated buffers, MASCODEC reuses audiobuf. */
  65
  66
  67 /***************** Constants *****************/
  68
  69 #define QUEUE_SIZE 64 /* must be a power of two */
  70 #define QUEUE_MASK (QUEUE_SIZE-1)
  71 const char* const dir_thumbnail_name = "_dirname.talk";
  72 const char* const file_thumbnail_ext = ".talk";
  73
  74 /***************** Functional Macros *****************/
  75
  76 #define QUEUE_LEVEL ((queue_write - queue_read) & QUEUE_MASK)
  77
  78 #define LOADED_MASK 0x80000000 /* MSB */
  79
  80 #if CONFIG_CODEC == SWCODEC
  81 #define MAX_THUMBNAIL_BUFSIZE 0x10000
  82 #endif
  83
  84 /***************** Data types *****************/
  85
  86 struct clip_entry /* one entry of the index table */
  87 {
  88     int offset; /* offset from start of voicefile file */
  89     int size; /* size of the clip */
  90 };
  91
  92 struct voicefile /* file format of our voice file */
  93 {
  94     int version; /* version of the voicefile */
  95     int target_id; /* the rockbox target the file was made for */
  96     int table;   /* offset to index table, (=header size) */
  97     int id1_max; /* number of "normal" clips contained in above index */
  98     int id2_max; /* number of "voice only" clips contained in above index */
  99     struct clip_entry index[]; /* followed by the index tables */
 100     /* and finally the mp3 clips, not visible here, bitswapped
 101        for SH based players */
 102 };
 103
 104 struct queue_entry /* one entry of the internal queue */
 105 {
 106     unsigned char* buf;
 107     long len;
 108 };
 109
 110
 111 /***************** Globals *****************/
 112
 113 static unsigned char* p_thumbnail = NULL; /* buffer for thumbnails */
 114 /* Multiple thumbnails can be loaded back-to-back in this buffer. */
 115 static volatile int thumbnail_buf_used SHAREDBSS_ATTR; /* length of data in
 116                                                           thumbnail buffer */
 117 static long size_for_thumbnail; /* total thumbnail buffer size */
 118 static struct voicefile* p_voicefile; /* loaded voicefile */
 119 static bool has_voicefile; /* a voicefile file is present */
 120 static bool need_shutup; /* is there possibly any voice playing to be shutup */
 121 static struct queue_entry queue[QUEUE_SIZE]; /* queue of scheduled clips */
 122 static bool force_enqueue_next; /* enqueue next utterance even if enqueue is false */
 123 static int queue_write; /* write index of queue, by application */
 124 static int queue_read; /* read index of queue, by ISR context */
 125 #if CONFIG_CODEC == SWCODEC
 126 struct mutex queue_mutex SHAREDBSS_ATTR; /* protects queue_read, queue_write
 127                                             and thumbnail_buf_used */
 128 #define talk_queue_lock() ({ mutex_lock(&queue_mutex); })
 129 #define talk_queue_unlock() ({ mutex_unlock(&queue_mutex); })
 130 #else
 131 #define talk_queue_lock() ({ })
 132 #define talk_queue_unlock() ({ })
 133 #endif /* CONFIG_CODEC */
 134 static int sent; /* how many bytes handed over to playback, owned by ISR */
 135 static unsigned char curr_hd[3]; /* current frame header, for re-sync */
 136 static int filehandle = -1; /* global, so the MMC variant can keep the file open */
 137 static unsigned char* p_silence; /* VOICE_PAUSE clip, used for termination */
 138 static long silence_len; /* length of the VOICE_PAUSE clip */
 139 static unsigned char* p_lastclip; /* address of latest clip, for silence add */
 140 static unsigned long voicefile_size = 0; /* size of the loaded voice file */
 141 static unsigned char last_lang[MAX_FILENAME+1]; /* name of last used lang file (in talk_init) */
 142 static bool talk_initialized; /* true if talk_init has been called */
 143 static int talk_temp_disable_count; /* if positive, temporarily disable voice UI (not saved) */
 144
 145
 146 /***************** Private implementation *****************/
 147
 148 static int open_voicefile(void)
 149 {
 150     char buf[64];
 151     char* p_lang = "english"; /* default */
 152
 153     if ( global_settings.lang_file[0] &&
 154          global_settings.lang_file[0] != 0xff )
 155     {   /* try to open the voice file of the selected language */
 156         p_lang = (char *)global_settings.lang_file;
 157     }
 158
 159     snprintf(buf, sizeof(buf), LANG_DIR "/%s.voice", p_lang);
 160
 161     return open(buf, O_RDONLY);
 162 }
 163
 164
 165 /* fetch a clip from the voice file */
 166 static unsigned char* get_clip(long id, long* p_size)
