[sgt/charset] / iso2022.c

/*
 * iso2022.c - support for ISO/IEC 2022 (alias ECMA-35).
 *
 * This isn't a complete implementation of ISO/IEC 2022, but it's
 * close.  It only handles decoding, because a fully general encoder
 * isn't really useful.  It can decode 8-bit and 7-bit versions, with
 * support for single-byte and multi-byte character sets, all four
 * containers (G0, G1, G2, and G3), using both single-shift and
 * locking-shift sequences.
 *
 * The general principle is that any valid ISO/IEC 2022 sequence
 * should either be correctly decoded or should emit an ERROR.  The
 * only exception to this is that the C0 and C1 sets are fixed as
 * those of ISO/IEC 6429.  Escape sequences for designating control
 * sets are passed through, so a post-processor could fix them up if
 * necessary.
 *
 * DOCS to UTF-8 works.  Other DOCS sequences are ignored, which will
 * produce surprising results.
 */

#ifndef ENUM_CHARSETS

#include <assert.h>

#include "charset.h"
#include "internal.h"
#include "sbcsdat.h"

#define LS1 (0x0E)
#define LS0 (0x0F)
#define ESC (0x1B)
#define SS2 (0x8E)
#define SS3 (0x8F)

enum {S4, S6, M4, M6};

static long int null_dbcs_to_unicode(int, int);

const struct iso2022_subcharset {
    char type, i, f;
    int offset;
    const sbcs_data *sbcs_base;
    long int (*dbcs_fn)(int, int);
} iso2022_subcharsets[] = {
    { S4, 0, '0', 0x00, &sbcsdata_CS_DEC_GRAPHICS },
    { S4, 0, '<', 0x80, &sbcsdata_CS_DEC_MCS },
    { S4, 0, 'A', 0x00, &sbcsdata_CS_BS4730 },
    { S4, 0, 'B', 0x00, &sbcsdata_CS_ASCII },
    { S4, 0, 'I', 0x80, &sbcsdata_CS_JISX0201 },
    { S4, 0, 'J', 0x00, &sbcsdata_CS_JISX0201 },
    { S4, 0, '~' },
    { S6, 0, 'A', 0x80, &sbcsdata_CS_ISO8859_1 },
    { S6, 0, 'B', 0x80, &sbcsdata_CS_ISO8859_2 },
    { S6, 0, 'C', 0x80, &sbcsdata_CS_ISO8859_3 },
    { S6, 0, 'D', 0x80, &sbcsdata_CS_ISO8859_4 },
    { S6, 0, 'F', 0x80, &sbcsdata_CS_ISO8859_7 },
    { S6, 0, 'G', 0x80, &sbcsdata_CS_ISO8859_6 },
    { S6, 0, 'H', 0x80, &sbcsdata_CS_ISO8859_8 },
    { S6, 0, 'L', 0x80, &sbcsdata_CS_ISO8859_5 },
    { S6, 0, 'M', 0x80, &sbcsdata_CS_ISO8859_9 },
    { S6, 0, 'T', 0x80, &sbcsdata_CS_ISO8859_11 },
    { S6, 0, 'V', 0x80, &sbcsdata_CS_ISO8859_10 },
    { S6, 0, 'Y', 0x80, &sbcsdata_CS_ISO8859_13 },
    { S6, 0, '_', 0x80, &sbcsdata_CS_ISO8859_14 },
    { S6, 0, 'b', 0x80, &sbcsdata_CS_ISO8859_15 },
    { S6, 0, 'f', 0x80, &sbcsdata_CS_ISO8859_16 },
    { S6, 0, '~' }, /* empty 96-set */
#if 0
    { M4, 0, '@' }, /* JIS C 6226-1978 */
#endif
    { M4, 0, 'A', -0x21, 0, &gb2312_to_unicode },
    { M4, 0, 'B', -0x21, 0, &jisx0208_to_unicode },
    { M4, 0, 'C', -0x21, 0, &ksx1001_to_unicode },
    { M4, 0, 'D', -0x21, 0, &jisx0212_to_unicode },
    { M4, 0, '~', 0, 0, &null_dbcs_to_unicode }, /* empty 94^n-set */
    { M6, 0, '~', 0, 0, &null_dbcs_to_unicode }, /* empty 96^n-set */
};

static long int null_dbcs_to_unicode(int r, int c)
{
    return ERROR;
}

/* States, or "what we're currently accumulating". */
enum {
    IDLE,	/* None of the below */
    SS2CHAR,	/* Accumulating a character after SS2 */
    SS3CHAR,	/* Accumulating a character after SS3 */
    ESCSEQ,	/* Accumulating an escape sequence */
    ESCDROP,	/* Discarding an escape sequence */
    ESCPASS,	/* Passing through an escape sequence */
    DOCSUTF8	/* DOCSed into UTF-8 */
};

#if 1
#include <stdio.h>
static void dump_state(charset_state *s)
{
    unsigned s0 = s->s0, s1 = s->s1;
    char const * const modes[] = { "IDLE", "SS2CHAR", "SS3CHAR",
				   "ESCSEQ", "ESCDROP", "ESCPASS",
				   "DOCSUTF8" };

    fprintf(stderr, "s0: %s", modes[s0 >> 29]);
    fprintf(stderr, " %02x %02x %02x   ", (s0 >> 16) & 0xff, (s0 >> 8) & 0xff,
	    s0 & 0xff);
    fprintf(stderr, "s1: LS%d LS%dR", (s1 >> 30) & 3, (s1 >> 28) & 3);
    fprintf(stderr, " %d %d %d %d\n", s1 & 0x7f, (s1 >> 7) & 0x7f,
	    (s1 >> 14) & 0x7f, (s1 >> 21) & 0x7f);
}
#endif

static void designate(charset_state *state, int container,
		      int type, int ibyte, int fbyte)
{
    unsigned long i;

    assert(container >= 0 && container <= 3);
    assert(type == S4 || type == S6 || type == M4 || type == M6);

    for (i = 0; i <= lenof(iso2022_subcharsets); i++) {
	if (iso2022_subcharsets[i].type == type &&
	    iso2022_subcharsets[i].i == ibyte &&
	    iso2022_subcharsets[i].f == fbyte) {
	    state->s1 &= ~(0x7fL << (container * 7));
	    state->s1 |= (i << (container * 7));
	    return;
	}
    }
    /*
     * If we don't find the charset, invoke the empty one, so we
     * output ERROR rather than garbage.
     */
    designate(state, container, type, 0, '~');
}

static void do_utf8(long int input_chr,
		    charset_state *state,
		    void (*emit)(void *ctx, long int output),
		    void *emitctx)
{
    charset_state ustate;
    charset_spec const *utf8;

    ustate.s1 = 0;
    ustate.s0 = state->s0 & 0x03ffffffL;
    read_utf8(NULL, input_chr, &ustate, emit, emitctx);
    state->s0 = (state->s0 & ~0x03ffffffL) | (ustate.s0 & 0x03ffffffL);
}

static void docs_utf8(long int input_chr,
		      charset_state *state,
		      void (*emit)(void *ctx, long int output),
		      void *emitctx)
{
    int retstate;

