1/* Conversion module for ISO-2022-CN-EXT.
2 Copyright (C) 2000-2016 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4 Contributed by Ulrich Drepper <drepper@cygnus.com>, 2000.
5
6 The GNU C Library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
10
11 The GNU C Library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
15
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, see
18 <http://www.gnu.org/licenses/>. */
19
20#include <dlfcn.h>
21#include <gconv.h>
22#include <stdint.h>
23#include <stdlib.h>
24#include <string.h>
25#include "gb2312.h"
26#include "iso-ir-165.h"
27#include "cns11643.h"
28#include "cns11643l1.h"
29#include "cns11643l2.h"
30
31#include <assert.h>
32
33/* This makes obvious what everybody knows: 0x1b is the Esc character. */
34#define ESC 0x1b
35
36/* We have single-byte shift-in and shift-out sequences, and the single
37 shift sequences SS2 and SS3 which replaces the SS2/SS3 designation for
38 the next two bytes. */
39#define SI 0x0f
40#define SO 0x0e
41#define SS2_0 ESC
42#define SS2_1 0x4e
43#define SS3_0 ESC
44#define SS3_1 0x4f
45
46/* Definitions used in the body of the `gconv' function. */
47#define CHARSET_NAME "ISO-2022-CN-EXT//"
48#define DEFINE_INIT 1
49#define DEFINE_FINI 1
50#define ONE_DIRECTION 0
51#define FROM_LOOP from_iso2022cn_ext_loop
52#define TO_LOOP to_iso2022cn_ext_loop
53#define FROM_LOOP_MIN_NEEDED_FROM 1
54#define FROM_LOOP_MAX_NEEDED_FROM 4
55#define FROM_LOOP_MIN_NEEDED_TO 4
56#define FROM_LOOP_MAX_NEEDED_TO 4
57#define TO_LOOP_MIN_NEEDED_FROM 4
58#define TO_LOOP_MAX_NEEDED_FROM 4
59#define TO_LOOP_MIN_NEEDED_TO 1
60#define TO_LOOP_MAX_NEEDED_TO 6
61#define PREPARE_LOOP \
62 int save_set; \
63 int *setp = &data->__statep->__count;
64#define EXTRA_LOOP_ARGS , setp
65
66
67/* The charsets GB/T 12345-90, GB 7589-87, GB/T 13131-9X, GB 7590-87,
68 and GB/T 13132-9X are not registered to the best of my knowledge and
69 therefore have no escape sequence assigned. We cannot handle them
70 for this reason. Tell the implementation about this. */
71#define X12345 '\0'
72#define X7589 '\0'
73#define X13131 '\0'
74#define X7590 '\0'
75#define X13132 '\0'
76
77
78/* The COUNT element of the state keeps track of the currently selected
79 character set. The possible values are: */
80enum
81{
82 ASCII_set = 0,
83 GB2312_set,
84 GB12345_set,
85 CNS11643_1_set,
86 ISO_IR_165_set,
87 SO_mask = 7,
88
89 GB7589_set = 1 << 3,
90 GB13131_set = 2 << 3,
91 CNS11643_2_set = 3 << 3,
92 SS2_mask = 3 << 3,
93
94 GB7590_set = 1 << 5,
95 GB13132_set = 2 << 5,
96 CNS11643_3_set = 3 << 5,
97 CNS11643_4_set = 4 << 5,
98 CNS11643_5_set = 5 << 5,
99 CNS11643_6_set = 6 << 5,
100 CNS11643_7_set = 7 << 5,
101 SS3_mask = 7 << 5,
102
103#define CURRENT_MASK (SO_mask | SS2_mask | SS3_mask)
104
105 GB2312_ann = 1 << 8,
106 GB12345_ann = 2 << 8,
107 CNS11643_1_ann = 3 << 8,
108 ISO_IR_165_ann = 4 << 8,
109 SO_ann = 7 << 8,
110
111 GB7589_ann = 1 << 11,
112 GB13131_ann = 2 << 11,
113 CNS11643_2_ann = 3 << 11,
