/*
* UCdomap.c
* =========
*
* This is a Lynx chartrans engine, its external calls are in UCMap.h
*
* Derived from code in the Linux kernel console driver.
* The GNU Public Licence therefore applies, see
* the file COPYING in the top-level directory
* which should come with every Lynx distribution.
*
* [ original comment: - KW ]
*
* Mapping from internal code (such as Latin-1 or Unicode or IBM PC code)
* to font positions.
*
* aeb, 950210
*/
#include <HTUtils.h>
#include <HTMLDTD.h>
#include <LYGlobalDefs.h>
#include <UCkd.h>
#include <UCdomap.h>
#include <UCMap.h>
#include <UCDefs.h>
#include <LYCharSets.h>
#include <LYLeaks.h>
/*
* Include chartrans tables:
*/
#include <cp1250_uni.h> /* WinLatin2 (cp1250) */
#include <cp1251_uni.h> /* WinCyrillic (cp1251) */
#include <cp1252_uni.h> /* WinLatin1 (cp1252) */
#include <cp1253_uni.h> /* WinGreek (cp1253) */
#include <cp1255_uni.h> /* WinHebrew (cp1255) */
#include <cp1256_uni.h> /* WinArabic (cp1256) */
#include <cp1257_uni.h> /* WinBaltRim (cp1257) */
#include <cp437_uni.h> /* DosLatinUS (cp437) */
#include <cp737_uni.h> /* DosGreek (cp737) */
#include <cp775_uni.h> /* DosBaltRim (cp775) */
#include <cp850_uni.h> /* DosLatin1 (cp850) */
#include <cp852_uni.h> /* DosLatin2 (cp852) */
#include <cp862_uni.h> /* DosHebrew (cp862) */
#include <cp864_uni.h> /* DosArabic (cp864) */
#include <cp866_uni.h> /* DosCyrillic (cp866) */
#include <cp869_uni.h> /* DosGreek2 (cp869) */
#include <def7_uni.h> /* 7 bit approximations */
#include <dmcs_uni.h> /* DEC Multinational */
#include <hp_uni.h> /* HP Roman8 */
#include <iso01_uni.h> /* ISO Latin 1 */
#include <iso02_uni.h> /* ISO Latin 2 */
#include <iso03_uni.h> /* ISO Latin 3 */
#include <iso04_uni.h> /* ISO Latin 4 */
#include <iso05_uni.h> /* ISO 8859-5 Cyrillic */
#include <iso06_uni.h> /* ISO 8859-6 Arabic */
#include <iso07_uni.h> /* ISO 8859-7 Greek */
#include <iso08_uni.h> /* ISO 8859-8 Hebrew */
#include <iso09_uni.h> /* ISO 8859-9 (Latin 5) */
#include <iso10_uni.h> /* ISO 8859-10 */
#include <iso15_uni.h> /* ISO 8859-15 (Latin 9)*/
#include <koi8r_uni.h> /* KOI8-R Cyrillic */
#include <mac_uni.h> /* Macintosh (8 bit) */
#include <mnem2_suni.h> /* RFC 1345 Mnemonic */
#include <next_uni.h> /* NeXT character set */
#include <rfc_suni.h> /* RFC 1345 w/o Intro */
/* #include <utf8_uni.h> */ /* UNICODE UTF 8 */
#include <viscii_uni.h> /* Vietnamese (VISCII) */
#include <cp866u_uni.h> /* Ukrainian Cyrillic (866) */
#include <koi8u_uni.h> /* Ukrainian Cyrillic (koi8-u */
#ifdef NOTDEFINED
#include <mnem_suni.h>
#endif /* NOTDEFINED */
/*
* Some of the code below, and some of the comments, are left in for
* historical reasons. Not all those tables below are currently
* really needed (and what with all those hardwired codepoints),
* but let's keep them around for now. They may come in handy if we
* decide to make more extended use of the mechanisms (including e.g.
* for chars < 127...). - KW
*/
PRIVATE u16 translations[][256] = {
/*
* 8-bit Latin-1 mapped to Unicode -- trivial mapping.