    /*
     * Bits [25:0] of s0 are reserved for read_utf8().
     * Bits [27:26] are a tiny state machine to recognise ESC % @.
     */
    retstate = (state->s0 & 0x0c000000L) >> 26;
    if (retstate == 1 && input_chr == '%')
	retstate = 2;
    else if (retstate == 2 && input_chr == '@') {
	/* If we've got a partial UTF-8 sequence, complain. */
	if (state->s0 & 0x03ffffffL)
	    emit(emitctx, ERROR);
	state->s0 = 0;
	return;
    } else {
	if (retstate >= 1) do_utf8(ESC, state, emit, emitctx);
	if (retstate >= 2) do_utf8('%', state, emit, emitctx);
	retstate = 0;
	if (input_chr == ESC)
	    retstate = 1;
	else {
	    do_utf8(input_chr, state, emit, emitctx);
	}
    }
    state->s0 = (state->s0 & ~0x0c000000L) | (retstate << 26);
}


static void read_iso2022(charset_spec const *charset, long int input_chr,
			  charset_state *state,
			  void (*emit)(void *ctx, long int output),
			  void *emitctx)
{

    /* dump_state(state); */
    /*
     * We have to make fairly efficient use of the 64 bits of state
     * available to us.  Long-term state goes in s1, and consists of
     * the identities of the character sets designated as G0/G1/G2/G3
     * and the locking-shift states for GL and GR.  Short-term state
     * goes in s0: The bottom half of s0 accumulates characters for an
     * escape sequence or a multi-byte character, while the top three
     * bits indicate what they're being accumulated for.  After DOCS,
     * the bottom 29 bits of state are available for the DOCS function
     * to use -- the UTF-8 one uses the bottom 26 for UTF-8 decoding
     * and the top two to recognised ESC % @.
     *
     * s0[31:29] = state enum
     * s0[24:0] = accumulated bytes
     * s1[31:30] = GL locking-shift state
     * s1[29:28] = GR locking-shift state
     * s1[27:21] = G3 charset
     * s1[20:14] = G2 charset
     * s1[13:7] = G1 charset
     * s1[6:0] = G0 charset
     */

#define LEFT 30
#define RIGHT 28
#define LOCKING_SHIFT(n,side) \
	(state->s1 = (state->s1 & ~(3L<<(side))) | ((n ## L)<<(side)))
#define MODE ((state->s0 & 0xe0000000L) >> 29)
#define ENTER_MODE(m) (state->s0 = (state->s0 & ~0xe0000000L) | ((m)<<29))
#define SINGLE_SHIFT(n) ENTER_MODE(SS2CHAR - 2 + (n))
#define ASSERT_IDLE do {						\
	if (state->s0 != 0) emit(emitctx, ERROR);			\
	state->s0 = 0;							\
} while (0)

    if (state->s1 == 0) {
	/*
	 * Since there's no LS0R, this means we must just have started.
	 * Set up a sane initial state (LS0, LS1R, ASCII in G0/G1/G2/G3).
	 */
	LOCKING_SHIFT(0, LEFT);
	LOCKING_SHIFT(1, RIGHT);
	designate(state, 0, S4, 0, 'B');
	designate(state, 1, S4, 0, 'B');
	designate(state, 2, S4, 0, 'B');
	designate(state, 3, S4, 0, 'B');
    }

    if (MODE == DOCSUTF8) {
	docs_utf8(input_chr, state, emit, emitctx);
	return;
    }

    if ((input_chr & 0x60) == 0x00) {
	/* C0 or C1 control */
	ASSERT_IDLE;
	switch (input_chr) {
	  case ESC:
	    ENTER_MODE(ESCSEQ);
	    break;
	  case LS0:
	    LOCKING_SHIFT(0, LEFT);
	    break;
	  case LS1:
	    LOCKING_SHIFT(1, LEFT);
	    break;
	  case SS2:
	    SINGLE_SHIFT(2);
	    break;
	  case SS3:
	    SINGLE_SHIFT(3);
	    break;
	  default:
	    emit(emitctx, input_chr);
	    break;
	}
    } else if ((input_chr & 0x80) || MODE < ESCSEQ) {
	int is_gl = 0;
	struct iso2022_subcharset const *subcs;
	unsigned container;
	long input_7bit;
	/*
	 * Actual data.
	 * Force idle state if we're in mid escape sequence, or in a
	 * multi-byte character with a different top bit.
	 */
	if (MODE >= ESCSEQ ||
	    ((state->s0 & 0x00ff0000L) != 0 && 
	     (((state->s0 >> 16) ^ input_chr) & 0x80)))
	    ASSERT_IDLE;
	if (MODE == SS2CHAR || MODE == SS3CHAR) /* Single-shift */
	    container = MODE - SS2CHAR + 2;
	else if (input_chr >= 0x80) /* GR */
	    container = (state->s1 >> 28) & 3;
	else { /* GL */
	    container = state->s1 >> 30;
	    is_gl = 1;
	}
	input_7bit = input_chr & ~0x80;
	subcs = &iso2022_subcharsets[(state->s1 >> (container * 7)) & 0x7f];
	if ((subcs->type == S4 || subcs->type == M4) &&
	    (input_7bit == 0x20 || input_7bit == 0x7f)) {
	    /* characters not in 94-char set */
	    if (is_gl) emit(emitctx, input_7bit);
	    else emit(emitctx, ERROR);
	} else if (subcs->type == M4 || subcs->type == M6) {
	    if ((state->s0 & 0x00ff0000L) == 0) {
		state->s0 |= input_chr << 16;
		return;
	    } else {
		emit(emitctx,
		     subcs->dbcs_fn(((state->s0 >> 16) & 0x7f) + subcs->offset,
				    input_7bit + subcs->offset));
	    }
	} else {
	    if ((state->s0 & 0x00ff0000L) != 0)
		emit(emitctx, ERROR);
	    emit(emitctx, subcs->sbcs_base ?
		 sbcs_to_unicode(subcs->sbcs_base, input_7bit + subcs->offset):
		 ERROR);
	}
	state->s0 = 0;
    } else {
	unsigned i1, i2;
	if (MODE == ESCPASS) {
	    emit(emitctx, input_chr);
	    if ((input_chr & 0xf0) != 0x20)
		ENTER_MODE(IDLE);
	    return;
	}