 167 {
 168     long clipsize;
 169     unsigned char* clipbuf;
 170
 171     if (id > VOICEONLY_DELIMITER)
 172     {   /* voice-only entries use the second part of the table */
 173         id -= VOICEONLY_DELIMITER + 1;
 174         if (id >= p_voicefile->id2_max)
 175             return NULL; /* must be newer than we have */
 176         id += p_voicefile->id1_max; /* table 2 is behind table 1 */
 177     }
 178     else
 179     {   /* normal use of the first table */
 180         if (id >= p_voicefile->id1_max)
 181             return NULL; /* must be newer than we have */
 182     }
 183
 184     clipsize = p_voicefile->index[id].size;
 185     if (clipsize == 0) /* clip not included in voicefile */
 186         return NULL;
 187     clipbuf = (unsigned char *) p_voicefile + p_voicefile->index[id].offset;
 188
 189 #if (CONFIG_STORAGE & STORAGE_MMC) /* dynamic loading, on demand */
 190     if (!(clipsize & LOADED_MASK))
 191     {   /* clip used for the first time, needs loading */
 192         lseek(filehandle, p_voicefile->index[id].offset, SEEK_SET);
 193         if (read(filehandle, clipbuf, clipsize) != clipsize)
 194             return NULL; /* read error */
 195
 196         p_voicefile->index[id].size |= LOADED_MASK; /* mark as loaded */
 197     }
 198     else
 199     {   /* clip is in memory already */
 200         clipsize &= ~LOADED_MASK; /* without the extra bit gives true size */
 201     }
 202 #endif
 203
 204     *p_size = clipsize;
 205     return clipbuf;
 206 }
 207
 208
 209 /* load the voice file into the mp3 buffer */
 210 static void load_voicefile(void)
 211 {
 212     int load_size;
 213     int got_size;
 214     int file_size;
 215 #ifdef ROCKBOX_LITTLE_ENDIAN
 216     int i;
 217 #endif
 218
 219     filehandle = open_voicefile();
 220     if (filehandle < 0) /* failed to open */
 221         goto load_err;
 222
 223     file_size = filesize(filehandle);
 224     if (file_size > audiobufend - audiobuf) /* won't fit? */
 225         goto load_err;
 226
 227 #if (CONFIG_STORAGE & STORAGE_MMC) /* load only the header for now */
 228     load_size = offsetof(struct voicefile, index);
 229 #else /* load the full file */
 230     load_size = file_size;
 231 #endif
 232
 233     got_size = read(filehandle, audiobuf, load_size);
 234     if (got_size != load_size /* failure */)
 235         goto load_err;
 236
 237 #ifdef ROCKBOX_LITTLE_ENDIAN
 238     logf("Byte swapping voice file");
 239     structec_convert(audiobuf, "lllll", 1, true);
 240 #endif
 241
 242     if (((struct voicefile*)audiobuf)->table /* format check */
 243            == offsetof(struct voicefile, index))
 244     {
 245         p_voicefile = (struct voicefile*)audiobuf;
 246
 247         if (p_voicefile->version != VOICE_VERSION ||
 248             p_voicefile->target_id != TARGET_ID)
 249         {
 250             logf("Incompatible voice file");
 251             goto load_err;
 252         }
 253 #if CONFIG_CODEC != SWCODEC
 254         /* MASCODEC: now use audiobuf for voice then thumbnail */
 255         p_thumbnail = audiobuf + file_size;
 256         p_thumbnail += (long)p_thumbnail % 2; /* 16-bit align */
 257         size_for_thumbnail = audiobufend - p_thumbnail;
 258 #endif
 259     }
 260     else
 261         goto load_err;
 262
 263 #ifdef ROCKBOX_LITTLE_ENDIAN
 264     for (i = 0; i < p_voicefile->id1_max + p_voicefile->id2_max; i++)
 265         structec_convert(&p_voicefile->index[i], "ll", 1, true);
 266 #endif
 267
 268 #if (CONFIG_STORAGE & STORAGE_MMC)
 269     /* load the index table, now that we know its size from the header */
 270     load_size = (p_voicefile->id1_max + p_voicefile->id2_max)
 271                 * sizeof(struct clip_entry);
 272     got_size = read(filehandle,
 273                     (unsigned char *) p_voicefile + offsetof(struct voicefile, index), load_size);
 274     if (got_size != load_size) /* read error */
 275         goto load_err;
 276 #else
 277     close(filehandle); /* only the MMC variant leaves it open */
 278     filehandle = -1;
 279 #endif
 280
 281     /* make sure to have the silence clip, if available */
 282     p_silence = get_clip(VOICE_PAUSE, &silence_len);
 283
 284     return;
 285
 286 load_err:
 287     p_voicefile = NULL;
 288     has_voicefile = false; /* don't try again */
 289     if (filehandle >= 0)
 290     {
 291         close(filehandle);
 292         filehandle = -1;
 293     }
 294     return;
 295 }
 296
