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-extra.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 /*#define LOGF_ENABLE*/
  41 #include "logf.h"
  42 #include "bitswap.h"
  43 #include "structec.h"
  44 #include "debug.h"
  45
  46
  47 /* Memory layout varies between targets because the
  48    Archos (MASCODEC) devices cannot mix voice and audio playback
  49
  50              MASCODEC  | MASCODEC  | SWCODEC
  51              (playing) | (stopped) |
  52     audiobuf-----------+-----------+------------
  53               audio    | voice     | thumbnail
  54                        |-----------|------------
  55                        | thumbnail | voice
  56                        |           |------------
  57                        |           | filebuf
  58                        |           |------------
  59                        |           | audio
  60   audiobufend----------+-----------+------------
  61
  62   SWCODEC allocates dedicated buffers, MASCODEC reuses audiobuf. */
  63
  64
  65 /***************** Constants *****************/
  66
  67 #define QUEUE_SIZE 64 /* must be a power of two */
  68 #define QUEUE_MASK (QUEUE_SIZE-1)
  69 const char* const dir_thumbnail_name = "_dirname.talk";
  70 const char* const file_thumbnail_ext = ".talk";
  71
  72 /***************** Functional Macros *****************/
  73
  74 #define QUEUE_LEVEL ((queue_write - queue_read) & QUEUE_MASK)
  75
  76 #define LOADED_MASK 0x80000000 /* MSB */
  77
  78 #if CONFIG_CODEC == SWCODEC
  79 #define MAX_THUMBNAIL_BUFSIZE 0x10000
  80 #endif
  81
  82 /***************** Data types *****************/
  83
  84 struct clip_entry /* one entry of the index table */
  85 {
  86     int offset; /* offset from start of voicefile file */
  87     int size; /* size of the clip */
  88 };
  89
  90 struct voicefile /* file format of our voice file */
  91 {
  92     int version; /* version of the voicefile */
  93     int target_id; /* the rockbox target the file was made for */
  94     int table;   /* offset to index table, (=header size) */
  95     int id1_max; /* number of "normal" clips contained in above index */
  96     int id2_max; /* number of "voice only" clips contained in above index */
  97     struct clip_entry index[]; /* followed by the index tables */
  98     /* and finally the mp3 clips, not visible here, bitswapped
  99        for SH based players */
 100 };
 101
 102 struct queue_entry /* one entry of the internal queue */
 103 {
 104     unsigned char* buf;
 105     long len;
 106 };
 107
 108
 109 /***************** Globals *****************/
 110
 111 static unsigned char* p_thumbnail = NULL; /* buffer for thumbnails */
 112 /* Multiple thumbnails can be loaded back-to-back in this buffer. */
 113 static volatile int thumbnail_buf_used SHAREDBSS_ATTR; /* length of data in
 114                                                           thumbnail buffer */
 115 static long size_for_thumbnail; /* total thumbnail buffer size */
 116 static struct voicefile* p_voicefile; /* loaded voicefile */
 117 static bool has_voicefile; /* a voicefile file is present */
 118 static bool need_shutup; /* is there possibly any voice playing to be shutup */
 119 static struct queue_entry queue[QUEUE_SIZE]; /* queue of scheduled clips */
 120 static bool force_enqueue_next; /* enqueue next utterance even if enqueue is false */
 121 static int queue_write; /* write index of queue, by application */
 122 static int queue_read; /* read index of queue, by ISR context */
 123 #if CONFIG_CODEC == SWCODEC
 124 /* protects queue_read, queue_write and thumbnail_buf_used */
 125 static struct mutex queue_mutex SHAREDBSS_ATTR;
 126 #define talk_queue_lock() ({ mutex_lock(&queue_mutex); })
 127 #define talk_queue_unlock() ({ mutex_unlock(&queue_mutex); })
 128 #else
 129 #define talk_queue_lock() ({ })
 130 #define talk_queue_unlock() ({ })
 131 #endif /* CONFIG_CODEC */
 132 static int sent; /* how many bytes handed over to playback, owned by ISR */
 133 static unsigned char curr_hd[3]; /* current frame header, for re-sync */
 134 static int filehandle = -1; /* global, so the MMC variant can keep the file open */
 135 static unsigned char* p_silence; /* VOICE_PAUSE clip, used for termination */
 136 static long silence_len; /* length of the VOICE_PAUSE clip */
