2 * iso2022.c - support for ISO/IEC 2022 (alias ECMA-35).
4 * This isn't a complete implementation of ISO/IEC 2022, but it's
5 * close. It can decode 8-bit and 7-bit versions, with support for
6 * single-byte and multi-byte character sets, all four containers
7 * (G0, G1, G2, and G3), using both single-shift and locking-shift
10 * The general principle is that any valid ISO/IEC 2022 sequence
11 * should either be correctly decoded or should emit an ERROR. The
12 * only exception to this is that the C0 and C1 sets are fixed as
13 * those of ISO/IEC 6429. Escape sequences for designating control
14 * sets are passed through, so a post-processor could fix them up if
17 * DOCS to UTF-8 works. Other DOCS sequences are ignored, which will
18 * produce surprising results.
36 enum {S4
, S6
, M4
, M6
};
38 static long int emacs_big5_1_to_unicode(int, int);
39 static long int emacs_big5_2_to_unicode(int, int);
40 static int unicode_to_emacs_big5(long int, int *, int *, int *);
41 static long int cns11643_1_to_unicode(int, int);
42 static long int cns11643_2_to_unicode(int, int);
43 static long int cns11643_3_to_unicode(int, int);
44 static long int cns11643_4_to_unicode(int, int);
45 static long int cns11643_5_to_unicode(int, int);
46 static long int cns11643_6_to_unicode(int, int);
47 static long int cns11643_7_to_unicode(int, int);
48 static long int null_dbcs_to_unicode(int, int);
49 static int unicode_to_null_dbcs(long int, int *, int *);
51 typedef int (*to_dbcs_t
)(long int, int *, int *);
52 typedef int (*to_dbcs_planar_t
)(long int, int *, int *, int *);
55 * These macros cast between to_dbcs_planar_t and to_dbcs_t, in
56 * such a way as to cause a compile-time error if the input is not
57 * of the appropriate type.
59 * Defining these portably is quite fiddly. My first effort was as
61 * #define DEPLANARISE(x) ( (x) == (to_dbcs_planar_t)NULL, (to_dbcs_t)(x) )
63 * so that the comparison on the left of the comma provokes the
64 * type check error, and the cast on the right is the actual
67 * gcc was entirely happy with this. However, when used in a static
68 * initialiser, MSVC objected - justifiably - that the first half
69 * of the comma expression wasn't constant and thus the expression
70 * as a whole was not a constant expression. We can get round this
71 * by enclosing the comparison in `sizeof', so that it isn't
74 * But then we run into a second problem, which is that C actually
75 * disallows the use of the comma operator within a constant
76 * expression for any purpose at all! Presumably this is on the
77 * basis that its purpose is to have side effects and constant
78 * expressions can't; unfortunately, this specific case is one in
79 * which the desired side effect is a compile-time rather than a
82 * We are permitted to use ?:, however, and that works quite well
83 * since the actual result of the sizeof expression _is_ evaluable
84 * at compile time. So here's my final answer, with the unfortunate
85 * remaining problem of evaluating its arguments multiple times:
87 #define TYPECHECK(x,y) ( sizeof((x)) == sizeof((x)) ? (y) : (y) )
88 #define DEPLANARISE(x) TYPECHECK((x) == (to_dbcs_planar_t)NULL, (to_dbcs_t)(x))
89 #define REPLANARISE(x) TYPECHECK((x) == (to_dbcs_t)NULL, (to_dbcs_planar_t)(x))
92 * Values used in the `enable' field. Each of these identifies a
93 * class of character sets; we then have a bitmask indicating which
94 * classes are allowable in a given mode.
96 * These values are currently only checked on output: for input,
97 * any ISO 2022 we can comprehend at all is considered acceptable.
99 #define CCS 1 /* CTEXT standard */
100 #define COS 2 /* other standard */
101 #define CPU 3 /* private use */
102 #define CDC 4 /* DOCS for CTEXT */
103 #define CDU 5 /* DOCS for UTF-8 */
104 #define CNU 31 /* never used */
106 struct iso2022_mode
{
108 char ltype
, li
, lf
, rtype
, ri
, rf
;
111 const struct iso2022_subcharset
{
112 char type
, i
, f
, enable
;
114 const sbcs_data
*sbcs_base
;
115 long int (*from_dbcs
)(int, int);
118 * If to_dbcs_plane < 0, then to_dbcs is used as expected.
119 * However, if to_dbcs_plane >= 0, then to_dbcs is expected to
120 * be cast to a to_dbcs_planar_t before use, and the returned
121 * plane value (the first int *) must equal to_dbcs_plane.
123 * I'd have preferred to do this by means of a union, but you
124 * can't initialise a selected field of a union at compile
125 * time. Function pointer casts are guaranteed to work sensibly
126 * in ISO C (that is, it's undefined what happens if you call a
127 * function via the wrong type of pointer, but if you cast it
128 * back to the right type before calling it then it must work),
129 * so this is safe if ugly.
132 int to_dbcs_plane
; /* use to_dbcs_planar iff >= 0 */
133 } iso2022_subcharsets
[] = {
135 * We list these subcharsets in preference order for output.
136 * Since the best-defined use of ISO 2022 output is compound
137 * text, we'll use a preference order which matches that. So we
138 * begin with the charsets defined in the compound text spec.
140 { S4
, 0, 'B', CCS
, 0x00, &sbcsdata_CS_ASCII
},
141 { S6
, 0, 'A', CCS
, 0x80, &sbcsdata_CS_ISO8859_1
},
142 { S6
, 0, 'B', CCS
, 0x80, &sbcsdata_CS_ISO8859_2
},
143 { S6
, 0, 'C', CCS
, 0x80, &sbcsdata_CS_ISO8859_3
},
144 { S6
, 0, 'D', CCS
, 0x80, &sbcsdata_CS_ISO8859_4
},
145 { S6
, 0, 'F', CCS
, 0x80, &sbcsdata_CS_ISO8859_7
},
146 { S6
, 0, 'G', CCS
, 0x80, &sbcsdata_CS_ISO8859_6
},
147 { S6
, 0, 'H', CCS
, 0x80, &sbcsdata_CS_ISO8859_8
},
148 { S6
, 0, 'L', CCS
, 0x80, &sbcsdata_CS_ISO8859_5
},
149 { S6
, 0, 'M', CCS
, 0x80, &sbcsdata_CS_ISO8859_9
},
150 { S4
, 0, 'I', CCS
, 0x80, &sbcsdata_CS_JISX0201
},
151 { S4
, 0, 'J', CCS
, 0x00, &sbcsdata_CS_JISX0201
},
152 { M4
, 0, 'A', CCS
, -0x21, 0, &gb2312_to_unicode
, &unicode_to_gb2312
, -1 },
153 { M4
, 0, 'B', CCS
, -0x21, 0, &jisx0208_to_unicode
, &unicode_to_jisx0208
, -1 },
154 { M4
, 0, 'C', CCS
, -0x21, 0, &ksx1001_to_unicode
, &unicode_to_ksx1001
, -1 },
155 { M4
, 0, 'D', CCS
, -0x21, 0, &jisx0212_to_unicode
, &unicode_to_jisx0212
, -1 },
158 * Next, other reasonably standard things: the rest of the ISO
159 * 8859 sets, UK-ASCII, and CNS 11643.