 297
 298 /* called in ISR context if mp3 data got consumed */
 299 static void mp3_callback(unsigned char** start, size_t* size)
 300 {
 301     queue[queue_read].len -= sent; /* we completed this */
 302     queue[queue_read].buf += sent;
 303
 304     if (queue[queue_read].len > 0) /* current clip not finished? */
 305     {   /* feed the next 64K-1 chunk */
 306 #if CONFIG_CODEC != SWCODEC
 307         sent = MIN(queue[queue_read].len, 0xFFFF);
 308 #else
 309         sent = queue[queue_read].len;
 310 #endif
 311         *start = queue[queue_read].buf;
 312         *size = sent;
 313         return;
 314     }
 315     talk_queue_lock();
 316     if(p_thumbnail
 317        && queue[queue_read].buf == p_thumbnail +thumbnail_buf_used)
 318         thumbnail_buf_used = 0;
 319     if (sent > 0) /* go to next entry */
 320     {
 321         queue_read = (queue_read + 1) & QUEUE_MASK;
 322     }
 323
 324 re_check:
 325
 326     if (QUEUE_LEVEL != 0) /* queue is not empty? */
 327     {   /* start next clip */
 328 #if CONFIG_CODEC != SWCODEC
 329         sent = MIN(queue[queue_read].len, 0xFFFF);
 330 #else
 331         sent = queue[queue_read].len;
 332 #endif
 333         *start = p_lastclip = queue[queue_read].buf;
 334         *size = sent;
 335         curr_hd[0] = p_lastclip[1];
 336         curr_hd[1] = p_lastclip[2];
 337         curr_hd[2] = p_lastclip[3];
 338     }
 339     else if (p_silence != NULL             /* silence clip available */
 340              && p_lastclip != p_silence    /* previous clip wasn't silence */
 341              && !(p_lastclip >= p_thumbnail /* ..or thumbnail */
 342                   && p_lastclip < p_thumbnail +size_for_thumbnail))
 343     {   /* add silence clip when queue runs empty playing a voice clip */
 344         queue[queue_write].buf = p_silence;
 345         queue[queue_write].len = silence_len;
 346         queue_write = (queue_write + 1) & QUEUE_MASK;
 347
 348         goto re_check;
 349     }
 350     else
 351     {
 352         *size = 0; /* end of data */
 353     }
 354     talk_queue_unlock();
 355 }
 356
 357 /***************** Public routines *****************/
 358
 359 /* stop the playback and the pending clips */
 360 void talk_force_shutup(void)
 361 {
 362     /* Most of this is MAS only */
 363 #if CONFIG_CODEC != SWCODEC
 364 #ifdef SIMULATOR
 365     return;
 366 #endif
 367     unsigned char* pos;
 368     unsigned char* search;
 369     unsigned char* end;
 370     if (QUEUE_LEVEL == 0) /* has ended anyway */
 371         return;
 372
 373 #if CONFIG_CPU == SH7034
 374     CHCR3 &= ~0x0001; /* disable the DMA (and therefore the interrupt also) */
 375 #endif /* CONFIG_CPU == SH7034 */
 376     /* search next frame boundary and continue up to there */
 377     pos = search = mp3_get_pos();
 378     end = queue[queue_read].buf + queue[queue_read].len;
 379
 380     if (pos >= queue[queue_read].buf
 381         && pos <= end) /* really our clip? */
 382     { /* (for strange reasons this isn't nesessarily the case) */
 383         /* find the next frame boundary */
 384         while (search < end) /* search the remaining data */
 385         {
 386             if (*search++ != 0xFF) /* quick search for frame sync byte */
 387                 continue; /* (this does the majority of the job) */
 388
 389             /* look at the (bitswapped) rest of header candidate */
 390             if (search[0] == curr_hd[0] /* do the quicker checks first */
 391              && search[2] == curr_hd[2]
 392              && (search[1] & 0x30) == (curr_hd[1] & 0x30)) /* sample rate */
 393             {
 394                 search--; /* back to the sync byte */
 395                 break; /* From looking at it, this is our header. */
 396             }
 397         }
 398
 399         if (search-pos)
 400         {   /* play old data until the frame end, to keep the MAS in sync */
 401             sent = search-pos;
 402
 403             queue_write = (queue_read + 1) & QUEUE_MASK; /* will be empty after next callback */
 404             queue[queue_read].len = sent; /* current one ends after this */
 405
 406 #if CONFIG_CPU == SH7034
 407             DTCR3 = sent; /* let the DMA finish this frame */
 408             CHCR3 |= 0x0001; /* re-enable DMA */
 409 #endif /* CONFIG_CPU == SH7034 */
 410             thumbnail_buf_used = 0;
 411             return;
 412         }
 413     }
 414 #endif /* CONFIG_CODEC != SWCODEC */
 415
