summaryrefslogtreecommitdiffstats
path: root/apps/talk.c
blob: 551d9672bc5bf0c9a82bace7f130c63a27b23d76 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
/***************************************************************************
 *             __________               __   ___.
 *   Open      \______   \ ____   ____ |  | _\_ |__   _______  ___
 *   Source     |       _//  _ \_/ ___\|  |/ /| __ \ /  _ \  \/  /
 *   Jukebox    |    |   (  <_> )  \___|    < | \_\ (  <_> > <  <
 *   Firmware   |____|_  /\____/ \___  >__|_ \|___  /\____/__/\_ \
 *                     \/            \/     \/    \/            \/
 * $Id$
 *
 * Copyright (C) 2004 Jörg Hohensohn
 *
 * This module collects the Talkbox and voice UI functions.
 * (Talkbox reads directory names from mp3 clips called thumbnails,
 *  the voice UI lets menus and screens "talk" from a voicefile in memory.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version 2
 * of the License, or (at your option) any later version.
 *
 * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
 * KIND, either express or implied.
 *
 ****************************************************************************/

#include <stdio.h>
#include <stddef.h>
#include "string-extra.h"
#include "file.h"
#include "system.h"
#include "kernel.h"
#include "settings.h"
#include "settings_list.h"
#include "splash.h"
#include "voice_thread.h"
#include "audio.h"
#include "lang.h"
#include "talk.h"
#include "metadata.h"
/*#define LOGF_ENABLE*/
#include "logf.h"
#include "bitswap.h"
#include "structec.h"
#include "plugin.h" /* plugin_get_buffer() */
#include "debug.h"
#include "panic.h"
#include "misc.h" /* time_split_units() */

/***************** Constants *****************/

#define QUEUE_SIZE 64 /* must be a power of two */
#define QUEUE_MASK (QUEUE_SIZE-1)
const char* const dir_thumbnail_name = "_dirname.talk";
const char* const file_thumbnail_ext = ".talk";

/***************** Functional Macros *****************/

#define QUEUE_LEVEL ((queue_write - queue_read) & QUEUE_MASK)

#define LOADED_MASK 0x80000000 /* MSB */

#ifndef DEFAULT_VOICE_LANG
#define DEFAULT_VOICE_LANG "english"
#endif

/***************** Data types *****************/

struct clip_entry /* one entry of the index table */
{
    int offset; /* offset from start of voicefile file */
    int size; /* size of the clip */
};

struct voicefile_header /* file format of our voice file */
{
    int version; /* version of the voicefile */
    int target_id; /* the rockbox target the file was made for */
    int table;   /* offset to index table, (=header size) */
    int id1_max; /* number of "normal" clips contained in above index */
    int id2_max; /* number of "voice only" clips contained in above index */
    /* The header is folled by the index tables (n*struct clip_entry),
     * which is followed by the mp3/speex encoded clip data */
};

/***************** Globals *****************/

#if MEMORYSIZE <= 2
/* On low memory swcodec targets the entire voice file wouldn't fit in memory
 * together with codecs, so we load clips each time they are accessed. */
#define TALK_PROGRESSIVE_LOAD
/* 70+ clips should fit into 100k */
#define MAX_CLIP_BUFFER_SIZE (100000)
#endif

#ifndef MAX_CLIP_BUFFER_SIZE
/* 1GB should be enough for everybody; will be capped to voicefile size */
#define MAX_CLIP_BUFFER_SIZE (1<<30)
#endif
#define THUMBNAIL_RESERVE (50000)

/* Multiple thumbnails can be loaded back-to-back in this buffer. */
static volatile int thumbnail_buf_used SHAREDBSS_ATTR; /* length of data in
                                                          thumbnail buffer */
static struct voicefile_header voicefile; /* loaded voicefile */
static bool has_voicefile; /* a voicefile file is present */
static bool need_shutup; /* is there possibly any voice playing to be shutup */
static bool force_enqueue_next; /* enqueue next utterance even if enqueue is false */
static int queue_write; /* write index of queue, by application */
static int queue_read; /* read index of queue, by ISR context */
static enum talk_status talk_status = TALK_STATUS_OK;
/* protects queue_read, queue_write and thumbnail_buf_used */
static struct mutex queue_mutex SHAREDBSS_ATTR;
#define talk_queue_lock() ({ mutex_lock(&queue_mutex); })
#define talk_queue_unlock() ({ mutex_unlock(&queue_mutex); })
static int sent; /* how many bytes handed over to playback, owned by ISR */
static unsigned char curr_hd[3]; /* current frame header, for re-sync */
static unsigned char last_lang[MAX_FILENAME+1]; /* name of last used lang file (in talk_init) */
static bool talk_initialized; /* true if talk_init has been called */
static bool give_buffer_away; /* true if we should give the buffers away in shrink_callback if requested */
static int talk_temp_disable_count; /* if positive, temporarily disable voice UI (not saved) */
 /* size of the voice data in the voice file and the actually allocated buffer
  * for it. voicebuf_size is always smaller or equal to voicefile_size */
static unsigned long voicefile_size, voicebuf_size;

struct queue_entry /* one entry of the internal queue */
{
    int handle;    /* buflib handle to the clip data */
    int length;    /* total length of the clip */
    int remaining; /* bytes that still need to be deoded */
};

static struct buflib_context clip_ctx;

struct clip_cache_metadata {
    long tick;
    int handle, voice_id;
};

static int metadata_table_handle;
static unsigned max_clips;
static int cache_hits, cache_misses;

static struct queue_entry queue[QUEUE_SIZE]; /* queue of scheduled clips */
static struct queue_entry silence, *last_clip;

