1 #include <iconv.h>
2 #include <errno.h>
3 #include <wchar.h>
4 #include <string.h>
5 #include <stdlib.h>
6 #include <limits.h>
7 #include <stdint.h>
8 #include "locale_impl.h"
9
10 #define UTF_32BE 0300
11 #define UTF_16LE 0301
12 #define UTF_16BE 0302
13 #define UTF_32LE 0303
14 #define UCS2BE 0304
15 #define UCS2LE 0305
16 #define WCHAR_T 0306
17 #define US_ASCII 0307
18 #define UTF_8 0310
19 #define UTF_16 0312
20 #define UTF_32 0313
21 #define UCS2 0314
22 #define EUC_JP 0320
23 #define SHIFT_JIS 0321
24 #define ISO2022_JP 0322
25 #define GB18030 0330
26 #define GBK 0331
27 #define GB2312 0332
28 #define BIG5 0340
29 #define EUC_KR 0350
30
31 /* Definitions of charmaps. Each charmap consists of:
32 * 1. Empty-string-terminated list of null-terminated aliases.
33 * 2. Special type code or number of elided quads of entries.
34 * 3. Character table (size determined by field 2), consisting
35 * of 5 bytes for every 4 characters, interpreted as 10-bit
36 * indices into the legacy_chars table. */
37
38 static const unsigned char charmaps[] =
39 "utf8\0char\0\0\310"
40 "wchart\0\0\306"
41 "ucs2be\0\0\304"
42 "ucs2le\0\0\305"
43 "utf16be\0\0\302"
44 "utf16le\0\0\301"
45 "ucs4be\0utf32be\0\0\300"
46 "ucs4le\0utf32le\0\0\303"
47 "ascii\0usascii\0iso646\0iso646us\0\0\307"
48 "utf16\0\0\312"
49 "ucs4\0utf32\0\0\313"
50 "ucs2\0\0\314"
51 "eucjp\0\0\320"
52 "shiftjis\0sjis\0\0\321"
53 "iso2022jp\0\0\322"
54 "gb18030\0\0\330"
55 "gbk\0\0\331"
56 "gb2312\0\0\332"
57 "big5\0bigfive\0cp950\0big5hkscs\0\0\340"
58 "euckr\0ksc5601\0ksx1001\0cp949\0\0\350"
59 #include "codepages.h"
60 ;
61
62 /* Table of characters that appear in legacy 8-bit codepages,
63 * limited to 1024 slots (10 bit indices). The first 256 entries
64 * are elided since those characters are obviously all included. */
65 static const unsigned short legacy_chars[] = {
66 #include "legacychars.h"
67 };
68
69 static const unsigned short jis0208[84][94] = {
70 #include "jis0208.h"
71 };
72
73 static const unsigned short gb18030[126][190] = {
74 #include "gb18030.h"
75 };
76
77 static const unsigned short big5[89][157] = {
78 #include "big5.h"
79 };
80
81 static const unsigned short hkscs[] = {
82 #include "hkscs.h"
83 };
84
85 static const unsigned short ksc[93][94] = {
86 #include "ksc.h"
87 };
88
89 static const unsigned short rev_jis[] = {
90 #include "revjis.h"
91 };
92
fuzzycmp(const unsigned char * a,const unsigned char * b)93 static int fuzzycmp(const unsigned char *a, const unsigned char *b)
94 {
95 for (; *a && *b; a++, b++) {
96 while (*a && (*a|32U)-'a'>26 && *a-'0'>10U) a++;
97 if ((*a|32U) != *b) return 1;
98 }
99 return *a != *b;
100 }
101
find_charmap(const void * name)102 static size_t find_charmap(const void *name)
