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
9 #define UTF_32BE 0300
10 #define UTF_16LE 0301
11 #define UTF_16BE 0302
12 #define UTF_32LE 0303
13 #define UCS2BE 0304
14 #define UCS2LE 0305
15 #define WCHAR_T 0306
16 #define US_ASCII 0307
17 #define UTF_8 0310
18 #define EUC_JP 0320
19 #define SHIFT_JIS 0321
20 #define GB18030 0330
21 #define GBK 0331
22 #define GB2312 0332
23 #define BIG5 0340
24 #define EUC_KR 0350
25
26 /* FIXME: these are not implemented yet
27 * EUC: A1-FE A1-FE
28 * GBK: 81-FE 40-7E,80-FE
29 * Big5: A1-FE 40-7E,A1-FE
30 */
31
32 /* Definitions of charmaps. Each charmap consists of:
33 * 1. Empty-string-terminated list of null-terminated aliases.
34 * 2. Special type code or number of elided entries.
35 * 3. Character table (size determined by field 2). */
36
37 static const unsigned char charmaps[] =
38 "utf8\0\0\310"
39 "wchart\0\0\306"
40 "ucs2\0ucs2be\0\0\304"
41 "ucs2le\0\0\305"
42 "utf16\0utf16be\0\0\302"
43 "utf16le\0\0\301"
44 "ucs4\0ucs4be\0utf32\0utf32be\0\0\300"
45 "ucs4le\0utf32le\0\0\303"
46 "ascii\0usascii\0iso646\0iso646us\0\0\307"
47 "eucjp\0\0\320"
48 "shiftjis\0sjis\0\0\321"
49 "gb18030\0\0\330"
50 "gbk\0\0\331"
51 "gb2312\0\0\332"
52 "big5\0bigfive\0cp950\0big5hkscs\0\0\340"
53 "euckr\0ksc5601\0ksx1001\0cp949\0\0\350"
54 #include "codepages.h"
55 ;
56
57 static const unsigned short legacy_chars[] = {
58 #include "legacychars.h"
59 };
60
61 static const unsigned short jis0208[84][94] = {
62 #include "jis0208.h"
63 };
64
65 static const unsigned short gb18030[126][190] = {
66 #include "gb18030.h"
67 };
68
69 static const unsigned short big5[89][157] = {
70 #include "big5.h"
71 };
72
73 static const unsigned short hkscs[] = {
74 #include "hkscs.h"
75 };
76
77 static const unsigned short ksc[93][94] = {
78 #include "ksc.h"
79 };
80
fuzzycmp(const unsigned char * a,const unsigned char * b)81 static int fuzzycmp(const unsigned char *a, const unsigned char *b)
82 {
83 for (; *a && *b; a++, b++) {
84 while (*a && (*a|32U)-'a'>26 && *a-'0'>10U) a++;
85 if ((*a|32U) != *b) return 1;
86 }
87 return *a != *b;
88 }
89
find_charmap(const void * name)90 static size_t find_charmap(const void *name)
91 {
92 const unsigned char *s;
93 for (s=charmaps; *s; ) {
94 if (!fuzzycmp(name, s)) {
95 for (; *s; s+=strlen((void *)s)+1);
96 return s+1-charmaps;
97 }
98 s += strlen((void *)s)+1;
99 if (!*s) {
100 if (s[1] > 0200) s+=2;
101 else s+=2+(128U-s[1])/4*5;
102 }
103 }
104 return -1;
105 }
106
iconv_open(const char * to,const char * from)107 iconv_t iconv_open(const char *to, const char *from)
108 {
109 size_t f, t;
110
111 if ((t = find_charmap(to))==-1
112 || (f = find_charmap(from))==-1
113 || (charmaps[t] >= 0320)) {