161 { S6
, 0, 'T', COS
, 0x80, &sbcsdata_CS_ISO8859_11
},
162 { S6
, 0, 'V', COS
, 0x80, &sbcsdata_CS_ISO8859_10
},
163 { S6
, 0, 'Y', COS
, 0x80, &sbcsdata_CS_ISO8859_13
},
164 { S6
, 0, '_', COS
, 0x80, &sbcsdata_CS_ISO8859_14
},
165 { S6
, 0, 'b', COS
, 0x80, &sbcsdata_CS_ISO8859_15
},
166 { S6
, 0, 'f', COS
, 0x80, &sbcsdata_CS_ISO8859_16
},
167 { S4
, 0, 'A', COS
, 0x00, &sbcsdata_CS_BS4730
},
168 { M4
, 0, 'G', COS
, -0x21, 0, &cns11643_1_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 0 },
169 { M4
, 0, 'H', COS
, -0x21, 0, &cns11643_2_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 1 },
170 { M4
, 0, 'I', COS
, -0x21, 0, &cns11643_3_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 2 },
171 { M4
, 0, 'J', COS
, -0x21, 0, &cns11643_4_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 3 },
172 { M4
, 0, 'K', COS
, -0x21, 0, &cns11643_5_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 4 },
173 { M4
, 0, 'L', COS
, -0x21, 0, &cns11643_6_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 5 },
174 { M4
, 0, 'M', COS
, -0x21, 0, &cns11643_7_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 6 },
177 * Private-use designations: DEC private sets and Emacs's Big5
180 { S4
, 0, '0', CPU
, 0x00, &sbcsdata_CS_DEC_GRAPHICS
},
181 { S4
, 0, '<', CPU
, 0x80, &sbcsdata_CS_DEC_MCS
},
182 { M4
, 0, '0', CPU
, -0x21, 0, &emacs_big5_1_to_unicode
, DEPLANARISE(&unicode_to_emacs_big5
), 1 },
183 { M4
, 0, '1', CPU
, -0x21, 0, &emacs_big5_2_to_unicode
, DEPLANARISE(&unicode_to_emacs_big5
), 2 },
186 * Ben left this conditioned out without explanation,
187 * presumably on the grounds that we don't have a translation
191 { M4
, 0, '@', CNU
}, /* JIS C 6226-1978 */
195 * Finally, fallback entries for null character sets.
198 { S6
, 0, '~', CNU
}, /* empty 96-set */
199 { M4
, 0, '~', CNU
, 0, 0, &null_dbcs_to_unicode
, &unicode_to_null_dbcs
, -1 }, /* empty 94^n-set */
200 { M6
, 0, '~', CNU
, 0, 0, &null_dbcs_to_unicode
, &unicode_to_null_dbcs
, -1 }, /* empty 96^n-set */
203 static long int null_dbcs_to_unicode(int r
, int c
)
209 static int unicode_to_null_dbcs(long int unicode
, int *r
, int *c
)
214 return 0; /* failed to convert anything */
218 * Emacs encodes Big5 in COMPOUND_TEXT as two 94x94 character sets.
219 * We treat Big5 as a 94x191 character set with a bunch of undefined
220 * columns in the middle, so we have to mess around a bit to make
224 static long int emacs_big5_1_to_unicode(int r
, int c
)
230 if (c
>= 64) c
+= 34; /* Skip over the gap */
231 return big5_to_unicode(r
, c
);
234 static long int emacs_big5_2_to_unicode(int r
, int c
)
240 if (c
>= 64) c
+= 34; /* Skip over the gap */
241 return big5_to_unicode(r
, c
);
244 static int unicode_to_emacs_big5(long int unicode
, int *p
, int *r
, int *c
)
247 if (!unicode_to_big5(unicode
, &rr
, &cc
))
265 /* Wrappers for cns11643_to_unicode() */
266 static long int cns11643_1_to_unicode(int r
, int c
)
268 return cns11643_to_unicode(0, r
, c
);
270 static long int cns11643_2_to_unicode(int r
, int c
)
272 return cns11643_to_unicode(1, r
, c
);
274 static long int cns11643_3_to_unicode(int r
, int c
)
276 return cns11643_to_unicode(2, r
, c
);
278 static long int cns11643_4_to_unicode(int r
, int c
)
280 return cns11643_to_unicode(3, r
, c
);
282 static long int cns11643_5_to_unicode(int r
, int c
)
284 return cns11643_to_unicode(4, r
, c
);
286 static long int cns11643_6_to_unicode(int r
, int c
)
288 return cns11643_to_unicode(5, r
, c
);
290 static long int cns11643_7_to_unicode(int r
, int c
)
292 return cns11643_to_unicode(6, r
, c
);
295 /* States, or "what we're currently accumulating". */
297 IDLE
, /* None of the below */
298 SS2CHAR
, /* Accumulating a character after SS2 */
299 SS3CHAR
, /* Accumulating a character after SS3 */
300 ESCSEQ
, /* Accumulating an escape sequence */
301 ESCDROP
, /* Discarding an escape sequence */
302 ESCPASS
, /* Passing through an escape sequence */
303 DOCSUTF8
, /* DOCSed into UTF-8 */
304 DOCSCTEXT
/* DOCSed into a COMPOUND_TEXT extended segment */
309 static void dump_state(charset_state
*s
)
311 unsigned s0
= s
->s0
, s1
= s
->s1
;
312 char const * const modes
[] = { "IDLE", "SS2CHAR", "SS3CHAR",
313 "ESCSEQ", "ESCDROP", "ESCPASS",
316 fprintf(stderr
, "s0: %s", modes
[s0
>> 29]);
317 fprintf(stderr
, " %02x %02x %02x ", (s0
>> 16) & 0xff, (s0
>> 8) & 0xff,
319 fprintf(stderr
, "s1: LS%d LS%dR", (s1
>> 30) & 3, (s1
>> 28) & 3);
320 fprintf(stderr
, " %d %d %d %d\n", s1
& 0x7f, (s1
>> 7) & 0x7f,
321 (s1
>> 14) & 0x7f, (s1
>> 21) & 0x7f);
325 static void designate(charset_state
*state
, int container
,
326 int type
, int ibyte
, int fbyte
)
330 assert(container
>= 0 && container
<= 3);
331 assert(type
== S4
|| type
== S6
|| type
== M4
|| type
== M6
);
333 for (i
= 0; i
< lenof(iso2022_subcharsets
); i
++) {
334 if (iso2022_subcharsets
[i
].type
== type
&&
335 iso2022_subcharsets
[i
].i
== ibyte
&&
336 iso2022_subcharsets
[i
].f
== fbyte
) {
337 state
->s1
&= ~(0x7fL
<< (container
* 7));
338 state
->s1
|= (i
<< (container
* 7));
343 * If we don't find the charset, invoke the empty one, so we
344 * output ERROR rather than garbage.