 416     /* Either SWCODEC, or MAS had nothing to do (was frame boundary or not our clip) */
 417     mp3_play_stop();
 418     talk_queue_lock();
 419     queue_write = queue_read = 0; /* reset the queue */
 420     thumbnail_buf_used = 0;
 421     talk_queue_unlock();
 422     need_shutup = false;
 423 }
 424
 425 /* Shutup the voice, except if force_enqueue_next is set. */
 426 void talk_shutup(void)
 427 {
 428     if (need_shutup && !force_enqueue_next)
 429         talk_force_shutup();
 430 }
 431
 432 /* schedule a clip, at the end or discard the existing queue */
 433 static void queue_clip(unsigned char* buf, long size, bool enqueue)
 434 {
 435     int queue_level;
 436
 437     if (!enqueue)
 438         talk_shutup(); /* cut off all the pending stuff */
 439     /* Something is being enqueued, force_enqueue_next override is no
 440        longer in effect. */
 441     force_enqueue_next = false;
 442
 443     if (!size)
 444         return; /* safety check */
 445 #if CONFIG_CPU == SH7034
 446     /* disable the DMA temporarily, to be safe of race condition */
 447     CHCR3 &= ~0x0001;
 448 #endif
 449     talk_queue_lock();
 450     queue_level = QUEUE_LEVEL; /* check old level */
 451
 452     if (queue_level < QUEUE_SIZE - 1) /* space left? */
 453     {
 454         queue[queue_write].buf = buf; /* populate an entry */
 455         queue[queue_write].len = size;
 456         queue_write = (queue_write + 1) & QUEUE_MASK;
 457     }
 458     talk_queue_unlock();
 459
 460     if (queue_level == 0)
 461     {   /* queue was empty, we have to do the initial start */
 462         p_lastclip = buf;
 463 #if CONFIG_CODEC != SWCODEC
 464         sent = MIN(size, 0xFFFF); /* DMA can do no more */
 465 #else
 466         sent = size;
 467 #endif
 468         mp3_play_data(buf, sent, mp3_callback);
 469         curr_hd[0] = buf[1];
 470         curr_hd[1] = buf[2];
 471         curr_hd[2] = buf[3];
 472         mp3_play_pause(true); /* kickoff audio */
 473     }
 474     else
 475     {
 476 #if CONFIG_CPU == SH7034
 477         CHCR3 |= 0x0001; /* re-enable DMA */
 478 #endif
 479     }
 480
 481     need_shutup = true;
 482
 483     return;
 484 }
 485
 486
 487 /* common code for talk_init() and talk_buffer_steal() */
 488 static void reset_state(void)
 489 {
 490     queue_write = queue_read = 0; /* reset the queue */
 491     p_voicefile = NULL; /* indicate no voicefile (trashed) */
 492 #if CONFIG_CODEC == SWCODEC
 493     /* Allocate a dedicated thumbnail buffer - once */
 494     if (p_thumbnail == NULL)
 495     {
 496         size_for_thumbnail = audiobufend - audiobuf;
 497         if (size_for_thumbnail > MAX_THUMBNAIL_BUFSIZE)
 498             size_for_thumbnail = MAX_THUMBNAIL_BUFSIZE;
 499         p_thumbnail = buffer_alloc(size_for_thumbnail);
 500     }
 501 #else
 502     /* Just use the audiobuf, without allocating anything */
 503     p_thumbnail = audiobuf;
 504     size_for_thumbnail = audiobufend - audiobuf;
 505 #endif
 506     thumbnail_buf_used = 0;
 507     p_silence = NULL; /* pause clip not accessible */
 508 }
 509
 510
 511 /***************** Public implementation *****************/
 512
 513 void talk_init(void)
 514 {
 515     talk_temp_disable_count = 0;
 516     if (talk_initialized && !strcasecmp(last_lang, global_settings.lang_file))
 517     {
 518         /* not a new file, nothing to do */
 519         return;
 520     }
 521
 522 #if (CONFIG_STORAGE & STORAGE_MMC)
 523     if (filehandle >= 0) /* MMC: An old voice file might still be open */
 524     {
 525         close(filehandle);
 526         filehandle = -1;
 527     }
 528 #endif
 529
 530 #if CONFIG_CODEC == SWCODEC
 531     if(!talk_initialized)
 532         mutex_init(&queue_mutex);
 533 #endif /* CONFIG_CODEC == SWCODEC */
 534
 535     talk_initialized = true;
 536     strlcpy((char *)last_lang, (char *)global_settings.lang_file,
 537         MAX_FILENAME);
 538
 539 #if CONFIG_CODEC == SWCODEC
 540     audio_get_buffer(false, NULL);  /* Must tell audio to reinitialize */
 541 #endif
 542     reset_state(); /* use this for most of our inits */
 543
 544     filehandle = open_voicefile();
 545     has_voicefile = (filehandle >= 0); /* test if we can open it */
 546     voicefile_size = 0;
 547
 548     if (has_voicefile)
 549     {
 550         voicefile_size = filesize(filehandle);
 551         close(filehandle); /* close again, this was just to detect presence */
 552         filehandle = -1;
 553     }
 554
 555 }
 556