 137 static unsigned char* p_lastclip; /* address of latest clip, for silence add */
 138 static unsigned long voicefile_size = 0; /* size of the loaded voice file */
 139 static unsigned char last_lang[MAX_FILENAME+1]; /* name of last used lang file (in talk_init) */
 140 static bool talk_initialized; /* true if talk_init has been called */
 141 static int talk_temp_disable_count; /* if positive, temporarily disable voice UI (not saved) */
 142
 143
 144 /***************** Private implementation *****************/
 145
 146 static int open_voicefile(void)
 147 {
 148     char buf[64];
 149     char* p_lang = "english"; /* default */
 150
 151     if ( global_settings.lang_file[0] &&
 152          global_settings.lang_file[0] != 0xff )
 153     {   /* try to open the voice file of the selected language */
 154         p_lang = (char *)global_settings.lang_file;
 155     }
 156
 157     snprintf(buf, sizeof(buf), LANG_DIR "/%s.voice", p_lang);
 158
 159     return open(buf, O_RDONLY);
 160 }
 161
 162
 163 /* fetch a clip from the voice file */
 164 static unsigned char* get_clip(long id, long* p_size)
 165 {
 166     long clipsize;
 167     unsigned char* clipbuf;
 168
 169     if (id > VOICEONLY_DELIMITER)
 170     {   /* voice-only entries use the second part of the table */
 171         id -= VOICEONLY_DELIMITER + 1;
 172         if (id >= p_voicefile->id2_max)
 173             return NULL; /* must be newer than we have */
 174         id += p_voicefile->id1_max; /* table 2 is behind table 1 */
 175     }
 176     else
 177     {   /* normal use of the first table */
 178         if (id >= p_voicefile->id1_max)
 179             return NULL; /* must be newer than we have */
 180     }
 181
 182     clipsize = p_voicefile->index[id].size;
 183     if (clipsize == 0) /* clip not included in voicefile */
 184         return NULL;
 185     clipbuf = (unsigned char *) p_voicefile + p_voicefile->index[id].offset;
 186
 187 #if (CONFIG_STORAGE & STORAGE_MMC) /* dynamic loading, on demand */
 188     if (!(clipsize & LOADED_MASK))
 189     {   /* clip used for the first time, needs loading */
 190         lseek(filehandle, p_voicefile->index[id].offset, SEEK_SET);
 191         if (read(filehandle, clipbuf, clipsize) != clipsize)
 192             return NULL; /* read error */
 193
 194         p_voicefile->index[id].size |= LOADED_MASK; /* mark as loaded */
 195     }
 196     else
 197     {   /* clip is in memory already */
 198         clipsize &= ~LOADED_MASK; /* without the extra bit gives true size */
 199     }
 200 #endif
 201
 202     *p_size = clipsize;
 203     return clipbuf;
 204 }
 205
 206
 207 /* load the voice file into the mp3 buffer */
 208 static void load_voicefile(void)
 209 {
 210     int load_size;
 211     int got_size;
 212     int file_size;
 213 #ifdef ROCKBOX_LITTLE_ENDIAN
 214     int i;
 215 #endif
 216
 217     filehandle = open_voicefile();
 218     if (filehandle < 0) /* failed to open */
 219         goto load_err;
 220
 221     file_size = filesize(filehandle);
 222     if (file_size > audiobufend - audiobuf) /* won't fit? */
 223         goto load_err;
 224
 225 #if (CONFIG_STORAGE & STORAGE_MMC) /* load only the header for now */
 226     load_size = offsetof(struct voicefile, index);
 227 #else /* load the full file */
 228     load_size = file_size;
 229 #endif
 230
 231     got_size = read(filehandle, audiobuf, load_size);
 232     if (got_size != load_size /* failure */)
 233         goto load_err;
 234
 235 #ifdef ROCKBOX_LITTLE_ENDIAN
 236     logf("Byte swapping voice file");
 237     structec_convert(audiobuf, "lllll", 1, true);
 238 #endif
 239
 240     if (((struct voicefile*)audiobuf)->table /* format check */
 241            == offsetof(struct voicefile, index))
 242     {
 243         p_voicefile = (struct voicefile*)audiobuf;
 244
 245         if (p_voicefile->version != VOICE_VERSION ||
 246             p_voicefile->target_id != TARGET_ID)
 247         {
 248             logf("Incompatible voice file");
 249             goto load_err;
 250         }
 251 #if CONFIG_CODEC != SWCODEC
 252         /* MASCODEC: now use audiobuf for voice then thumbnail */
 253         p_thumbnail = audiobuf + file_size;
 254         p_thumbnail += (long)p_thumbnail % 2; /* 16-bit align */
 255         size_for_thumbnail = audiobufend - p_thumbnail;
 256 #endif