*/
{
0x0000, 0x0001, 0x0002, 0x0003, 0x0004, 0x0005, 0x0006, 0x0007,
0x0008, 0x0009, 0x000a, 0x000b, 0x000c, 0x000d, 0x000e, 0x000f,
0x0010, 0x0011, 0x0012, 0x0013, 0x0014, 0x0015, 0x0016, 0x0017,
0x0018, 0x0019, 0x001a, 0x001b, 0x001c, 0x001d, 0x001e, 0x001f,
0x0020, 0x0021, 0x0022, 0x0023, 0x0024, 0x0025, 0x0026, 0x0027,
0x0028, 0x0029, 0x002a, 0x002b, 0x002c, 0x002d, 0x002e, 0x002f,
0x0030, 0x0031, 0x0032, 0x0033, 0x0034, 0x0035, 0x0036, 0x0037,
0x0038, 0x0039, 0x003a, 0x003b, 0x003c, 0x003d, 0x003e, 0x003f,
0x0040, 0x0041, 0x0042, 0x0043, 0x0044, 0x0045, 0x0046, 0x0047,
0x0048, 0x0049, 0x004a, 0x004b, 0x004c, 0x004d, 0x004e, 0x004f,
0x0050, 0x0051, 0x0052, 0x0053, 0x0054, 0x0055, 0x0056, 0x0057,
0x0058, 0x0059, 0x005a, 0x005b, 0x005c, 0x005d, 0x005e, 0x005f,
0x0060, 0x0061, 0x0062, 0x0063, 0x0064, 0x0065, 0x0066, 0x0067,
0x0068, 0x0069, 0x006a, 0x006b, 0x006c, 0x006d, 0x006e, 0x006f,
0x0070, 0x0071, 0x0072, 0x0073, 0x0074, 0x0075, 0x0076, 0x0077,
0x0078, 0x0079, 0x007a, 0x007b, 0x007c, 0x007d, 0x007e, 0x007f,
0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087,
0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f,
0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097,
0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f,
0x00a0, 0x00a1, 0x00a2, 0x00a3, 0x00a4, 0x00a5, 0x00a6, 0x00a7,
0x00a8, 0x00a9, 0x00aa, 0x00ab, 0x00ac, 0x00ad, 0x00ae, 0x00af,
0x00b0, 0x00b1, 0x00b2, 0x00b3, 0x00b4, 0x00b5, 0x00b6, 0x00b7,
0x00b8, 0x00b9, 0x00ba, 0x00bb, 0x00bc, 0x00bd, 0x00be, 0x00bf,
0x00c0, 0x00c1, 0x00c2, 0x00c3, 0x00c4, 0x00c5, 0x00c6, 0x00c7,
0x00c8, 0x00c9, 0x00ca, 0x00cb, 0x00cc, 0x00cd, 0x00ce, 0x00cf,
0x00d0, 0x00d1, 0x00d2, 0x00d3, 0x00d4, 0x00d5, 0x00d6, 0x00d7,
0x00d8, 0x00d9, 0x00da, 0x00db, 0x00dc, 0x00dd, 0x00de, 0x00df,
0x00e0, 0x00e1, 0x00e2, 0x00e3, 0x00e4, 0x00e5, 0x00e6, 0x00e7,
0x00e8, 0x00e9, 0x00ea, 0x00eb, 0x00ec, 0x00ed, 0x00ee, 0x00ef,
0x00f0, 0x00f1, 0x00f2, 0x00f3, 0x00f4, 0x00f5, 0x00f6, 0x00f7,
0x00f8, 0x00f9, 0x00fa, 0x00fb, 0x00fc, 0x00fd, 0x00fe, 0x00ff
},
/*
* VT100 graphics mapped to Unicode.