/***************** Private implementation *****************/

static int index_handle, talk_handle;

static int move_callback(int handle, void *current, void *new)
{
    (void)current;
    if (handle == talk_handle && !buflib_context_relocate(&clip_ctx, new))
        return BUFLIB_CB_CANNOT_MOVE;
    return BUFLIB_CB_OK;
}


static struct mutex read_buffer_mutex;

static inline bool check_audio_status(void)
{
    return true;
}

/* ISR (voice_callback()) must not run during moving of the clip buffer,
 * because the MAS may get out-of-sync */
static void sync_callback(int handle, bool sync_on)
{
    (void) handle;
    if (sync_on)
        mutex_lock(&read_buffer_mutex);
    else
        mutex_unlock(&read_buffer_mutex);
}

static ssize_t read_to_handle_ex(int fd, struct buflib_context *ctx, int handle,
                                 int handle_offset, size_t count)
{
    unsigned char *buf;
    ssize_t ret;
    mutex_lock(&read_buffer_mutex);

    if (!ctx)
        buf = core_get_data(handle);
    else
        buf = buflib_get_data(ctx, handle);

    buf += handle_offset;
    ret = read(fd, buf, count);

    mutex_unlock(&read_buffer_mutex);

    return ret;
}

static ssize_t read_to_handle(int fd, int handle, int handle_offset, size_t count)
{
    return read_to_handle_ex(fd, NULL, handle, handle_offset, count);
}


static int shrink_callback(int handle, unsigned hints, void *start, size_t old_size)
{
    (void)start;(void)old_size;
    int *h;
#if (MAX_CLIP_BUFFER_SIZE < (MEMORYSIZE<<20) || (MEMORYSIZE > 2))
    /* on low-mem and when the voice buffer size is not limited (i.e.
     * on 2MB HWCODEC) we effectively own the entire buffer because
     * the voicefile takes up all RAM. This blocks other Rockbox parts
     * from allocating, especially during bootup. Therefore always give
     * up the buffer and reload when clips are played back. On high-mem
     * or when the clip buffer is limited to a few 100K this provision is
     * not necessary. */
    if (give_buffer_away
            && (hints & BUFLIB_SHRINK_POS_MASK) == BUFLIB_SHRINK_POS_MASK)
#else
    (void)hints;
#endif
    {
        if (handle == talk_handle)
            h = &talk_handle;
        else if (handle == index_handle)
            h = &index_handle;
        else h = NULL;

        mutex_lock(&read_buffer_mutex);
        /* the clip buffer isn't usable without index table */
        if (handle == index_handle && talk_handle > 0)
            talk_handle = core_free(talk_handle);
        if (h)
            *h = core_free(handle);
        mutex_unlock(&read_buffer_mutex);

        return BUFLIB_CB_OK;
    }
    return BUFLIB_CB_CANNOT_SHRINK;
}

static struct buflib_callbacks talk_ops = {
    .move_callback = move_callback,
    .sync_callback = sync_callback,
    .shrink_callback = shrink_callback,
};


static int open_voicefile(void)
{
    char buf[64];
    char* p_lang = DEFAULT_VOICE_LANG; /* default */

    if ( global_settings.lang_file[0] &&
         global_settings.lang_file[0] != 0xff )
    {   /* try to open the voice file of the selected language */
        p_lang = (char *)global_settings.lang_file;
    }

    snprintf(buf, sizeof(buf), LANG_DIR "/%s.voice", p_lang);

    return open(buf, O_RDONLY);
}


static int id2index(int id)
{
    int index = id;
    if (id > VOICEONLY_DELIMITER)
    {   /* voice-only entries use the second part of the table.
           The first string comes after VOICEONLY_DELIMITER so we need to
           substract VOICEONLY_DELIMITER + 1 */
        index -= VOICEONLY_DELIMITER + 1;
        if (index >= voicefile.id2_max)
            return -1; /* must be newer than we have */
        index += voicefile.id1_max; /* table 2 is behind table 1 */
    }
    else
    {   /* normal use of the first table */
        if (id >= voicefile.id1_max)
            return -1; /* must be newer than we have */
    }
    return index;
}

#ifndef TALK_PROGRESSIVE_LOAD
static int index2id(int index)
{
    int id = index;

    if (index >= voicefile.id2_max + voicefile.id1_max)
        return -1;

    if (index >= voicefile.id1_max)
    {   /* must be voice-only if it exceeds table 1 */
        id -= voicefile.id1_max;
        /* The first string comes after VOICEONLY_DELIMITER so we need to
           add VOICEONLY_DELIMITER + 1 */
        id += VOICEONLY_DELIMITER + 1;
    }

    return id;
}
#endif

static int free_oldest_clip(void)
{
    unsigned i;
    int oldest = 0;
    bool thumb = false;
    long age, now;
    struct clip_entry* clipbuf;
    struct clip_cache_metadata *cc = buflib_get_data(&clip_ctx, metadata_table_handle);
    for(age = i = 0, now = current_tick; i < max_clips; i++)
    {
        if (cc[i].handle)
        {
            if (thumb && cc[i].voice_id == VOICEONLY_DELIMITER && (now - cc[i].tick) > age)
            {
                /* thumb clips are freed first */
                age = now - cc[i].tick;
                oldest = i;
            }
            else if (!thumb)
            {
                if (cc[i].voice_id == VOICEONLY_DELIMITER)
                {
                    age = now - cc[i].tick;
                    oldest = i;
                    thumb = true;
                }
                else if ((now - cc[i].tick) > age && cc[i].voice_id != VOICE_PAUSE)
                {
                    /* find the last-used clip but never consider silence */
                    age = now - cc[i].tick;
                    oldest = i;
                }
            }
        }
    }
    /* free the last one if no oldest one could be determined */
    cc = &cc[oldest];
    cc->handle = buflib_free(&clip_ctx, cc->handle);
    /* need to clear the LOADED bit too (not for thumb clips) */
    if (cc->voice_id != VOICEONLY_DELIMITER)
    {
        clipbuf = core_get_data(index_handle);
        clipbuf[id2index(cc->voice_id)].size &= ~LOADED_MASK;
    }
    return oldest;
}


/* common code for load_initial_clips() and get_clip() */
static void add_cache_entry(int clip_handle, int table_index, int id)
{
    unsigned i;
    struct clip_cache_metadata *cc = buflib_get_data(&clip_ctx, metadata_table_handle);

    if (table_index != -1)
    {
        /* explicit slot; use that */
        cc = &cc[table_index];
        if (cc->handle > 0) panicf("%s(): Slot already used", __func__);
    }
    else
    {   /* find an empty slot */
        for(i = 0; cc[i].handle && i < max_clips; i++) ;
        if (i == max_clips) /* no free slot in the cache table? */
            i = free_oldest_clip();
        cc = &cc[i];
    }
    cc->handle = clip_handle;
    cc->tick = current_tick;
    cc->voice_id = id;
}

static ssize_t read_clip_data(int fd, int index, int clip_handle)
{
    struct clip_entry* clipbuf;
    size_t clipsize;
    ssize_t ret;

    if (fd < 0)
    {
        buflib_free(&clip_ctx, clip_handle);
        return -1; /* open error */
    }