 257     }
 258     else
 259         goto load_err;
 260
 261 #ifdef ROCKBOX_LITTLE_ENDIAN
 262     for (i = 0; i < p_voicefile->id1_max + p_voicefile->id2_max; i++)
 263         structec_convert(&p_voicefile->index[i], "ll", 1, true);
 264 #endif
 265
 266 #if (CONFIG_STORAGE & STORAGE_MMC)
 267     /* load the index table, now that we know its size from the header */
 268     load_size = (p_voicefile->id1_max + p_voicefile->id2_max)
 269                 * sizeof(struct clip_entry);
 270     got_size = read(filehandle,
 271                     (unsigned char *) p_voicefile + offsetof(struct voicefile, index), load_size);
 272     if (got_size != load_size) /* read error */
 273         goto load_err;
 274 #else
 275     close(filehandle); /* only the MMC variant leaves it open */
 276     filehandle = -1;
 277 #endif
 278
 279     /* make sure to have the silence clip, if available */
 280     p_silence = get_clip(VOICE_PAUSE, &silence_len);
 281
 282     return;
 283
 284 load_err:
 285     p_voicefile = NULL;
 286     has_voicefile = false; /* don't try again */
 287     if (filehandle >= 0)
 288     {
 289         close(filehandle);
 290         filehandle = -1;
 291     }
 292     return;
 293 }
 294
 295
 296 /* called in ISR context if mp3 data got consumed */
 297 static void mp3_callback(unsigned char** start, size_t* size)
 298 {
 299     queue[queue_read].len -= sent; /* we completed this */
 300     queue[queue_read].buf += sent;
 301
 302     if (queue[queue_read].len > 0) /* current clip not finished? */
 303     {   /* feed the next 64K-1 chunk */
 304 #if CONFIG_CODEC != SWCODEC
 305         sent = MIN(queue[queue_read].len, 0xFFFF);
 306 #else
 307         sent = queue[queue_read].len;
 308 #endif
 309         *start = queue[queue_read].buf;
 310         *size = sent;
 311         return;
 312     }
 313     talk_queue_lock();
 314     if(p_thumbnail
 315        && queue[queue_read].buf == p_thumbnail +thumbnail_buf_used)
 316         thumbnail_buf_used = 0;
 317     if (sent > 0) /* go to next entry */
 318     {
 319         queue_read = (queue_read + 1) & QUEUE_MASK;
 320     }
 321
 322 re_check:
 323
 324     if (QUEUE_LEVEL != 0) /* queue is not empty? */
 325     {   /* start next clip */
 326 #if CONFIG_CODEC != SWCODEC
 327         sent = MIN(queue[queue_read].len, 0xFFFF);
 328 #else
 329         sent = queue[queue_read].len;
 330 #endif
 331         *start = p_lastclip = queue[queue_read].buf;
 332         *size = sent;
 333         curr_hd[0] = p_lastclip[1];
 334         curr_hd[1] = p_lastclip[2];
 335         curr_hd[2] = p_lastclip[3];
 336     }
 337     else if (p_silence != NULL             /* silence clip available */
 338              && p_lastclip != p_silence    /* previous clip wasn't silence */
 339              && !(p_lastclip >= p_thumbnail /* ..or thumbnail */
 340                   && p_lastclip < p_thumbnail +size_for_thumbnail))
 341     {   /* add silence clip when queue runs empty playing a voice clip */
 342         queue[queue_write].buf = p_silence;
 343         queue[queue_write].len = silence_len;
 344         queue_write = (queue_write + 1) & QUEUE_MASK;
 345
 346         goto re_check;
 347     }
 348     else
 349     {
 350         *size = 0; /* end of data */
 351     }
 352     talk_queue_unlock();
 353 }
 354
 355 /***************** Public routines *****************/
 356
 357 /* stop the playback and the pending clips */
 358 void talk_force_shutup(void)
 359 {
 360     /* Most of this is MAS only */
 361 #if CONFIG_CODEC != SWCODEC
 362 #ifdef SIMULATOR
 363     return;
 364 #endif
 365     unsigned char* pos;
 366     unsigned char* search;
 367     unsigned char* end;
 368     if (QUEUE_LEVEL == 0) /* has ended anyway */
 369         return;
 370
 371 #if CONFIG_CPU == SH7034
 372     CHCR3 &= ~0x0001; /* disable the DMA (and therefore the interrupt also) */
 373 #endif /* CONFIG_CPU == SH7034 */
 374     /* search next frame boundary and continue up to there */
 375     pos = search = mp3_get_pos();
 376     end = queue[queue_read].buf + queue[queue_read].len;
 377
 378     if (pos >= queue[queue_read].buf
 379         && pos <= end) /* really our clip? */
 380     { /* (for strange reasons this isn't nesessarily the case) */
 381         /* find the next frame boundary */