114 SS2_ann = 3 << 11,
115
116 GB7590_ann = 1 << 13,
117 GB13132_ann = 2 << 13,
118 CNS11643_3_ann = 3 << 13,
119 CNS11643_4_ann = 4 << 13,
120 CNS11643_5_ann = 5 << 13,
121 CNS11643_6_ann = 6 << 13,
122 CNS11643_7_ann = 7 << 13,
123 SS3_ann = 7 << 13
124};
125
126
127/* Since this is a stateful encoding we have to provide code which resets
128 the output state to the initial state. This has to be done during the
129 flushing. */
130#define EMIT_SHIFT_TO_INIT \
131 if (data->__statep->__count >> 3 != ASCII_set) \
132 { \
133 if (FROM_DIRECTION) \
134 /* It's easy, we don't have to emit anything, we just reset the \
135 state for the input. */ \
136 data->__statep->__count = ASCII_set << 3; \
137 else \
138 { \
139 /* We are not in the initial state. To switch back we have \
140 to emit `SI'. */ \
141 if (__glibc_unlikely (outbuf == outend)) \
142 /* We don't have enough room in the output buffer. */ \
143 status = __GCONV_FULL_OUTPUT; \
144 else \
145 { \
146 /* Write out the shift sequence. */ \
147 *outbuf++ = SI; \
148 if (data->__flags & __GCONV_IS_LAST) \
149 *irreversible += 1; \
150 data->__statep->__count = ASCII_set << 3; \
151 } \
152 } \
153 }
154
155
156/* Since we might have to reset input pointer we must be able to save
157 and retore the state. */
158#define SAVE_RESET_STATE(Save) \
159 if (Save) \
160 save_set = *setp; \
161 else \
162 *setp = save_set
163
164
165/* First define the conversion function from ISO-2022-CN to UCS4. */
166#define MIN_NEEDED_INPUT FROM_LOOP_MIN_NEEDED_FROM
167#define MAX_NEEDED_INPUT FROM_LOOP_MAX_NEEDED_FROM
168#define MIN_NEEDED_OUTPUT FROM_LOOP_MIN_NEEDED_TO
169#define MAX_NEEDED_OUTPUT FROM_LOOP_MAX_NEEDED_TO
170#define LOOPFCT FROM_LOOP
171#define BODY \
172 { \
173 uint32_t ch = *inptr; \
174 \
175 /* This is a 7bit character set, disallow all 8bit characters. */ \
176 if (ch > 0x7f) \
177 STANDARD_FROM_LOOP_ERR_HANDLER (1); \
178 \
179 /* Recognize escape sequences. */ \
180 if (ch == ESC) \
181 { \
182 /* There are three kinds of escape sequences we have to handle: \
183 - those announcing the use of GB and CNS characters on the \
184 line; we can simply ignore them \
185 - the initial byte of the SS2 sequence. \
186 - the initial byte of the SS3 sequence. \
187 */ \
188 if (inptr + 2 > inend \
189 || (inptr[1] == '$' \
190 && (inptr + 3 > inend \
191 || (inptr[2] == ')' && inptr + 4 > inend) \
192 || (inptr[2] == '*' && inptr + 4 > inend) \
193 || (inptr[2] == '+' && inptr + 4 > inend))) \
194 || (inptr[1] == SS2_1 && inptr + 4 > inend) \
195 || (inptr[1] == SS3_1 && inptr + 4 > inend)) \
196 { \
197 result = __GCONV_INCOMPLETE_INPUT; \
198 break; \
199 } \
200 if (inptr[1] == '$' \
201 && ((inptr[2] == ')' \
202 && (inptr[3] == 'A' \
203 || (X12345 != '\0' && inptr[3] == X12345) \
204 || inptr[3] == 'E' || inptr[3] == 'G')) \
205 || (inptr[2] == '*' \
206 && ((X7589 != '\0' && inptr[3] == X7589) \
207 || (X13131 != '\0' && inptr[3] == X13131) \
208 || inptr[3] == 'H')) \
