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 only handles decoding, because a fully general encoder
6 * isn't really useful. It can decode 8-bit and 7-bit versions, with
7 * support for single-byte and multi-byte character sets, all four
8 * containers (G0, G1, G2, and G3), using both single-shift and
9 * locking-shift sequences.
11 * The general principle is that any valid ISO/IEC 2022 sequence
12 * should either be correctly decoded or should emit an ERROR. The
13 * only exception to this is that the C0 and C1 sets are fixed as
14 * those of ISO/IEC 6429. Escape sequences for designating control
15 * sets are passed through, so a post-processor could fix them up if
18 * DOCS to UTF-8 works. Other DOCS sequences are ignored, which will
19 * produce surprising results.
37 enum {S4
, S6
, M4
, M6
};
39 static long int emacs_big5_1_to_unicode(int, int);
40 static long int emacs_big5_2_to_unicode(int, int);
41 static int unicode_to_emacs_big5(long int, int *, int *, int *);
42 static long int cns11643_1_to_unicode(int, int);
43 static long int cns11643_2_to_unicode(int, int);
44 static long int cns11643_3_to_unicode(int, int);
45 static long int cns11643_4_to_unicode(int, int);
46 static long int cns11643_5_to_unicode(int, int);
47 static long int cns11643_6_to_unicode(int, int);
48 static long int cns11643_7_to_unicode(int, int);
49 static long int null_dbcs_to_unicode(int, int);
50 static int unicode_to_null_dbcs(long int, int *, int *);
52 typedef int (*to_dbcs_t
)(long int, int *, int *);
53 typedef int (*to_dbcs_planar_t
)(long int, int *, int *, int *);
56 * These macros cast between to_dbcs_planar_t and to_dbcs_t, in
57 * such a way as to cause a compile-time error if the input is not
58 * of the appropriate type.
60 * Defining these portably is quite fiddly. My first effort was as
62 * #define DEPLANARISE(x) ( (x) == (to_dbcs_planar_t)NULL, (to_dbcs_t)(x) )
64 * so that the comparison on the left of the comma provokes the
65 * type check error, and the cast on the right is the actual
68 * gcc was entirely happy with this. However, when used in a static
69 * initialiser, MSVC objected - justifiably - that the first half
70 * of the comma expression wasn't constant and thus the expression
71 * as a whole was not a constant expression. We can get round this
72 * by enclosing the comparison in `sizeof', so that it isn't
75 * But then we run into a second problem, which is that C actually
76 * disallows the use of the comma operator within a constant
77 * expression for any purpose at all! Presumably this is on the
78 * basis that its purpose is to have side effects and constant
79 * expressions can't; unfortunately, this specific case is one in
80 * which the desired side effect is a compile-time rather than a
83 * We are permitted to use ?:, however, and that works quite well
84 * since the actual result of the sizeof expression _is_ evaluable
85 * at compile time. So here's my final answer, with the unfortunate
86 * remaining problem of evaluating its arguments multiple times:
88 #define TYPECHECK(x,y) ( sizeof((x)) == sizeof((x)) ? (y) : (y) )
89 #define DEPLANARISE(x) TYPECHECK((x) == (to_dbcs_planar_t)NULL, (to_dbcs_t)(x))
90 #define REPLANARISE(x) TYPECHECK((x) == (to_dbcs_t)NULL, (to_dbcs_planar_t)(x))
93 * Values used in the `enable' field. Each of these identifies a
94 * class of character sets; we then have a bitmask indicating which
95 * classes are allowable in a given mode.
97 * These values are currently only checked on output: for input,
98 * any ISO 2022 we can comprehend at all is considered acceptable.
100 #define CCS 1 /* CTEXT standard */
101 #define COS 2 /* other standard */
102 #define CPU 3 /* private use */
103 #define CDC 4 /* DOCS for CTEXT */
104 #define CDU 5 /* DOCS for UTF-8 */
105 #define CNU 31 /* never used */
107 struct iso2022_mode
{
109 char ltype
, li
, lf
, rtype
, ri
, rf
;
112 const struct iso2022_subcharset
{
113 char type
, i
, f
, enable
;
115 const sbcs_data
*sbcs_base
;
116 long int (*from_dbcs
)(int, int);
119 * If to_dbcs_plane < 0, then to_dbcs is used as expected.
120 * However, if to_dbcs_plane >= 0, then to_dbcs is expected to
121 * be cast to a to_dbcs_planar_t before use, and the returned
122 * plane value (the first int *) must equal to_dbcs_plane.
124 * I'd have preferred to do this by means of a union, but you
125 * can't initialise a selected field of a union at compile
126 * time. Function pointer casts are guaranteed to work sensibly
127 * in ISO C (that is, it's undefined what happens if you call a
128 * function via the wrong type of pointer, but if you cast it
129 * back to the right type before calling it then it must work),
130 * so this is safe if ugly.
133 int to_dbcs_plane
; /* use to_dbcs_planar iff >= 0 */
134 } iso2022_subcharsets
[] = {
136 * We list these subcharsets in preference order for output.
137 * Since the best-defined use of ISO 2022 output is compound
138 * text, we'll use a preference order which matches that. So we
139 * begin with the charsets defined in the compound text spec.
141 { S4
, 0, 'B', CCS
, 0x00, &sbcsdata_CS_ASCII
},
142 { S6
, 0, 'A', CCS
, 0x80, &sbcsdata_CS_ISO8859_1
},
143 { S6
, 0, 'B', CCS
, 0x80, &sbcsdata_CS_ISO8859_2
},
144 { S6
, 0, 'C', CCS
, 0x80, &sbcsdata_CS_ISO8859_3
},
145 { S6
, 0, 'D', CCS
, 0x80, &sbcsdata_CS_ISO8859_4
},
146 { S6
, 0, 'F', CCS
, 0x80, &sbcsdata_CS_ISO8859_7
},
147 { S6
, 0, 'G', CCS
, 0x80, &sbcsdata_CS_ISO8859_6
},
148 { S6
, 0, 'H', CCS
, 0x80, &sbcsdata_CS_ISO8859_8
},
149 { S6
, 0, 'L', CCS
, 0x80, &sbcsdata_CS_ISO8859_5
},
150 { S6
, 0, 'M', CCS
, 0x80, &sbcsdata_CS_ISO8859_9
},
151 { S4
, 0, 'I', CCS
, 0x80, &sbcsdata_CS_JISX0201
},
152 { S4
, 0, 'J', CCS
, 0x00, &sbcsdata_CS_JISX0201
},
153 { M4
, 0, 'A', CCS
, -0x21, 0, &gb2312_to_unicode
, &unicode_to_gb2312
, -1 },
154 { M4
, 0, 'B', CCS
, -0x21, 0, &jisx0208_to_unicode
, &unicode_to_jisx0208
, -1 },
155 { M4
, 0, 'C', CCS
, -0x21, 0, &ksx1001_to_unicode
, &unicode_to_ksx1001
, -1 },
156 { M4
, 0, 'D', CCS
, -0x21, 0, &jisx0212_to_unicode
, &unicode_to_jisx0212
, -1 },
159 * Next, other reasonably standard things: the rest of the ISO
160 * 8859 sets, UK-ASCII, and CNS 11643.