    clipbuf = core_get_data(index_handle);
    /* this must not be called with LOADED_MASK set in clipsize */
    clipsize = clipbuf[index].size;
    lseek(fd, clipbuf[index].offset, SEEK_SET);
    ret = read_to_handle_ex(fd, &clip_ctx, clip_handle, 0, clipsize);

    if (ret < 0 || clipsize != (size_t)ret)
    {
        buflib_free(&clip_ctx, clip_handle);
        return -2; /* read error */
    }

    clipbuf = core_get_data(index_handle);
    clipbuf[index].size |= LOADED_MASK; /* mark as loaded */

    return ret;
}

static void load_initial_clips(int fd)
{
#if defined(TALK_PROGRESSIVE_LOAD)
    (void) fd;
#else
    unsigned index, i;
    unsigned num_clips = voicefile.id1_max + voicefile.id2_max;

    for(index = i = 0; index < num_clips && i < max_clips; index++)
    {
        int handle;
        struct clip_entry* clipbuf = core_get_data(index_handle);
        size_t clipsize = clipbuf[index].size;
        ssize_t ret;

        if (clipsize == 0) /* clip not included in voicefile */
            continue;

        handle = buflib_alloc(&clip_ctx, clipsize);
        if (handle < 0)
            break;

        ret = read_clip_data(fd, index, handle);
        if (ret < 0)
            break;

        add_cache_entry(handle, i++, index2id(index));
    }
#endif
}

/* fetch a clip from the voice file */
static int get_clip(long id, struct queue_entry *q)
{
    int index;
    int retval = -1;
    struct clip_entry* clipbuf;
    size_t clipsize;

    index = id2index(id);
    if (index == -1 || index_handle <= 0)
        return -1;

    clipbuf = core_get_data(index_handle);
    clipsize = clipbuf[index].size;
    if (clipsize == 0) /* clip not included in voicefile */
        return -1;

    if (!(clipsize & LOADED_MASK))
    {   /* clip needs loading */
        int fd, handle, oldest = -1;
        ssize_t ret;
        cache_misses++;
        /* free clips from cache until this one succeeds to allocate */
        while ((handle = buflib_alloc(&clip_ctx, clipsize)) < 0)
            oldest = free_oldest_clip();
        /* handle should now hold a valid alloc. Load from disk
         * and insert into cache */
        fd = open_voicefile();
        ret = read_clip_data(fd, index, handle);
        close(fd);
        if (ret < 0)
            return ret;
        /* finally insert into metadata table */
        add_cache_entry(handle, oldest, id);
        retval = handle;
    }
    else
    {   /* clip is in memory already; find where it was loaded */
        cache_hits++;
        struct clip_cache_metadata *cc;
        static int i;
        cc = buflib_get_data(&clip_ctx, metadata_table_handle);
        for (i = 0; cc[i].voice_id != id || !cc[i].handle; i++) ;
        cc[i].tick = current_tick; /* reset age */
        clipsize &= ~LOADED_MASK; /* without the extra bit gives true size */
        retval = cc[i].handle;
    }

    q->handle    = retval;
    q->length    = clipsize;
    q->remaining = clipsize;
    return 0;
}

static bool load_index_table(int fd, const struct voicefile_header *hdr)
{
    ssize_t ret;

    if (index_handle > 0) /* nothing to do? */
        return true;

    ssize_t alloc_size = (hdr->id1_max + hdr->id2_max) * sizeof(struct clip_entry);
    index_handle = core_alloc_ex("voice index", alloc_size, &talk_ops);
    if (index_handle < 0)
        return false;

    ret = read_to_handle(fd, index_handle, 0, alloc_size);

    if (ret == alloc_size)
    {
#ifdef ROCKBOX_LITTLE_ENDIAN
        struct clip_entry *buf;
        buf = core_get_data(index_handle);
        for (int i = 0; i < hdr->id1_max + hdr->id2_max; i++)
        {
            /* doesn't yield() */
            structec_convert(&buf[i], "ll", 1, true);
        }
#endif
    }
    else
        index_handle = core_free(index_handle);

    return ret == alloc_size;
}

static bool load_header(int fd, struct voicefile_header *hdr)
{
    ssize_t got_size = read(fd, hdr, sizeof(*hdr));
    if (got_size != sizeof(*hdr))
        return false;

#ifdef ROCKBOX_LITTLE_ENDIAN
    logf("Byte swapping voice file");
    structec_convert(&voicefile, "lllll", 1, true);
#endif
    return true;
}

static bool create_clip_buffer(size_t max_size)
{
    /* just allocate, populate on an as-needed basis later */
    talk_handle = core_alloc_ex("voice data", max_size, &talk_ops);
    if (talk_handle < 0)
        goto alloc_err;

    buflib_init(&clip_ctx, core_get_data(talk_handle), max_size);

    return true;

alloc_err:
    talk_status = TALK_STATUS_ERR_ALLOC;
    if (index_handle > 0)
        index_handle = core_free(index_handle);
    return false;
}

/* load the voice file into the mp3 buffer */
static bool load_voicefile_index(int fd)
{
    if (fd < 0) /* failed to open */
    {
        talk_status = TALK_STATUS_ERR_NOFILE;
        return false;
    }

    /* load the header first */
    if (!load_header(fd, &voicefile))
    {
        talk_status = TALK_STATUS_ERR_INCOMPATIBLE;
        return false;
    }

    /* format check */
    if (voicefile.table == sizeof(struct voicefile_header))
    {
        if (voicefile.version == VOICE_VERSION &&
            voicefile.target_id == TARGET_ID &&
            voicefile.id1_max == TALK_FINAL_ID &&
            voicefile.id2_max == TALK_FINAL_ID_VOICEONLY - VOICEONLY_DELIMITER - 1)
        {
            if (load_index_table(fd, &voicefile))
                return true;
        }
    }

    talk_status = TALK_STATUS_ERR_INCOMPATIBLE;
    logf("Incompatible voice file");
    logf("version %d expected %d", voicefile.version, VOICE_VERSION);
    logf("target_id %d expected %d", voicefile.target_id, TARGET_ID);
    logf("id1_max %d expected %d", voicefile.id1_max, TALK_FINAL_ID);
    logf("id2_max %d expected %d",
               voicefile.id2_max, TALK_FINAL_ID_VOICEONLY - VOICEONLY_DELIMITER);
    return false;
}