 557 #if CONFIG_CODEC == SWCODEC
 558 /* return if a voice codec is required or not */
 559 bool talk_voice_required(void)
 560 {
 561     return (voicefile_size != 0) /* Voice file is available */
 562         || (global_settings.talk_dir_clip)  /* Thumbnail clips are required */
 563         || (global_settings.talk_file_clip);
 564 }
 565 #endif
 566
 567 /* return size of voice file */
 568 int talk_get_bufsize(void)
 569 {
 570     return voicefile_size;
 571 }
 572
 573 /* somebody else claims the mp3 buffer, e.g. for regular play/record */
 574 void talk_buffer_steal(void)
 575 {
 576 #if CONFIG_CODEC != SWCODEC
 577     mp3_play_stop();
 578 #endif
 579 #if (CONFIG_STORAGE & STORAGE_MMC)
 580     if (filehandle >= 0) /* only relevant for MMC */
 581     {
 582         close(filehandle);
 583         filehandle = -1;
 584     }
 585 #endif
 586     reset_state();
 587 }
 588
 589
 590 /* play a voice ID from voicefile */
 591 int talk_id(int32_t id, bool enqueue)
 592 {
 593     long clipsize;
 594     unsigned char* clipbuf;
 595     int32_t unit;
 596     int decimals;
 597
 598     if (talk_temp_disable_count > 0)
 599         return -1;  /* talking has been disabled */
 600 #if CONFIG_CODEC != SWCODEC
 601     if (audio_status()) /* busy, buffer in use */
 602         return -1;
 603 #endif
 604
 605     if (p_voicefile == NULL && has_voicefile)
 606         load_voicefile(); /* reload needed */
 607
 608     if (p_voicefile == NULL) /* still no voices? */
 609         return -1;
 610
 611     if (id == -1) /* -1 is an indication for silence */
 612         return -1;
 613
 614     decimals = (((uint32_t)id) >> DECIMAL_SHIFT) & 0x7;
 615
 616     /* check if this is a special ID, with a value */
 617     unit = ((uint32_t)id) >> UNIT_SHIFT;
 618     if (unit || decimals)
 619     {   /* sign-extend the value */
 620         id = (uint32_t)id << (32-DECIMAL_SHIFT);
 621         id >>= (32-DECIMAL_SHIFT);
 622
 623         talk_value_decimal(id, unit, decimals, enqueue); /* speak it */
 624         return 0; /* and stop, end of special case */
 625     }
 626
 627     clipbuf = get_clip(id, &clipsize);
 628     if (clipbuf == NULL)
 629         return -1; /* not present */
 630
 631     queue_clip(clipbuf, clipsize, enqueue);
 632
 633     return 0;
 634 }
 635 /* Speaks zero or more IDs (from an array). */
 636 int talk_idarray(const long *ids, bool enqueue)
 637 {
 638     int r;
 639     if(!ids)
 640         return 0;
 641     while(*ids != TALK_FINAL_ID)
 642     {
 643         if((r = talk_id(*ids++, enqueue)) <0)
 644             return r;
 645         enqueue = true;
 646     }
 647     return 0;
 648 }
 649
 650 /* Make sure the current utterance is not interrupted by the next one. */
 651 void talk_force_enqueue_next(void)
 652 {
 653     force_enqueue_next = true;
 654 }
 655
 656 /* play a thumbnail from file */
 657 /* Returns size of spoken thumbnail, so >0 means something is spoken,
 658    <=0 means something went wrong. */
 659 static int _talk_file(const char* filename,
 660                       const long *prefix_ids, bool enqueue)
 661 {
 662     int fd;
 663     int size;
 664     int thumb_used;
 665 #if CONFIG_CODEC != SWCODEC
 666     struct mp3entry info;
 667 #endif
 668
 669     if (talk_temp_disable_count > 0)
 670         return -1;  /* talking has been disabled */
 671 #if CONFIG_CODEC != SWCODEC
 672     if (audio_status()) /* busy, buffer in use */
 673         return -1;
 674 #endif
 675
 676     if (p_thumbnail == NULL || size_for_thumbnail <= 0)
 677         return -1;
 678
 679 #if CONFIG_CODEC != SWCODEC
 680     if(mp3info(&info, filename)) /* use this to find real start */
 681     {
 682         return 0; /* failed to open, or invalid */
 683     }
 684 #endif
 685
 686     if (!enqueue)
 687         /* shutup now to free the thumbnail buffer */
 688         talk_shutup();
 689
 690     fd = open(filename, O_RDONLY);
 691     if (fd < 0) /* failed to open */
 692     {
 693         return 0;
 694     }
 695
 696     thumb_used = thumbnail_buf_used;
 697     if(filesize(fd) > size_for_thumbnail -thumb_used)
 698     {   /* Don't play truncated  clips */
 699         close(fd);
 700         return 0;
 701     }
 702
 703 #if CONFIG_CODEC != SWCODEC
 704     lseek(fd, info.first_frame_offset, SEEK_SET); /* behind ID data */
 705 #endif
 706
 707     size = read(fd, p_thumbnail +thumb_used,
 708                 size_for_thumbnail -thumb_used);
 709     close(fd);