	/*
	 * Intermediate bytes shall be any of the 16 positions of
	 * column 02 of the code table; they are denoted by the symbol
	 * I.
	 */
	if ((input_chr & 0xf0) == 0x20) {
	    if (((state->s0 >> 16) & 0xff) == 0)
		state->s0 |= input_chr << 16;
	    else if (((state->s0 >> 8) & 0xff) == 0)
		state->s0 |= input_chr << 8;
	    else {
		/* Long escape sequence.  Switch to ESCPASS or ESCDROP. */
		i1 = (state->s0 >> 16) & 0xff;
		i2 = (state->s0 >> 8) & 0xff;
		switch (i1) {
		  case '(': case ')': case '*': case '+':
		  case '-': case '.': case '/':
		  case '$':
		    ENTER_MODE(ESCDROP);
		    break;
		  default:
		    emit(emitctx, ESC);
		    emit(emitctx, i1);
		    emit(emitctx, i2);
		    emit(emitctx, input_chr);
		    state->s0 = 0;
		    ENTER_MODE(ESCPASS);
		    break;
		}
	    }
	    return;
	}

	/*
	 * Final bytes shall be any of the 79 positions of columns 03
	 * to 07 of the code table excluding position 07/15; they are
	 * denoted by the symbol F.
	 */
	i1 = (state->s0 >> 16) & 0xff;
	i2 = (state->s0 >> 8) & 0xff;
	if (MODE == ESCDROP)
	    input_chr = 0; /* Make sure it won't match. */
	state->s0 = 0;
	switch (i1) {
	  case 0: /* No intermediate bytes */
	    switch (input_chr) {
	      case 'N': /* SS2 */
		SINGLE_SHIFT(2);
		break;
	      case 'O': /* SS3 */
		SINGLE_SHIFT(3);
		break;
	      case 'n': /* LS2 */
		LOCKING_SHIFT(2, LEFT);
		break;
	      case 'o': /* LS3 */
		LOCKING_SHIFT(3, LEFT);
		break;
	      case '|': /* LS3R */
		LOCKING_SHIFT(3, RIGHT);
		break;
	      case '}': /* LS2R */
		LOCKING_SHIFT(2, RIGHT);
		break;
	      case '~': /* LS1R */
		LOCKING_SHIFT(1, RIGHT);
		break;
	      default:
		/* Unsupported escape sequence.  Spit it back out. */
		emit(emitctx, ESC);
		emit(emitctx, input_chr);
	    }
	    break;
	  case ' ': /* ACS */
	    /*
	     * Various coding structure facilities specify that designating
	     * a code element also invokes it.  As far as I can see, invoking
	     * it now will have the same practical effect, since those
	     * facilities also ban the use of locking shifts.
	     */
	    switch (input_chr) {
	      case 'A': /* G0 element used and invoked into GL */
		LOCKING_SHIFT(0, LEFT);
		break;
	      case 'C': /* G0 in GL, G1 in GR */
	      case 'D': /* Ditto, at least for 8-bit codes */
	      case 'L': /* ISO 4873 (ECMA-43) level 1 */
	      case 'M': /* ISO 4873 (ECMA-43) level 2 */
		LOCKING_SHIFT(0, LEFT);
		LOCKING_SHIFT(1, RIGHT);
		break;
	    }
	    break;
	  case '&': /* IRR */
	    /*
	     * IRR (Identify Revised Registration) is ignored here,
	     * since any revised registration must be
	     * upward-compatible with the old one, so either we'll
	     * support the new one or we'll emit ERROR when we run
	     * into a new character.  In either case, there's nothing
	     * to be done here.
	     */
	    break;
	  case '(': /* GZD4 */  case ')': /* G1D4 */
	  case '*': /* G2D4 */  case '+': /* G3D4 */
	    designate(state, i1 - '(', S4, i2, input_chr);
	    break;
	  case '-': /* G1D6 */  case '.': /* G2D6 */  case '/': /* G3D6 */
	    designate(state, i1 - ',', S6, i2, input_chr);
	    break;
	  case '$': /* G?DM? */
	    switch (i2) {
	      case 0: /* Obsolete version of GZDM4 */
		i2 = '(';
	      case '(': /* GZDM4 */  case ')': /* G1DM4 */
	      case '*': /* G2DM4 */  case '+': /* G3DM4 */
		designate(state, i2 - '(', M4, 0, input_chr);
		break;
	      case '-': /* G1DM6 */
	      case '.': /* G2DM6 */  case '/': /* G3DM6 */
		designate(state, i2 - ',', M6, 0, input_chr);
		break;
	      default:
		emit(emitctx, ERROR);
		break;
	    }
	  case '%': /* DOCS */
	    /* XXX What's a reasonable way to handle an unrecognised DOCS? */
	    switch (i2) {
	      case 0:
		switch (input_chr) {
		  case 'G':
		    ENTER_MODE(DOCSUTF8);
		    break;
		}
		break;
	    }
	    break;
	  default:
	    /* Unsupported nF escape sequence.  Re-emit it. */
	    emit(emitctx, ESC);
	    emit(emitctx, i1);
	    if (i2) emit(emitctx, i2);
	    emit(emitctx, input_chr);
	    break;
	}
    }
}

static int write_iso2022(charset_spec const *charset, long int input_chr,
			 charset_state *state,
			 void (*emit)(void *ctx, long int output),
			 void *emitctx)
{
    return FALSE;
}

const charset_spec charset_CS_ISO2022 = {
    CS_ISO2022, read_iso2022, write_iso2022, NULL
};

#ifdef TESTMODE

#include <stdio.h>
#include <stdarg.h>
#include <string.h>

int total_errs = 0;

void iso2022_emit(void *ctx, long output)
{
    wchar_t **p = (wchar_t **)ctx;
    *(*p)++ = output;
}

void iso2022_read_test(int line, char *input, int inlen, ...)
{
    va_list ap;
    wchar_t *p, str[512];
    int i;
    charset_state state;
    unsigned long l;

    state.s0 = state.s1 = 0;
    p = str;

    for (i = 0; i < inlen; i++)
	read_iso2022(NULL, input[i] & 0xFF, &state, iso2022_emit, &p);

    va_start(ap, inlen);
    l = 0;
    for (i = 0; i < p - str; i++) {
	l = va_arg(ap, long int);
	if (l == -1) {
	    printf("%d: correct string shorter than output\n", line);
	    total_errs++;
	    break;
	}
	if (l != str[i]) {
	    printf("%d: char %d came out as %08x, should be %08lx\n",
		    line, i, str[i], l);
	    total_errs++;
	}
    }
    if (l != -1) {
	l = va_arg(ap, long int);
	if (l != -1) {
	    printf("%d: correct string longer than output\n", line);
	    total_errs++;
	}
    }
    va_end(ap);
}