/* this function caps the voicefile buffer and allocates it. It can
 * be called after talk_init(), e.g. when the voice was temporarily disabled.
 * The buffer size has to be capped again each time because the available
 * audio buffer changes over time */
static bool load_voicefile_data(int fd)
{
    voicebuf_size = voicefile_size;
    /* cap to the max. number of clips or the size of the available audio
     * buffer which we grab. We leave some to the rest of the system.
     * While that reduces our buffer size it improves the chance that
     * other allocs succeed without disabling voice which would require
     * reloading the voice from disk (as we do not shrink our buffer when
     * other code attempts new allocs these would fail) */
    size_t metadata_alloc_size;
    ssize_t cap = MIN(MAX_CLIP_BUFFER_SIZE, audio_buffer_available() - (64<<10));
    if (UNLIKELY(cap < 0))
    {
        logf("Not enough memory for voice. Disabling...\n");
        talk_status = TALK_STATUS_ERR_OOM;
        return false;
    }
    else if (voicebuf_size > (size_t)cap)
        voicebuf_size = cap;

    /* just allocate, populate on an as-needed basis later
     * re-create the clip buffer to ensure clip_ctx is up-to-date */
    if (talk_handle > 0)
        talk_handle = core_free(talk_handle);
    if (!create_clip_buffer(voicebuf_size))
        return false;

    /* the first alloc is the clip metadata table */
    metadata_alloc_size = max_clips * sizeof(struct clip_cache_metadata);
    metadata_table_handle = buflib_alloc(&clip_ctx, metadata_alloc_size);
    if (metadata_table_handle <= 0)
    {
        talk_status = TALK_STATUS_ERR_OOM;
        return false;
    }
    memset(buflib_get_data(&clip_ctx, metadata_table_handle), 0, metadata_alloc_size);

    load_initial_clips(fd);
    /* make sure to have the silence clip, if available return value can
     * be cached globally even for TALK_PROGRESSIVE_LOAD because the
     * VOICE_PAUSE clip is specially handled */
    get_clip(VOICE_PAUSE, &silence);

    return true;
}

/* Use a static buffer to avoid difficulties with buflib during
 * buffer passing to the voice_thread (swcodec). Clips can be played
   in chunks so the size is not that important */
static unsigned char commit_buffer[2<<10];

static void* commit_transfer(struct queue_entry *qe, size_t *size)
{
    void *buf = NULL; /* shut up gcc */
    static unsigned char *bufpos = commit_buffer;
    sent = qe->remaining;
    sent = MIN((size_t)sent, sizeof(commit_buffer));
    buf = buflib_get_data(&clip_ctx, qe->handle);
    /* adjust buffer position to what has been played already */
    buf += (qe->length - qe->remaining);
    memcpy(bufpos, buf, sent);
    *size = sent;

    return commit_buffer;
}

static inline bool is_silence(struct queue_entry *qe)
{
    if (silence.handle > 0) /* silence clip available? */
        return (qe->handle == silence.handle);
    else
        return false;
}

/* called in ISR context (on HWCODEC) if mp3 data got consumed */
static void mp3_callback(const void** start, size_t* size)
{
    struct queue_entry *qe = &queue[queue_read];
    /* voice_thread.c hints us how many of the buffer we provided it actually
     * consumed. Because buffers have to be frame-aligned for speex
     * it might be less than what we presented */
    if (*size)
        sent = *size;
    qe->remaining -= sent; /* we completed this */

    if (qe->remaining > 0) /* current clip not finished? */
    {   /* feed the next 64K-1 chunk */
        *start = commit_transfer(qe, size);
        return;
    }

    talk_queue_lock();

    /* increment read position for the just played clip */
    queue_read = (queue_read + 1) & QUEUE_MASK;

    if (QUEUE_LEVEL == 0)
    {
        if (!is_silence(last_clip))
        {   /* add silence clip when queue runs empty playing a voice clip,
             * only if the previous clip wasn't already silence */
            queue[queue_write] = silence;
            queue_write = (queue_write + 1) & QUEUE_MASK;
        }
        else
        {
            *size = 0; /* end of data */
        }
    }

    if (QUEUE_LEVEL != 0) /* queue is not empty? */
    {   /* start next clip */
        last_clip = &queue[queue_read];
        *start = commit_transfer(last_clip, size);
        curr_hd[0] = commit_buffer[1];
        curr_hd[1] = commit_buffer[2];
        curr_hd[2] = commit_buffer[3];
    }

    talk_queue_unlock();
}

/***************** Public routines *****************/

/* stop the playback and the pending clips */
void talk_force_shutup(void)
{
    /* Had nothing to do (was frame boundary or not our clip) */
    voice_play_stop();
    talk_queue_lock();
    queue_write = queue_read = 0; /* reset the queue */
    thumbnail_buf_used = 0;
    talk_queue_unlock();
    need_shutup = false;
}

/* Shutup the voice, except if force_enqueue_next is set. */
void talk_shutup(void)
{
    if (need_shutup && !force_enqueue_next)
        talk_force_shutup();
}

/* schedule a clip, at the end or discard the existing queue */
static void queue_clip(struct queue_entry *clip, bool enqueue)
{
    struct queue_entry *qe;
    int queue_level;

    if (!enqueue)
        talk_shutup(); /* cut off all the pending stuff */
    /* Something is being enqueued, force_enqueue_next override is no
       longer in effect. */
    force_enqueue_next = false;

    if (!clip->length)
        return; /* safety check */
    talk_queue_lock();
    queue_level = QUEUE_LEVEL; /* check old level */
    qe = &queue[queue_write];

    if (queue_level < QUEUE_SIZE - 1) /* space left? */
    {
        queue[queue_write] = *clip;
        queue_write = (queue_write + 1) & QUEUE_MASK;
    }
    talk_queue_unlock();

    if (queue_level == 0)
    {   /* queue was empty, we have to do the initial start */
        size_t size;
        void *buf = commit_transfer(qe, &size);
        last_clip = qe;
        voice_play_data(buf, size, mp3_callback);
        curr_hd[0] = commit_buffer[1];
        curr_hd[1] = commit_buffer[2];
        curr_hd[2] = commit_buffer[3];
    }

    need_shutup = true;

    return;
}

/* return if a voice codec is required or not */
static bool talk_voice_required(void)
{
    return (has_voicefile) /* Voice file is available */
        || (global_settings.talk_dir_clip)  /* Thumbnail clips are required */
        || (global_settings.talk_file_clip);
}