162 { S6
, 0, 'T', COS
, 0x80, &sbcsdata_CS_ISO8859_11
},
163 { S6
, 0, 'V', COS
, 0x80, &sbcsdata_CS_ISO8859_10
},
164 { S6
, 0, 'Y', COS
, 0x80, &sbcsdata_CS_ISO8859_13
},
165 { S6
, 0, '_', COS
, 0x80, &sbcsdata_CS_ISO8859_14
},
166 { S6
, 0, 'b', COS
, 0x80, &sbcsdata_CS_ISO8859_15
},
167 { S6
, 0, 'f', COS
, 0x80, &sbcsdata_CS_ISO8859_16
},
168 { S4
, 0, 'A', COS
, 0x00, &sbcsdata_CS_BS4730
},
169 { M4
, 0, 'G', COS
, -0x21, 0, &cns11643_1_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 0 },
170 { M4
, 0, 'H', COS
, -0x21, 0, &cns11643_2_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 1 },
171 { M4
, 0, 'I', COS
, -0x21, 0, &cns11643_3_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 2 },
172 { M4
, 0, 'J', COS
, -0x21, 0, &cns11643_4_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 3 },
173 { M4
, 0, 'K', COS
, -0x21, 0, &cns11643_5_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 4 },
174 { M4
, 0, 'L', COS
, -0x21, 0, &cns11643_6_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 5 },
175 { M4
, 0, 'M', COS
, -0x21, 0, &cns11643_7_to_unicode
, DEPLANARISE(&unicode_to_cns11643
), 6 },
178 * Private-use designations: DEC private sets and Emacs's Big5
181 { S4
, 0, '0', CPU
, 0x00, &sbcsdata_CS_DEC_GRAPHICS
},
182 { S4
, 0, '<', CPU
, 0x80, &sbcsdata_CS_DEC_MCS
},
183 { M4
, 0, '0', CPU
, -0x21, 0, &emacs_big5_1_to_unicode
, DEPLANARISE(&unicode_to_emacs_big5
), 1 },
184 { M4
, 0, '1', CPU
, -0x21, 0, &emacs_big5_2_to_unicode
, DEPLANARISE(&unicode_to_emacs_big5
), 2 },
187 * Ben left this conditioned out without explanation,
188 * presumably on the grounds that we don't have a translation
192 { M4
, 0, '@', CNU
}, /* JIS C 6226-1978 */
196 * Finally, fallback entries for null character sets.
199 { S6
, 0, '~', CNU
}, /* empty 96-set */
200 { M4
, 0, '~', CNU
, 0, 0, &null_dbcs_to_unicode
, &unicode_to_null_dbcs
, -1 }, /* empty 94^n-set */
201 { M6
, 0, '~', CNU
, 0, 0, &null_dbcs_to_unicode
, &unicode_to_null_dbcs
, -1 }, /* empty 96^n-set */
204 static long int null_dbcs_to_unicode(int r
, int c
)
210 static int unicode_to_null_dbcs(long int unicode
, int *r
, int *c
)
215 return 0; /* failed to convert anything */
219 * Emacs encodes Big5 in COMPOUND_TEXT as two 94x94 character sets.
220 * We treat Big5 as a 94x191 character set with a bunch of undefined
221 * columns in the middle, so we have to mess around a bit to make
225 static long int emacs_big5_1_to_unicode(int r
, int c
)
231 if (c
>= 64) c
+= 34; /* Skip over the gap */
232 return big5_to_unicode(r
, c
);
235 static long int emacs_big5_2_to_unicode(int r
, int c
)
241 if (c
>= 64) c
+= 34; /* Skip over the gap */
242 return big5_to_unicode(r
, c
);
245 static int unicode_to_emacs_big5(long int unicode
, int *p
, int *r
, int *c
)
248 if (!unicode_to_big5(unicode
, &rr
, &cc
))
266 /* Wrappers for cns11643_to_unicode() */
267 static long int cns11643_1_to_unicode(int r
, int c
)
269 return cns11643_to_unicode(0, r
, c
);
271 static long int cns11643_2_to_unicode(int r
, int c
)
273 return cns11643_to_unicode(1, r
, c
);
275 static long int cns11643_3_to_unicode(int r
, int c
)
277 return cns11643_to_unicode(2, r
, c
);
279 static long int cns11643_4_to_unicode(int r
, int c
)
281 return cns11643_to_unicode(3, r
, c
);
283 static long int cns11643_5_to_unicode(int r
, int c
)
285 return cns11643_to_unicode(4, r
, c
);
287 static long int cns11643_6_to_unicode(int r
, int c
)
289 return cns11643_to_unicode(5, r
, c
);
291 static long int cns11643_7_to_unicode(int r
, int c
)
293 return cns11643_to_unicode(6, r
, c
);
296 /* States, or "what we're currently accumulating". */
298 IDLE
, /* None of the below */
299 SS2CHAR
, /* Accumulating a character after SS2 */
300 SS3CHAR
, /* Accumulating a character after SS3 */
301 ESCSEQ
, /* Accumulating an escape sequence */
302 ESCDROP
, /* Discarding an escape sequence */
303 ESCPASS
, /* Passing through an escape sequence */
304 DOCSUTF8
, /* DOCSed into UTF-8 */
305 DOCSCTEXT
/* DOCSed into a COMPOUND_TEXT extended segment */
310 static void dump_state(charset_state
*s
)
312 unsigned s0
= s
->s0
, s1
= s
->s1
;
313 char const * const modes
[] = { "IDLE", "SS2CHAR", "SS3CHAR",
314 "ESCSEQ", "ESCDROP", "ESCPASS",
317 fprintf(stderr
, "s0: %s", modes
[s0
>> 29]);
318 fprintf(stderr
, " %02x %02x %02x ", (s0
>> 16) & 0xff, (s0
>> 8) & 0xff,
320 fprintf(stderr
, "s1: LS%d LS%dR", (s1
>> 30) & 3, (s1
>> 28) & 3);
321 fprintf(stderr
, " %d %d %d %d\n", s1
& 0x7f, (s1
>> 7) & 0x7f,
322 (s1
>> 14) & 0x7f, (s1
>> 21) & 0x7f);
326 static void designate(charset_state
*state
, int container
,
327 int type
, int ibyte
, int fbyte
)
331 assert(container
>= 0 && container
<= 3);
332 assert(type
== S4
|| type
== S6
|| type
== M4
|| type
== M6
);
334 for (i
= 0; i
< lenof(iso2022_subcharsets
); i
++) {
335 if (iso2022_subcharsets
[i
].type
== type
&&
336 iso2022_subcharsets
[i
].i
== ibyte
&&
337 iso2022_subcharsets
[i
].f
== fbyte
) {
338 state
->s1
&= ~(0x7fL
<< (container
* 7));
339 state
->s1
|= (i
<< (container
* 7));
344 * If we don't find the charset, invoke the empty one, so we
345 * output ERROR rather than garbage.