/* Macro to concoct the first three parameters of iso2022_read_test. */
#define TESTSTR(x) __LINE__, x, lenof(x)

int main(void)
{
    printf("read tests beginning\n");
    /* Simple test (Emacs sample text for Japanese, in ISO-2022-JP) */
    iso2022_read_test(TESTSTR("Japanese (\x1b$BF|K\\8l\x1b(B)\t"
			      "\x1b$B$3$s$K$A$O\x1b(B, "
			      "\x1b$B%3%s%K%A%O\x1b(B\n"),
		      'J','a','p','a','n','e','s','e',' ','(',
		      0x65E5, 0x672C, 0x8A9E, ')', '\t',
		      0x3053, 0x3093, 0x306b, 0x3061, 0x306f, ',', ' ',
		      0x30b3, 0x30f3, 0x30cb, 0x30c1, 0x30cf, '\n', 0, -1);
    /* Same thing in EUC-JP (with designations, and half-width katakana) */
    iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D"
			      "Japanese (\xc6\xfc\xcb\xdc\xb8\xec)\t"
			      "\xa4\xb3\xa4\xf3\xa4\xcb\xa4\xc1\xa4\xcf, "
			      "\x8e\xba\x8e\xdd\x8e\xc6\x8e\xc1\x8e\xca\n"),
		      'J','a','p','a','n','e','s','e',' ','(',
		      0x65E5, 0x672C, 0x8A9E, ')', '\t',
		      0x3053, 0x3093, 0x306b, 0x3061, 0x306f, ',', ' ',
		      0xff7a, 0xff9d, 0xff86, 0xff81, 0xff8a, '\n', 0, -1);
    /* Multibyte single-shift */
    iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D\x8f\"/!"),
		      0x02D8, '!', 0, -1);
    /* Non-existent SBCS */
    iso2022_read_test(TESTSTR("\x1b(!Zfnord\n"),
		      ERROR, ERROR, ERROR, ERROR, ERROR, '\n', 0, -1);
    /* Pass-through of ordinary escape sequences, including a long one */
    iso2022_read_test(TESTSTR("\x1b""b\x1b#5\x1b#!!!5"),
		      0x1B, 'b', 0x1B, '#', '5',
		      0x1B, '#', '!', '!', '!', '5', 0, -1);
    /* Non-existent DBCS (also 5-byte escape sequence) */
    iso2022_read_test(TESTSTR("\x1b$(!Bfnord!"),
		      ERROR, ERROR, ERROR, 0, -1);
    /* Incomplete DB characters */
    iso2022_read_test(TESTSTR("\x1b$B(,(\x1b(BHi\x1b$B(,(\n"),
		      0x2501, ERROR, 'H', 'i', 0x2501, ERROR, '\n', 0, -1);
    iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D\xa4""B"),
		      ERROR, 'B', 0, -1);
    iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D\x0e\x1b|$\xa2\xaf"),
		      ERROR, 0x02D8, 0, -1);
    /* Incomplete escape sequence */
    iso2022_read_test(TESTSTR("\x1b\n"), ERROR, '\n', 0, -1);
    iso2022_read_test(TESTSTR("\x1b-A\x1b~\x1b\xa1"), ERROR, 0xa1, 0, -1);
    /* Incomplete single-shift */
    iso2022_read_test(TESTSTR("\x8e\n"), ERROR, '\n', 0, -1);
    iso2022_read_test(TESTSTR("\x1b$*B\x8e(\n"), ERROR, '\n', 0, -1);
    /* Corner cases (02/00 and 07/15) */
    iso2022_read_test(TESTSTR("\x1b(B\x20\x7f"), 0x20, 0x7f, 0, -1);
    iso2022_read_test(TESTSTR("\x1b(I\x20\x7f"), 0x20, 0x7f, 0, -1);
    iso2022_read_test(TESTSTR("\x1b$B\x20\x7f"), 0x20, 0x7f, 0, -1);
    iso2022_read_test(TESTSTR("\x1b-A\x0e\x20\x7f"), 0xa0, 0xff, 0, -1);
    iso2022_read_test(TESTSTR("\x1b$-~\x0e\x20\x7f"), ERROR, 0, -1);
    iso2022_read_test(TESTSTR("\x1b)B\xa0\xff"), ERROR, ERROR, 0, -1);
    iso2022_read_test(TESTSTR("\x1b)I\xa0\xff"), ERROR, ERROR, 0, -1);
    iso2022_read_test(TESTSTR("\x1b$)B\xa0\xff"), ERROR, ERROR, 0, -1);
    iso2022_read_test(TESTSTR("\x1b-A\x1b~\xa0\xff"), 0xa0, 0xff, 0, -1);
    iso2022_read_test(TESTSTR("\x1b$-~\x1b~\xa0\xff"), ERROR, 0, -1);
    /* Designate control sets */
    iso2022_read_test(TESTSTR("\x1b!@"), 0x1b, '!', '@', 0, -1);
    /* Designate other coding system */
    iso2022_read_test(TESTSTR("\x1b%G"
			      "\xCE\xBA\xE1\xBD\xB9\xCF\x83\xCE\xBC\xCE\xB5"),
		      0x03BA, 0x1F79, 0x03C3, 0x03BC, 0x03B5, 0, -1);
    iso2022_read_test(TESTSTR("\x1b-A\x1b%G\xCE\xBA\x1b%@\xa0"),
		      0x03BA, 0xA0, 0, -1);
    iso2022_read_test(TESTSTR("\x1b%G\xCE\x1b%@"), ERROR, 0, -1);
    iso2022_read_test(TESTSTR("\x1b%G\xCE\xBA\x1b%\x1b%@"),
		      0x03BA, 0x1B, '%', 0, -1);
    printf("read tests completed\n");
    printf("total: %d errors\n", total_errs);
    return (total_errs != 0);
}

#endif /* TESTMODE */

#else /* ENUM_CHARSETS */

ENUM_CHARSET(CS_ISO2022)