/***************** Public implementation *****************/

void talk_init(void)
{
    int filehandle;
    talk_temp_disable_count = 0;
    if (talk_initialized && !strcasecmp(last_lang, global_settings.lang_file))
    {
        /* not a new file, nothing to do */
        return;
    }

    if(!talk_initialized)
    {
        mutex_init(&queue_mutex);
        mutex_init(&read_buffer_mutex);
    }

    talk_force_shutup();  /* In case we have something speaking! */

    talk_initialized = true;
    strlcpy((char *)last_lang, (char *)global_settings.lang_file,
        MAX_FILENAME);

    /* reset some states */
    queue_write = queue_read = 0; /* reset the queue */
    memset(&voicefile, 0, sizeof(voicefile));

    silence.handle = -1; /* pause clip not accessible */
    voicefile_size = has_voicefile = 0;
    /* need to free these as their size depends on the voice file, and
     * this function is called when the talk voice file changes */
    if (index_handle > 0) index_handle = core_free(index_handle);
    if (talk_handle  > 0) talk_handle  = core_free(talk_handle);
    /* don't free thumb handle, it doesn't depend on the actual voice file
     * and so we can re-use it if it's already allocated in any event */

    filehandle = open_voicefile();
    if (filehandle > -1)
    {
        if (!load_voicefile_index(filehandle))
        {
            if (global_settings.talk_menu)
                splashf(HZ, str(LANG_READ_FAILED), ".voice");
            goto out;
        }
        /* Now determine the maximum buffer size needed for the voicefile.
         * The below pretends the entire voicefile would be loaded. The buffer
         * size is eventually capped later on in load_voicefile_data() */
        int num_clips = voicefile.id1_max + voicefile.id2_max;
        int non_empty = num_clips;
        int total_size = 0, avg_size;
        struct clip_entry *clips = core_get_data(index_handle);
        /* check for the average clip size to estimate the maximum number of
         * clips the buffer can hold */
        for (int i = 0; i<num_clips; i++)
        {
            if (clips[i].size)
                total_size += ALIGN_UP(clips[i].size, sizeof(void *));
            else
                non_empty -= 1;
        }
        avg_size = total_size / non_empty;
        max_clips = MIN((int)(MAX_CLIP_BUFFER_SIZE/avg_size) + 1, non_empty);
        /* account for possible thumb clips */
        total_size += THUMBNAIL_RESERVE;
        max_clips += 16;
        voicefile_size = total_size;
        has_voicefile = true;
    }
    else if (talk_voice_required())
    {
        /* create buffer just for thumb clips */
        max_clips = 16;
        voicefile_size = THUMBNAIL_RESERVE;
    }
    /* additionally to the clip we need a table to record the age of the clips
     * so that, when memory is tight, only the most recently used ones are kept */
    voicefile_size += sizeof(struct clip_cache_metadata) * max_clips;
    /* compensate a bit for buflib alloc overhead. */
    voicefile_size += BUFLIB_ALLOC_OVERHEAD * max_clips + 64;

    load_voicefile_data(filehandle);

    /* Initialize the actual voice clip playback engine as well */
    if (talk_voice_required())
        voice_thread_init();

out:
    close(filehandle); /* close again, this was just to detect presence */
}

/* somebody else claims the mp3 buffer, e.g. for regular play/record */
void talk_buffer_set_policy(int policy)
{
    switch(policy)
    {
        case TALK_BUFFER_DEFAULT:
        case TALK_BUFFER_HOLD:  give_buffer_away = false;            break;
        case TALK_BUFFER_LOOSE: give_buffer_away = true;             break;
        default:           DEBUGF("Ignoring unknown policy\n"); break;
    }
}

/* play a voice ID from voicefile */
int talk_id(int32_t id, bool enqueue)
{
    int32_t unit;
    int decimals;
    struct queue_entry clip;
    bool isloaded = true;

    if (!has_voicefile)
        return 0; /* no voicefile loaded, not an error -> pretent success */
    if (talk_temp_disable_count > 0)
        return -1;  /* talking has been disabled */
    if (!check_audio_status())
        return -1;

    if (talk_handle <= 0 || index_handle <= 0) /* reload needed? */
    {
        int fd = open_voicefile();
        if (fd < 0 || !load_voicefile_index(fd))
            return -1;
        isloaded = load_voicefile_data(fd);
        close(fd);
    }

    if (id == -1 || !isloaded) /* -1 is an indication for silence */
        return -1;

    decimals = (((uint32_t)id) >> DECIMAL_SHIFT) & 0x7;

    /* check if this is a special ID, with a value */
    unit = ((uint32_t)id) >> UNIT_SHIFT;
    if (unit || decimals)
    {   /* sign-extend the value */
        id = (uint32_t)id << (32-DECIMAL_SHIFT);
        id >>= (32-DECIMAL_SHIFT);

        talk_value_decimal(id, unit, decimals, enqueue); /* speak it */
        return 0; /* and stop, end of special case */
    }

    if (get_clip(id, &clip) < 0)
        return -1; /* not present */

#ifdef LOGF_ENABLE
    if (id > VOICEONLY_DELIMITER)
        logf("\ntalk_id: Say voice clip 0x%x\n", id);
    else
        logf("\ntalk_id: Say '%s'\n", str(id));
#endif

    queue_clip(&clip, enqueue);

    return 0;
}
/* Speaks zero or more IDs (from an array). */
int talk_idarray(const long *ids, bool enqueue)
{
    int r;
    if(!ids)
        return 0;
    while(*ids != TALK_FINAL_ID)
    {
        if((r = talk_id(*ids++, enqueue)) <0)
            return r;
        enqueue = true;
    }
    return 0;
}

/* Make sure the current utterance is not interrupted by the next one. */
void talk_force_enqueue_next(void)
{
    force_enqueue_next = true;
}

/* play a thumbnail from file */
/* Returns size of spoken thumbnail, so >0 means something is spoken,
   <=0 means something went wrong. */
static int _talk_file(const char* filename,
                      const long *prefix_ids, bool enqueue)
{
    int fd;
    int size;
    int handle, oldest = -1;

    /* reload needed? */
    if (talk_temp_disable_count > 0)
        return -1;  /* talking has been disabled */
    if (!check_audio_status())
        return -1;
    if (talk_handle <= 0 || index_handle <= 0)
    {
        int fd = open_voicefile();
        if (fd < 0 || !load_voicefile_index(fd))
            return -1;
        load_voicefile_data(fd);
        close(fd);
    }

    if (!enqueue)
        /* shutup now to free the thumbnail buffer */
        talk_shutup();

    fd = open(filename, O_RDONLY);
    if (fd < 0) /* failed to open */
    {
        return 0;
    }
    size = filesize(fd);