346 designate(state
, container
, type
, 0, '~');
349 static void do_utf8(long int input_chr
,
350 charset_state
*state
,
351 void (*emit
)(void *ctx
, long int output
),
354 charset_state ustate
;
357 ustate
.s0
= state
->s0
& 0x03ffffffL
;
358 read_utf8(NULL
, input_chr
, &ustate
, emit
, emitctx
);
359 state
->s0
= (state
->s0
& ~0x03ffffffL
) | (ustate
.s0
& 0x03ffffffL
);
362 static void docs_utf8(long int input_chr
,
363 charset_state
*state
,
364 void (*emit
)(void *ctx
, long int output
),
370 * Bits [25:0] of s0 are reserved for read_utf8().
371 * Bits [27:26] are a tiny state machine to recognise ESC % @.
373 retstate
= (state
->s0
& 0x0c000000L
) >> 26;
374 if (retstate
== 1 && input_chr
== '%')
376 else if (retstate
== 2 && input_chr
== '@') {
377 /* If we've got a partial UTF-8 sequence, complain. */
378 if (state
->s0
& 0x03ffffffL
)
379 emit(emitctx
, ERROR
);
383 if (retstate
>= 1) do_utf8(ESC
, state
, emit
, emitctx
);
384 if (retstate
>= 2) do_utf8('%', state
, emit
, emitctx
);
386 if (input_chr
== ESC
)
389 do_utf8(input_chr
, state
, emit
, emitctx
);
392 state
->s0
= (state
->s0
& ~0x0c000000L
) | (retstate
<< 26);
395 struct ctext_encoding
{
397 char octets_per_char
, enable
;
398 charset_spec
const *subcs
;
402 * In theory, this list is in <ftp://ftp.x.org/pub/DOCS/registry>,
403 * but XLib appears to have its own ideas, and encodes these three
407 extern charset_spec
const charset_CS_ISO8859_14
;
408 extern charset_spec
const charset_CS_ISO8859_15
;
409 extern charset_spec
const charset_CS_BIG5
;
411 static struct ctext_encoding
const ctext_encodings
[] = {
412 { "big5-0\2", 0 /* variable */, CDC
, &charset_CS_BIG5
},
413 { "iso8859-14\2", 1, CDC
, &charset_CS_ISO8859_14
},
414 { "iso8859-15\2", 1, CDC
, &charset_CS_ISO8859_15
}
417 static void docs_ctext(long int input_chr
,
418 charset_state
*state
,
419 void (*emit
)(void *ctx
, long int output
),
423 * s0[27:26] = first entry in ctext_encodings that matches
424 * s0[25:22] = number of characters successfully matched, 0xf if all
425 * s0[21:8] count the number of octets left in the segment
426 * s0[7:0] are for sub-charset use
428 int n
= (state
->s0
>> 22) & 0xf, i
= (state
->s0
>> 26) & 3, oi
= i
, j
;
429 int length
= (state
->s0
>> 8) & 0x3fff;
432 * Note that we do not bother checking the octets-per-character
433 * byte against the selected charset when reading. It's
434 * extremely unlikely that this code will ever have to deal
435 * with two charset identifiers with the same name and
436 * different octets-per-character values! If it ever happens,
437 * we'll have to edit this file anyway so we can modify the
442 /* Haven't read length yet */
443 if ((state
->s0
& 0xff) == 0)
444 /* ... or even the first byte */
445 state
->s0
|= input_chr
;
447 length
= (state
->s0
& 0x7f) * 0x80 + (input_chr
& 0x7f);
451 state
->s0
= (state
->s0
& 0xf0000000) | (length
<< 8);
458 /* Skipping unknown encoding. Look out for STX. */
460 state
->s0
= (state
->s0
& 0xf0000000) | (i
<< 26) | (0xf << 22);
461 } else if (n
!= 0xf) {
462 while ((unsigned)j
< lenof(ctext_encodings
) &&
463 !memcmp(ctext_encodings
[j
].name
,
464 ctext_encodings
[oi
].name
, n
)) {
465 if (ctext_encodings
[j
].name
[n
] < input_chr
)
470 if ((unsigned)i
>= lenof(ctext_encodings
) ||
471 memcmp(ctext_encodings
[i
].name
,
472 ctext_encodings
[oi
].name
, n
) ||
473 ctext_encodings
[i
].name
[n
] != input_chr
) {
474 /* Doom! We haven't heard of this encoding */
475 i
= lenof(ctext_encodings
);
479 * Otherwise, we have found an additional character in our
480 * encoding name. See if we have reached the _end_ of our
484 if (!ctext_encodings
[i
].name
[n
])
488 * Failing _that_, we simply update our encoding-name-
491 assert(i
< 4 && n
< 16);
492 state
->s0
= (state
->s0
& 0xf0000000) | (i
<< 26) | (n
<< 22);
494 if ((unsigned)i
>= lenof(ctext_encodings
))
495 emit(emitctx
, ERROR
);
497 charset_state substate
;
498 charset_spec
const *subcs
= ctext_encodings
[i
].subcs
;
500 substate
.s0
= state
->s0
& 0xff;
501 subcs
->read(subcs
, input_chr
, &substate
, emit
, emitctx
);
502 state
->s0
= (state
->s0
& ~0xff) | (substate
.s0
& 0xff);
508 state
->s0
= (state
->s0
&~0x003fff00) | (length
<< 8);
511 static void read_iso2022(charset_spec
const *charset
, long int input_chr
,
512 charset_state
*state
,
513 void (*emit
)(void *ctx
, long int output
),
516 struct iso2022_mode
const *mode
= (struct iso2022_mode
*)charset
->data
;
518 /* dump_state(state); */
520 * We have to make fairly efficient use of the 64 bits of state
521 * available to us. Long-term state goes in s1, and consists of
522 * the identities of the character sets designated as G0/G1/G2/G3
523 * and the locking-shift states for GL and GR. Short-term state
524 * goes in s0: The bottom half of s0 accumulates characters for an
525 * escape sequence or a multi-byte character, while the top three
526 * bits indicate what they're being accumulated for. After DOCS,
527 * the bottom 29 bits of state are available for the DOCS function
528 * to use -- the UTF-8 one uses the bottom 26 for UTF-8 decoding
529 * and the top two to recognised ESC % @.