103 {
104 const unsigned char *s;
105 if (!*(char *)name) name=charmaps; /* "utf8" */
106 for (s=charmaps; *s; ) {
107 if (!fuzzycmp(name, s)) {
108 for (; *s; s+=strlen((void *)s)+1);
109 return s+1-charmaps;
110 }
111 s += strlen((void *)s)+1;
112 if (!*s) {
113 if (s[1] > 0200) s+=2;
114 else s+=2+(64U-s[1])*5;
115 }
116 }
117 return -1;
118 }
119
120 struct stateful_cd {
121 iconv_t base_cd;
122 unsigned state;
123 };
124
combine_to_from(size_t t,size_t f)125 static iconv_t combine_to_from(size_t t, size_t f)
126 {
127 return (void *)(f<<16 | t<<1 | 1);
128 }
129
extract_from(iconv_t cd)130 static size_t extract_from(iconv_t cd)
131 {
132 return (size_t)cd >> 16;
133 }
134
extract_to(iconv_t cd)135 static size_t extract_to(iconv_t cd)
136 {
137 return (size_t)cd >> 1 & 0x7fff;
138 }
139
iconv_open(const char * to,const char * from)140 iconv_t iconv_open(const char *to, const char *from)
141 {
142 size_t f, t;
143 struct stateful_cd *scd;
144
145 if ((t = find_charmap(to))==-1
146 || (f = find_charmap(from))==-1
147 || (charmaps[t] >= 0330)) {
148 errno = EINVAL;
149 return (iconv_t)-1;
150 }
151 iconv_t cd = combine_to_from(t, f);
152
153 switch (charmaps[f]) {
154 case UTF_16:
155 case UTF_32:
156 case UCS2:
157 case ISO2022_JP:
158 scd = malloc(sizeof *scd);
159 if (!scd) return (iconv_t)-1;
160 scd->base_cd = cd;
161 scd->state = 0;
162 cd = (iconv_t)scd;
163 }
164
165 return cd;
166 }
167
get_16(const unsigned char * s,int e)168 static unsigned get_16(const unsigned char *s, int e)
169 {
170 e &= 1;
171 return s[e]<<8 | s[1-e];
172 }
173
put_16(unsigned char * s,unsigned c,int e)174 static void put_16(unsigned char *s, unsigned c, int e)
175 {
176 e &= 1;
177 s[e] = c>>8;
178 s[1-e] = c;
179 }
180
get_32(const unsigned char * s,int e)181 static unsigned get_32(const unsigned char *s, int e)
182 {
183 e &= 3;
184 return s[e]+0U<<24 | s[e^1]<<16 | s[e^2]<<8 | s[e^3];
185 }
186
put_32(unsigned char * s,unsigned c,int e)187 static void put_32(unsigned char *s, unsigned c, int e)
188 {
189 e &= 3;
190 s[e^0] = c>>24;
191 s[e^1] = c>>16;
192 s[e^2] = c>>8;
193 s[e^3] = c;
194 }
195
196 /* Adapt as needed */
197 #define mbrtowc_utf8 mbrtowc
198 #define wctomb_utf8 wctomb
199
legacy_map(const unsigned char * map,unsigned c)200 static unsigned legacy_map(const unsigned char *map, unsigned c)
201 {
202 if (c < 4*map[-1]) return c;
203 unsigned x = c - 4*map[-1];
204 x = map[x*5/4]>>2*x%8 | map[x*5/4+1]<<8-2*x%8 & 1023;
205 return x < 256 ? x : legacy_chars[x-256];
206 }
207
uni_to_jis(unsigned c)208 static unsigned uni_to_jis(unsigned c)
209 {
210 unsigned nel = sizeof rev_jis / sizeof *rev_jis;
211 unsigned d, j, i, b = 0;
212 for (;;) {
213 i = nel/2;