114 errno = EINVAL;
115 return (iconv_t)-1;
116 }
117
118 return (void *)(f<<16 | t);
119 }
120
iconv_close(iconv_t cd)121 int iconv_close(iconv_t cd)
122 {
123 return 0;
124 }
125
get_16(const unsigned char * s,int e)126 static unsigned get_16(const unsigned char *s, int e)
127 {
128 e &= 1;
129 return s[e]<<8 | s[1-e];
130 }
131
put_16(unsigned char * s,unsigned c,int e)132 static void put_16(unsigned char *s, unsigned c, int e)
133 {
134 e &= 1;
135 s[e] = c>>8;
136 s[1-e] = c;
137 }
138
get_32(const unsigned char * s,int e)139 static unsigned get_32(const unsigned char *s, int e)
140 {
141 e &= 3;
142 return s[e]+0U<<24 | s[e^1]<<16 | s[e^2]<<8 | s[e^3];
143 }
144
put_32(unsigned char * s,unsigned c,int e)145 static void put_32(unsigned char *s, unsigned c, int e)
146 {
147 e &= 3;
148 s[e^0] = c>>24;
149 s[e^1] = c>>16;
150 s[e^2] = c>>8;
151 s[e^3] = c;
152 }
153
154 /* Adapt as needed */
155 #define mbrtowc_utf8 mbrtowc
156 #define wctomb_utf8 wctomb
157
iconv(iconv_t cd0,char ** restrict in,size_t * restrict inb,char ** restrict out,size_t * restrict outb)158 size_t iconv(iconv_t cd0, char **restrict in, size_t *restrict inb, char **restrict out, size_t *restrict outb)
159 {
160 size_t x=0;
161 unsigned long cd = (unsigned long)cd0;
162 unsigned to = cd & 0xffff;
163 unsigned from = cd >> 16;
164 const unsigned char *map = charmaps+from+1;
165 const unsigned char *tomap = charmaps+to+1;
166 mbstate_t st = {0};
167 wchar_t wc;
168 unsigned c, d;
169 size_t k, l;
170 int err;
171 unsigned char type = map[-1];
172 unsigned char totype = tomap[-1];
173
174 if (!in || !*in || !*inb) return 0;
175
176 for (; *inb; *in+=l, *inb-=l) {
177 c = *(unsigned char *)*in;
178 l = 1;
179
180 if (c >= 128 || type-UTF_32BE < 7U) switch (type) {
181 case UTF_8:
182 l = mbrtowc_utf8(&wc, *in, *inb, &st);
183 if (!l) l++;
184 else if (l == (size_t)-1) goto ilseq;
185 else if (l == (size_t)-2) goto starved;
186 c = wc;
187 break;
188 case US_ASCII:
189 goto ilseq;
190 case WCHAR_T:
191 l = sizeof(wchar_t);
192 if (*inb < l) goto starved;
193 c = *(wchar_t *)*in;
194 if (0) {
195 case UTF_32BE:
196 case UTF_32LE:
197 l = 4;
198 if (*inb < 4) goto starved;
199 c = get_32((void *)*in, type);
200 }
201 if (c-0xd800u < 0x800u || c >= 0x110000u) goto ilseq;
202 break;
203 case UCS2BE:
204 case UCS2LE:
205 case UTF_16BE:
206 case UTF_16LE:
207 l = 2;
208 if (*inb < 2) goto starved;
209 c = get_16((void *)*in, type);
210 if ((unsigned)(c-0xdc00) < 0x400) goto ilseq;
211 if ((unsigned)(c-0xd800) < 0x400) {
212 if (type-UCS2BE < 2U) goto ilseq;
213 l = 4;
214 if (*inb < 4) goto starved;
215 d = get_16((void *)(*in + 2), type);
216 if ((unsigned)(d-0xdc00) >= 0x400) goto ilseq;