    /* free clips from cache until this one succeeds to allocate */
    while ((handle = buflib_alloc(&clip_ctx, size)) < 0)
        oldest = free_oldest_clip();

    size = read_to_handle_ex(fd, &clip_ctx, handle, 0, size);
    close(fd);

    /* ToDo: find audio, skip ID headers and trailers */

    if (size > 0)    /* Don't play missing clips */
    {
        struct queue_entry clip;
        clip.handle = handle;
        clip.length = clip.remaining = size;
        if(prefix_ids)
            /* prefix thumbnail by speaking these ids, but only now
               that we know there's actually a thumbnail to be
               spoken. */
            talk_idarray(prefix_ids, true);
        /* finally insert into metadata table. thumb clips go under the
         * VOICEONLY_DELIMITER id so the cache can distinguish them from
         * normal clips */
        add_cache_entry(handle, oldest, VOICEONLY_DELIMITER);
        queue_clip(&clip, true);
    }
    else
        buflib_free(&clip_ctx, handle);

    return size;
}

int talk_file(const char *root, const char *dir, const char *file,
              const char *ext, const long *prefix_ids, bool enqueue)
/* Play a thumbnail file */
{
    char buf[MAX_PATH];
    /* Does root end with a slash */
    char *slash = (root && root[0]
                   && root[strlen(root)-1] != '/') ? "/" : "";
    snprintf(buf, MAX_PATH, "%s%s%s%s%s%s",
             root ? root : "", slash,
             dir ? dir : "", dir ? "/" : "",
             file ? file : "",
             ext ? ext : "");
    return _talk_file(buf, prefix_ids, enqueue);
}

static int talk_spell_basename(const char *path,
                               const long *prefix_ids, bool enqueue)
{
    if(prefix_ids)
    {
        talk_idarray(prefix_ids, enqueue);
        enqueue = true;
    }
    char buf[MAX_PATH];
    /* Spell only the path component after the last slash */
    strlcpy(buf, path, sizeof(buf));
    if(strlen(buf) >1 && buf[strlen(buf)-1] == '/')
        /* strip trailing slash */
        buf[strlen(buf)-1] = '\0';
    char *ptr = strrchr(buf, '/');
    if(ptr && strlen(buf) >1)
        ++ptr;
    else ptr = buf;
    return talk_spell(ptr, enqueue);
}

/* Play a file's .talk thumbnail, fallback to spelling the filename, or
   go straight to spelling depending on settings. */
int talk_file_or_spell(const char *dirname, const char *filename,
                       const long *prefix_ids, bool enqueue)
{
    if (global_settings.talk_file_clip)
    {   /* .talk clips enabled */
        if(talk_file(dirname, NULL, filename, file_thumbnail_ext,
                              prefix_ids, enqueue) >0)
            return 0;
    }
    if (global_settings.talk_file == 2)
        /* Either .talk clips are disabled, or as a fallback */
        return talk_spell_basename(filename, prefix_ids, enqueue);
    return 0;
}

/* Play a directory's .talk thumbnail, fallback to spelling the filename, or
   go straight to spelling depending on settings. */
int talk_dir_or_spell(const char* dirname,
                      const long *prefix_ids, bool enqueue)
{
    if (global_settings.talk_dir_clip)
    {   /* .talk clips enabled */
        if(talk_file(dirname, NULL, dir_thumbnail_name, NULL,
                              prefix_ids, enqueue) >0)
            return 0;
    }
    if (global_settings.talk_dir == 2)
        /* Either .talk clips disabled or as a fallback */
        return talk_spell_basename(dirname, prefix_ids, enqueue);
    return 0;
}

/* Speak thumbnail for each component of a full path, again falling
   back or going straight to spelling depending on settings. */
int talk_fullpath(const char* path, bool enqueue)
{
    if (!enqueue)
        talk_shutup();
    if(path[0] != '/')
        /* path ought to start with /... */
        return talk_spell(path, true);
    talk_id(VOICE_CHAR_SLASH, true);
    char buf[MAX_PATH];
    strlcpy(buf, path, MAX_PATH);
    char *start = buf+1; /* start of current component */
    char *ptr = strchr(start, '/'); /* end of current component */
    while(ptr) { /* There are more slashes ahead */
        /* temporarily poke a NULL at end of component to truncate string */
        *ptr = '\0';
        talk_dir_or_spell(buf, NULL, true);
        *ptr = '/'; /* restore string */
        talk_id(VOICE_CHAR_SLASH, true);
        start = ptr+1; /* setup for next component */
        ptr = strchr(start, '/');
    }
    /* no more slashes, final component is a filename */
    return talk_file_or_spell(NULL, buf, NULL, true);
}

/* say a numeric value, this word ordering works for english,
   but not necessarily for other languages (e.g. german) */
int talk_number(long n, bool enqueue)
{
    int level = 2; /* mille count */
    long mil = 1000000000; /* highest possible "-illion" */

    if (talk_temp_disable_count > 0)
        return -1;  /* talking has been disabled */
    if (!check_audio_status())
        return -1;

    if (!enqueue)
        talk_shutup(); /* cut off all the pending stuff */

    if (n==0)
    {   /* special case */
        talk_id(VOICE_ZERO, true);
        return 0;
    }

    if (n<0)
    {
        talk_id(VOICE_MINUS, true);
        n = -n;
    }

    while (n)
    {
        int segment = n / mil; /* extract in groups of 3 digits */
        n -= segment * mil; /* remove the used digits from number */
        mil /= 1000; /* digit place for next round */

        if (segment)
        {
            int hundreds = segment / 100;
            int ones = segment % 100;

            if (hundreds)
            {
                talk_id(VOICE_ZERO + hundreds, true);
                talk_id(VOICE_HUNDRED, true);
            }

            /* combination indexing */
            if (ones > 20)
            {
               int tens = ones/10 + 18;
               talk_id(VOICE_ZERO + tens, true);
               ones %= 10;
            }

            /* direct indexing */
            if (ones)
                talk_id(VOICE_ZERO + ones, true);