347 designate(state
, container
, type
, 0, '~');
350 static void do_utf8(long int input_chr
,
351 charset_state
*state
,
352 void (*emit
)(void *ctx
, long int output
),
355 charset_state ustate
;
358 ustate
.s0
= state
->s0
& 0x03ffffffL
;
359 read_utf8(NULL
, input_chr
, &ustate
, emit
, emitctx
);
360 state
->s0
= (state
->s0
& ~0x03ffffffL
) | (ustate
.s0
& 0x03ffffffL
);
363 static void docs_utf8(long int input_chr
,
364 charset_state
*state
,
365 void (*emit
)(void *ctx
, long int output
),
371 * Bits [25:0] of s0 are reserved for read_utf8().
372 * Bits [27:26] are a tiny state machine to recognise ESC % @.
374 retstate
= (state
->s0
& 0x0c000000L
) >> 26;
375 if (retstate
== 1 && input_chr
== '%')
377 else if (retstate
== 2 && input_chr
== '@') {
378 /* If we've got a partial UTF-8 sequence, complain. */
379 if (state
->s0
& 0x03ffffffL
)
380 emit(emitctx
, ERROR
);
384 if (retstate
>= 1) do_utf8(ESC
, state
, emit
, emitctx
);
385 if (retstate
>= 2) do_utf8('%', state
, emit
, emitctx
);
387 if (input_chr
== ESC
)
390 do_utf8(input_chr
, state
, emit
, emitctx
);
393 state
->s0
= (state
->s0
& ~0x0c000000L
) | (retstate
<< 26);
396 struct ctext_encoding
{
398 char octets_per_char
, enable
;
399 charset_spec
const *subcs
;
403 * In theory, this list is in <ftp://ftp.x.org/pub/DOCS/registry>,
404 * but XLib appears to have its own ideas, and encodes these three
408 extern charset_spec
const charset_CS_ISO8859_14
;
409 extern charset_spec
const charset_CS_ISO8859_15
;
410 extern charset_spec
const charset_CS_BIG5
;
412 static struct ctext_encoding
const ctext_encodings
[] = {
413 { "big5-0\2", 0 /* variable */, CDC
, &charset_CS_BIG5
},
414 { "iso8859-14\2", 1, CDC
, &charset_CS_ISO8859_14
},
415 { "iso8859-15\2", 1, CDC
, &charset_CS_ISO8859_15
}
418 static void docs_ctext(long int input_chr
,
419 charset_state
*state
,
420 void (*emit
)(void *ctx
, long int output
),
424 * s0[27:26] = first entry in ctext_encodings that matches
425 * s0[25:22] = number of characters successfully matched, 0xf if all
426 * s0[21:8] count the number of octets left in the segment
427 * s0[7:0] are for sub-charset use
429 int n
= (state
->s0
>> 22) & 0xf, i
= (state
->s0
>> 26) & 3, oi
= i
, j
;
430 int length
= (state
->s0
>> 8) & 0x3fff;
433 * Note that we do not bother checking the octets-per-character
434 * byte against the selected charset when reading. It's
435 * extremely unlikely that this code will ever have to deal
436 * with two charset identifiers with the same name and
437 * different octets-per-character values! If it ever happens,
438 * we'll have to edit this file anyway so we can modify the
443 /* Haven't read length yet */
444 if ((state
->s0
& 0xff) == 0)
445 /* ... or even the first byte */
446 state
->s0
|= input_chr
;
448 length
= (state
->s0
& 0x7f) * 0x80 + (input_chr
& 0x7f);
452 state
->s0
= (state
->s0
& 0xf0000000) | (length
<< 8);
459 /* Skipping unknown encoding. Look out for STX. */
461 state
->s0
= (state
->s0
& 0xf0000000) | (i
<< 26) | (0xf << 22);
462 } else if (n
!= 0xf) {
463 while ((unsigned)j
< lenof(ctext_encodings
) &&
464 !memcmp(ctext_encodings
[j
].name
,
465 ctext_encodings
[oi
].name
, n
)) {
466 if (ctext_encodings
[j
].name
[n
] < input_chr
)
471 if ((unsigned)i
>= lenof(ctext_encodings
) ||
472 memcmp(ctext_encodings
[i
].name
,
473 ctext_encodings
[oi
].name
, n
) ||
474 ctext_encodings
[i
].name
[n
] != input_chr
) {
475 /* Doom! We haven't heard of this encoding */
476 i
= lenof(ctext_encodings
);
480 * Otherwise, we have found an additional character in our
481 * encoding name. See if we have reached the _end_ of our
485 if (!ctext_encodings
[i
].name
[n
])
489 * Failing _that_, we simply update our encoding-name-
492 assert(i
< 4 && n
< 16);
493 state
->s0
= (state
->s0
& 0xf0000000) | (i
<< 26) | (n
<< 22);
495 if ((unsigned)i
>= lenof(ctext_encodings
))
496 emit(emitctx
, ERROR
);
498 charset_state substate
;
499 charset_spec
const *subcs
= ctext_encodings
[i
].subcs
;
501 substate
.s0
= state
->s0
& 0xff;
502 subcs
->read(subcs
, input_chr
, &substate
, emit
, emitctx
);
503 state
->s0
= (state
->s0
& ~0xff) | (substate
.s0
& 0xff);
509 state
->s0
= (state
->s0
&~0x003fff00) | (length
<< 8);
512 static void read_iso2022(charset_spec
const *charset
, long int input_chr
,
513 charset_state
*state
,
514 void (*emit
)(void *ctx
, long int output
),
517 struct iso2022_mode
const *mode
= (struct iso2022_mode
*)charset
->data
;
519 /* dump_state(state); */
521 * We have to make fairly efficient use of the 64 bits of state
522 * available to us. Long-term state goes in s1, and consists of
523 * the identities of the character sets designated as G0/G1/G2/G3
524 * and the locking-shift states for GL and GR. Short-term state
525 * goes in s0: The bottom half of s0 accumulates characters for an
526 * escape sequence or a multi-byte character, while the top three
527 * bits indicate what they're being accumulated for. After DOCS,
528 * the bottom 29 bits of state are available for the DOCS function
529 * to use -- the UTF-8 one uses the bottom 26 for UTF-8 decoding
530 * and the top two to recognised ESC % @.