209 || (inptr[2] == '+' \
210 && ((X7590 != '\0' && inptr[3] == X7590) \
211 || (X13132 != '\0' && inptr[3] == X13132) \
212 || inptr[3] == 'I' || inptr[3] == 'J' \
213 || inptr[3] == 'K' || inptr[3] == 'L' \
214 || inptr[3] == 'M')))) \
215 { \
216 /* OK, we accept those character sets. */ \
217 if (inptr[3] == 'A') \
218 ann = (ann & ~SO_ann) | GB2312_ann; \
219 else if (inptr[3] == 'G') \
220 ann = (ann & ~SO_ann) | CNS11643_1_ann; \
221 else if (inptr[3] == 'E') \
222 ann = (ann & ~SO_ann) | ISO_IR_165_ann; \
223 else if (X12345 != '\0' && inptr[3] == X12345) \
224 ann = (ann & ~SO_ann) | GB12345_ann; \
225 else if (inptr[3] == 'H') \
226 ann = (ann & ~SS2_ann) | CNS11643_2_ann; \
227 else if (X7589 != '\0' && inptr[3] == X7589) \
228 ann = (ann & ~SS2_ann) | GB7589_ann; \
229 else if (X13131 != '\0' && inptr[3] == X13131) \
230 ann = (ann & ~SS2_ann) | GB13131_ann; \
231 else if (inptr[3] == 'I') \
232 ann = (ann & ~SS3_ann) | CNS11643_3_ann; \
233 else if (inptr[3] == 'J') \
234 ann = (ann & ~SS3_ann) | CNS11643_4_ann; \
235 else if (inptr[3] == 'K') \
236 ann = (ann & ~SS3_ann) | CNS11643_5_ann; \
237 else if (inptr[3] == 'L') \
238 ann = (ann & ~SS3_ann) | CNS11643_6_ann; \
239 else if (inptr[3] == 'M') \
240 ann = (ann & ~SS3_ann) | CNS11643_7_ann; \
241 else if (X7590 != '\0' && inptr[3] == X7590) \
242 ann = (ann & ~SS3_ann) | GB7590_ann; \
243 else if (X13132 != '\0' && inptr[3] == X13132) \
244 ann = (ann & ~SS3_ann) | GB13132_ann; \
245 inptr += 4; \
246 continue; \
247 } \
248 } \
249 else if (ch == SO) \
250 { \
251 /* Switch to use GB2312, GB12345, CNS 11643 plane 1, or ISO-IR-165, \
252 depending on which S0 designation came last. The only problem \
253 is what to do with faulty input files where no designator came. \
254 XXX For now I'll default to use GB2312. If this is not the \
255 best behavior (e.g., we should flag an error) let me know. */ \
256 ++inptr; \
257 if ((ann & SO_ann) != 0) \
258 switch (ann & SO_ann) \
259 { \
260 case GB2312_ann: \
261 set = GB2312_set; \
262 break; \
263 case GB12345_ann: \
264 set = GB12345_set; \
265 break; \
266 case CNS11643_1_ann: \
267 set = CNS11643_1_set; \
268 break; \
269 case ISO_IR_165_ann: \
270 set = ISO_IR_165_set; \
271 break; \
272 default: \
273 abort (); \
274 } \
275 else \
276 { \
277 STANDARD_FROM_LOOP_ERR_HANDLER (1); \
278 } \
279 continue; \
280 } \
281 else if (ch == SI) \
282 { \
283 /* Switch to use ASCII. */ \
284 ++inptr; \
285 set = ASCII_set; \
286 continue; \
287 } \
288 \
289 if (ch == ESC && inptr[1] == SS2_1) \
290 { \
291 /* This is a character from CNS 11643 plane 2. \
292 XXX We could test here whether the use of this character \
293 set was announced. \
294 XXX Currently GB7589 and GB13131 are not supported. */ \
295 inptr += 2; \
296 ch = cns11643l2_to_ucs4 (&inptr, 2, 0); \
297 if (ch == __UNKNOWN_10646_CHAR) \
298 STANDARD_FROM_LOOP_ERR_HANDLER (2); \
299 } \
300 /* Note that we can assume here that at least 4 bytes are available if \