 382         while (search < end) /* search the remaining data */
 383         {
 384             if (*search++ != 0xFF) /* quick search for frame sync byte */
 385                 continue; /* (this does the majority of the job) */
 386
 387             /* look at the (bitswapped) rest of header candidate */
 388             if (search[0] == curr_hd[0] /* do the quicker checks first */
 389              && search[2] == curr_hd[2]
 390              && (search[1] & 0x30) == (curr_hd[1] & 0x30)) /* sample rate */
 391             {
 392                 search--; /* back to the sync byte */
 393                 break; /* From looking at it, this is our header. */
 394             }
 395         }
 396
 397         if (search-pos)
 398         {   /* play old data until the frame end, to keep the MAS in sync */
 399             sent = search-pos;
 400
 401             queue_write = (queue_read + 1) & QUEUE_MASK; /* will be empty after next callback */
 402             queue[queue_read].len = sent; /* current one ends after this */
 403
 404 #if CONFIG_CPU == SH7034
 405             DTCR3 = sent; /* let the DMA finish this frame */
 406             CHCR3 |= 0x0001; /* re-enable DMA */
 407 #endif /* CONFIG_CPU == SH7034 */
 408             thumbnail_buf_used = 0;
 409             return;
 410         }
 411     }
 412 #endif /* CONFIG_CODEC != SWCODEC */
 413
 414     /* Either SWCODEC, or MAS had nothing to do (was frame boundary or not our clip) */
 415     mp3_play_stop();
 416     talk_queue_lock();
 417     queue_write = queue_read = 0; /* reset the queue */
 418     thumbnail_buf_used = 0;
 419     talk_queue_unlock();
 420     need_shutup = false;
 421 }
 422
 423 /* Shutup the voice, except if force_enqueue_next is set. */
 424 void talk_shutup(void)
 425 {
 426     if (need_shutup && !force_enqueue_next)
 427         talk_force_shutup();
 428 }
 429
 430 /* schedule a clip, at the end or discard the existing queue */
 431 static void queue_clip(unsigned char* buf, long size, bool enqueue)
 432 {
 433     int queue_level;
 434
 435     if (!enqueue)
 436         talk_shutup(); /* cut off all the pending stuff */
 437     /* Something is being enqueued, force_enqueue_next override is no
 438        longer in effect. */
 439     force_enqueue_next = false;
 440
 441     if (!size)
 442         return; /* safety check */
 443 #if CONFIG_CPU == SH7034
 444     /* disable the DMA temporarily, to be safe of race condition */
 445     CHCR3 &= ~0x0001;
 446 #endif
 447     talk_queue_lock();
 448     queue_level = QUEUE_LEVEL; /* check old level */
 449
 450     if (queue_level < QUEUE_SIZE - 1) /* space left? */
 451     {
 452         queue[queue_write].buf = buf; /* populate an entry */
 453         queue[queue_write].len = size;
 454         queue_write = (queue_write + 1) & QUEUE_MASK;
 455     }
 456     talk_queue_unlock();
 457
 458     if (queue_level == 0)
 459     {   /* queue was empty, we have to do the initial start */
 460         p_lastclip = buf;
 461 #if CONFIG_CODEC != SWCODEC
 462         sent = MIN(size, 0xFFFF); /* DMA can do no more */
 463 #else
 464         sent = size;
 465 #endif
 466         mp3_play_data(buf, sent, mp3_callback);
 467         curr_hd[0] = buf[1];
 468         curr_hd[1] = buf[2];
 469         curr_hd[2] = buf[3];
 470         mp3_play_pause(true); /* kickoff audio */
 471     }
 472     else
 473     {
 474 #if CONFIG_CPU == SH7034
 475         CHCR3 |= 0x0001; /* re-enable DMA */
 476 #endif
 477     }
 478
 479     need_shutup = true;
 480
 481     return;
 482 }
 483
 484
 485 /* common code for talk_init() and talk_buffer_steal() */
 486 static void reset_state(void)
 487 {
 488     queue_write = queue_read = 0; /* reset the queue */
 489     p_voicefile = NULL; /* indicate no voicefile (trashed) */
 490 #if CONFIG_CODEC == SWCODEC
 491     /* Allocate a dedicated thumbnail buffer - once */
 492     if (p_thumbnail == NULL)
 493     {
 494         size_for_thumbnail = audiobufend - audiobuf;
 495         if (size_for_thumbnail > MAX_THUMBNAIL_BUFSIZE)
 496             size_for_thumbnail = MAX_THUMBNAIL_BUFSIZE;
 497         p_thumbnail = buffer_alloc(size_for_thumbnail);
 498     }
 499 #else
 500     /* Just use the audiobuf, without allocating anything */
 501     p_thumbnail = audiobuf;
 502     size_for_thumbnail = audiobufend - audiobuf;