532 * s0[31:29] = state enum
533 * s0[24:0] = accumulated bytes
534 * s1[31:30] = GL locking-shift state
535 * s1[29:28] = GR locking-shift state
536 * s1[27:21] = G3 charset
537 * s1[20:14] = G2 charset
538 * s1[13:7] = G1 charset
539 * s1[6:0] = G0 charset
544 #define LOCKING_SHIFT(n,side) \
545 (state->s1 = (state->s1 & ~(3UL<<(side))) | ((n ## UL)<<(side)))
546 #define MODE ((state->s0 & 0xe0000000UL) >> 29)
547 #define ENTER_MODE(m) (state->s0 = (state->s0 & ~0xe0000000UL) | ((unsigned long)(m)<<29))
548 #define SINGLE_SHIFT(n) ENTER_MODE(SS2CHAR - 2 + (n))
549 #define ASSERT_IDLE do { \
550 if (state->s0 != 0) emit(emitctx, ERROR); \
554 if (state
->s1
== 0) {
556 * Since there's no LS0R, this means we must just have started.
557 * Set up a sane initial state (LS0, LS1R, ASCII in G0/G1/G2/G3).
559 LOCKING_SHIFT(0, LEFT
);
560 LOCKING_SHIFT(1, RIGHT
);
561 designate(state
, 0, mode
->ltype
, mode
->li
, mode
->lf
);
562 designate(state
, 1, mode
->rtype
, mode
->ri
, mode
->rf
);
563 designate(state
, 2, S4
, 0, 'B');
564 designate(state
, 3, S4
, 0, 'B');
567 if (MODE
== DOCSUTF8
) {
568 docs_utf8(input_chr
, state
, emit
, emitctx
);
571 if (MODE
== DOCSCTEXT
) {
572 docs_ctext(input_chr
, state
, emit
, emitctx
);
576 if ((input_chr
& 0x60) == 0x00) {
577 /* C0 or C1 control */
584 LOCKING_SHIFT(0, LEFT
);
587 LOCKING_SHIFT(1, LEFT
);
596 emit(emitctx
, input_chr
);
599 } else if ((input_chr
& 0x80) || MODE
< ESCSEQ
) {
601 struct iso2022_subcharset
const *subcs
;
606 * Force idle state if we're in mid escape sequence, or in a
607 * multi-byte character with a different top bit.
609 if (MODE
>= ESCSEQ
||
610 ((state
->s0
& 0x00ff0000L
) != 0 &&
611 (((state
->s0
>> 16) ^ input_chr
) & 0x80)))
613 if (MODE
== SS2CHAR
|| MODE
== SS3CHAR
) /* Single-shift */
614 container
= MODE
- SS2CHAR
+ 2;
615 else if (input_chr
>= 0x80) /* GR */
616 container
= (state
->s1
>> 28) & 3;
618 container
= state
->s1
>> 30;
621 input_7bit
= input_chr
& ~0x80;
622 subcs
= &iso2022_subcharsets
[(state
->s1
>> (container
* 7)) & 0x7f];
623 if ((subcs
->type
== S4
|| subcs
->type
== M4
) &&
624 (input_7bit
== 0x20 || input_7bit
== 0x7f)) {
625 /* characters not in 94-char set */
626 if (is_gl
) emit(emitctx
, input_7bit
);
627 else emit(emitctx
, ERROR
);
628 } else if (subcs
->type
== M4
|| subcs
->type
== M6
) {
629 if ((state
->s0
& 0x00ff0000L
) == 0) {
630 state
->s0
|= input_chr
<< 16;
634 subcs
->from_dbcs(((state
->s0
>> 16) & 0x7f) +
636 input_7bit
+ subcs
->offset
));
639 if ((state
->s0
& 0x00ff0000L
) != 0)
640 emit(emitctx
, ERROR
);
641 emit(emitctx
, subcs
->sbcs_base ?
642 sbcs_to_unicode(subcs
->sbcs_base
, input_7bit
+ subcs
->offset
):
648 if (MODE
== ESCPASS
) {
649 emit(emitctx
, input_chr
);
650 if ((input_chr
& 0xf0) != 0x20)
656 * Intermediate bytes shall be any of the 16 positions of
657 * column 02 of the code table; they are denoted by the symbol
660 if ((input_chr
& 0xf0) == 0x20) {
661 if (((state
->s0
>> 16) & 0xff) == 0)
662 state
->s0
|= input_chr
<< 16;
663 else if (((state
->s0
>> 8) & 0xff) == 0)
664 state
->s0
|= input_chr
<< 8;
666 /* Long escape sequence. Switch to ESCPASS or ESCDROP. */
667 i1
= (state
->s0
>> 16) & 0xff;
668 i2
= (state
->s0
>> 8) & 0xff;
670 case '(': case ')': case '*': case '+':
671 case '-': case '.': case '/':
679 emit(emitctx
, input_chr
);
689 * Final bytes shall be any of the 79 positions of columns 03
690 * to 07 of the code table excluding position 07/15; they are
691 * denoted by the symbol F.