#endif
Commit	Line	Data
b97e5427	1	/*
	2	* iso2022.c - support for ISO/IEC 2022 (alias ECMA-35).
	3	*
	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.
	10	*
	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
	16	* necessary.
	17	*
c69732bb	18	* DOCS to UTF-8 works. Other DOCS sequences are ignored, which will
c69732bb	19	* produce surprising results.
b97e5427	20	*/
	21
	22	#ifndef ENUM_CHARSETS
	23
	24	#include <assert.h>
	25
	26	#include "charset.h"
	27	#include "internal.h"
	28	#include "sbcsdat.h"
	29
	30	#define LS1 (0x0E)
	31	#define LS0 (0x0F)
	32	#define ESC (0x1B)
	33	#define SS2 (0x8E)
	34	#define SS3 (0x8F)
	35
	36	enum {S4, S6, M4, M6};
	37
	38	static long int null_dbcs_to_unicode(int, int);
	39
	40	const struct iso2022_subcharset {
	41	char type, i, f;
	42	int offset;
	43	const sbcs_data *sbcs_base;
	44	long int (*dbcs_fn)(int, int);
	45	} iso2022_subcharsets[] = {
294941fa	46	{ S4, 0, '0', 0x00, &sbcsdata_CS_DEC_GRAPHICS },
b97e5427	47	{ S4, 0, '<', 0x80, &sbcsdata_CS_DEC_MCS },
294941fa	48	{ S4, 0, 'A', 0x00, &sbcsdata_CS_BS4730 },
294941fa	49	{ S4, 0, 'B', 0x00, &sbcsdata_CS_ASCII },
b97e5427	50	{ S4, 0, 'I', 0x80, &sbcsdata_CS_JISX0201 },
	51	{ S4, 0, 'J', 0x00, &sbcsdata_CS_JISX0201 },
	52	{ S4, 0, '~' },
	53	{ S6, 0, 'A', 0x80, &sbcsdata_CS_ISO8859_1 },
	54	{ S6, 0, 'B', 0x80, &sbcsdata_CS_ISO8859_2 },
	55	{ S6, 0, 'C', 0x80, &sbcsdata_CS_ISO8859_3 },
	56	{ S6, 0, 'D', 0x80, &sbcsdata_CS_ISO8859_4 },
	57	{ S6, 0, 'F', 0x80, &sbcsdata_CS_ISO8859_7 },
	58	{ S6, 0, 'G', 0x80, &sbcsdata_CS_ISO8859_6 },
	59	{ S6, 0, 'H', 0x80, &sbcsdata_CS_ISO8859_8 },
	60	{ S6, 0, 'L', 0x80, &sbcsdata_CS_ISO8859_5 },
	61	{ S6, 0, 'M', 0x80, &sbcsdata_CS_ISO8859_9 },
	62	{ S6, 0, 'T', 0x80, &sbcsdata_CS_ISO8859_11 },
	63	{ S6, 0, 'V', 0x80, &sbcsdata_CS_ISO8859_10 },
	64	{ S6, 0, 'Y', 0x80, &sbcsdata_CS_ISO8859_13 },
	65	{ S6, 0, '_', 0x80, &sbcsdata_CS_ISO8859_14 },
	66	{ S6, 0, 'b', 0x80, &sbcsdata_CS_ISO8859_15 },
	67	{ S6, 0, 'f', 0x80, &sbcsdata_CS_ISO8859_16 },
	68	{ S6, 0, '~' }, /* empty 96-set */
	69	#if 0
	70	{ M4, 0, '@' }, /* JIS C 6226-1978 */
	71	#endif
	72	{ M4, 0, 'A', -0x21, 0, &gb2312_to_unicode },
	73	{ M4, 0, 'B', -0x21, 0, &jisx0208_to_unicode },
	74	{ M4, 0, 'C', -0x21, 0, &ksx1001_to_unicode },
	75	{ M4, 0, 'D', -0x21, 0, &jisx0212_to_unicode },
	76	{ M4, 0, '~', 0, 0, &null_dbcs_to_unicode }, /* empty 94^n-set */
	77	{ M6, 0, '~', 0, 0, &null_dbcs_to_unicode }, /* empty 96^n-set */
	78	};
	79
	80	static long int null_dbcs_to_unicode(int r, int c)
	81	{
	82	return ERROR;
	83	}
	84
	85	/* States, or "what we're currently accumulating". */
	86	enum {
	87	IDLE, /* None of the below */
	88	SS2CHAR, /* Accumulating a character after SS2 */
	89	SS3CHAR, /* Accumulating a character after SS3 */
	90	ESCSEQ, /* Accumulating an escape sequence */
	91	ESCDROP, /* Discarding an escape sequence */
a89fe3cf	92	ESCPASS, /* Passing through an escape sequence */
a89fe3cf	93	DOCSUTF8 /* DOCSed into UTF-8 */
b97e5427	94	};
b97e5427	95
a89fe3cf	96	#if 1
b97e5427	97	#include <stdio.h>
	98	static void dump_state(charset_state *s)
	99	{
	100	unsigned s0 = s->s0, s1 = s->s1;
	101	char const * const modes[] = { "IDLE", "SS2CHAR", "SS3CHAR",
a89fe3cf	102	"ESCSEQ", "ESCDROP", "ESCPASS",
a89fe3cf	103	"DOCSUTF8" };
b97e5427	104
	105	fprintf(stderr, "s0: %s", modes[s0 >> 29]);
	106	fprintf(stderr, " %02x %02x %02x ", (s0 >> 16) & 0xff, (s0 >> 8) & 0xff,
	107	s0 & 0xff);
	108	fprintf(stderr, "s1: LS%d LS%dR", (s1 >> 30) & 3, (s1 >> 28) & 3);
	109	fprintf(stderr, " %d %d %d %d\n", s1 & 0x7f, (s1 >> 7) & 0x7f,
	110	(s1 >> 14) & 0x7f, (s1 >> 21) & 0x7f);
	111	}
	112	#endif
	113
	114	static void designate(charset_state *state, int container,
	115	int type, int ibyte, int fbyte)
	116	{
	117	unsigned long i;
	118
	119	assert(container >= 0 && container <= 3);
	120	assert(type == S4 \|\| type == S6 \|\| type == M4 \|\| type == M6);
	121
	122	for (i = 0; i <= lenof(iso2022_subcharsets); i++) {
	123	if (iso2022_subcharsets[i].type == type &&
	124	iso2022_subcharsets[i].i == ibyte &&
	125	iso2022_subcharsets[i].f == fbyte) {
	126	state->s1 &= ~(0x7fL << (container * 7));
	127	state->s1 \|= (i << (container * 7));
	128	return;
	129	}
	130	}
	131	/*
	132	* If we don't find the charset, invoke the empty one, so we
	133	* output ERROR rather than garbage.
	134	*/
	135	designate(state, container, type, 0, '~');
	136	}
	137
a89fe3cf	138	static void do_utf8(long int input_chr,
	139	charset_state *state,
	140	void (emit)(void ctx, long int output),
	141	void *emitctx)
	142	{
	143	charset_state ustate;
	144	charset_spec const *utf8;
	145
	146	ustate.s1 = 0;
	147	ustate.s0 = state->s0 & 0x03ffffffL;
7a7dc0a7	148	read_utf8(NULL, input_chr, &ustate, emit, emitctx);
a89fe3cf	149	state->s0 = (state->s0 & ~0x03ffffffL) \| (ustate.s0 & 0x03ffffffL);
	150	}
	151
	152	static void docs_utf8(long int input_chr,
	153	charset_state *state,
	154	void (emit)(void ctx, long int output),