214 j = rev_jis[b+i];
215 d = jis0208[j/256][j%256];
216 if (d==c) return j + 0x2121;
217 else if (nel == 1) return 0;
218 else if (c < d)
219 nel /= 2;
220 else {
221 b += i;
222 nel -= nel/2;
223 }
224 }
225 }
226
iconv(iconv_t cd,char ** restrict in,size_t * restrict inb,char ** restrict out,size_t * restrict outb)227 size_t iconv(iconv_t cd, char **restrict in, size_t *restrict inb, char **restrict out, size_t *restrict outb)
228 {
229 size_t x=0;
230 struct stateful_cd *scd=0;
231 if (!((size_t)cd & 1)) {
232 scd = (void *)cd;
233 cd = scd->base_cd;
234 }
235 unsigned to = extract_to(cd);
236 unsigned from = extract_from(cd);
237 const unsigned char *map = charmaps+from+1;
238 const unsigned char *tomap = charmaps+to+1;
239 mbstate_t st = {0};
240 wchar_t wc;
241 unsigned c, d;
242 size_t k, l;
243 int err;
244 unsigned char type = map[-1];
245 unsigned char totype = tomap[-1];
246 locale_t locale = CURRENT_LOCALE;
247 locale_t *ploc = &locale, loc = *ploc;
248
249 if (!in || !*in || !*inb) return 0;
250
251 *ploc = UTF8_LOCALE;
252
253 for (; *inb; *in+=l, *inb-=l) {
254 c = *(unsigned char *)*in;
255 l = 1;
256
257 switch (type) {
258 case UTF_8:
259 if (c < 128) break;
260 l = mbrtowc_utf8(&wc, *in, *inb, &st);
261 if (l == (size_t)-1) goto ilseq;
262 if (l == (size_t)-2) goto starved;
263 c = wc;
264 break;
265 case US_ASCII:
266 if (c >= 128) goto ilseq;
267 break;
268 case WCHAR_T:
269 l = sizeof(wchar_t);
270 if (*inb < l) goto starved;
271 c = *(wchar_t *)*in;
272 if (0) {
273 case UTF_32BE:
274 case UTF_32LE:
275 l = 4;
276 if (*inb < 4) goto starved;
277 c = get_32((void *)*in, type);
278 }
279 if (c-0xd800u < 0x800u || c >= 0x110000u) goto ilseq;
280 break;
281 case UCS2BE:
282 case UCS2LE:
283 case UTF_16BE:
284 case UTF_16LE:
285 l = 2;
286 if (*inb < 2) goto starved;
287 c = get_16((void *)*in, type);
288 if ((unsigned)(c-0xdc00) < 0x400) goto ilseq;
289 if ((unsigned)(c-0xd800) < 0x400) {
290 if (type-UCS2BE < 2U) goto ilseq;
291 l = 4;
292 if (*inb < 4) goto starved;
293 d = get_16((void *)(*in + 2), type);
294 if ((unsigned)(d-0xdc00) >= 0x400) goto ilseq;
295 c = ((c-0xd7c0)<<10) + (d-0xdc00);
296 }
297 break;
298 case UCS2:
299 case UTF_16:
300 l = 0;
301 if (!scd->state) {
302 if (*inb < 2) goto starved;
303 c = get_16((void *)*in, 0);
304 scd->state = type==UCS2
305 ? c==0xfffe ? UCS2LE : UCS2BE
306 : c==0xfffe ? UTF_16LE : UTF_16BE;
307 if (c == 0xfffe || c == 0xfeff)
308 l = 2;
309 }
310 type = scd->state;
311 continue;
312 case UTF_32:
313 l = 0;
314 if (!scd->state) {
315 if (*inb < 4) goto starved;
316 c = get_32((void *)*in, 0);
317 scd->state = c==0xfffe0000 ? UTF_32LE : UTF_32BE;
318 if (c == 0xfffe0000 || c == 0xfeff)
319 l = 4;
320 }