 710
 711     /* ToDo: find audio, skip ID headers and trailers */
 712
 713     if (size > 0)    /* Don't play missing clips */
 714     {
 715 #if CONFIG_CODEC != SWCODEC && !defined(SIMULATOR)
 716         bitswap(p_thumbnail, size);
 717 #endif
 718         if(prefix_ids)
 719             /* prefix thumbnail by speaking these ids, but only now
 720                that we know there's actually a thumbnail to be
 721                spoken. */
 722             talk_idarray(prefix_ids, true);
 723         talk_queue_lock();
 724         thumbnail_buf_used = thumb_used +size;
 725         talk_queue_unlock();
 726         queue_clip(p_thumbnail +thumb_used, size, true);
 727     }
 728
 729     return size;
 730 }
 731
 732 int talk_file(const char *root, const char *dir, const char *file,
 733               const char *ext, const long *prefix_ids, bool enqueue)
 734 /* Play a thumbnail file */
 735 {
 736     char buf[MAX_PATH];
 737     /* Does root end with a slash */
 738     char *slash = (root && root[0]
 739                    && root[strlen(root)-1] != '/') ? "/" : "";
 740     snprintf(buf, MAX_PATH, "%s%s%s%s%s%s",
 741              root ? root : "", slash,
 742              dir ? dir : "", dir ? "/" : "",
 743              file ? file : "",
 744              ext ? ext : "");
 745     return _talk_file(buf, prefix_ids, enqueue);
 746 }
 747
 748 static int talk_spell_basename(const char *path,
 749                                const long *prefix_ids, bool enqueue)
 750 {
 751     if(prefix_ids)
 752     {
 753         talk_idarray(prefix_ids, enqueue);
 754         enqueue = true;
 755     }
 756     char buf[MAX_PATH];
 757     /* Spell only the path component after the last slash */
 758     strlcpy(buf, path, sizeof(buf));
 759     if(strlen(buf) >1 && buf[strlen(buf)-1] == '/')
 760         /* strip trailing slash */
 761         buf[strlen(buf)-1] = '\0';
 762     char *ptr = strrchr(buf, '/');
 763     if(ptr && strlen(buf) >1)
 764         ++ptr;
 765     else ptr = buf;
 766     return talk_spell(ptr, enqueue);
 767 }
 768
 769 /* Play a file's .talk thumbnail, fallback to spelling the filename, or
 770    go straight to spelling depending on settings. */
 771 int talk_file_or_spell(const char *dirname, const char *filename,
 772                        const long *prefix_ids, bool enqueue)
 773 {
 774     if (global_settings.talk_file_clip)
 775     {   /* .talk clips enabled */
 776         if(talk_file(dirname, NULL, filename, file_thumbnail_ext,
 777                               prefix_ids, enqueue) >0)
 778             return 0;
 779     }
 780     if (global_settings.talk_file == 2)
 781         /* Either .talk clips are disabled, or as a fallback */
 782         return talk_spell_basename(filename, prefix_ids, enqueue);
 783     return 0;
 784 }
 785
 786 /* Play a directory's .talk thumbnail, fallback to spelling the filename, or
 787    go straight to spelling depending on settings. */
 788 int talk_dir_or_spell(const char* dirname,
 789                       const long *prefix_ids, bool enqueue)
 790 {
 791     if (global_settings.talk_dir_clip)
 792     {   /* .talk clips enabled */
 793         if(talk_file(dirname, NULL, dir_thumbnail_name, NULL,
 794                               prefix_ids, enqueue) >0)
 795             return 0;
 796     }
 797     if (global_settings.talk_dir == 2)
 798         /* Either .talk clips disabled or as a fallback */
 799         return talk_spell_basename(dirname, prefix_ids, enqueue);
 800     return 0;
 801 }
 802
 803 /* say a numeric value, this word ordering works for english,
 804    but not necessarily for other languages (e.g. german) */
 805 int talk_number(long n, bool enqueue)
 806 {
 807     int level = 2; /* mille count */
 808     long mil = 1000000000; /* highest possible "-illion" */
 809
 810     if (talk_temp_disable_count > 0)
 811         return -1;  /* talking has been disabled */
 812 #if CONFIG_CODEC != SWCODEC
 813     if (audio_status()) /* busy, buffer in use */
 814         return -1;
 815 #endif
 816
 817     if (!enqueue)
 818         talk_shutup(); /* cut off all the pending stuff */
 819
 820     if (n==0)
 821     {   /* special case */
 822         talk_id(VOICE_ZERO, true);
 823         return 0;
 824     }
 825
 826     if (n<0)
 827     {
 828         talk_id(VOICE_MINUS, true);
 829         n = -n;
 830     }
 831
 832     while (n)
 833     {
 834         int segment = n / mil; /* extract in groups of 3 digits */