531 * s0[31:29] = state enum
532 * s0[24:0] = accumulated bytes
533 * s1[31:30] = GL locking-shift state
534 * s1[29:28] = GR locking-shift state
535 * s1[27:21] = G3 charset
536 * s1[20:14] = G2 charset
537 * s1[13:7] = G1 charset
538 * s1[6:0] = G0 charset
543 #define LOCKING_SHIFT(n,side) \
544 (state->s1 = (state->s1 & ~(3UL<<(side))) | ((n ## UL)<<(side)))
545 #define MODE ((state->s0 & 0xe0000000UL) >> 29)
546 #define ENTER_MODE(m) (state->s0 = (state->s0 & ~0xe0000000UL) | ((unsigned long)(m)<<29))
547 #define SINGLE_SHIFT(n) ENTER_MODE(SS2CHAR - 2 + (n))
548 #define ASSERT_IDLE do { \
549 if (state->s0 != 0) emit(emitctx, ERROR); \
553 if (state
->s1
== 0) {
555 * Since there's no LS0R, this means we must just have started.
556 * Set up a sane initial state (LS0, LS1R, ASCII in G0/G1/G2/G3).
558 LOCKING_SHIFT(0, LEFT
);
559 LOCKING_SHIFT(1, RIGHT
);
560 designate(state
, 0, mode
->ltype
, mode
->li
, mode
->lf
);
561 designate(state
, 1, mode
->rtype
, mode
->ri
, mode
->rf
);
562 designate(state
, 2, S4
, 0, 'B');
563 designate(state
, 3, S4
, 0, 'B');
566 if (MODE
== DOCSUTF8
) {
567 docs_utf8(input_chr
, state
, emit
, emitctx
);
570 if (MODE
== DOCSCTEXT
) {
571 docs_ctext(input_chr
, state
, emit
, emitctx
);
575 if ((input_chr
& 0x60) == 0x00) {
576 /* C0 or C1 control */
583 LOCKING_SHIFT(0, LEFT
);
586 LOCKING_SHIFT(1, LEFT
);
595 emit(emitctx
, input_chr
);
598 } else if ((input_chr
& 0x80) || MODE
< ESCSEQ
) {
600 struct iso2022_subcharset
const *subcs
;
605 * Force idle state if we're in mid escape sequence, or in a
606 * multi-byte character with a different top bit.
608 if (MODE
>= ESCSEQ
||
609 ((state
->s0
& 0x00ff0000L
) != 0 &&
610 (((state
->s0
>> 16) ^ input_chr
) & 0x80)))
612 if (MODE
== SS2CHAR
|| MODE
== SS3CHAR
) /* Single-shift */
613 container
= MODE
- SS2CHAR
+ 2;
614 else if (input_chr
>= 0x80) /* GR */
615 container
= (state
->s1
>> 28) & 3;
617 container
= state
->s1
>> 30;
620 input_7bit
= input_chr
& ~0x80;
621 subcs
= &iso2022_subcharsets
[(state
->s1
>> (container
* 7)) & 0x7f];
622 if ((subcs
->type
== S4
|| subcs
->type
== M4
) &&
623 (input_7bit
== 0x20 || input_7bit
== 0x7f)) {
624 /* characters not in 94-char set */
625 if (is_gl
) emit(emitctx
, input_7bit
);
626 else emit(emitctx
, ERROR
);
627 } else if (subcs
->type
== M4
|| subcs
->type
== M6
) {
628 if ((state
->s0
& 0x00ff0000L
) == 0) {
629 state
->s0
|= input_chr
<< 16;
633 subcs
->from_dbcs(((state
->s0
>> 16) & 0x7f) +
635 input_7bit
+ subcs
->offset
));
638 if ((state
->s0
& 0x00ff0000L
) != 0)
639 emit(emitctx
, ERROR
);
640 emit(emitctx
, subcs
->sbcs_base ?
641 sbcs_to_unicode(subcs
->sbcs_base
, input_7bit
+ subcs
->offset
):
647 if (MODE
== ESCPASS
) {
648 emit(emitctx
, input_chr
);
649 if ((input_chr
& 0xf0) != 0x20)
655 * Intermediate bytes shall be any of the 16 positions of
656 * column 02 of the code table; they are denoted by the symbol
659 if ((input_chr
& 0xf0) == 0x20) {
660 if (((state
->s0
>> 16) & 0xff) == 0)
661 state
->s0
|= input_chr
<< 16;
662 else if (((state
->s0
>> 8) & 0xff) == 0)
663 state
->s0
|= input_chr
<< 8;
665 /* Long escape sequence. Switch to ESCPASS or ESCDROP. */
666 i1
= (state
->s0
>> 16) & 0xff;
667 i2
= (state
->s0
>> 8) & 0xff;
669 case '(': case ')': case '*': case '+':
670 case '-': case '.': case '/':
678 emit(emitctx
, input_chr
);
688 * Final bytes shall be any of the 79 positions of columns 03
689 * to 07 of the code table excluding position 07/15; they are
690 * denoted by the symbol F.