            /* add billion, million, thousand */
            if (mil)
                talk_id(VOICE_THOUSAND + level, true);
        }
        level--;
    }

    return 0;
}

/* Say year like "nineteen ninety nine" instead of "one thousand 9
   hundred ninety nine". */
static int talk_year(long year, bool enqueue)
{
    int rem;
    if(year < 1100 || year >=2000)
        /* just say it as a regular number */
        return talk_number(year, enqueue);
    /* Say century */
    talk_number(year/100, enqueue);
    rem = year%100;
    if(rem == 0)
        /* as in 1900 */
        return talk_id(VOICE_HUNDRED, true);
    if(rem <10)
        /* as in 1905 */
        talk_id(VOICE_ZERO, true);
    /* sub-century year */
    return talk_number(rem, true);
}

/* Say time duration/interval. Input is time in seconds,
   say hours,minutes,seconds. */
static int talk_time_unit(long secs, bool enqueue)
{
    return talk_time_intervals(secs, UNIT_SEC, enqueue);
}

void talk_fractional(char *tbuf, int value, int unit)
{
    int i;
    /* strip trailing zeros from the fraction */
    for (i = strlen(tbuf) - 1; (i >= 0) && (tbuf[i] == '0'); i--)
        tbuf[i] = '\0';

    talk_number(value, true);
    if (tbuf[0] != 0)
    {
        talk_id(LANG_POINT, true);
        talk_spell(tbuf, true);
    }
    talk_id(unit, true);
}

int talk_value(long n, int unit, bool enqueue)
{
    return talk_value_decimal(n, unit, 0, enqueue);
}

/* singular/plural aware saying of a value */
int talk_value_decimal(long n, int unit, int decimals, bool enqueue)
{
    int unit_id;
    static const int unit_voiced[] =
    {   /* lookup table for the voice ID of the units */
        [0 ... UNIT_LAST-1] = -1, /* regular ID, int, signed */
        [UNIT_MS]
            = VOICE_MILLISECONDS, /* here come the "real" units */
        [UNIT_SEC]
            = VOICE_SECONDS,
        [UNIT_MIN]
            = VOICE_MINUTES,
        [UNIT_HOUR]
            = VOICE_HOURS,
        [UNIT_KHZ]
            = VOICE_KHZ,
        [UNIT_DB]
            = VOICE_DB,
        [UNIT_PERCENT]
            = VOICE_PERCENT,
        [UNIT_MAH]
            = VOICE_MILLIAMPHOURS,
        [UNIT_PIXEL]
            = VOICE_PIXEL,
        [UNIT_PER_SEC]
            = VOICE_PER_SEC,
        [UNIT_HERTZ]
            = VOICE_HERTZ,
        [UNIT_MB]
            = LANG_MEBIBYTE,
        [UNIT_KBIT]
            = VOICE_KBIT_PER_SEC,
        [UNIT_PM_TICK]
            = VOICE_PM_UNITS_PER_TICK,
    };

    static const int pow10[] = { /* 10^0 - 10^7 */
        1, 10, 100, 1000, 10000, 100000, 1000000, 10000000
    };

    char tbuf[8];
    char fmt[] = "%0nd";

    if (talk_temp_disable_count > 0)
        return -1;  /* talking has been disabled */
    if (!check_audio_status())
        return -1;

    /* special pronounciation for year number */
    if (unit == UNIT_DATEYEAR)
        return talk_year(n, enqueue);
    /* special case for time duration */
    if (unit == UNIT_TIME)
        return talk_time_unit(n, enqueue);

    if (unit < 0 || unit >= UNIT_LAST)
        unit_id = -1;
    else
        unit_id = unit_voiced[unit];

    if ((n==1 || n==-1) /* singular? */
        && unit_id >= VOICE_SECONDS && unit_id <= VOICE_HOURS)
    {
        unit_id--; /* use the singular for those units which have */
    }

    /* special case with a "plus" before */
    if (n > 0 && (unit == UNIT_SIGNED || unit == UNIT_DB))
    {
        talk_id(VOICE_PLUS, enqueue);
        enqueue = true;
    }

    if (decimals)
    {
        /* needed for the "-0.5" corner case */
        if (n < 0)
        {
            talk_id(VOICE_MINUS, enqueue);
            n = -n;
        }

        fmt[2] = '0' + decimals;

        snprintf(tbuf, sizeof(tbuf), fmt, n % pow10[decimals]);
        talk_fractional(tbuf, n / pow10[decimals], unit_id);

        return 0;
    }

    talk_number(n, enqueue); /* say the number */
    talk_id(unit_id, true); /* say the unit, if any */

    return 0;
}

/* Say time duration/interval. Input is time unit specifies base unit,
   say hours,minutes,seconds, milliseconds. or any combination thereof */
int talk_time_intervals(long time, int unit_idx, bool enqueue)
{
    unsigned long units_in[UNIT_IDX_TIME_COUNT];

    if (talk_temp_disable_count > 0)
        return -1;  /* talking has been disabled */
    if (!check_audio_status())
        return -1;

    if (talk_handle <= 0 || index_handle <= 0) /* reload needed? */
    {
        int fd = open_voicefile();
        if (fd < 0 || !load_voicefile_index(fd))
            return -1;
        load_voicefile_data(fd);
        close(fd);
    }

    if (!enqueue)
        talk_shutup(); /* cut off all the pending stuff */

    time_split_units(unit_idx, labs(time), &units_in);

    if (time < 0)
        talk_id(VOICE_MINUS, true);

    if (time == 0)
        talk_value(0, unit_idx, true);
    else
    {
        if (units_in[UNIT_IDX_HR] != 0)
        {
            talk_value(units_in[UNIT_IDX_HR], UNIT_HOUR, true);
        }
        if (units_in[UNIT_IDX_MIN] != 0)
        {
            talk_value(units_in[UNIT_IDX_MIN], UNIT_MIN, true);
        }
        if (units_in[UNIT_IDX_SEC] != 0)
        {
            talk_value(units_in[UNIT_IDX_SEC], UNIT_SEC, true);
        }
        if (units_in[UNIT_IDX_MS] != 0)
        {
            talk_value(units_in[UNIT_IDX_MS], UNIT_MS, true);
        }
    }

    return -1;
}

/* spell a string */
int talk_spell(const char* spell, bool enqueue)
{
    char c; /* currently processed char */

    if (talk_temp_disable_count > 0)
        return -1;  /* talking has been disabled */
    if (!check_audio_status())
        return -1;