321 type = scd->state;
322 continue;
323 case SHIFT_JIS:
324 if (c < 128) break;
325 if (c-0xa1 <= 0xdf-0xa1) {
326 c += 0xff61-0xa1;
327 break;
328 }
329 l = 2;
330 if (*inb < 2) goto starved;
331 d = *((unsigned char *)*in + 1);
332 if (c-129 <= 159-129) c -= 129;
333 else if (c-224 <= 239-224) c -= 193;
334 else goto ilseq;
335 c *= 2;
336 if (d-64 <= 158-64) {
337 if (d==127) goto ilseq;
338 if (d>127) d--;
339 d -= 64;
340 } else if (d-159 <= 252-159) {
341 c++;
342 d -= 159;
343 }
344 c = jis0208[c][d];
345 if (!c) goto ilseq;
346 break;
347 case EUC_JP:
348 if (c < 128) break;
349 l = 2;
350 if (*inb < 2) goto starved;
351 d = *((unsigned char *)*in + 1);
352 if (c==0x8e) {
353 c = d;
354 if (c-0xa1 > 0xdf-0xa1) goto ilseq;
355 c += 0xff61 - 0xa1;
356 break;
357 }
358 c -= 0xa1;
359 d -= 0xa1;
360 if (c >= 84 || d >= 94) goto ilseq;
361 c = jis0208[c][d];
362 if (!c) goto ilseq;
363 break;
364 case ISO2022_JP:
365 if (c >= 128) goto ilseq;
366 if (c == '\033') {
367 l = 3;
368 if (*inb < 3) goto starved;
369 c = *((unsigned char *)*in + 1);
370 d = *((unsigned char *)*in + 2);
371 if (c != '(' && c != '$') goto ilseq;
372 switch (128*(c=='$') + d) {
373 case 'B': scd->state=0; continue;
374 case 'J': scd->state=1; continue;
375 case 'I': scd->state=4; continue;
376 case 128+'@': scd->state=2; continue;
377 case 128+'B': scd->state=3; continue;
378 }
379 goto ilseq;
380 }
381 switch (scd->state) {
382 case 1:
383 if (c=='\\') c = 0xa5;
384 if (c=='~') c = 0x203e;
385 break;
386 case 2:
387 case 3:
388 l = 2;
389 if (*inb < 2) goto starved;
390 d = *((unsigned char *)*in + 1);
391 c -= 0x21;
392 d -= 0x21;
393 if (c >= 84 || d >= 94) goto ilseq;
394 c = jis0208[c][d];
395 if (!c) goto ilseq;
396 break;
397 case 4:
398 if (c-0x60 < 0x1f) goto ilseq;
399 if (c-0x21 < 0x5e) c += 0xff61-0x21;
400 break;
401 }
402 break;
403 case GB2312:
404 if (c < 128) break;
405 if (c < 0xa1) goto ilseq;
406 case GBK:
407 case GB18030:
408 if (c < 128) break;
409 c -= 0x81;
410 if (c >= 126) goto ilseq;
411 l = 2;
412 if (*inb < 2) goto starved;
413 d = *((unsigned char *)*in + 1);
414 if (d < 0xa1 && type == GB2312) goto ilseq;
415 if (d-0x40>=191 || d==127) {
416 if (d-'0'>9 || type != GB18030)
417 goto ilseq;
418 l = 4;
419 if (*inb < 4) goto starved;
420 c = (10*c + d-'0') * 1260;
421 d = *((unsigned char *)*in + 2);
422 if (d-0x81>126) goto ilseq;
423 c += 10*(d-0x81);
424 d = *((unsigned char *)*in + 3);
425 if (d-'0'>9) goto ilseq;
426 c += d-'0';
427 c += 128;
428 for (d=0; d<=c; ) {
429 k = 0;
430 for (int i=0; i<126; i++)
431 for (int j=0; j<190; j++)
432 if (gb18030[i][j]-d <= c-d)
433 k++;
434 d = c+1;
435 c += k;
436 }
437 break;
438 }
439 d -= 0x40;
440 if (d>63) d--;