692 i1
= (state
->s0
>> 16) & 0xff;
693 i2
= (state
->s0
>> 8) & 0xff;
695 input_chr
= 0; /* Make sure it won't match. */
698 case 0: /* No intermediate bytes */
707 LOCKING_SHIFT(2, LEFT
);
710 LOCKING_SHIFT(3, LEFT
);
713 LOCKING_SHIFT(3, RIGHT
);
716 LOCKING_SHIFT(2, RIGHT
);
719 LOCKING_SHIFT(1, RIGHT
);
722 /* Unsupported escape sequence. Spit it back out. */
724 emit(emitctx
, input_chr
);
729 * Various coding structure facilities specify that designating
730 * a code element also invokes it. As far as I can see, invoking
731 * it now will have the same practical effect, since those
732 * facilities also ban the use of locking shifts.
735 case 'A': /* G0 element used and invoked into GL */
736 LOCKING_SHIFT(0, LEFT
);
738 case 'C': /* G0 in GL, G1 in GR */
739 case 'D': /* Ditto, at least for 8-bit codes */
740 case 'L': /* ISO 4873 (ECMA-43) level 1 */
741 case 'M': /* ISO 4873 (ECMA-43) level 2 */
742 LOCKING_SHIFT(0, LEFT
);
743 LOCKING_SHIFT(1, RIGHT
);
749 * IRR (Identify Revised Registration) is ignored here,
750 * since any revised registration must be
751 * upward-compatible with the old one, so either we'll
752 * support the new one or we'll emit ERROR when we run
753 * into a new character. In either case, there's nothing
757 case '(': /* GZD4 */ case ')': /* G1D4 */
758 case '*': /* G2D4 */ case '+': /* G3D4 */
759 designate(state
, i1
- '(', S4
, i2
, input_chr
);
761 case '-': /* G1D6 */ case '.': /* G2D6 */ case '/': /* G3D6 */
762 designate(state
, i1
- ',', S6
, i2
, input_chr
);
764 case '$': /* G?DM? */
766 case 0: /* Obsolete version of GZDM4 */
768 case '(': /* GZDM4 */ case ')': /* G1DM4 */
769 case '*': /* G2DM4 */ case '+': /* G3DM4 */
770 designate(state
, i2
- '(', M4
, 0, input_chr
);
772 case '-': /* G1DM6 */
773 case '.': /* G2DM6 */ case '/': /* G3DM6 */
774 designate(state
, i2
- ',', M6
, 0, input_chr
);
777 emit(emitctx
, ERROR
);
781 /* XXX What's a reasonable way to handle an unrecognised DOCS? */
786 ENTER_MODE(DOCSUTF8
);
793 ENTER_MODE(DOCSCTEXT
);
800 /* Unsupported nF escape sequence. Re-emit it. */
803 if (i2
) emit(emitctx
, i2
);
804 emit(emitctx
, input_chr
);
810 static void oselect(charset_state
*state
, int i
, int right
,
811 void (*emit
)(void *ctx
, long int output
),
814 int shift
= (right ?
31-7 : 31-7-7);
815 struct iso2022_subcharset
const *subcs
= &iso2022_subcharsets
[i
];
817 if (((state
->s1
>> shift
) & 0x7F) != (unsigned)i
) {
818 state
->s1
&= ~(0x7FL
<< shift
);
819 state
->s1
|= (i
<< shift
);
823 if (subcs
->type
== M4
|| subcs
->type
== M6
)
825 if (subcs
->type
== S6
|| subcs
->type
== M6
) {
834 emit(emitctx
, subcs
->i
);
835 emit(emitctx
, subcs
->f
);
840 static void docs_char(charset_state
*state
,
841 void (*emit
)(void *ctx
, long int output
),
842 void *emitctx
, int cset
, char *data
, int datalen
)
844 int curr_cset
, currlen
, i
;
847 * cset is the index into ctext_encodings[]. It can also be -1
848 * to mean DOCS UTF-8, or -2 to mean no DOCS (ordinary 2022).
849 * In the latter case, `chr' is ignored.
853 * First, terminate a DOCS segment if necessary. We always have
854 * to terminate a DOCS segment if one is active and we're about
855 * to switch to a different one; we might also have to
856 * terminate a length-encoded DOCS segment if we've run out of
857 * storage space to accumulate characters in it.
859 curr_cset
= ((state
->s1
>> 14) & 7) - 2;
860 currlen
= ((state
->s1
>> 11) & 7);
861 if ((curr_cset
!= -2 && curr_cset
!= cset
) ||
862 (curr_cset
>= 0 && currlen
+ datalen
> 5)) {
863 if (curr_cset
== -1) {
865 * Terminating DOCS UTF-8 is easy.
874 * To terminate a length-encoded DOCS segment we must
875 * actually output the whole thing.
880 emit(emitctx
, '0' + ctext_encodings
[curr_cset
].octets_per_char
);
881 len
= currlen
+ datalen
+
882 strlen(ctext_encodings
[curr_cset
].name
);
883 assert(len
< (1 << 14));
884 emit(emitctx
, 0x80 | ((len
>> 7) & 0x7F));
885 emit(emitctx
, 0x80 | ((len
) & 0x7F));
886 /* The name stored in ctext_encodings[] includes the trailing \2 */
887 for (i
= 0; ctext_encodings
[curr_cset
].name
[i
]; i
++)
888 emit(emitctx
, ctext_encodings
[curr_cset
].name
[i
]);
889 for (i
= 0; i
< currlen
; i
++)
891 (i
== 0 ? state
->s1
: state
->s0
>> (8*(4-i
))) & 0xFF);
892 for (i
= 0; i
< datalen
; i
++)
893 emit(emitctx
, data
[i
]);
896 * We've now dealt with the input data, so clear it so
897 * we don't try to do so again below.
905 * Now, start a DOCS segment if necessary.
907 if (curr_cset
!= cset
) {
918 * Starting a length-encoded DOCS segment is simply a
919 * matter of setting our stored length counter to zero.