217 c = ((c-0xd7c0)<<10) + (d-0xdc00);
218 }
219 break;
220 case SHIFT_JIS:
221 if (c-0xa1 <= 0xdf-0xa1) {
222 c += 0xff61-0xa1;
223 break;
224 }
225 l = 2;
226 if (*inb < 2) goto starved;
227 d = *((unsigned char *)*in + 1);
228 if (c-129 <= 159-129) c -= 129;
229 else if (c-224 <= 239-224) c -= 193;
230 else goto ilseq;
231 c *= 2;
232 if (d-64 <= 158-64) {
233 if (d==127) goto ilseq;
234 if (d>127) d--;
235 d -= 64;
236 } else if (d-159 <= 252-159) {
237 c++;
238 d -= 159;
239 }
240 c = jis0208[c][d];
241 if (!c) goto ilseq;
242 break;
243 case EUC_JP:
244 l = 2;
245 if (*inb < 2) goto starved;
246 d = *((unsigned char *)*in + 1);
247 if (c==0x8e) {
248 c = d;
249 if (c-0xa1 > 0xdf-0xa1) goto ilseq;
250 c += 0xff61 - 0xa1;
251 break;
252 }
253 c -= 0xa1;
254 d -= 0xa1;
255 if (c >= 84 || d >= 94) goto ilseq;
256 c = jis0208[c][d];
257 if (!c) goto ilseq;
258 break;
259 case GB2312:
260 if (c < 0xa1) goto ilseq;
261 case GBK:
262 case GB18030:
263 c -= 0x81;
264 if (c >= 126) goto ilseq;
265 l = 2;
266 if (*inb < 2) goto starved;
267 d = *((unsigned char *)*in + 1);
268 if (d < 0xa1 && type == GB2312) goto ilseq;
269 if (d-0x40>=191 || d==127) {
270 if (d-'0'>9 || type != GB18030)
271 goto ilseq;
272 l = 4;
273 if (*inb < 4) goto starved;
274 c = (10*c + d-'0') * 1260;
275 d = *((unsigned char *)*in + 2);
276 if (d-0x81>126) goto ilseq;
277 c += 10*(d-0x81);
278 d = *((unsigned char *)*in + 3);
279 if (d-'0'>9) goto ilseq;
280 c += d-'0';
281 c += 128;
282 for (d=0; d<=c; ) {
283 k = 0;
284 int i, j;
285 for (i=0; i<126; i++)
286 for (j=0; j<190; j++)
287 if (gb18030[i][j]-d <= c-d)
288 k++;
289 d = c+1;
290 c += k;
291 }
292 break;
293 }
294 d -= 0x40;
295 if (d>63) d--;
296 c = gb18030[c][d];
297 break;
298 case BIG5:
299 l = 2;
300 if (*inb < 2) goto starved;
301 d = *((unsigned char *)*in + 1);
302 if (d-0x40>=0xff-0x40 || d-0x7f<0xa1-0x7f) goto ilseq;
303 d -= 0x40;
304 if (d > 0x3e) d -= 0x22;
305 if (c-0xa1>=0xfa-0xa1) {
306 if (c-0x87>=0xff-0x87) goto ilseq;
307 if (c < 0xa1) c -= 0x87;
308 else c -= 0x87 + (0xfa-0xa1);
309 c = (hkscs[4867+(c*157+d)/16]>>(c*157+d)%16)%2<<17
310 | hkscs[c*157+d];
311 /* A few HKSCS characters map to pairs of UCS
312 * characters. These are mapped to surrogate
313 * range in the hkscs table then hard-coded
314 * here. Ugly, yes. */
315 if (c/256 == 0xdc) {
316 if (totype-0300U > 8) k = 2;
317 else k = "\10\4\4\10\4\4\10\2\4"[totype-0300];
318 if (k > *outb) goto toobig;
319 x += iconv((iconv_t)(uintptr_t)to,
320 &(char *){"\303\212\314\204"
321 "\303\212\314\214"
322 "\303\252\314\204"
323 "\303\252\314\214"
324 +c%256}, &(size_t){4},
325 out, outb);
326 continue;
327 }
328 if (!c) goto ilseq;