*/
{
0x0000, 0x0001, 0x0002, 0x0003, 0x0004, 0x0005, 0x0006, 0x0007,
0x0008, 0x0009, 0x000a, 0x000b, 0x000c, 0x000d, 0x000e, 0x000f,
0x0010, 0x0011, 0x0012, 0x0013, 0x0014, 0x0015, 0x0016, 0x0017,
0x0018, 0x0019, 0x001a, 0x001b, 0x001c, 0x001d, 0x001e, 0x001f,
0x0020, 0x0021, 0x0022, 0x0023, 0x0024, 0x0025, 0x0026, 0x0027,
0x0028, 0x0029, 0x002a, 0x002b, 0x002c, 0x002d, 0x002e, 0x002f,
0x0030, 0x0031, 0x0032, 0x0033, 0x0034, 0x0035, 0x0036, 0x0037,
0x0038, 0x0039, 0x003a, 0x003b, 0x003c, 0x003d, 0x003e, 0x003f,
0x0040, 0x0041, 0x0042, 0x0043, 0x0044, 0x0045, 0x0046, 0x0047,
0x0048, 0x0049, 0x004a, 0x004b, 0x004c, 0x004d, 0x004e, 0x004f,
0x0050, 0x0051, 0x0052, 0x0053, 0x0054, 0x0055, 0x0056, 0x0057,
0x0058, 0x0059, 0x005a, 0x005b, 0x005c, 0x005d, 0x005e, 0x00a0,
0x25c6, 0x2592, 0x2409, 0x240c, 0x240d, 0x240a, 0x00b0, 0x00b1,
0x2424, 0x240b, 0x2518, 0x2510, 0x250c, 0x2514, 0x253c, 0xf800,
0xf801, 0x2500, 0xf803, 0xf804, 0x251c, 0x2524, 0x2534, 0x252c,
0x2502, 0x2264, 0x2265, 0x03c0, 0x2260, 0x00a3, 0x00b7, 0x007f,
0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087,
0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f,
0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097,
0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f,
0x00a0, 0x00a1, 0x00a2, 0x00a3, 0x00a4, 0x00a5, 0x00a6, 0x00a7,
0x00a8, 0x00a9, 0x00aa, 0x00ab, 0x00ac, 0x00ad, 0x00ae, 0x00af,
0x00b0, 0x00b1, 0x00b2, 0x00b3, 0x00b4, 0x00b5, 0x00b6, 0x00b7,
0x00b8, 0x00b9, 0x00ba, 0x00bb, 0x00bc, 0x00bd, 0x00be, 0x00bf,
0x00c0, 0x00c1, 0x00c2, 0x00c3, 0x00c4, 0x00c5, 0x00c6, 0x00c7,
0x00c8, 0x00c9, 0x00ca, 0x00cb, 0x00cc, 0x00cd, 0x00ce, 0x00cf,
0x00d0, 0x00d1, 0x00d2, 0x00d3, 0x00d4, 0x00d5, 0x00d6, 0x00d7,
0x00d8, 0x00d9, 0x00da, 0x00db, 0x00dc, 0x00dd, 0x00de, 0x00df,
0x00e0, 0x00e1, 0x00e2, 0x00e3, 0x00e4, 0x00e5, 0x00e6, 0x00e7,
0x00e8, 0x00e9, 0x00ea, 0x00eb, 0x00ec, 0x00ed, 0x00ee, 0x00ef,
0x00f0, 0x00f1, 0x00f2, 0x00f3, 0x00f4, 0x00f5, 0x00f6, 0x00f7,
0x00f8, 0x00f9, 0x00fa, 0x00fb, 0x00fc, 0x00fd, 0x00fe, 0x00ff
},
/*
* IBM Codepage 437 mapped to Unicode.