	155	void *emitctx)
	156	{
	157	int retstate;
	158
	159	/*
	160	* Bits [25:0] of s0 are reserved for read_utf8().
	161	* Bits [27:26] are a tiny state machine to recognise ESC % @.
	162	*/
	163	retstate = (state->s0 & 0x0c000000L) >> 26;
	164	if (retstate == 1 && input_chr == '%')
	165	retstate = 2;
	166	else if (retstate == 2 && input_chr == '@') {
	167	/* If we've got a partial UTF-8 sequence, complain. */
	168	if (state->s0 & 0x03ffffffL)
	169	emit(emitctx, ERROR);
	170	state->s0 = 0;
	171	return;
	172	} else {
	173	if (retstate >= 1) do_utf8(ESC, state, emit, emitctx);
	174	if (retstate >= 2) do_utf8('%', state, emit, emitctx);
	175	retstate = 0;
	176	if (input_chr == ESC)
	177	retstate = 1;
	178	else {
	179	do_utf8(input_chr, state, emit, emitctx);
	180	}
	181	}
	182	state->s0 = (state->s0 & ~0x0c000000L) \| (retstate << 26);
	183	}
	184
	185
b97e5427	186	static void read_iso2022(charset_spec const *charset, long int input_chr,
	187	charset_state *state,
	188	void (emit)(void ctx, long int output),
	189	void *emitctx)
	190	{
	191
a89fe3cf	192	/* dump_state(state); */
b97e5427	193	/*
04c24cbb	194	* We have to make fairly efficient use of the 64 bits of state
0fab6a2b	195	* available to us. Long-term state goes in s1, and consists of
04c24cbb	196	* the identities of the character sets designated as G0/G1/G2/G3
04c24cbb	197	* and the locking-shift states for GL and GR. Short-term state
0fab6a2b	198	* goes in s0: The bottom half of s0 accumulates characters for an
04c24cbb	199	* escape sequence or a multi-byte character, while the top three
	200	* bits indicate what they're being accumulated for. After DOCS,
	201	* the bottom 29 bits of state are available for the DOCS function
	202	* to use -- the UTF-8 one uses the bottom 26 for UTF-8 decoding
	203	* and the top two to recognised ESC % @.
b97e5427	204	*
	205	* s0[31:29] = state enum
	206	* s0[24:0] = accumulated bytes
	207	* s1[31:30] = GL locking-shift state
	208	* s1[29:28] = GR locking-shift state
	209	* s1[27:21] = G3 charset
	210	* s1[20:14] = G2 charset
	211	* s1[13:7] = G1 charset
	212	* s1[6:0] = G0 charset
	213	*/
	214
	215	#define LEFT 30
	216	#define RIGHT 28
	217	#define LOCKING_SHIFT(n,side) \
	218	(state->s1 = (state->s1 & ~(3L<<(side))) \| ((n ## L)<<(side)))
	219	#define MODE ((state->s0 & 0xe0000000L) >> 29)
	220	#define ENTER_MODE(m) (state->s0 = (state->s0 & ~0xe0000000L) \| ((m)<<29))
	221	#define SINGLE_SHIFT(n) ENTER_MODE(SS2CHAR - 2 + (n))
	222	#define ASSERT_IDLE do { \
	223	if (state->s0 != 0) emit(emitctx, ERROR); \
	224	state->s0 = 0; \
	225	} while (0)
	226
	227	if (state->s1 == 0) {
	228	/*
	229	* Since there's no LS0R, this means we must just have started.
	230	* Set up a sane initial state (LS0, LS1R, ASCII in G0/G1/G2/G3).
	231	*/
	232	LOCKING_SHIFT(0, LEFT);
	233	LOCKING_SHIFT(1, RIGHT);
	234	designate(state, 0, S4, 0, 'B');
	235	designate(state, 1, S4, 0, 'B');
	236	designate(state, 2, S4, 0, 'B');
	237	designate(state, 3, S4, 0, 'B');
	238	}
	239
a89fe3cf	240	if (MODE == DOCSUTF8) {
	241	docs_utf8(input_chr, state, emit, emitctx);
	242	return;
	243	}
	244
b97e5427	245	if ((input_chr & 0x60) == 0x00) {
	246	/* C0 or C1 control */
	247	ASSERT_IDLE;
	248	switch (input_chr) {
	249	case ESC:
	250	ENTER_MODE(ESCSEQ);
	251	break;
	252	case LS0:
	253	LOCKING_SHIFT(0, LEFT);
	254	break;
	255	case LS1:
	256	LOCKING_SHIFT(1, LEFT);
	257	break;
	258	case SS2:
	259	SINGLE_SHIFT(2);
	260	break;
	261	case SS3:
	262	SINGLE_SHIFT(3);
	263	break;
	264	default:
	265	emit(emitctx, input_chr);
	266	break;
	267	}
	268	} else if ((input_chr & 0x80) \|\| MODE < ESCSEQ) {
	269	int is_gl = 0;
	270	struct iso2022_subcharset const *subcs;
	271	unsigned container;
	272	long input_7bit;
	273	/*
	274	* Actual data.
	275	* Force idle state if we're in mid escape sequence, or in a
	276	* multi-byte character with a different top bit.
	277	*/
	278	if (MODE >= ESCSEQ \|\|
	279	((state->s0 & 0x00ff0000L) != 0 &&
	280	(((state->s0 >> 16) ^ input_chr) & 0x80)))
	281	ASSERT_IDLE;
	282	if (MODE == SS2CHAR \|\| MODE == SS3CHAR) /* Single-shift */
	283	container = MODE - SS2CHAR + 2;
	284	else if (input_chr >= 0x80) /* GR */
	285	container = (state->s1 >> 28) & 3;
	286	else { /* GL */
	287	container = state->s1 >> 30;
	288	is_gl = 1;
	289	}
	290	input_7bit = input_chr & ~0x80;
	291	subcs = &iso2022_subcharsets[(state->s1 >> (container * 7)) & 0x7f];
	292	if ((subcs->type == S4 \|\| subcs->type == M4) &&
	293	(input_7bit == 0x20 \|\| input_7bit == 0x7f)) {
	294	/* characters not in 94-char set */
	295	if (is_gl) emit(emitctx, input_7bit);
	296	else emit(emitctx, ERROR);
	297	} else if (subcs->type == M4 \|\| subcs->type == M6) {
	298	if ((state->s0 & 0x00ff0000L) == 0) {
	299	state->s0 \|= input_chr << 16;
	300	return;
	301	} else {
	302	emit(emitctx,
	303	subcs->dbcs_fn(((state->s0 >> 16) & 0x7f) + subcs->offset,
	304	input_7bit + subcs->offset));
	305	}
	306	} else {