 835         n -= segment * mil; /* remove the used digits from number */
 836         mil /= 1000; /* digit place for next round */
 837
 838         if (segment)
 839         {
 840             int hundreds = segment / 100;
 841             int ones = segment % 100;
 842
 843             if (hundreds)
 844             {
 845                 talk_id(VOICE_ZERO + hundreds, true);
 846                 talk_id(VOICE_HUNDRED, true);
 847             }
 848
 849             /* combination indexing */
 850             if (ones > 20)
 851             {
 852                int tens = ones/10 + 18;
 853                talk_id(VOICE_ZERO + tens, true);
 854                ones %= 10;
 855             }
 856
 857             /* direct indexing */
 858             if (ones)
 859                 talk_id(VOICE_ZERO + ones, true);
 860
 861             /* add billion, million, thousand */
 862             if (mil)
 863                 talk_id(VOICE_THOUSAND + level, true);
 864         }
 865         level--;
 866     }
 867
 868     return 0;
 869 }
 870
 871 /* Say time duration/interval. Input is time in seconds,
 872    say hours,minutes,seconds. */
 873 static int talk_time_unit(long secs, bool exact, bool enqueue)
 874 {
 875     int hours, mins;
 876     if (!enqueue)
 877         talk_shutup();
 878     if((hours = secs/3600)) {
 879         secs %= 3600;
 880         talk_value(hours, UNIT_HOUR, true);
 881     }
 882     if((mins = secs/60)) {
 883         secs %= 60;
 884         if(exact || !hours)
 885             talk_value(mins, UNIT_MIN, true);
 886         else talk_number(mins, true); /* don't say "minutes" */
 887     }
 888     if((exact && secs) || (!hours && !mins))
 889         talk_value(secs, UNIT_SEC, true);
 890     else if(!hours && secs)
 891         talk_number(secs, true);
 892     return 0;
 893 }
 894
 895 void talk_fractional(char *tbuf, int value, int unit)
 896 {
 897     int i;
 898     /* strip trailing zeros from the fraction */
 899     for (i = strlen(tbuf) - 1; (i >= 0) && (tbuf[i] == '0'); i--)
 900         tbuf[i] = '\0';
 901
 902     talk_number(value, true);
 903     if (tbuf[0] != 0)
 904     {
 905         talk_id(LANG_POINT, true);
 906         talk_spell(tbuf, true);
 907     }
 908     talk_id(unit, true);
 909 }
 910
 911 int talk_value(long n, int unit, bool enqueue)
 912 {
 913     return talk_value_decimal(n, unit, 0, enqueue);
 914 }
 915
 916 /* singular/plural aware saying of a value */
 917 int talk_value_decimal(long n, int unit, int decimals, bool enqueue)
 918 {
 919     int unit_id;
 920     static const int unit_voiced[] =
 921     {   /* lookup table for the voice ID of the units */
 922         [0 ... UNIT_LAST-1] = -1, /* regular ID, int, signed */
 923         [UNIT_MS]
 924             = VOICE_MILLISECONDS, /* here come the "real" units */
 925         [UNIT_SEC]
 926             = VOICE_SECONDS,
 927         [UNIT_MIN]
 928             = VOICE_MINUTES,
 929         [UNIT_HOUR]
 930             = VOICE_HOURS,
 931         [UNIT_KHZ]
 932             = VOICE_KHZ,
 933         [UNIT_DB]
 934             = VOICE_DB,
 935         [UNIT_PERCENT]
 936             = VOICE_PERCENT,
 937         [UNIT_MAH]
 938             = VOICE_MILLIAMPHOURS,
 939         [UNIT_PIXEL]
 940             = VOICE_PIXEL,
 941         [UNIT_PER_SEC]
 942             = VOICE_PER_SEC,
 943         [UNIT_HERTZ]
 944             = VOICE_HERTZ,
 945         [UNIT_MB]
 946             = LANG_MEGABYTE,
 947         [UNIT_KBIT]
 948             = VOICE_KBIT_PER_SEC,
 949         [UNIT_PM_TICK]
 950             = VOICE_PM_UNITS_PER_TICK,
 951     };
 952
 953     static const int pow10[] = { /* 10^0 - 10^7 */
 954         1, 10, 100, 1000, 10000, 100000, 1000000, 10000000
 955     };
 956
 957     char tbuf[8];
 958     char fmt[] = "%0nd";
 959
 960     if (talk_temp_disable_count > 0)
 961         return -1;  /* talking has been disabled */
 962 #if CONFIG_CODEC != SWCODEC
 963     if (audio_status()) /* busy, buffer in use */
 964         return -1;
 965 #endif
 966
 967     /* special case for time duration */
 968     if (unit == UNIT_TIME || unit == UNIT_TIME_EXACT)
 969         return talk_time_unit(n, unit == UNIT_TIME_EXACT, enqueue);
 970
 971     if (unit < 0 || unit >= UNIT_LAST)
 972         unit_id = -1;
 973     else
 974         unit_id = unit_voiced[unit];
 975
 976     if ((n==1 || n==-1) /* singular? */
 977         && unit_id >= VOICE_SECONDS && unit_id <= VOICE_HOURS)