922 state
->s1
&= ~(7 << 11);
927 state
->s1
&= ~(7 << 14);
928 assert((cset
+2) >= 0 && (cset
+2) < 8);
929 state
->s1
|= ((cset
+2) << 14);
932 * Now we're in the right DOCS state. Actually deal with the
933 * input data, if we haven't already done so above.
939 * In DOCS UTF-8, we output data as soon as we get it.
941 for (i
= 0; i
< datalen
; i
++)
942 emit(emitctx
, data
[i
]);
945 * In length-encoded DOCS, we just store our data and
946 * bide our time. It'll all be output when we fill up
947 * or switch to another character set.
949 assert(currlen
+ datalen
<= 5); /* overflow handled already */
950 for (i
= 0; i
< datalen
; i
++) {
951 if (currlen
+ i
== 0)
952 state
->s1
|= data
[i
] & 0xFF;
954 state
->s0
|= (data
[i
] & 0xFF) << (8*(4-(currlen
+i
)));
957 assert(currlen
>= 0 && currlen
< 8);
958 state
->s1
&= ~(7 << 11);
959 state
->s1
|= (currlen
<< 11);
964 static void write_to_pointer(void *ctx
, long int output
)
966 char **ptr
= (char **)ctx
;
971 * Writing full ISO-2022 is not useful in very many circumstances.
972 * One of the few situations in which it _is_ useful is generating
973 * X11 COMPOUND_TEXT; therefore, this writing function will obey
974 * the compound text restrictions and hence output the subset of
975 * ISO-2022 that's usable in that context.
977 * The subset in question is roughly that we use GL/GR for G0/G1
978 * always, and that the _only_ escape sequences we output (other
979 * than the occasional DOCS) are those which designate different
980 * subcharsets into G0 and G1. There are additional constraints
981 * about which things go in which container; see below.
983 * FIXME: this wants some decent tests to be written, and also the
984 * exact output policy for compound text wants thinking about more
987 static int write_iso2022(charset_spec
const *charset
, long int input_chr
,
988 charset_state
*state
,
989 void (*emit
)(void *ctx
, long int output
),
993 struct iso2022_subcharset
const *subcs
;
994 struct iso2022_mode
const *mode
= (struct iso2022_mode
*)charset
->data
;
995 to_dbcs_planar_t last_planar_dbcs
= NULL
;
996 int last_p
, last_r
, last_c
;
1000 * For output, I allocate the state variables as follows:
1002 * s1[31] == 1 if output state has been initialised
1003 * s1[30:24] == G1 charset (always in GR)
1004 * s1[23:17] == G0 charset (always in GL)
1005 * s1[16:14] == DOCS index plus 2 (because -1 and -2 are special)
1006 * s1[13:11] == number of DOCS accumulated characters (up to five)
1007 * s1[7:0] + s0[31:0] == DOCS collected characters
1011 state
->s0
= 0x00000000UL
;
1012 state
->s1
= 0x80000000UL
;
1014 * Start with US-ASCII in GL and also in GR.
1016 for (i
= 0; (unsigned)i
< lenof(iso2022_subcharsets
); i
++) {
1017 subcs
= &iso2022_subcharsets
[i
];
1018 if (subcs
->type
== mode
->ltype
&&
1019 subcs
->i
== mode
->li
&&
1020 subcs
->f
== mode
->lf
)
1021 oselect(state
, i
, FALSE
, NULL
, NULL
);
1022 if (subcs
->type
== mode
->rtype
&&
1023 subcs
->i
== mode
->ri
&&
1024 subcs
->f
== mode
->rf
)
1025 oselect(state
, i
, TRUE
, NULL
, NULL
);
1029 if (input_chr
== -1) {
1031 * Special case: reset encoding state.
1033 docs_char(state
, emit
, emitctx
, -2, NULL
, 0); /* leave DOCS */
1035 for (i
= 0; (unsigned)i
< lenof(iso2022_subcharsets
); i
++) {
1036 subcs
= &iso2022_subcharsets
[i
];
1037 if (subcs
->type
== mode
->ltype
&&
1038 subcs
->i
== mode
->li
&&
1039 subcs
->f
== mode
->lf
)
1040 oselect(state
, i
, FALSE
, emit
, emitctx
);
1041 if (subcs
->type
== mode
->rtype
&&
1042 subcs
->i
== mode
->ri
&&
1043 subcs
->f
== mode
->rf
)
1044 oselect(state
, i
, TRUE
, emit
, emitctx
);
1050 * Special-case characters: Space, Delete, and anything in C0
1051 * or C1 are output unchanged.
1053 if (input_chr
<= 0x20 || (input_chr
>= 0x7F && input_chr
< 0xA0)) {
1054 emit(emitctx
, input_chr
);
1059 * Analyse the input character and work out which subcharset it
1062 for (i
= 0; (unsigned)i
< lenof(iso2022_subcharsets
); i
++) {
1063 subcs
= &iso2022_subcharsets
[i
];
1064 if (!(mode
->enable_mask
& (1 << subcs
->enable
)))
1065 continue; /* this charset is disabled */
1066 if (subcs
->sbcs_base
) {
1067 c1
= sbcs_from_unicode(subcs
->sbcs_base
, input_chr
);
1068 c1
-= subcs
->offset
;
1069 if (c1
>= 0x20 && c1
<= 0x7f) {
1073 } else if (subcs
->to_dbcs
) {
1074 if (subcs
->to_dbcs_plane
>= 0) {
1076 * Since multiplanar DBCSes almost by definition
1077 * involve several entries in iso2022_subcharsets
1078 * with the same to_dbcs function and different
1079 * plane values, we remember the last such function
1080 * we called and what its result was, so that we
1081 * don't (for example) have to call
1082 * unicode_to_cns11643 seven times.