	307	if ((state->s0 & 0x00ff0000L) != 0)
	308	emit(emitctx, ERROR);
309	emit(emitctx, subcs->sbcs_base ?
310	sbcs_to_unicode(subcs->sbcs_base, input_7bit + subcs->offset):
311	ERROR);
312	}
313	state->s0 = 0;
314	} else {
315	unsigned i1, i2;
316	if (MODE == ESCPASS) {
317	emit(emitctx, input_chr);
318	if ((input_chr & 0xf0) != 0x20)
319	ENTER_MODE(IDLE);
320	return;
321	}
322
323	/*
324	* Intermediate bytes shall be any of the 16 positions of
325	* column 02 of the code table; they are denoted by the symbol
326	* I.
327	*/
328	if ((input_chr & 0xf0) == 0x20) {
329	if (((state->s0 >> 16) & 0xff) == 0)
330	state->s0 \|= input_chr << 16;
331	else if (((state->s0 >> 8) & 0xff) == 0)
332	state->s0 \|= input_chr << 8;
333	else {
334	/* Long escape sequence. Switch to ESCPASS or ESCDROP. */
335	i1 = (state->s0 >> 16) & 0xff;
336	i2 = (state->s0 >> 8) & 0xff;
337	switch (i1) {
338	case '(': case ')': case '*': case '+':
339	case '-': case '.': case '/':
340	case '$':
341	ENTER_MODE(ESCDROP);
342	break;
343	default:
344	emit(emitctx, ESC);
345	emit(emitctx, i1);
346	emit(emitctx, i2);
347	emit(emitctx, input_chr);
348	state->s0 = 0;
349	ENTER_MODE(ESCPASS);
350	break;
351	}
352	}
353	return;
354	}
355
356	/*
357	* Final bytes shall be any of the 79 positions of columns 03
358	* to 07 of the code table excluding position 07/15; they are
359	* denoted by the symbol F.
360	*/
361	i1 = (state->s0 >> 16) & 0xff;
362	i2 = (state->s0 >> 8) & 0xff;
363	if (MODE == ESCDROP)
364	input_chr = 0; /* Make sure it won't match. */
365	state->s0 = 0;
366	switch (i1) {
367	case 0: /* No intermediate bytes */
368	switch (input_chr) {
369	case 'N': /* SS2 */
370	SINGLE_SHIFT(2);
371	break;
372	case 'O': /* SS3 */
373	SINGLE_SHIFT(3);
374	break;
375	case 'n': /* LS2 */
376	LOCKING_SHIFT(2, LEFT);
377	break;
378	case 'o': /* LS3 */
379	LOCKING_SHIFT(3, LEFT);
380	break;
381	case '\|': /* LS3R */
382	LOCKING_SHIFT(3, RIGHT);
383	break;
384	case '}': /* LS2R */
385	LOCKING_SHIFT(2, RIGHT);
386	break;
387	case '~': /* LS1R */
388	LOCKING_SHIFT(1, RIGHT);
389	break;
390	default:
391	/* Unsupported escape sequence. Spit it back out. */
392	emit(emitctx, ESC);
393	emit(emitctx, input_chr);
394	}
395	break;
396	case ' ': /* ACS */
397	/*
398	* Various coding structure facilities specify that designating
399	* a code element also invokes it. As far as I can see, invoking
400	* it now will have the same practical effect, since those
401	* facilities also ban the use of locking shifts.
402	*/
403	switch (input_chr) {
404	case 'A': /* G0 element used and invoked into GL */
405	LOCKING_SHIFT(0, LEFT);
406	break;
407	case 'C': /* G0 in GL, G1 in GR */
408	case 'D': /* Ditto, at least for 8-bit codes */
409	case 'L': /* ISO 4873 (ECMA-43) level 1 */
410	case 'M': /* ISO 4873 (ECMA-43) level 2 */
411	LOCKING_SHIFT(0, LEFT);
412	LOCKING_SHIFT(1, RIGHT);
413	break;
414	}
415	break;
416	case '&': /* IRR */
417	/*
418	* IRR (Identify Revised Registration) is ignored here,
419	* since any revised registration must be
420	* upward-compatible with the old one, so either we'll
421	* support the new one or we'll emit ERROR when we run
422	* into a new character. In either case, there's nothing
423	* to be done here.
424	*/
425	break;
426	case '(': /* GZD4 / case ')': / G1D4 */
427	case '': / G2D4 / case '+': / G3D4 */
428	designate(state, i1 - '(', S4, i2, input_chr);
429	break;
430	case '-': /* G1D6 / case '.': / G2D6 / case '/': / G3D6 */
431	designate(state, i1 - ',', S6, i2, input_chr);
432	break;
433	case '$': /* G?DM? */
434	switch (i2) {
435	case 0: /* Obsolete version of GZDM4 */
436	i2 = '(';
437	case '(': /* GZDM4 / case ')': / G1DM4 */
438	case '': / G2DM4 / case '+': / G3DM4 */
439	designate(state, i2 - '(', M4, 0, input_chr);
440	break;
441	case '-': /* G1DM6 */
442	case '.': /* G2DM6 / case '/': / G3DM6 */
443	designate(state, i2 - ',', M6, 0, input_chr);
444	break;
445	default:
446	emit(emitctx, ERROR);
447	break;
448	}
449	case '%': /* DOCS */
a89fe3cf	450	/* XXX What's a reasonable way to handle an unrecognised DOCS? */
	451	switch (i2) {
	452	case 0:
	453	switch (input_chr) {
	454	case 'G':
	455	ENTER_MODE(DOCSUTF8);
	456	break;
	457	}
	458	break;
	459	}
b97e5427	460	break;
	461	default:
	462	/* Unsupported nF escape sequence. Re-emit it. */
	463	emit(emitctx, ESC);
	464	emit(emitctx, i1);
	465	if (i2) emit(emitctx, i2);
	466	emit(emitctx, input_chr);
	467	break;
	468	}
	469	}
	470	}
	471
04c24cbb	472	static int write_iso2022(charset_spec const *charset, long int input_chr,
	473	charset_state *state,
	474	void (emit)(void ctx, long int output),
	475	void *emitctx)
	476	{
	477	return FALSE;
	478	}
	479
b97e5427	480	const charset_spec charset_CS_ISO2022 = {
04c24cbb	481	CS_ISO2022, read_iso2022, write_iso2022, NULL
b97e5427	482	};
	483
	484	#ifdef TESTMODE
	485
	486	#include <stdio.h>
	487	#include <stdarg.h>
	488	#include <string.h>
	489
	490	int total_errs = 0;
	491