441 c = gb18030[c][d];
442 break;
443 case BIG5:
444 if (c < 128) break;
445 l = 2;
446 if (*inb < 2) goto starved;
447 d = *((unsigned char *)*in + 1);
448 if (d-0x40>=0xff-0x40 || d-0x7f<0xa1-0x7f) goto ilseq;
449 d -= 0x40;
450 if (d > 0x3e) d -= 0x22;
451 if (c-0xa1>=0xfa-0xa1) {
452 if (c-0x87>=0xff-0x87) goto ilseq;
453 if (c < 0xa1) c -= 0x87;
454 else c -= 0x87 + (0xfa-0xa1);
455 c = (hkscs[4867+(c*157+d)/16]>>(c*157+d)%16)%2<<17
456 | hkscs[c*157+d];
457 /* A few HKSCS characters map to pairs of UCS
458 * characters. These are mapped to surrogate
459 * range in the hkscs table then hard-coded
460 * here. Ugly, yes. */
461 if (c/256 == 0xdc) {
462 union {
463 char c[8];
464 wchar_t wc[2];
465 } tmp;
466 char *ptmp = tmp.c;
467 size_t tmpx = iconv(combine_to_from(to, find_charmap("utf8")),
468 &(char *){"\303\212\314\204"
469 "\303\212\314\214"
470 "\303\252\314\204"
471 "\303\252\314\214"
472 +c%256}, &(size_t){4},
473 &ptmp, &(size_t){sizeof tmp});
474 size_t tmplen = ptmp - tmp.c;
475 if (tmplen > *outb) goto toobig;
476 if (tmpx) x++;
477 memcpy(*out, &tmp, tmplen);
478 *out += tmplen;
479 *outb -= tmplen;
480 continue;
481 }
482 if (!c) goto ilseq;
483 break;
484 }
485 c -= 0xa1;
486 c = big5[c][d]|(c==0x27&&(d==0x3a||d==0x3c||d==0x42))<<17;
487 if (!c) goto ilseq;
488 break;
489 case EUC_KR:
490 if (c < 128) break;
491 l = 2;
492 if (*inb < 2) goto starved;
493 d = *((unsigned char *)*in + 1);
494 c -= 0xa1;
495 d -= 0xa1;
496 if (c >= 93 || d >= 94) {
497 c += (0xa1-0x81);
498 d += 0xa1;
499 if (c >= 93 || c>=0xc6-0x81 && d>0x52)
500 goto ilseq;
501 if (d-'A'<26) d = d-'A';
502 else if (d-'a'<26) d = d-'a'+26;
503 else if (d-0x81<0xff-0x81) d = d-0x81+52;
504 else goto ilseq;
505 if (c < 0x20) c = 178*c + d;
506 else c = 178*0x20 + 84*(c-0x20) + d;
507 c += 0xac00;
508 for (d=0xac00; d<=c; ) {
509 k = 0;
510 for (int i=0; i<93; i++)
511 for (int j=0; j<94; j++)
512 if (ksc[i][j]-d <= c-d)
513 k++;
514 d = c+1;
515 c += k;
516 }
517 break;
518 }
519 c = ksc[c][d];
520 if (!c) goto ilseq;
521 break;
522 default:
523 if (!c) break;
524 c = legacy_map(map, c);
525 if (!c) goto ilseq;
526 }
527
528 switch (totype) {
529 case WCHAR_T:
530 if (*outb < sizeof(wchar_t)) goto toobig;
531 *(wchar_t *)*out = c;
532 *out += sizeof(wchar_t);
533 *outb -= sizeof(wchar_t);
534 break;
535 case UTF_8:
536 if (*outb < 4) {
537 char tmp[4];
538 k = wctomb_utf8(tmp, c);
539 if (*outb < k) goto toobig;
540 memcpy(*out, tmp, k);
541 } else k = wctomb_utf8(*out, c);
542 *out += k;
543 *outb -= k;
544 break;
545 case US_ASCII:
546 if (c > 0x7f) subst: x++, c='*';
547 default:
548 if (*outb < 1) goto toobig;
549 if (c<256 && c==legacy_map(tomap, c)) {
550 revout:
551 if (*outb < 1) goto toobig;