1084 if (last_planar_dbcs
!= REPLANARISE(subcs
->to_dbcs
)) {
1085 last_planar_dbcs
= REPLANARISE(subcs
->to_dbcs
);
1086 if (!last_planar_dbcs(input_chr
,
1087 &last_p
, &last_r
, &last_c
))
1091 last_p
= subcs
->to_dbcs_plane
;
1092 if (!subcs
->to_dbcs(input_chr
, &last_r
, &last_c
))
1093 last_p
= 0; /* cannot match since to_dbcs_plane<0 */
1096 if (last_p
== subcs
->to_dbcs_plane
) {
1097 c1
= last_r
- subcs
->offset
;
1098 c2
= last_c
- subcs
->offset
;
1099 assert(c1
>= 0x20 && c1
<= 0x7f);
1100 assert(c2
>= 0x20 && c2
<= 0x7f);
1106 if ((unsigned)i
< lenof(iso2022_subcharsets
)) {
1110 * Our character is represented by c1 (and possibly also
1111 * c2) in subcharset `subcs'. So now we must decide whether
1112 * to designate that character set into G0/GL or G1/GR.
1114 * Any S6 or M6 subcharset has to go in GR because it won't
1115 * fit in GL. In addition, the compound text rules state
1116 * that any single-byte subcharset defined as the
1117 * right-hand half of some SBCS must go in GR.
1119 * M4 subcharsets can go in either half according to the
1120 * rules. I choose to put them in GR always because it's a
1121 * simple policy with reasonable behaviour (facilitates
1122 * switching between them and ASCII).
1124 right
= (subcs
->type
== S6
|| subcs
->type
== M6
|| subcs
->type
== M4
||
1125 (subcs
->sbcs_base
&& subcs
->offset
== 0x80));
1128 * If we're in a DOCS mode, leave it.
1130 docs_char(state
, emit
, emitctx
, -2, NULL
, 0);
1133 * If this subcharset is not already selected in that
1134 * container, select it.
1136 oselect(state
, i
, right
, emit
, emitctx
);
1139 * Now emit the actual characters.
1142 assert(c1
>= 0x20 && c1
<= 0x7f);
1143 emit(emitctx
, c1
| 0x80);
1145 assert(c2
>= 0x20 && c2
<= 0x7f);
1146 emit(emitctx
, c2
| 0x80);
1149 assert(c1
> 0x20 && c1
< 0x7f);
1152 assert(c2
> 0x20 && c2
< 0x7f);
1161 * Fall back to DOCS.
1168 cs
= -2; /* means failure */
1170 for (i
= 0; (unsigned)i
<= lenof(ctext_encodings
); i
++) {
1171 charset_state substate
;
1172 charset_spec
const *subcs
= ctext_encodings
[i
].subcs
;
1175 * We assume that all character sets dealt with by DOCS
1176 * are stateless for output purposes.
1178 substate
.s1
= substate
.s0
= 0;
1181 if ((unsigned)i
< lenof(ctext_encodings
)) {
1182 if ((mode
->enable_mask
& (1 << ctext_encodings
[i
].enable
)) &&
1183 subcs
->write(subcs
, input_chr
, &substate
,
1184 write_to_pointer
, &p
)) {
1189 if ((mode
->enable_mask
& (1 << CDU
)) &&
1190 write_utf8(NULL
, input_chr
, NULL
, write_to_pointer
, &p
)) {
1198 docs_char(state
, emit
, emitctx
, cs
, data
, p
- data
);
1207 * Full ISO 2022 output with all options on. Not entirely sure what
1208 * if anything this is useful for, but here it is anyway. All
1209 * output character sets and DOCS variants are permitted; all
1210 * containers start out with ASCII in them.
1212 static const struct iso2022_mode iso2022_all
= {
1213 (1<<CCS
) | (1<<COS
) | (1<<CPU
) | (1<<CDC
) | (1<<CDU
),
1214 S4
, 0, 'B', S4
, 0, 'B',
1217 const charset_spec charset_CS_ISO2022
= {
1218 CS_ISO2022
, read_iso2022
, write_iso2022
, &iso2022_all
1222 * X11 compound text. A subset of output charsets is permitted, and
1223 * G1/GR starts off in ISO8859-1.
1225 static const struct iso2022_mode iso2022_ctext
= {
1226 (1<<CCS
) | (1<<CDC
),
1227 S4
, 0, 'B', S6
, 0, 'A',
1230 const charset_spec charset_CS_CTEXT
= {
1231 CS_CTEXT
, read_iso2022
, write_iso2022
, &iso2022_ctext
1242 void iso2022_emit(void *ctx
, long output
)
1244 wchar_t **p
= (wchar_t **)ctx
;
1248 void iso2022_read_test(int line
, char *input
, int inlen
, ...)
1251 wchar_t *p
, str
[512];
1253 charset_state state
;
1256 state
.s0
= state
.s1
= 0;
1259 for (i
= 0; i
< inlen
; i
++)
1260 read_iso2022(NULL
, input
[i
] & 0xFF, &state
, iso2022_emit
, &p
);
1262 va_start(ap
, inlen
);
1264 for (i
= 0; i
< p
- str
; i
++) {
1265 l
= va_arg(ap
, long int);
1267 printf("%d: correct string shorter than output\n", line
);
1272 printf("%d: char %d came out as %08x, should be %08lx\n",
1273 line
, i
, str
[i
], l
);
1278 l
= va_arg(ap
, long int);
1280 printf("%d: correct string longer than output\n", line
);
1287 /* Macro to concoct the first three parameters of iso2022_read_test. */
1288 #define TESTSTR(x) __LINE__, x, lenof(x)
1292 printf("read tests beginning\n");
1293 /* Simple test (Emacs sample text for Japanese, in ISO-2022-JP) */