693 i1
= (state
->s0
>> 16) & 0xff;
694 i2
= (state
->s0
>> 8) & 0xff;
696 input_chr
= 0; /* Make sure it won't match. */
699 case 0: /* No intermediate bytes */
708 LOCKING_SHIFT(2, LEFT
);
711 LOCKING_SHIFT(3, LEFT
);
714 LOCKING_SHIFT(3, RIGHT
);
717 LOCKING_SHIFT(2, RIGHT
);
720 LOCKING_SHIFT(1, RIGHT
);
723 /* Unsupported escape sequence. Spit it back out. */
725 emit(emitctx
, input_chr
);
730 * Various coding structure facilities specify that designating
731 * a code element also invokes it. As far as I can see, invoking
732 * it now will have the same practical effect, since those
733 * facilities also ban the use of locking shifts.
736 case 'A': /* G0 element used and invoked into GL */
737 LOCKING_SHIFT(0, LEFT
);
739 case 'C': /* G0 in GL, G1 in GR */
740 case 'D': /* Ditto, at least for 8-bit codes */
741 case 'L': /* ISO 4873 (ECMA-43) level 1 */
742 case 'M': /* ISO 4873 (ECMA-43) level 2 */
743 LOCKING_SHIFT(0, LEFT
);
744 LOCKING_SHIFT(1, RIGHT
);
750 * IRR (Identify Revised Registration) is ignored here,
751 * since any revised registration must be
752 * upward-compatible with the old one, so either we'll
753 * support the new one or we'll emit ERROR when we run
754 * into a new character. In either case, there's nothing
758 case '(': /* GZD4 */ case ')': /* G1D4 */
759 case '*': /* G2D4 */ case '+': /* G3D4 */
760 designate(state
, i1
- '(', S4
, i2
, input_chr
);
762 case '-': /* G1D6 */ case '.': /* G2D6 */ case '/': /* G3D6 */
763 designate(state
, i1
- ',', S6
, i2
, input_chr
);
765 case '$': /* G?DM? */
767 case 0: /* Obsolete version of GZDM4 */
769 case '(': /* GZDM4 */ case ')': /* G1DM4 */
770 case '*': /* G2DM4 */ case '+': /* G3DM4 */
771 designate(state
, i2
- '(', M4
, 0, input_chr
);
773 case '-': /* G1DM6 */
774 case '.': /* G2DM6 */ case '/': /* G3DM6 */
775 designate(state
, i2
- ',', M6
, 0, input_chr
);
778 emit(emitctx
, ERROR
);
782 /* XXX What's a reasonable way to handle an unrecognised DOCS? */
787 ENTER_MODE(DOCSUTF8
);
794 ENTER_MODE(DOCSCTEXT
);
801 /* Unsupported nF escape sequence. Re-emit it. */
804 if (i2
) emit(emitctx
, i2
);
805 emit(emitctx
, input_chr
);
811 static void oselect(charset_state
*state
, int i
, int right
,
812 void (*emit
)(void *ctx
, long int output
),
815 int shift
= (right ?
31-7 : 31-7-7);
816 struct iso2022_subcharset
const *subcs
= &iso2022_subcharsets
[i
];
818 if (((state
->s1
>> shift
) & 0x7F) != (unsigned)i
) {
819 state
->s1
&= ~(0x7FL
<< shift
);
820 state
->s1
|= (i
<< shift
);
824 if (subcs
->type
== M4
|| subcs
->type
== M6
)
826 if (subcs
->type
== S6
|| subcs
->type
== M6
) {
835 emit(emitctx
, subcs
->i
);
836 emit(emitctx
, subcs
->f
);
841 static void docs_char(charset_state
*state
,
842 void (*emit
)(void *ctx
, long int output
),
843 void *emitctx
, int cset
, char *data
, int datalen
)
845 int curr_cset
, currlen
, i
;
848 * cset is the index into ctext_encodings[]. It can also be -1
849 * to mean DOCS UTF-8, or -2 to mean no DOCS (ordinary 2022).
850 * In the latter case, `chr' is ignored.
854 * First, terminate a DOCS segment if necessary. We always have
855 * to terminate a DOCS segment if one is active and we're about
856 * to switch to a different one; we might also have to
857 * terminate a length-encoded DOCS segment if we've run out of
858 * storage space to accumulate characters in it.
860 curr_cset
= ((state
->s1
>> 14) & 7) - 2;
861 currlen
= ((state
->s1
>> 11) & 7);
862 if ((curr_cset
!= -2 && curr_cset
!= cset
) ||
863 (curr_cset
>= 0 && currlen
+ datalen
> 5)) {
864 if (curr_cset
== -1) {
866 * Terminating DOCS UTF-8 is easy.
875 * To terminate a length-encoded DOCS segment we must
876 * actually output the whole thing.
881 emit(emitctx
, '0' + ctext_encodings
[curr_cset
].octets_per_char
);
882 len
= currlen
+ datalen
+
883 strlen(ctext_encodings
[curr_cset
].name
);
884 assert(len
< (1 << 14));
885 emit(emitctx
, 0x80 | ((len
>> 7) & 0x7F));
886 emit(emitctx
, 0x80 | ((len
) & 0x7F));
887 /* The name stored in ctext_encodings[] includes the trailing \2 */
888 for (i
= 0; ctext_encodings
[curr_cset
].name
[i
]; i
++)
889 emit(emitctx
, ctext_encodings
[curr_cset
].name
[i
]);
890 for (i
= 0; i
< currlen
; i
++)
892 (i
== 0 ? state
->s1
: state
->s0
>> (8*(4-i
))) & 0xFF);
893 for (i
= 0; i
< datalen
; i
++)
894 emit(emitctx
, data
[i
]);
897 * We've now dealt with the input data, so clear it so
898 * we don't try to do so again below.
906 * Now, start a DOCS segment if necessary.
908 if (curr_cset
!= cset
) {
919 * Starting a length-encoded DOCS segment is simply a
920 * matter of setting our stored length counter to zero.