	492	void iso2022_emit(void *ctx, long output)
	493	{
	494	wchar_t p = (wchar_t )ctx;
	495	(p)++ = output;
	496	}
	497
	498	void iso2022_read_test(int line, char *input, int inlen, ...)
	499	{
	500	va_list ap;
	501	wchar_t *p, str[512];
	502	int i;
	503	charset_state state;
	504	unsigned long l;
	505
	506	state.s0 = state.s1 = 0;
	507	p = str;
	508
	509	for (i = 0; i < inlen; i++)
	510	read_iso2022(NULL, input[i] & 0xFF, &state, iso2022_emit, &p);
	511
	512	va_start(ap, inlen);
	513	l = 0;
	514	for (i = 0; i < p - str; i++) {
	515	l = va_arg(ap, long int);
	516	if (l == -1) {
	517	printf("%d: correct string shorter than output\n", line);
	518	total_errs++;
	519	break;
	520	}
	521	if (l != str[i]) {
	522	printf("%d: char %d came out as %08x, should be %08lx\n",
	523	line, i, str[i], l);
	524	total_errs++;
	525	}
	526	}
	527	if (l != -1) {
	528	l = va_arg(ap, long int);
	529	if (l != -1) {
	530	printf("%d: correct string longer than output\n", line);
	531	total_errs++;
	532	}
	533	}
	534	va_end(ap);
	535	}
	536
	537	/* Macro to concoct the first three parameters of iso2022_read_test. */
	538	#define TESTSTR(x) __LINE__, x, lenof(x)
	539
	540	int main(void)
	541	{
	542	printf("read tests beginning\n");
	543	/* Simple test (Emacs sample text for Japanese, in ISO-2022-JP) */
	544	iso2022_read_test(TESTSTR("Japanese (\x1b$BF\|K\\8l\x1b(B)\t"
	545	"\x1b$B$3$s$K$A$O\x1b(B, "
546	"\x1b$B%3%s%K%A%O\x1b(B\n"),
547	'J','a','p','a','n','e','s','e',' ','(',
548	0x65E5, 0x672C, 0x8A9E, ')', '\t',
549	0x3053, 0x3093, 0x306b, 0x3061, 0x306f, ',', ' ',
550	0x30b3, 0x30f3, 0x30cb, 0x30c1, 0x30cf, '\n', 0, -1);
551	/* Same thing in EUC-JP (with designations, and half-width katakana) */
552	iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D"
553	"Japanese (\xc6\xfc\xcb\xdc\xb8\xec)\t"
554	"\xa4\xb3\xa4\xf3\xa4\xcb\xa4\xc1\xa4\xcf, "
555	"\x8e\xba\x8e\xdd\x8e\xc6\x8e\xc1\x8e\xca\n"),
556	'J','a','p','a','n','e','s','e',' ','(',
557	0x65E5, 0x672C, 0x8A9E, ')', '\t',
558	0x3053, 0x3093, 0x306b, 0x3061, 0x306f, ',', ' ',
559	0xff7a, 0xff9d, 0xff86, 0xff81, 0xff8a, '\n', 0, -1);
560	/* Multibyte single-shift */
561	iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D\x8f\"/!"),
562	0x02D8, '!', 0, -1);
563	/* Non-existent SBCS */
564	iso2022_read_test(TESTSTR("\x1b(!Zfnord\n"),
565	ERROR, ERROR, ERROR, ERROR, ERROR, '\n', 0, -1);
566	/* Pass-through of ordinary escape sequences, including a long one */
567	iso2022_read_test(TESTSTR("\x1b""b\x1b#5\x1b#!!!5"),
568	0x1B, 'b', 0x1B, '#', '5',
569	0x1B, '#', '!', '!', '!', '5', 0, -1);
570	/* Non-existent DBCS (also 5-byte escape sequence) */
571	iso2022_read_test(TESTSTR("\x1b$(!Bfnord!"),
572	ERROR, ERROR, ERROR, 0, -1);
573	/* Incomplete DB characters */
574	iso2022_read_test(TESTSTR("\x1b$B(,(\x1b(BHi\x1b$B(,(\n"),
575	0x2501, ERROR, 'H', 'i', 0x2501, ERROR, '\n', 0, -1);
576	iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D\xa4""B"),
577	ERROR, 'B', 0, -1);
578	iso2022_read_test(TESTSTR("\x1b$)B\x1b*I\x1b$+D\x0e\x1b\|$\xa2\xaf"),
579	ERROR, 0x02D8, 0, -1);
580	/* Incomplete escape sequence */
581	iso2022_read_test(TESTSTR("\x1b\n"), ERROR, '\n', 0, -1);
582	iso2022_read_test(TESTSTR("\x1b-A\x1b~\x1b\xa1"), ERROR, 0xa1, 0, -1);
583	/* Incomplete single-shift */
584	iso2022_read_test(TESTSTR("\x8e\n"), ERROR, '\n', 0, -1);
585	iso2022_read_test(TESTSTR("\x1b$*B\x8e(\n"), ERROR, '\n', 0, -1);
586	/* Corner cases (02/00 and 07/15) */
587	iso2022_read_test(TESTSTR("\x1b(B\x20\x7f"), 0x20, 0x7f, 0, -1);
588	iso2022_read_test(TESTSTR("\x1b(I\x20\x7f"), 0x20, 0x7f, 0, -1);
589	iso2022_read_test(TESTSTR("\x1b$B\x20\x7f"), 0x20, 0x7f, 0, -1);
590	iso2022_read_test(TESTSTR("\x1b-A\x0e\x20\x7f"), 0xa0, 0xff, 0, -1);
591	iso2022_read_test(TESTSTR("\x1b$-~\x0e\x20\x7f"), ERROR, 0, -1);
592	iso2022_read_test(TESTSTR("\x1b)B\xa0\xff"), ERROR, ERROR, 0, -1);
593	iso2022_read_test(TESTSTR("\x1b)I\xa0\xff"), ERROR, ERROR, 0, -1);
594	iso2022_read_test(TESTSTR("\x1b$)B\xa0\xff"), ERROR, ERROR, 0, -1);
595	iso2022_read_test(TESTSTR("\x1b-A\x1b~\xa0\xff"), 0xa0, 0xff, 0, -1);
596	iso2022_read_test(TESTSTR("\x1b$-~\x1b~\xa0\xff"), ERROR, 0, -1);
597	/* Designate control sets */
598	iso2022_read_test(TESTSTR("\x1b!@"), 0x1b, '!', '@', 0, -1);
a89fe3cf	599	/* Designate other coding system */
	600	iso2022_read_test(TESTSTR("\x1b%G"
	601	"\xCE\xBA\xE1\xBD\xB9\xCF\x83\xCE\xBC\xCE\xB5"),
	602	0x03BA, 0x1F79, 0x03C3, 0x03BC, 0x03B5, 0, -1);
	603	iso2022_read_test(TESTSTR("\x1b-A\x1b%G\xCE\xBA\x1b%@\xa0"),
	604	0x03BA, 0xA0, 0, -1);
	605	iso2022_read_test(TESTSTR("\x1b%G\xCE\x1b%@"), ERROR, 0, -1);
	606	iso2022_read_test(TESTSTR("\x1b%G\xCE\xBA\x1b%\x1b%@"),
	607	0x03BA, 0x1B, '%', 0, -1);
b97e5427	608	printf("read tests completed\n");
	609	printf("total: %d errors\n", total_errs);
	610	return (total_errs != 0);
	611	}
	612
	613	#endif /* TESTMODE */
	614
	615	#else /* ENUM_CHARSETS */
	616
	617	ENUM_CHARSET(CS_ISO2022)
	618
	619	#endif