301 the first byte is ESC since otherwise the first if would have been \
302 true. */ \
303 else if (ch == ESC && inptr[1] == SS3_1) \
304 { \
305 /* This is a character from CNS 11643 plane 3 or higher. \
306 XXX Currently GB7590 and GB13132 are not supported. */ \
307 unsigned char buf[3]; \
308 const unsigned char *tmp = buf; \
309 \
310 buf[1] = inptr[2]; \
311 buf[2] = inptr[3]; \
312 switch (ann & SS3_ann) \
313 { \
314 case CNS11643_3_ann: \
315 buf[0] = 0x23; \
316 ch = cns11643_to_ucs4 (&tmp, 3, 0); \
317 break; \
318 case CNS11643_4_ann: \
319 buf[0] = 0x24; \
320 ch = cns11643_to_ucs4 (&tmp, 3, 0); \
321 break; \
322 case CNS11643_5_ann: \
323 buf[0] = 0x25; \
324 ch = cns11643_to_ucs4 (&tmp, 3, 0); \
325 break; \
326 case CNS11643_6_ann: \
327 buf[0] = 0x26; \
328 ch = cns11643_to_ucs4 (&tmp, 3, 0); \
329 break; \
330 case CNS11643_7_ann: \
331 buf[0] = 0x27; \
332 ch = cns11643_to_ucs4 (&tmp, 3, 0); \
333 break; \
334 default: \
335 /* XXX Currently GB7590 and GB13132 are not supported. */ \
336 ch = __UNKNOWN_10646_CHAR; \
337 break; \
338 } \
339 if (ch == __UNKNOWN_10646_CHAR) \
340 { \
341 STANDARD_FROM_LOOP_ERR_HANDLER (4); \
342 } \
343 assert (tmp == buf + 3); \
344 inptr += 4; \
345 } \
346 else if (set == ASCII_set) \
347 { \
348 /* Almost done, just advance the input pointer. */ \
349 ++inptr; \
350 } \
351 else \
352 { \
353 /* That's pretty easy, we have a dedicated functions for this. */ \
354 if (inend - inptr < 2) \
355 { \
356 result = __GCONV_INCOMPLETE_INPUT; \
357 break; \
358 } \
359 if (set == GB2312_set) \
360 ch = gb2312_to_ucs4 (&inptr, inend - inptr, 0); \
361 else if (set == ISO_IR_165_set) \
362 ch = isoir165_to_ucs4 (&inptr, inend - inptr); \
363 else \
364 { \
365 assert (set == CNS11643_1_set); \
366 ch = cns11643l1_to_ucs4 (&inptr, inend - inptr, 0); \
367 } \
368 \
369 if (ch == 0) \
370 { \
371 result = __GCONV_INCOMPLETE_INPUT; \
372 break; \
373 } \
374 else if (ch == __UNKNOWN_10646_CHAR) \
375 { \
376 STANDARD_FROM_LOOP_ERR_HANDLER (2); \
377 } \
378 } \
379 \
380 *((uint32_t *) outptr) = ch; \
381 outptr += sizeof (uint32_t); \
382 }
383#define EXTRA_LOOP_DECLS , int *setp
384#define INIT_PARAMS int set = (*setp >> 3) & CURRENT_MASK; \
385 int ann = (*setp >> 3) & ~CURRENT_MASK
386#define UPDATE_PARAMS *setp = (set | ann) << 3
387#define LOOP_NEED_FLAGS
388#include <iconv/loop.c>
389
390
391/* Next, define the other direction. */
392#define MIN_NEEDED_INPUT TO_LOOP_MIN_NEEDED_FROM
393#define MAX_NEEDED_INPUT TO_LOOP_MAX_NEEDED_FROM
394#define MIN_NEEDED_OUTPUT TO_LOOP_MIN_NEEDED_TO
395#define MAX_NEEDED_OUTPUT TO_LOOP_MAX_NEEDED_TO
396#define LOOPFCT TO_LOOP
397#define BODY \
398 { \
399 uint32_t ch; \
400 size_t written = 0; \
401 \
402 ch = *((const uint32_t *) inptr); \
403 \
404 /* First see whether we can write the character using the currently \
405 selected character set. */ \
406 if (ch < 0x80) \
407 { \
408 if (set != ASCII_set) \
409 { \