923 state
->s1
&= ~(7 << 11);
928 state
->s1
&= ~(7 << 14);
929 assert((cset
+2) >= 0 && (cset
+2) < 8);
930 state
->s1
|= ((cset
+2) << 14);
933 * Now we're in the right DOCS state. Actually deal with the
934 * input data, if we haven't already done so above.
940 * In DOCS UTF-8, we output data as soon as we get it.
942 for (i
= 0; i
< datalen
; i
++)
943 emit(emitctx
, data
[i
]);
946 * In length-encoded DOCS, we just store our data and
947 * bide our time. It'll all be output when we fill up
948 * or switch to another character set.
950 assert(currlen
+ datalen
<= 5); /* overflow handled already */
951 for (i
= 0; i
< datalen
; i
++) {
952 if (currlen
+ i
== 0)
953 state
->s1
|= data
[i
] & 0xFF;
955 state
->s0
|= (data
[i
] & 0xFF) << (8*(4-(currlen
+i
)));
958 assert(currlen
>= 0 && currlen
< 8);
959 state
->s1
&= ~(7 << 11);
960 state
->s1
|= (currlen
<< 11);
965 static void write_to_pointer(void *ctx
, long int output
)
967 char **ptr
= (char **)ctx
;
972 * Writing full ISO-2022 is not useful in very many circumstances.
973 * One of the few situations in which it _is_ useful is generating
974 * X11 COMPOUND_TEXT; therefore, this writing function will obey
975 * the compound text restrictions and hence output the subset of
976 * ISO-2022 that's usable in that context.
978 * The subset in question is roughly that we use GL/GR for G0/G1
979 * always, and that the _only_ escape sequences we output (other
980 * than the occasional DOCS) are those which designate different
981 * subcharsets into G0 and G1. There are additional constraints
982 * about which things go in which container; see below.
984 * FIXME: this wants some decent tests to be written, and also the
985 * exact output policy for compound text wants thinking about more
988 static int write_iso2022(charset_spec
const *charset
, long int input_chr
,
989 charset_state
*state
,
990 void (*emit
)(void *ctx
, long int output
),
994 struct iso2022_subcharset
const *subcs
;
995 struct iso2022_mode
const *mode
= (struct iso2022_mode
*)charset
->data
;
996 to_dbcs_planar_t last_planar_dbcs
= NULL
;
997 int last_p
, last_r
, last_c
;
1001 * For output, I allocate the state variables as follows:
1003 * s1[31] == 1 if output state has been initialised
1004 * s1[30:24] == G1 charset (always in GR)
1005 * s1[23:17] == G0 charset (always in GL)
1006 * s1[16:14] == DOCS index plus 2 (because -1 and -2 are special)
1007 * s1[13:11] == number of DOCS accumulated characters (up to five)
1008 * s1[7:0] + s0[31:0] == DOCS collected characters
1012 state
->s0
= 0x00000000UL
;
1013 state
->s1
= 0x80000000UL
;
1015 * Start with US-ASCII in GL and also in GR.
1017 for (i
= 0; (unsigned)i
< lenof(iso2022_subcharsets
); i
++) {
1018 subcs
= &iso2022_subcharsets
[i
];
1019 if (subcs
->type
== mode
->ltype
&&
1020 subcs
->i
== mode
->li
&&
1021 subcs
->f
== mode
->lf
)
1022 oselect(state
, i
, FALSE
, NULL
, NULL
);
1023 if (subcs
->type
== mode
->rtype
&&
1024 subcs
->i
== mode
->ri
&&
1025 subcs
->f
== mode
->rf
)
1026 oselect(state
, i
, TRUE
, NULL
, NULL
);
1030 if (input_chr
== -1) {
1032 * Special case: reset encoding state.
1034 docs_char(state
, emit
, emitctx
, -2, NULL
, 0); /* leave DOCS */
1036 for (i
= 0; (unsigned)i
< lenof(iso2022_subcharsets
); i
++) {
1037 subcs
= &iso2022_subcharsets
[i
];
1038 if (subcs
->type
== mode
->ltype
&&
1039 subcs
->i
== mode
->li
&&
1040 subcs
->f
== mode
->lf
)
1041 oselect(state
, i
, FALSE
, emit
, emitctx
);
1042 if (subcs
->type
== mode
->rtype
&&
1043 subcs
->i
== mode
->ri
&&
1044 subcs
->f
== mode
->rf
)
1045 oselect(state
, i
, TRUE
, emit
, emitctx
);
1051 * Special-case characters: Space, Delete, and anything in C0
1052 * or C1 are output unchanged.
1054 if (input_chr
<= 0x20 || (input_chr
>= 0x7F && input_chr
< 0xA0)) {
1055 emit(emitctx
, input_chr
);
1060 * Analyse the input character and work out which subcharset it
1063 for (i
= 0; (unsigned)i
< lenof(iso2022_subcharsets
); i
++) {
1064 subcs
= &iso2022_subcharsets
[i
];
1065 if (!(mode
->enable_mask
& (1 << subcs
->enable
)))
1066 continue; /* this charset is disabled */
1067 if (subcs
->sbcs_base
) {
1068 c1
= sbcs_from_unicode(subcs
->sbcs_base
, input_chr
);
1069 c1
-= subcs
->offset
;
1070 if (c1
>= 0x20 && c1
<= 0x7f) {
1074 } else if (subcs
->to_dbcs
) {
1075 if (subcs
->to_dbcs_plane
>= 0) {
1077 * Since multiplanar DBCSes almost by definition
1078 * involve several entries in iso2022_subcharsets
1079 * with the same to_dbcs function and different
1080 * plane values, we remember the last such function
1081 * we called and what its result was, so that we
1082 * don't (for example) have to call
1083 * unicode_to_cns11643 seven times.