329 break;
330 }
331 c -= 0xa1;
332 c = big5[c][d]|(c==0x27&&(d==0x3a||d==0x3c||d==0x42))<<17;
333 if (!c) goto ilseq;
334 break;
335 case EUC_KR:
336 l = 2;
337 if (*inb < 2) goto starved;
338 d = *((unsigned char *)*in + 1);
339 c -= 0xa1;
340 d -= 0xa1;
341 if (c >= 93 || d >= 94) {
342 c += (0xa1-0x81);
343 d += 0xa1;
344 if (c >= 93 || c>=0xc6-0x81 && d>0x52)
345 goto ilseq;
346 if (d-'A'<26) d = d-'A';
347 else if (d-'a'<26) d = d-'a'+26;
348 else if (d-0x81<0xff-0x81) d = d-0x81+52;
349 else goto ilseq;
350 if (c < 0x20) c = 178*c + d;
351 else c = 178*0x20 + 84*(c-0x20) + d;
352 c += 0xac00;
353 for (d=0xac00; d<=c; ) {
354 k = 0;
355 int i, j;
356 for (i=0; i<93; i++)
357 for (j=0; j<94; j++)
358 if (ksc[i][j]-d <= c-d)
359 k++;
360 d = c+1;
361 c += k;
362 }
363 break;
364 }
365 c = ksc[c][d];
366 if (!c) goto ilseq;
367 break;
368 default:
369 if (c < 128+type) break;
370 c -= 128+type;
371 c = legacy_chars[ map[c*5/4]>>2*c%8 |
372 map[c*5/4+1]<<8-2*c%8 & 1023 ];
373 if (!c) c = *(unsigned char *)*in;
374 if (c==1) goto ilseq;
375 }
376
377 switch (totype) {
378 case WCHAR_T:
379 if (*outb < sizeof(wchar_t)) goto toobig;
380 *(wchar_t *)*out = c;
381 *out += sizeof(wchar_t);
382 *outb -= sizeof(wchar_t);
383 break;
384 case UTF_8:
385 if (*outb < 4) {
386 char tmp[4];
387 k = wctomb_utf8(tmp, c);
388 if (*outb < k) goto toobig;
389 memcpy(*out, tmp, k);
390 } else k = wctomb_utf8(*out, c);
391 *out += k;
392 *outb -= k;
393 break;
394 case US_ASCII:
395 if (c > 0x7f) subst: x++, c='*';
396 default:
397 if (*outb < 1) goto toobig;
398 if (c < 128+totype) {
399 revout:
400 *(*out)++ = c;
401 *outb -= 1;
402 break;
403 }
404 d = c;
405 for (c=0; c<128-totype; c++) {
406 if (d == legacy_chars[ tomap[c*5/4]>>2*c%8 |
407 tomap[c*5/4+1]<<8-2*c%8 & 1023 ]) {
408 c += 128;
409 goto revout;
410 }
411 }
412 goto subst;
413 case UCS2BE:
414 case UCS2LE:
415 case UTF_16BE:
416 case UTF_16LE:
417 if (c < 0x10000 || type-UCS2BE < 2U) {
418 if (c >= 0x10000) c = 0xFFFD;
419 if (*outb < 2) goto toobig;
420 put_16((void *)*out, c, totype);
421 *out += 2;
422 *outb -= 2;
423 break;
424 }
425 if (*outb < 4) goto toobig;
426 c -= 0x10000;
427 put_16((void *)*out, (c>>10)|0xd800, totype);
428 put_16((void *)(*out + 2), (c&0x3ff)|0xdc00, totype);
429 *out += 4;
430 *outb -= 4;
431 break;
432 case UTF_32BE:
433 case UTF_32LE:
434 if (*outb < 4) goto toobig;
435 put_32((void *)*out, c, totype);
436 *out += 4;
437 *outb -= 4;
438 break;
439 }
440 }
441 return x;
442 ilseq:
443 err = EILSEQ;
444 x = -1;
445 goto end;
446 toobig:
447 err = E2BIG;
448 x = -1;
449 goto end;
450 starved:
451 err = EINVAL;
452 x = -1;
453 end:
454 errno = err;
455 return x;
456 }
457