552 *(*out)++ = c;
553 *outb -= 1;
554 break;
555 }
556 d = c;
557 for (c=4*totype; c<256; c++) {
558 if (d == legacy_map(tomap, c)) {
559 goto revout;
560 }
561 }
562 goto subst;
563 case SHIFT_JIS:
564 if (c < 128) goto revout;
565 if (c == 0xa5) {
566 x++;
567 c = '\\';
568 goto revout;
569 }
570 if (c == 0x203e) {
571 x++;
572 c = '~';
573 goto revout;
574 }
575 if (c-0xff61 <= 0xdf-0xa1) {
576 c += 0xa1 - 0xff61;
577 goto revout;
578 }
579 c = uni_to_jis(c);
580 if (!c) goto subst;
581 if (*outb < 2) goto toobig;
582 d = c%256;
583 c = c/256;
584 *(*out)++ = (c+1)/2 + (c<95 ? 112 : 176);
585 *(*out)++ = c%2 ? d + 31 + d/96 : d + 126;
586 *outb -= 2;
587 break;
588 case EUC_JP:
589 if (c < 128) goto revout;
590 if (c-0xff61 <= 0xdf-0xa1) {
591 c += 0x0e00 + 0x21 - 0xff61;
592 } else {
593 c = uni_to_jis(c);
594 }
595 if (!c) goto subst;
596 if (*outb < 2) goto toobig;
597 *(*out)++ = c/256 + 0x80;
598 *(*out)++ = c%256 + 0x80;
599 *outb -= 2;
600 break;
601 case ISO2022_JP:
602 if (c < 128) goto revout;
603 if (c-0xff61 <= 0xdf-0xa1 || c==0xa5 || c==0x203e) {
604 if (*outb < 7) goto toobig;
605 *(*out)++ = '\033';
606 *(*out)++ = '(';
607 if (c==0xa5) {
608 *(*out)++ = 'J';
609 *(*out)++ = '\\';
610 } else if (c==0x203e) {
611 *(*out)++ = 'J';
612 *(*out)++ = '~';
613 } else {
614 *(*out)++ = 'I';
615 *(*out)++ = c-0xff61+0x21;
616 }
617 *(*out)++ = '\033';
618 *(*out)++ = '(';
619 *(*out)++ = 'B';
620 *outb -= 7;
621 break;
622 }
623 c = uni_to_jis(c);
624 if (!c) goto subst;
625 if (*outb < 8) goto toobig;
626 *(*out)++ = '\033';
627 *(*out)++ = '$';
628 *(*out)++ = 'B';
629 *(*out)++ = c/256;
630 *(*out)++ = c%256;
631 *(*out)++ = '\033';
632 *(*out)++ = '(';
633 *(*out)++ = 'B';
634 *outb -= 8;
635 break;
636 case UCS2:
637 totype = UCS2BE;
638 case UCS2BE:
639 case UCS2LE:
640 case UTF_16:
641 case UTF_16BE:
642 case UTF_16LE:
643 if (c < 0x10000 || totype-UCS2BE < 2U) {
644 if (c >= 0x10000) c = 0xFFFD;
645 if (*outb < 2) goto toobig;
646 put_16((void *)*out, c, totype);
647 *out += 2;
648 *outb -= 2;
649 break;
650 }
651 if (*outb < 4) goto toobig;
652 c -= 0x10000;
653 put_16((void *)*out, (c>>10)|0xd800, totype);
654 put_16((void *)(*out + 2), (c&0x3ff)|0xdc00, totype);
655 *out += 4;
656 *outb -= 4;
657 break;
658 case UTF_32:
659 totype = UTF_32BE;
660 case UTF_32BE:
661 case UTF_32LE:
662 if (*outb < 4) goto toobig;
663 put_32((void *)*out, c, totype);
664 *out += 4;
665 *outb -= 4;
666 break;
667 }
668 }
669 *ploc = loc;
670 return x;
671 ilseq:
672 err = EILSEQ;
673 x = -1;
674 goto end;
675 toobig:
676 err = E2BIG;
677 x = -1;
678 goto end;
679 starved:
680 err = EINVAL;
681 x = -1;
682 end:
683 errno = err;
684 *ploc = loc;
685 return x;
686 }
687