 503 #endif
 504     thumbnail_buf_used = 0;
 505     p_silence = NULL; /* pause clip not accessible */
 506 }
 507
 508
 509 /***************** Public implementation *****************/
 510
 511 void talk_init(void)
 512 {
 513     talk_temp_disable_count = 0;
 514     if (talk_initialized && !strcasecmp(last_lang, global_settings.lang_file))
 515     {
 516         /* not a new file, nothing to do */
 517         return;
 518     }
 519
 520 #if (CONFIG_STORAGE & STORAGE_MMC)
 521     if (filehandle >= 0) /* MMC: An old voice file might still be open */
 522     {
 523         close(filehandle);
 524         filehandle = -1;
 525     }
 526 #endif
 527
 528 #if CONFIG_CODEC == SWCODEC
 529     if(!talk_initialized)
 530         mutex_init(&queue_mutex);
 531 #endif /* CONFIG_CODEC == SWCODEC */
 532
 533     talk_initialized = true;
 534     strlcpy((char *)last_lang, (char *)global_settings.lang_file,
 535         MAX_FILENAME);
 536
 537 #if CONFIG_CODEC == SWCODEC
 538     audio_get_buffer(false, NULL);  /* Must tell audio to reinitialize */
 539 #endif
 540     reset_state(); /* use this for most of our inits */
 541
 542     filehandle = open_voicefile();
 543     size_t audiobufsz = audiobufend - audiobuf;
 544     /* test if we can open and if it fits in the audiobuffer */
 545     has_voicefile = filehandle >= 0 && filesize(filehandle) <= (off_t)audiobufsz;
 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 #ifdef LOGF_ENABLE
 632     if (id > VOICEONLY_DELIMITER)
 633         logf("\ntalk_id: Say voice clip 0x%x\n", id);
 634     else
 635         logf("\ntalk_id: Say '%s'\n", str(id));
 636 #endif
 637
 638     queue_clip(clipbuf, clipsize, enqueue);
 639
 640     return 0;
 641 }
 642 /* Speaks zero or more IDs (from an array). */
 643 int talk_idarray(const long *ids, bool enqueue)
 644 {
 645     int r;
 646     if(!ids)
 647         return 0;
 648     while(*ids != TALK_FINAL_ID)
 649     {
 650         if((r = talk_id(*ids++, enqueue)) <0)
 651             return r;
 652         enqueue = true;
 653     }
 654     return 0;
 655 }
 656
 657 /* Make sure the current utterance is not interrupted by the next one. */
 658 void talk_force_enqueue_next(void)
 659 {
 660     force_enqueue_next = true;
 661 }
 662
 663 /* play a thumbnail from file */
 664 /* Returns size of spoken thumbnail, so >0 means something is spoken,
 665    <=0 means something went wrong. */
 666 static int _talk_file(const char* filename,
 667                       const long *prefix_ids, bool enqueue)
 668 {
 669     int fd;
 670     int size;
 671     int thumb_used;
 672 #if CONFIG_CODEC != SWCODEC
 673     struct mp3entry info;
 674 #endif
 675
 676     if (talk_temp_disable_count > 0)
 677         return -1;  /* talking has been disabled */
 678 #if CONFIG_CODEC != SWCODEC
 679     if (audio_status()) /* busy, buffer in use */
 680         return -1;
 681 #endif
 682
 683     if (p_thumbnail == NULL || size_for_thumbnail <= 0)
 684         return -1;
 685
 686 #if CONFIG_CODEC != SWCODEC
 687     if(mp3info(&info, filename)) /* use this to find real start */
 688     {
 689         return 0; /* failed to open, or invalid */
 690     }
 691 #endif
 692
 693     if (!enqueue)
 694         /* shutup now to free the thumbnail buffer */
 695         talk_shutup();
 696
 697     fd = open(filename, O_RDONLY);
 698     if (fd < 0) /* failed to open */
 699     {
 700         return 0;
 701     }
 702
 703     thumb_used = thumbnail_buf_used;
 704     if(filesize(fd) > size_for_thumbnail -thumb_used)
 705     {   /* Don't play truncated  clips */
 706         close(fd);
 707         return 0;
 708     }
 709
 710 #if CONFIG_CODEC != SWCODEC
 711     lseek(fd, info.first_frame_offset, SEEK_SET); /* behind ID data */
 712 #endif
 713
 714     size = read(fd, p_thumbnail +thumb_used,
 715                 size_for_thumbnail -thumb_used);
 716     close(fd);
 717
 718     /* ToDo: find audio, skip ID headers and trailers */
 719
 720     if (size > 0)    /* Don't play missing clips */
 721     {
 722 #if CONFIG_CODEC != SWCODEC && !defined(SIMULATOR)
 723         bitswap(p_thumbnail, size);
 724 #endif
 725         if(prefix_ids)
 726             /* prefix thumbnail by speaking these ids, but only now
 727                that we know there's actually a thumbnail to be
 728                spoken. */