410 *outptr++ = SI; \
411 set = ASCII_set; \
412 if (outptr == outend) \
413 { \
414 result = __GCONV_FULL_OUTPUT; \
415 break; \
416 } \
417 } \
418 \
419 *outptr++ = ch; \
420 written = 1; \
421 \
422 /* At the end of the line we have to clear the `ann' flags since \
423 every line must contain this information again. */ \
424 if (ch == L'\n') \
425 ann = 0; \
426 } \
427 else \
428 { \
429 unsigned char buf[2] = { 0, 0 }; \
430 int used; \
431 \
432 if (set == GB2312_set || ((ann & SO_ann) != CNS11643_1_ann \
433 && (ann & SO_ann) != ISO_IR_165_ann)) \
434 { \
435 written = ucs4_to_gb2312 (ch, buf, 2); \
436 used = GB2312_set; \
437 } \
438 else if (set == ISO_IR_165_set || (ann & SO_ann) == ISO_IR_165_set) \
439 { \
440 written = ucs4_to_isoir165 (ch, buf, 2); \
441 used = ISO_IR_165_set; \
442 } \
443 else \
444 { \
445 written = ucs4_to_cns11643l1 (ch, buf, 2); \
446 used = CNS11643_1_set; \
447 } \
448 \
449 if (written == __UNKNOWN_10646_CHAR) \
450 { \
451 /* Cannot convert it using the currently selected SO set. \
452 Next try the SS2 set. */ \
453 written = ucs4_to_cns11643l2 (ch, buf, 2); \
454 if (written != __UNKNOWN_10646_CHAR) \
455 /* Yep, that worked. */ \
456 used = CNS11643_2_set; \
457 else \
458 { \
459 unsigned char tmpbuf[3]; \
460 \
461 switch (0) \
462 { \
463 default: \
464 /* Well, see whether we have to change the SO set. */ \
465 \
466 if (used != GB2312_set) \
467 { \
468 written = ucs4_to_gb2312 (ch, buf, 2); \
469 if (written != __UNKNOWN_10646_CHAR) \
470 { \
471 used = GB2312_set; \
472 break; \
473 } \
474 } \
475 \
476 if (used != ISO_IR_165_set) \
477 { \
478 written = ucs4_to_isoir165 (ch, buf, 2); \
479 if (written != __UNKNOWN_10646_CHAR) \
480 { \
481 used = ISO_IR_165_set; \
482 break; \
483 } \
484 } \
485 \
486 if (used != CNS11643_1_set) \
487 { \
488 written = ucs4_to_cns11643l1 (ch, buf, 2); \
489 if (written != __UNKNOWN_10646_CHAR) \
490 { \
491 used = CNS11643_1_set; \
492 break; \
493 } \
494 } \
495 \
496 written = ucs4_to_cns11643 (ch, tmpbuf, 3); \
497 if (written == 3 && tmpbuf[0] >= 3 && tmpbuf[0] <= 7) \
498 { \
499 buf[0] = tmpbuf[1]; \
500 buf[1] = tmpbuf[2]; \
501 switch (tmpbuf[0]) \
502 { \
503 case 3: \
504 used = CNS11643_3_set; \
505 break; \
506 case 4: \
507 used = CNS11643_4_set; \
508 break; \
509 case 5: \
510 used = CNS11643_5_set; \
511 break; \
512 case 6: \
513 used = CNS11643_6_set; \
514 break; \
515 case 7: \
516 used = CNS11643_7_set; \
517 break; \
518 default: \
519 abort (); \
520 } \
521 written = 2; \
522 break; \
523 } \
524 \
525 /* XXX Currently GB7590 and GB13132 are not supported. */\
526 \
527 /* Even this does not work. Error. */ \
528 used = ASCII_set; \
529 } \
530 if (used == ASCII_set) \
531 { \
532 UNICODE_TAG_HANDLER (ch, 4); \
533 STANDARD_TO_LOOP_ERR_HANDLER (4); \
534 } \
535 } \
536 } \
537 assert (written == 2); \
538 \
539 /* See whether we have to emit an escape sequence. */ \
540 if (set != used) \
541 { \