1085 if (last_planar_dbcs
!= REPLANARISE(subcs
->to_dbcs
)) {
1086 last_planar_dbcs
= REPLANARISE(subcs
->to_dbcs
);
1087 if (!last_planar_dbcs(input_chr
,
1088 &last_p
, &last_r
, &last_c
))
1092 last_p
= subcs
->to_dbcs_plane
;
1093 if (!subcs
->to_dbcs(input_chr
, &last_r
, &last_c
))
1094 last_p
= 0; /* cannot match since to_dbcs_plane<0 */
1097 if (last_p
== subcs
->to_dbcs_plane
) {
1098 c1
= last_r
- subcs
->offset
;
1099 c2
= last_c
- subcs
->offset
;
1100 assert(c1
>= 0x20 && c1
<= 0x7f);
1101 assert(c2
>= 0x20 && c2
<= 0x7f);
1107 if ((unsigned)i
< lenof(iso2022_subcharsets
)) {
1111 * Our character is represented by c1 (and possibly also
1112 * c2) in subcharset `subcs'. So now we must decide whether
1113 * to designate that character set into G0/GL or G1/GR.
1115 * Any S6 or M6 subcharset has to go in GR because it won't
1116 * fit in GL. In addition, the compound text rules state
1117 * that any single-byte subcharset defined as the
1118 * right-hand half of some SBCS must go in GR.
1120 * M4 subcharsets can go in either half according to the
1121 * rules. I choose to put them in GR always because it's a
1122 * simple policy with reasonable behaviour (facilitates
1123 * switching between them and ASCII).
1125 right
= (subcs
->type
== S6
|| subcs
->type
== M6
|| subcs
->type
== M4
||
1126 (subcs
->sbcs_base
&& subcs
->offset
== 0x80));
1129 * If we're in a DOCS mode, leave it.
1131 docs_char(state
, emit
, emitctx
, -2, NULL
, 0);
1134 * If this subcharset is not already selected in that
1135 * container, select it.
1137 oselect(state
, i
, right
, emit
, emitctx
);
1140 * Now emit the actual characters.
1143 assert(c1
>= 0x20 && c1
<= 0x7f);
1144 emit(emitctx
, c1
| 0x80);
1146 assert(c2
>= 0x20 && c2
<= 0x7f);
1147 emit(emitctx
, c2
| 0x80);
1150 assert(c1
> 0x20 && c1
< 0x7f);
1153 assert(c2
> 0x20 && c2
< 0x7f);
1162 * Fall back to DOCS.
1169 cs
= -2; /* means failure */
1171 for (i
= 0; (unsigned)i
<= lenof(ctext_encodings
); i
++) {
1172 charset_state substate
;
1173 charset_spec
const *subcs
= ctext_encodings
[i
].subcs
;
1176 * We assume that all character sets dealt with by DOCS
1177 * are stateless for output purposes.
1179 substate
.s1
= substate
.s0
= 0;
1182 if ((unsigned)i
< lenof(ctext_encodings
)) {
1183 if ((mode
->enable_mask
& (1 << ctext_encodings
[i
].enable
)) &&
1184 subcs
->write(subcs
, input_chr
, &substate
,
1185 write_to_pointer
, &p
)) {
1190 if ((mode
->enable_mask
& (1 << CDU
)) &&
1191 write_utf8(NULL
, input_chr
, NULL
, write_to_pointer
, &p
)) {
1199 docs_char(state
, emit
, emitctx
, cs
, data
, p
- data
);
1208 * Full ISO 2022 output with all options on. Not entirely sure what
1209 * if anything this is useful for, but here it is anyway. All
1210 * output character sets and DOCS variants are permitted; all
1211 * containers start out with ASCII in them.
1213 static const struct iso2022_mode iso2022_all
= {
1214 (1<<CCS
) | (1<<COS
) | (1<<CPU
) | (1<<CDC
) | (1<<CDU
),
1215 S4
, 0, 'B', S4
, 0, 'B',
1218 const charset_spec charset_CS_ISO2022
= {
1219 CS_ISO2022
, read_iso2022
, write_iso2022
, &iso2022_all
1223 * X11 compound text. A subset of output charsets is permitted, and
1224 * G1/GR starts off in ISO8859-1.
1226 static const struct iso2022_mode iso2022_ctext
= {
1227 (1<<CCS
) | (1<<CDC
),
1228 S4
, 0, 'B', S6
, 0, 'A',
1231 const charset_spec charset_CS_CTEXT
= {
1232 CS_CTEXT
, read_iso2022
, write_iso2022
, &iso2022_ctext
1243 void iso2022_emit(void *ctx
, long output
)
1245 wchar_t **p
= (wchar_t **)ctx
;
1249 void iso2022_read_test(int line
, char *input
, int inlen
, ...)
1252 wchar_t *p
, str
[512];
1254 charset_state state
;
1257 state
.s0
= state
.s1
= 0;
1260 for (i
= 0; i
< inlen
; i
++)
1261 read_iso2022(NULL
, input
[i
] & 0xFF, &state
, iso2022_emit
, &p
);
1263 va_start(ap
, inlen
);
1265 for (i
= 0; i
< p
- str
; i
++) {
1266 l
= va_arg(ap
, long int);
1268 printf("%d: correct string shorter than output\n", line
);
1273 printf("%d: char %d came out as %08x, should be %08lx\n",
1274 line
, i
, str
[i
], l
);
1279 l
= va_arg(ap
, long int);
1281 printf("%d: correct string longer than output\n", line
);
1288 /* Macro to concoct the first three parameters of iso2022_read_test. */
1289 #define TESTSTR(x) __LINE__, x, lenof(x)
1293 printf("read tests beginning\n");
1294 /* Simple test (Emacs sample text for Japanese, in ISO-2022-JP) */