 729             talk_idarray(prefix_ids, true);
 730         talk_queue_lock();
 731         thumbnail_buf_used = thumb_used +size;
 732         talk_queue_unlock();
 733         queue_clip(p_thumbnail +thumb_used, size, true);
 734     }
 735
 736     return size;
 737 }
 738
 739 int talk_file(const char *root, const char *dir, const char *file,
 740               const char *ext, const long *prefix_ids, bool enqueue)
 741 /* Play a thumbnail file */
 742 {
 743     char buf[MAX_PATH];
 744     /* Does root end with a slash */
 745     char *slash = (root && root[0]
 746                    && root[strlen(root)-1] != '/') ? "/" : "";
 747     snprintf(buf, MAX_PATH, "%s%s%s%s%s%s",
 748              root ? root : "", slash,
 749              dir ? dir : "", dir ? "/" : "",
 750              file ? file : "",
 751              ext ? ext : "");
 752     return _talk_file(buf, prefix_ids, enqueue);
 753 }
 754
 755 static int talk_spell_basename(const char *path,
 756                                const long *prefix_ids, bool enqueue)
 757 {
 758     if(prefix_ids)
 759     {
 760         talk_idarray(prefix_ids, enqueue);
 761         enqueue = true;
 762     }
 763     char buf[MAX_PATH];
 764     /* Spell only the path component after the last slash */
 765     strlcpy(buf, path, sizeof(buf));
 766     if(strlen(buf) >1 && buf[strlen(buf)-1] == '/')
 767         /* strip trailing slash */
 768         buf[strlen(buf)-1] = '\0';
 769     char *ptr = strrchr(buf, '/');
 770     if(ptr && strlen(buf) >1)
 771         ++ptr;
 772     else ptr = buf;
 773     return talk_spell(ptr, enqueue);
 774 }
 775
 776 /* Play a file's .talk thumbnail, fallback to spelling the filename, or
 777    go straight to spelling depending on settings. */
 778 int talk_file_or_spell(const char *dirname, const char *filename,
 779                        const long *prefix_ids, bool enqueue)
 780 {
 781     if (global_settings.talk_file_clip)
 782     {   /* .talk clips enabled */
 783         if(talk_file(dirname, NULL, filename, file_thumbnail_ext,
 784                               prefix_ids, enqueue) >0)
 785             return 0;
 786     }
 787     if (global_settings.talk_file == 2)
 788         /* Either .talk clips are disabled, or as a fallback */
 789         return talk_spell_basename(filename, prefix_ids, enqueue);
 790     return 0;
 791 }
 792
 793 /* Play a directory's .talk thumbnail, fallback to spelling the filename, or
 794    go straight to spelling depending on settings. */
 795 int talk_dir_or_spell(const char* dirname,
 796                       const long *prefix_ids, bool enqueue)
 797 {
 798     if (global_settings.talk_dir_clip)
 799     {   /* .talk clips enabled */
 800         if(talk_file(dirname, NULL, dir_thumbnail_name, NULL,
 801                               prefix_ids, enqueue) >0)
 802             return 0;
 803     }
 804     if (global_settings.talk_dir == 2)
 805         /* Either .talk clips disabled or as a fallback */
 806         return talk_spell_basename(dirname, prefix_ids, enqueue);
 807     return 0;
 808 }
 809
 810 /* say a numeric value, this word ordering works for english,
 811    but not necessarily for other languages (e.g. german) */
 812 int talk_number(long n, bool enqueue)
 813 {
 814     int level = 2; /* mille count */
 815     long mil = 1000000000; /* highest possible "-illion" */
 816
 817     if (talk_temp_disable_count > 0)
 818         return -1;  /* talking has been disabled */
 819 #if CONFIG_CODEC != SWCODEC
 820     if (audio_status()) /* busy, buffer in use */
 821         return -1;
 822 #endif
 823
 824     if (!enqueue)
 825         talk_shutup(); /* cut off all the pending stuff */
 826
 827     if (n==0)
 828     {   /* special case */
 829         talk_id(VOICE_ZERO, true);
 830         return 0;
 831     }
 832
 833     if (n<0)
 834     {
 835         talk_id(VOICE_MINUS, true);
 836         n = -n;
 837     }
 838
 839     while (n)
 840     {
 841         int segment = n / mil; /* extract in groups of 3 digits */
 842         n -= segment * mil; /* remove the used digits from number */
 843         mil /= 1000; /* digit place for next round */
 844
 845         if (segment)
 846         {
 847             int hundreds = segment / 100;
 848             int ones = segment % 100;
 849