1294 iso2022_read_test(TESTSTR("Japanese (\x1b$BF|K\\8l\x1b(B)\t"
1295 "\x1b$B$3$s$K$A$O\x1b(B, "
1296 "\x1b$B%3%s%K%A%O\x1b(B\n"),
1297 'J','a','p','a','n','e','s','e',' ','(',
1298 0x65E5, 0x672C, 0x8A9E, ')', '\t',
1299 0x3053, 0x3093, 0x306b, 0x3061, 0x306f, ',', ' ',
1300 0x30b3, 0x30f3, 0x30cb, 0x30c1, 0x30cf, '\n', 0, -1);
1301 /* Same thing in EUC-JP (with designations, and half-width katakana) */
1302 iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D"
1303 "Japanese (\xc6\xfc\xcb\xdc\xb8\xec)\t"
1304 "\xa4\xb3\xa4\xf3\xa4\xcb\xa4\xc1\xa4\xcf, "
1305 "\x8e\xba\x8e\xdd\x8e\xc6\x8e\xc1\x8e\xca\n"),
1306 'J','a','p','a','n','e','s','e',' ','(',
1307 0x65E5, 0x672C, 0x8A9E, ')', '\t',
1308 0x3053, 0x3093, 0x306b, 0x3061, 0x306f, ',', ' ',
1309 0xff7a, 0xff9d, 0xff86, 0xff81, 0xff8a, '\n', 0, -1);
1310 /* Multibyte single-shift */
1311 iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D\x8f\"/!"),
1312 0x02D8, '!', 0, -1);
1313 /* Non-existent SBCS */
1314 iso2022_read_test(TESTSTR("\x1b(!Zfnord\n"),
1315 ERROR
, ERROR
, ERROR
, ERROR
, ERROR
, '\n', 0, -1);
1316 /* Pass-through of ordinary escape sequences, including a long one */
1317 iso2022_read_test(TESTSTR("\x1b""b\x1b#5\x1b#!!!5"),
1318 0x1B, 'b', 0x1B, '#', '5',
1319 0x1B, '#', '!', '!', '!', '5', 0, -1);
1320 /* Non-existent DBCS (also 5-byte escape sequence) */
1321 iso2022_read_test(TESTSTR("\x1b$(!Bfnord!"),
1322 ERROR
, ERROR
, ERROR
, 0, -1);
1323 /* Incomplete DB characters */
1324 iso2022_read_test(TESTSTR("\x1b$B(,(\x1b(BHi\x1b$B(,(\n"),
1325 0x2501, ERROR
, 'H', 'i', 0x2501, ERROR
, '\n', 0, -1);
1326 iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D\xa4""B"),
1328 iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D\x0e\x1b|$\xa2\xaf"),
1329 ERROR
, 0x02D8, 0, -1);
1330 /* Incomplete escape sequence */
1331 iso2022_read_test(TESTSTR("\x1b\n"), ERROR
, '\n', 0, -1);
1332 iso2022_read_test(TESTSTR("\x1b-A\x1b~\x1b\xa1"), ERROR
, 0xa1, 0, -1);
1333 /* Incomplete single-shift */
1334 iso2022_read_test(TESTSTR("\x8e\n"), ERROR
, '\n', 0, -1);
1335 iso2022_read_test(TESTSTR("\x1b$*B\x8e(\n"), ERROR
, '\n', 0, -1);
1336 /* Corner cases (02/00 and 07/15) */
1337 iso2022_read_test(TESTSTR("\x1b(B\x20\x7f"), 0x20, 0x7f, 0, -1);
1338 iso2022_read_test(TESTSTR("\x1b(I\x20\x7f"), 0x20, 0x7f, 0, -1);
1339 iso2022_read_test(TESTSTR("\x1b$B\x20\x7f"), 0x20, 0x7f, 0, -1);
1340 iso2022_read_test(TESTSTR("\x1b-A\x0e\x20\x7f"), 0xa0, 0xff, 0, -1);
1341 iso2022_read_test(TESTSTR("\x1b$-~\x0e\x20\x7f"), ERROR
, 0, -1);
1342 iso2022_read_test(TESTSTR("\x1b)B\xa0\xff"), ERROR
, ERROR
, 0, -1);
1343 iso2022_read_test(TESTSTR("\x1b)I\xa0\xff"), ERROR
, ERROR
, 0, -1);
1344 iso2022_read_test(TESTSTR("\x1b$)B\xa0\xff"), ERROR
, ERROR
, 0, -1);
1345 iso2022_read_test(TESTSTR("\x1b-A\x1b~\xa0\xff"), 0xa0, 0xff, 0, -1);
1346 iso2022_read_test(TESTSTR("\x1b$-~\x1b~\xa0\xff"), ERROR
, 0, -1);
1347 /* Designate control sets */
1348 iso2022_read_test(TESTSTR("\x1b!@"), 0x1b, '!', '@', 0, -1);
1349 /* Designate other coding system (UTF-8) */
1350 iso2022_read_test(TESTSTR("\x1b%G"
1351 "\xCE\xBA\xE1\xBD\xB9\xCF\x83\xCE\xBC\xCE\xB5"),
1352 0x03BA, 0x1F79, 0x03C3, 0x03BC, 0x03B5, 0, -1);
1353 iso2022_read_test(TESTSTR("\x1b-A\x1b%G\xCE\xBA\x1b%@\xa0"),
1354 0x03BA, 0xA0, 0, -1);
1355 iso2022_read_test(TESTSTR("\x1b%G\xCE\x1b%@"), ERROR
, 0, -1);
1356 iso2022_read_test(TESTSTR("\x1b%G\xCE\xBA\x1b%\x1b%@"),
1357 0x03BA, 0x1B, '%', 0, -1);
1358 /* DOCS (COMPOUND_TEXT extended segment) */
1359 iso2022_read_test(TESTSTR("\x1b%/1\x80\x80"), 0, -1);
1360 iso2022_read_test(TESTSTR("\x1b%/1\x80\x8fiso-8859-15\2xyz\x1b(B"),
1361 ERROR
, ERROR
, ERROR
, 0, -1);
1362 iso2022_read_test(TESTSTR("\x1b%/1\x80\x8eiso8859-15\2xyz\x1b(B"),
1363 'x', 'y', 'z', 0, -1);
1364 iso2022_read_test(TESTSTR("\x1b-A\x1b%/2\x80\x89"
1365 "big5-0\2\xa1\x40\xa1\x40"),
1366 0x3000, 0xa1, 0x40, 0, -1);
1367 /* Emacs Big5-in-ISO-2022 mapping */
1368 iso2022_read_test(TESTSTR("\x1b$(0&x86\x1b(B \x1b$(0DeBv"),
1369 0x5143, 0x6c23, ' ', ' ', 0x958b, 0x767c, 0, -1);
1370 /* Test from RFC 1922 (ISO-2022-CN) */
1371 iso2022_read_test(TESTSTR("\x1b$)A\x0e=;;;\x1b$)GG(_P\x0f"),
1372 0x4EA4, 0x6362, 0x4EA4, 0x63db, 0, -1);
1374 printf("read tests completed\n");
1375 printf("total: %d errors\n", total_errs
);
1376 return (total_errs
!= 0);
1379 #endif /* TESTMODE */
1381 #else /* ENUM_CHARSETS */
1383 ENUM_CHARSET(CS_ISO2022
)