1295 iso2022_read_test(TESTSTR("Japanese (\x1b$BF|K\\8l\x1b(B)\t"
1296 "\x1b$B$3$s$K$A$O\x1b(B, "
1297 "\x1b$B%3%s%K%A%O\x1b(B\n"),
1298 'J','a','p','a','n','e','s','e',' ','(',
1299 0x65E5, 0x672C, 0x8A9E, ')', '\t',
1300 0x3053, 0x3093, 0x306b, 0x3061, 0x306f, ',', ' ',
1301 0x30b3, 0x30f3, 0x30cb, 0x30c1, 0x30cf, '\n', 0, -1);
1302 /* Same thing in EUC-JP (with designations, and half-width katakana) */
1303 iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D"
1304 "Japanese (\xc6\xfc\xcb\xdc\xb8\xec)\t"
1305 "\xa4\xb3\xa4\xf3\xa4\xcb\xa4\xc1\xa4\xcf, "
1306 "\x8e\xba\x8e\xdd\x8e\xc6\x8e\xc1\x8e\xca\n"),
1307 'J','a','p','a','n','e','s','e',' ','(',
1308 0x65E5, 0x672C, 0x8A9E, ')', '\t',
1309 0x3053, 0x3093, 0x306b, 0x3061, 0x306f, ',', ' ',
1310 0xff7a, 0xff9d, 0xff86, 0xff81, 0xff8a, '\n', 0, -1);
1311 /* Multibyte single-shift */
1312 iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D\x8f\"/!"),
1313 0x02D8, '!', 0, -1);
1314 /* Non-existent SBCS */
1315 iso2022_read_test(TESTSTR("\x1b(!Zfnord\n"),
1316 ERROR
, ERROR
, ERROR
, ERROR
, ERROR
, '\n', 0, -1);
1317 /* Pass-through of ordinary escape sequences, including a long one */
1318 iso2022_read_test(TESTSTR("\x1b""b\x1b#5\x1b#!!!5"),
1319 0x1B, 'b', 0x1B, '#', '5',
1320 0x1B, '#', '!', '!', '!', '5', 0, -1);
1321 /* Non-existent DBCS (also 5-byte escape sequence) */
1322 iso2022_read_test(TESTSTR("\x1b$(!Bfnord!"),
1323 ERROR
, ERROR
, ERROR
, 0, -1);
1324 /* Incomplete DB characters */
1325 iso2022_read_test(TESTSTR("\x1b$B(,(\x1b(BHi\x1b$B(,(\n"),
1326 0x2501, ERROR
, 'H', 'i', 0x2501, ERROR
, '\n', 0, -1);
1327 iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D\xa4""B"),
1329 iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D\x0e\x1b|$\xa2\xaf"),
1330 ERROR
, 0x02D8, 0, -1);
1331 /* Incomplete escape sequence */
1332 iso2022_read_test(TESTSTR("\x1b\n"), ERROR
, '\n', 0, -1);
1333 iso2022_read_test(TESTSTR("\x1b-A\x1b~\x1b\xa1"), ERROR
, 0xa1, 0, -1);
1334 /* Incomplete single-shift */
1335 iso2022_read_test(TESTSTR("\x8e\n"), ERROR
, '\n', 0, -1);
1336 iso2022_read_test(TESTSTR("\x1b$*B\x8e(\n"), ERROR
, '\n', 0, -1);
1337 /* Corner cases (02/00 and 07/15) */
1338 iso2022_read_test(TESTSTR("\x1b(B\x20\x7f"), 0x20, 0x7f, 0, -1);
1339 iso2022_read_test(TESTSTR("\x1b(I\x20\x7f"), 0x20, 0x7f, 0, -1);
1340 iso2022_read_test(TESTSTR("\x1b$B\x20\x7f"), 0x20, 0x7f, 0, -1);
1341 iso2022_read_test(TESTSTR("\x1b-A\x0e\x20\x7f"), 0xa0, 0xff, 0, -1);
1342 iso2022_read_test(TESTSTR("\x1b$-~\x0e\x20\x7f"), ERROR
, 0, -1);
1343 iso2022_read_test(TESTSTR("\x1b)B\xa0\xff"), ERROR
, ERROR
, 0, -1);
1344 iso2022_read_test(TESTSTR("\x1b)I\xa0\xff"), ERROR
, ERROR
, 0, -1);
1345 iso2022_read_test(TESTSTR("\x1b$)B\xa0\xff"), ERROR
, ERROR
, 0, -1);
1346 iso2022_read_test(TESTSTR("\x1b-A\x1b~\xa0\xff"), 0xa0, 0xff, 0, -1);
1347 iso2022_read_test(TESTSTR("\x1b$-~\x1b~\xa0\xff"), ERROR
, 0, -1);
1348 /* Designate control sets */
1349 iso2022_read_test(TESTSTR("\x1b!@"), 0x1b, '!', '@', 0, -1);
1350 /* Designate other coding system (UTF-8) */
1351 iso2022_read_test(TESTSTR("\x1b%G"
1352 "\xCE\xBA\xE1\xBD\xB9\xCF\x83\xCE\xBC\xCE\xB5"),
1353 0x03BA, 0x1F79, 0x03C3, 0x03BC, 0x03B5, 0, -1);
1354 iso2022_read_test(TESTSTR("\x1b-A\x1b%G\xCE\xBA\x1b%@\xa0"),
1355 0x03BA, 0xA0, 0, -1);
1356 iso2022_read_test(TESTSTR("\x1b%G\xCE\x1b%@"), ERROR
, 0, -1);
1357 iso2022_read_test(TESTSTR("\x1b%G\xCE\xBA\x1b%\x1b%@"),
1358 0x03BA, 0x1B, '%', 0, -1);
1359 /* DOCS (COMPOUND_TEXT extended segment) */
1360 iso2022_read_test(TESTSTR("\x1b%/1\x80\x80"), 0, -1);
1361 iso2022_read_test(TESTSTR("\x1b%/1\x80\x8fiso-8859-15\2xyz\x1b(B"),
1362 ERROR
, ERROR
, ERROR
, 0, -1);
1363 iso2022_read_test(TESTSTR("\x1b%/1\x80\x8eiso8859-15\2xyz\x1b(B"),
1364 'x', 'y', 'z', 0, -1);
1365 iso2022_read_test(TESTSTR("\x1b-A\x1b%/2\x80\x89"
1366 "big5-0\2\xa1\x40\xa1\x40"),
1367 0x3000, 0xa1, 0x40, 0, -1);
1368 /* Emacs Big5-in-ISO-2022 mapping */
1369 iso2022_read_test(TESTSTR("\x1b$(0&x86\x1b(B \x1b$(0DeBv"),
1370 0x5143, 0x6c23, ' ', ' ', 0x958b, 0x767c, 0, -1);
1371 /* Test from RFC 1922 (ISO-2022-CN) */
1372 iso2022_read_test(TESTSTR("\x1b$)A\x0e=;;;\x1b$)GG(_P\x0f"),
1373 0x4EA4, 0x6362, 0x4EA4, 0x63db, 0, -1);
1375 printf("read tests completed\n");
1376 printf("total: %d errors\n", total_errs
);
1377 return (total_errs
!= 0);
1380 #endif /* TESTMODE */
1382 #else /* ENUM_CHARSETS */
1384 ENUM_CHARSET(CS_ISO2022
)