 850             if (hundreds)
 851             {
 852                 talk_id(VOICE_ZERO + hundreds, true);
 853                 talk_id(VOICE_HUNDRED, true);
 854             }
 855
 856             /* combination indexing */
 857             if (ones > 20)
 858             {
 859                int tens = ones/10 + 18;
 860                talk_id(VOICE_ZERO + tens, true);
 861                ones %= 10;
 862             }
 863
 864             /* direct indexing */
 865             if (ones)
 866                 talk_id(VOICE_ZERO + ones, true);
 867
 868             /* add billion, million, thousand */
 869             if (mil)
 870                 talk_id(VOICE_THOUSAND + level, true);
 871         }
 872         level--;
 873     }
 874
 875     return 0;
 876 }
 877
 878 /* Say time duration/interval. Input is time in seconds,
 879    say hours,minutes,seconds. */
 880 static int talk_time_unit(long secs, bool exact, bool enqueue)
 881 {
 882     int hours, mins;
 883     if (!enqueue)
 884         talk_shutup();
 885     if((hours = secs/3600)) {
 886         secs %= 3600;
 887         talk_value(hours, UNIT_HOUR, true);
 888     }
 889     if((mins = secs/60)) {
 890         secs %= 60;
 891         if(exact || !hours)
 892             talk_value(mins, UNIT_MIN, true);
 893         else talk_number(mins, true); /* don't say "minutes" */
 894     }
 895     if((exact && secs) || (!hours && !mins))
 896         talk_value(secs, UNIT_SEC, true);
 897     else if(!hours && secs)
 898         talk_number(secs, true);
 899     return 0;
 900 }
 901
 902 void talk_fractional(char *tbuf, int value, int unit)
 903 {
 904     int i;
 905     /* strip trailing zeros from the fraction */
 906     for (i = strlen(tbuf) - 1; (i >= 0) && (tbuf[i] == '0'); i--)
 907         tbuf[i] = '\0';
 908
 909     talk_number(value, true);
 910     if (tbuf[0] != 0)
 911     {
 912         talk_id(LANG_POINT, true);
 913         talk_spell(tbuf, true);
 914     }
 915     talk_id(unit, true);
 916 }
 917
 918 int talk_value(long n, int unit, bool enqueue)
 919 {
 920     return talk_value_decimal(n, unit, 0, enqueue);
 921 }
 922
 923 /* singular/plural aware saying of a value */
 924 int talk_value_decimal(long n, int unit, int decimals, bool enqueue)
 925 {
 926     int unit_id;
 927     static const int unit_voiced[] =
 928     {   /* lookup table for the voice ID of the units */
 929         [0 ... UNIT_LAST-1] = -1, /* regular ID, int, signed */
 930         [UNIT_MS]
 931             = VOICE_MILLISECONDS, /* here come the "real" units */
 932         [UNIT_SEC]
 933             = VOICE_SECONDS,
 934         [UNIT_MIN]
 935             = VOICE_MINUTES,
 936         [UNIT_HOUR]
 937             = VOICE_HOURS,
 938         [UNIT_KHZ]
 939             = VOICE_KHZ,
 940         [UNIT_DB]
 941             = VOICE_DB,
 942         [UNIT_PERCENT]
 943             = VOICE_PERCENT,
 944         [UNIT_MAH]
 945             = VOICE_MILLIAMPHOURS,
 946         [UNIT_PIXEL]
 947             = VOICE_PIXEL,
 948         [UNIT_PER_SEC]
 949             = VOICE_PER_SEC,
 950         [UNIT_HERTZ]
 951             = VOICE_HERTZ,
 952         [UNIT_MB]
 953             = LANG_MEGABYTE,
 954         [UNIT_KBIT]
 955             = VOICE_KBIT_PER_SEC,
 956         [UNIT_PM_TICK]
 957             = VOICE_PM_UNITS_PER_TICK,
 958     };
 959
 960     static const int pow10[] = { /* 10^0 - 10^7 */
 961         1, 10, 100, 1000, 10000, 100000, 1000000, 10000000
 962     };
 963
 964     char tbuf[8];
 965     char fmt[] = "%0nd";
 966
 967     if (talk_temp_disable_count > 0)
 968         return -1;  /* talking has been disabled */
 969 #if CONFIG_CODEC != SWCODEC
 970     if (audio_status()) /* busy, buffer in use */
 971         return -1;
 972 #endif
 973
 974     /* special case for time duration */
 975     if (unit == UNIT_TIME || unit == UNIT_TIME_EXACT)
 976         return talk_time_unit(n, unit == UNIT_TIME_EXACT, enqueue);
 977
 978     if (unit < 0 || unit >= UNIT_LAST)
 979         unit_id = -1;
 980     else
 981         unit_id = unit_voiced[unit];
 982
 983     if ((n==1 || n==-1) /* singular? */
 984         && unit_id >= VOICE_SECONDS && unit_id <= VOICE_HOURS)
 985     {
 986         unit_id--; /* use the singular for those units which have */
 987     }
 988
 989     /* special case with a "plus" before */