 978     {
 979         unit_id--; /* use the singular for those units which have */
 980     }
 981
 982     /* special case with a "plus" before */
 983     if (n > 0 && (unit == UNIT_SIGNED || unit == UNIT_DB))
 984     {
 985         talk_id(VOICE_PLUS, enqueue);
 986         enqueue = true;
 987     }
 988
 989     if (decimals)
 990     {
 991         /* needed for the "-0.5" corner case */
 992         if (n < 0)
 993         {
 994             talk_id(VOICE_MINUS, enqueue);
 995             n = -n;
 996         }
 997
 998         fmt[2] = '0' + decimals;
 999
1000         snprintf(tbuf, sizeof(tbuf), fmt, n % pow10[decimals]);
1001         talk_fractional(tbuf, n / pow10[decimals], unit_id);
1002
1003         return 0;
1004     }
1005
1006     talk_number(n, enqueue); /* say the number */
1007     talk_id(unit_id, true); /* say the unit, if any */
1008
1009     return 0;
1010 }
1011
1012 /* spell a string */
1013 int talk_spell(const char* spell, bool enqueue)
1014 {
1015     char c; /* currently processed char */
1016
1017     if (talk_temp_disable_count > 0)
1018         return -1;  /* talking has been disabled */
1019 #if CONFIG_CODEC != SWCODEC
1020     if (audio_status()) /* busy, buffer in use */
1021         return -1;
1022 #endif
1023
1024     if (!enqueue)
1025         talk_shutup(); /* cut off all the pending stuff */
1026
1027     while ((c = *spell++) != '\0')
1028     {
1029         /* if this grows into too many cases, I should use a table */
1030         if (c >= 'A' && c <= 'Z')
1031             talk_id(VOICE_CHAR_A + c - 'A', true);
1032         else if (c >= 'a' && c <= 'z')
1033             talk_id(VOICE_CHAR_A + c - 'a', true);
1034         else if (c >= '0' && c <= '9')
1035             talk_id(VOICE_ZERO + c - '0', true);
1036         else if (c == '-')
1037             talk_id(VOICE_MINUS, true);
1038         else if (c == '+')
1039             talk_id(VOICE_PLUS, true);
1040         else if (c == '.')
1041             talk_id(VOICE_DOT, true);
1042         else if (c == ' ')
1043             talk_id(VOICE_PAUSE, true);
1044         else if (c == '/')
1045             talk_id(VOICE_CHAR_SLASH, true);
1046     }
1047
1048     return 0;
1049 }
1050
1051 void talk_disable(bool disable)
1052 {
1053     if (disable)
1054         talk_temp_disable_count++;
1055     else
1056         talk_temp_disable_count--;
1057 }
1058
1059 void talk_setting(const void *global_settings_variable)
1060 {
1061     const struct settings_list *setting;
1062     if (!global_settings.talk_menu)
1063         return;
1064     setting = find_setting(global_settings_variable, NULL);
1065     if (setting == NULL)
1066         return;
1067     if (setting->lang_id)
1068         talk_id(setting->lang_id,false);
1069 }
1070
1071
1072 #if CONFIG_RTC
1073 void talk_date(const struct tm *tm, bool enqueue)
1074 {
1075     talk_id(LANG_MONTH_JANUARY + tm->tm_mon, enqueue);
1076     talk_number(tm->tm_mday, true);
1077     talk_number(1900 + tm->tm_year, true);
1078 }
1079
1080 void talk_time(const struct tm *tm, bool enqueue)
1081 {
1082     if (global_settings.timeformat == 1)
1083     {
1084         /* Voice the hour */
1085         long am_pm_id = VOICE_AM;
1086         int hour = tm->tm_hour;
1087         if (hour >= 12)
1088         {
1089             am_pm_id = VOICE_PM;
1090             hour -= 12;
1091         }
1092         if (hour == 0)
1093             hour = 12;
1094         talk_number(hour, enqueue);
1095
1096         /* Voice the minutes */
1097         if (tm->tm_min == 0)
1098         {
1099             /* Say o'clock if the minute is 0. */
1100             talk_id(VOICE_OCLOCK, true);
1101         }
1102         else
1103         {
1104             /* Pronounce the leading 0 */
1105             if(tm->tm_min < 10)
1106                 talk_id(VOICE_OH, true);
1107             talk_number(tm->tm_min, true);
1108         }
1109         talk_id(am_pm_id, true);
1110     }
1111     else
1112     {
1113         /* Voice the time in 24 hour format */
1114         talk_number(tm->tm_hour, enqueue);
1115         if (tm->tm_min == 0)
1116         {
1117             talk_id(VOICE_HUNDRED, true);
1118             talk_id(VOICE_HOUR, true);
1119         }
1120         else
1121         {
1122             /* Pronounce the leading 0 */
1123             if(tm->tm_min < 10)
1124                 talk_id(VOICE_OH, true);
1125             talk_number(tm->tm_min, true);
1126         }
1127     }
1128 }
1129
1130 #endif