    if (!enqueue)
        talk_shutup(); /* cut off all the pending stuff */

    while ((c = *spell++) != '\0')
    {
        /* if this grows into too many cases, I should use a table */
        if (c >= 'A' && c <= 'Z')
            talk_id(VOICE_CHAR_A + c - 'A', true);
        else if (c >= 'a' && c <= 'z')
            talk_id(VOICE_CHAR_A + c - 'a', true);
        else if (c >= '0' && c <= '9')
            talk_id(VOICE_ZERO + c - '0', true);
        else if (c == '-')
            talk_id(VOICE_MINUS, true);
        else if (c == '+')
            talk_id(VOICE_PLUS, true);
        else if (c == '.')
            talk_id(VOICE_DOT, true);
        else if (c == ' ')
            talk_id(VOICE_PAUSE, true);
        else if (c == '/')
            talk_id(VOICE_CHAR_SLASH, true);
    }

    return 0;
}

void talk_disable(bool disable)
{
    if (disable)
        talk_temp_disable_count++;
    else
        talk_temp_disable_count--;
}

void talk_setting(const void *global_settings_variable)
{
    const struct settings_list *setting;
    if (!global_settings.talk_menu)
        return;
    setting = find_setting(global_settings_variable, NULL);
    if (setting == NULL)
        return;
    if (setting->lang_id)
        talk_id(setting->lang_id,false);
}


void talk_date(const struct tm *tm, bool enqueue)
{
    talk_id(LANG_MONTH_JANUARY + tm->tm_mon, enqueue);
    talk_number(tm->tm_mday, true);
    talk_number(1900 + tm->tm_year, true);
}

void talk_time(const struct tm *tm, bool enqueue)
{
    if (global_settings.timeformat == 1)
    {
        /* Voice the hour */
        long am_pm_id = VOICE_AM;
        int hour = tm->tm_hour;
        if (hour >= 12)
        {
            am_pm_id = VOICE_PM;
            hour -= 12;
        }
        if (hour == 0)
            hour = 12;
        talk_number(hour, enqueue);

        /* Voice the minutes */
        if (tm->tm_min == 0)
        {
            /* Say o'clock if the minute is 0. */
            talk_id(VOICE_OCLOCK, true);
        }
        else
        {
            /* Pronounce the leading 0 */
            if(tm->tm_min < 10)
                talk_id(VOICE_OH, true);
            talk_number(tm->tm_min, true);
        }
        talk_id(am_pm_id, true);
    }
    else
    {
        /* Voice the time in 24 hour format */
        talk_number(tm->tm_hour, enqueue);
        if (tm->tm_min == 0)
            talk_ids(true, VOICE_HUNDRED, VOICE_HOUR);
        else
        {
            /* Pronounce the leading 0 */
            if(tm->tm_min < 10)
                talk_id(VOICE_OH, true);
            talk_number(tm->tm_min, true);
        }
    }
}

void talk_announce_voice_invalid(void)
{
    int voice_fd;
    int voice_sz;
    int buf_handle;
    struct queue_entry qe;

    const char talkfile[] =
               LANG_DIR "/InvalidVoice_" DEFAULT_VOICE_LANG ".talk";

    if (global_settings.talk_menu && talk_status != TALK_STATUS_OK)
    {
        talk_temp_disable_count = 0xFF; /* don't let anyone else use voice sys */

        voice_fd = open(talkfile, O_RDONLY);
        if (voice_fd < 0)
            return; /* can't open */

        voice_sz= lseek(voice_fd, 0, SEEK_END);
        if (voice_sz == 0 || voice_sz > (64<<10))
            return; /* nothing here or too big */

        lseek(voice_fd, 0, SEEK_SET);
        /* add a bit extra for buflib overhead (2K) */
        if (!create_clip_buffer(ALIGN_UP(voice_sz, sizeof(long)) + (2<<10)))
            return;
        mutex_lock(&read_buffer_mutex);
        buf_handle = buflib_alloc(&clip_ctx, ALIGN_UP(voice_sz, sizeof(long)));

        if (buf_handle < 0)
            return;

        if (read_to_handle_ex(voice_fd, &clip_ctx, buf_handle, 0, voice_sz) > 0)
        {
            voice_thread_init();
            qe.handle = buf_handle;
            qe.length = qe.remaining = voice_sz;
            queue_clip(&qe, false);
            voice_wait();
            voice_thread_kill();
        }

        mutex_unlock(&read_buffer_mutex);
        close(voice_fd);

        buf_handle = buflib_free(&clip_ctx, buf_handle);
        talk_handle = core_free(talk_handle);
    }
}

bool talk_get_debug_data(struct talk_debug_data *data)
{
    char* p_lang = DEFAULT_VOICE_LANG; /* default */
    struct clip_cache_metadata *cc;

    memset(data, 0, sizeof(*data));

    data->status = talk_status;

    if (global_settings.lang_file[0] && global_settings.lang_file[0] != 0xff)
        p_lang = (char *)global_settings.lang_file;

    strlcpy(data->voicefile, p_lang, sizeof(data->voicefile));

    if (!has_voicefile || index_handle <= 0)
    {
        if (data->status == TALK_STATUS_OK)
            data->status = TALK_STATUS_ERR_NOFILE;

        return false;
    }

    struct clip_entry *clips = core_get_data(index_handle);
    int cached = 0;
    int real_clips = 0;

    data->num_clips = voicefile.id1_max + voicefile.id2_max;
    data->avg_clipsize = data->max_clipsize = 0;
    data->min_clipsize = INT_MAX;
    for(int i = 0; i < data->num_clips; i++)
    {
        int size = clips[i].size & (~LOADED_MASK);
        if (!size) continue;
        real_clips += 1;
        if (size < data->min_clipsize)
            data->min_clipsize = size;
        if (size > data->max_clipsize)
            data->max_clipsize = size;
        data->avg_clipsize += size;
    }
    cc = buflib_get_data(&clip_ctx, metadata_table_handle);
    for (int i = 0; i < (int) max_clips; i++)
    {
        if (cc[i].handle > 0)
            cached += 1;
    }
    data->avg_clipsize /= real_clips;
    data->num_empty_clips = data->num_clips - real_clips;
    data->memory_allocated = sizeof(commit_buffer) + sizeof(voicefile)
                         + data->num_clips * sizeof(struct clip_entry)
                         + voicebuf_size;
    data->memory_used = 0;
    if (talk_handle > 0)
        data->memory_used = data->memory_allocated - buflib_available(&clip_ctx);
    data->cached_clips = cached;
    data->cache_hits   = cache_hits;
    data->cache_misses = cache_misses;

    return true;
}