 990     if (n > 0 && (unit == UNIT_SIGNED || unit == UNIT_DB))
 991     {
 992         talk_id(VOICE_PLUS, enqueue);
 993         enqueue = true;
 994     }
 995
 996     if (decimals)
 997     {
 998         /* needed for the "-0.5" corner case */
 999         if (n < 0)
1000         {
1001             talk_id(VOICE_MINUS, enqueue);
1002             n = -n;
1003         }
1004
1005         fmt[2] = '0' + decimals;
1006
1007         snprintf(tbuf, sizeof(tbuf), fmt, n % pow10[decimals]);
1008         talk_fractional(tbuf, n / pow10[decimals], unit_id);
1009
1010         return 0;
1011     }
1012
1013     talk_number(n, enqueue); /* say the number */
1014     talk_id(unit_id, true); /* say the unit, if any */
1015
1016     return 0;
1017 }
1018
1019 /* spell a string */
1020 int talk_spell(const char* spell, bool enqueue)
1021 {
1022     char c; /* currently processed char */
1023
1024     if (talk_temp_disable_count > 0)
1025         return -1;  /* talking has been disabled */
1026 #if CONFIG_CODEC != SWCODEC
1027     if (audio_status()) /* busy, buffer in use */
1028         return -1;
1029 #endif
1030
1031     if (!enqueue)
1032         talk_shutup(); /* cut off all the pending stuff */
1033
1034     while ((c = *spell++) != '\0')
1035     {
1036         /* if this grows into too many cases, I should use a table */
1037         if (c >= 'A' && c <= 'Z')
1038             talk_id(VOICE_CHAR_A + c - 'A', true);
1039         else if (c >= 'a' && c <= 'z')
1040             talk_id(VOICE_CHAR_A + c - 'a', true);
1041         else if (c >= '0' && c <= '9')
1042             talk_id(VOICE_ZERO + c - '0', true);
1043         else if (c == '-')
1044             talk_id(VOICE_MINUS, true);
1045         else if (c == '+')
1046             talk_id(VOICE_PLUS, true);
1047         else if (c == '.')
1048             talk_id(VOICE_DOT, true);
1049         else if (c == ' ')
1050             talk_id(VOICE_PAUSE, true);
1051         else if (c == '/')
1052             talk_id(VOICE_CHAR_SLASH, true);
1053     }
1054
1055     return 0;
1056 }
1057
1058 void talk_disable(bool disable)
1059 {
1060     if (disable)
1061         talk_temp_disable_count++;
1062     else
1063         talk_temp_disable_count--;
1064 }
1065
1066 void talk_setting(const void *global_settings_variable)
1067 {
1068     const struct settings_list *setting;
1069     if (!global_settings.talk_menu)
1070         return;
1071     setting = find_setting(global_settings_variable, NULL);
1072     if (setting == NULL)
1073         return;
1074     if (setting->lang_id)
1075         talk_id(setting->lang_id,false);
1076 }
1077
1078
1079 #if CONFIG_RTC
1080 void talk_date(const struct tm *tm, bool enqueue)
1081 {
1082     talk_id(LANG_MONTH_JANUARY + tm->tm_mon, enqueue);
1083     talk_number(tm->tm_mday, true);
1084     talk_number(1900 + tm->tm_year, true);
1085 }
1086
1087 void talk_time(const struct tm *tm, bool enqueue)
1088 {
1089     if (global_settings.timeformat == 1)
1090     {
1091         /* Voice the hour */
1092         long am_pm_id = VOICE_AM;
1093         int hour = tm->tm_hour;
1094         if (hour >= 12)
1095         {
1096             am_pm_id = VOICE_PM;
1097             hour -= 12;
1098         }
1099         if (hour == 0)
1100             hour = 12;
1101         talk_number(hour, enqueue);
1102
1103         /* Voice the minutes */
1104         if (tm->tm_min == 0)
1105         {
1106             /* Say o'clock if the minute is 0. */
1107             talk_id(VOICE_OCLOCK, true);
1108         }
1109         else
1110         {
1111             /* Pronounce the leading 0 */
1112             if(tm->tm_min < 10)
1113                 talk_id(VOICE_OH, true);
1114             talk_number(tm->tm_min, true);
1115         }
1116         talk_id(am_pm_id, true);
1117     }
1118     else
1119     {
1120         /* Voice the time in 24 hour format */
1121         talk_number(tm->tm_hour, enqueue);
1122         if (tm->tm_min == 0)
1123         {
1124             talk_id(VOICE_HUNDRED, true);
1125             talk_id(VOICE_HOUR, true);
1126         }
1127         else
1128         {
1129             /* Pronounce the leading 0 */
1130             if(tm->tm_min < 10)
1131                 talk_id(VOICE_OH, true);
1132             talk_number(tm->tm_min, true);
1133         }
1134     }
1135 }
1136
1137 #endif