542 /* First see whether we announced that we use this \
543 character set. */ \
544 if ((used & SO_mask) != 0 && (ann & SO_ann) != (used << 8)) \
545 { \
546 const char *escseq; \
547 \
548 if (outptr + 4 > outend) \
549 { \
550 result = __GCONV_FULL_OUTPUT; \
551 break; \
552 } \
553 \
554 assert (used >= 1 && used <= 4); \
555 escseq = ")A\0\0)G)E" + (used - 1) * 2; \
556 *outptr++ = ESC; \
557 *outptr++ = '$'; \
558 *outptr++ = *escseq++; \
559 *outptr++ = *escseq++; \
560 \
561 ann = (ann & ~SO_ann) | (used << 8); \
562 } \
563 else if ((used & SS2_mask) != 0 && (ann & SS2_ann) != (used << 8))\
564 { \
565 const char *escseq; \
566 \
567 assert (used == CNS11643_2_set); /* XXX */ \
568 escseq = "*H"; \
569 *outptr++ = ESC; \
570 *outptr++ = '$'; \
571 *outptr++ = *escseq++; \
572 *outptr++ = *escseq++; \
573 \
574 ann = (ann & ~SS2_ann) | (used << 8); \
575 } \
576 else if ((used & SS3_mask) != 0 && (ann & SS3_ann) != (used << 8))\
577 { \
578 const char *escseq; \
579 \
580 assert ((used >> 5) >= 3 && (used >> 5) <= 7); \
581 escseq = "+I+J+K+L+M" + ((used >> 5) - 3) * 2; \
582 *outptr++ = ESC; \
583 *outptr++ = '$'; \
584 *outptr++ = *escseq++; \
585 *outptr++ = *escseq++; \
586 \
587 ann = (ann & ~SS3_ann) | (used << 8); \
588 } \
589 \
590 if (used == CNS11643_2_set) \
591 { \
592 if (outptr + 2 > outend) \
593 { \
594 result = __GCONV_FULL_OUTPUT; \
595 break; \
596 } \
597 *outptr++ = SS2_0; \
598 *outptr++ = SS2_1; \
599 } \
600 else if (used >= CNS11643_3_set && used <= CNS11643_7_set) \
601 { \
602 if (outptr + 2 > outend) \
603 { \
604 result = __GCONV_FULL_OUTPUT; \
605 break; \
606 } \
607 *outptr++ = SS3_0; \
608 *outptr++ = SS3_1; \
609 } \
610 else \
611 { \
612 /* We only have to emit something if currently ASCII is \
613 selected. Otherwise we are switching within the \
614 SO charset. */ \
615 if (set == ASCII_set) \
616 { \
617 if (outptr + 1 > outend) \
618 { \
619 result = __GCONV_FULL_OUTPUT; \
620 break; \
621 } \
622 *outptr++ = SO; \
623 } \
624 } \
625 \
626 /* Always test the length here since we have used up all the \
627 guaranteed output buffer slots. */ \
628 if (outptr + 2 > outend) \
629 { \
630 result = __GCONV_FULL_OUTPUT; \
631 break; \
632 } \
633 } \
634 else if (outptr + 2 > outend) \
635 { \
636 result = __GCONV_FULL_OUTPUT; \
637 break; \
638 } \
639 \
640 *outptr++ = buf[0]; \
641 *outptr++ = buf[1]; \
642 set = used; \
643 } \
644 \
645 /* Now that we wrote the output increment the input pointer. */ \
646 inptr += 4; \
647 }
648#define EXTRA_LOOP_DECLS , int *setp
649#define INIT_PARAMS int set = (*setp >> 3) & CURRENT_MASK; \
650 int ann = (*setp >> 3) & ~CURRENT_MASK
651#define REINIT_PARAMS do \
652 { \
653 set = (*setp >> 3) & CURRENT_MASK; \
654 ann = (*setp >> 3) & ~CURRENT_MASK; \
655 } \
656 while (0)
657#define UPDATE_PARAMS *setp = (set | ann) << 3
658#define LOOP_NEED_FLAGS
659#include <iconv/loop.c>
660
661
662/* Now define the toplevel functions. */
663#include <iconv/skeleton.c>
664