*/
{
0x0000, 0x263a, 0x263b, 0x2665, 0x2666, 0x2663, 0x2660, 0x2022,
0x25d8, 0x25cb, 0x25d9, 0x2642, 0x2640, 0x266a, 0x266b, 0x263c,
0x25ba, 0x25c4, 0x2195, 0x203c, 0x00b6, 0x00a7, 0x25ac, 0x21a8,
0x2191, 0x2193, 0x2192, 0x2190, 0x221f, 0x2194, 0x25b2, 0x25bc,
0x0020, 0x0021, 0x0022, 0x0023, 0x0024, 0x0025, 0x0026, 0x0027,
0x0028, 0x0029, 0x002a, 0x002b, 0x002c, 0x002d, 0x002e, 0x002f,
0x0030, 0x0031, 0x0032, 0x0033, 0x0034, 0x0035, 0x0036, 0x0037,
0x0038, 0x0039, 0x003a, 0x003b, 0x003c, 0x003d, 0x003e, 0x003f,
0x0040, 0x0041, 0x0042, 0x0043, 0x0044, 0x0045, 0x0046, 0x0047,
0x0048, 0x0049, 0x004a, 0x004b, 0x004c, 0x004d, 0x004e, 0x004f,
0x0050, 0x0051, 0x0052, 0x0053, 0x0054, 0x0055, 0x0056, 0x0057,
0x0058, 0x0059, 0x005a, 0x005b, 0x005c, 0x005d, 0x005e, 0x005f,
0x0060, 0x0061, 0x0062, 0x0063, 0x0064, 0x0065, 0x0066, 0x0067,
0x0068, 0x0069, 0x006a, 0x006b, 0x006c, 0x006d, 0x006e, 0x006f,
0x0070, 0x0071, 0x0072, 0x0073, 0x0074, 0x0075, 0x0076, 0x0077,
0x0078, 0x0079, 0x007a, 0x007b, 0x007c, 0x007d, 0x007e, 0x2302,
0x00c7, 0x00fc, 0x00e9, 0x00e2, 0x00e4, 0x00e0, 0x00e5, 0x00e7,
0x00ea, 0x00eb, 0x00e8, 0x00ef, 0x00ee, 0x00ec, 0x00c4, 0x00c5,
0x00c9, 0x00e6, 0x00c6, 0x00f4, 0x00f6, 0x00f2, 0x00fb, 0x00f9,
0x00ff, 0x00d6, 0x00dc, 0x00a2, 0x00a3, 0x00a5, 0x20a7, 0x0192,
0x00e1, 0x00ed, 0x00f3, 0x00fa, 0x00f1, 0x00d1, 0x00aa, 0x00ba,
0x00bf, 0x2310, 0x00ac, 0x00bd, 0x00bc, 0x00a1, 0x00ab, 0x00bb,
0x2591, 0x2592, 0x2593, 0x2502, 0x2524, 0x2561, 0x2562, 0x2556,
0x2555, 0x2563, 0x2551, 0x2557, 0x255d, 0x255c, 0x255b, 0x2510,
0x2514, 0x2534, 0x252c, 0x251c, 0x2500, 0x253c, 0x255e, 0x255f,
0x255a, 0x2554, 0x2569, 0x2566, 0x2560, 0x2550, 0x256c, 0x2567,
0x2568, 0x2564, 0x2565, 0x2559, 0x2558, 0x2552, 0x2553, 0x256b,
0x256a, 0x2518, 0x250c, 0x2588, 0x2584, 0x258c, 0x2590, 0x2580,
0x03b1, 0x00df, 0x0393, 0x03c0, 0x03a3, 0x03c3, 0x00b5, 0x03c4,
0x03a6, 0x0398, 0x03a9, 0x03b4, 0x221e, 0x03c6, 0x03b5, 0x2229,
0x2261, 0x00b1, 0x2265, 0x2264, 0x2320, 0x2321, 0x00f7, 0x2248,
0x00b0, 0x2219, 0x00b7, 0x221a, 0x207f, 0x00b2, 0x25a0, 0x00a0
},
/*
* User mapping -- default to codes for direct font mapping.
*/
{
0xf000, 0xf001, 0xf002, 0xf003, 0xf004, 0xf005, 0xf006, 0xf007,
0xf008, 0xf009, 0xf00a, 0xf00b, 0xf00c, 0xf00d, 0xf00e, 0xf00f,
0xf010, 0xf011, 0xf012, 0xf013, 0xf014, 0xf015, 0xf016, 0xf017,
0xf018, 0xf019, 0xf01a, 0xf01b, 0xf01c, 0xf01d, 0xf01e, 0xf01f,
0xf020, 0xf021, 0xf022, 0xf023, 0xf024, 0xf025, 0xf026, 0xf027,
0xf028, 0xf029, 0xf02a, 0xf02b, 0xf02c, 0xf02d, 0xf02e, 0xf02f,
0xf030, 0xf031, 0xf032, 0xf033, 0xf034, 0xf035, 0xf036, 0xf037,
0xf038, 0xf039, 0xf03a, 0xf03b, 0xf03c, 0xf03d, 0xf03e, 0xf03f,
0xf040, 0xf041, 0xf042, 0xf043, 0xf044, 0xf045, 0xf046, 0xf047,
0xf048, 0xf049, 0xf04a, 0xf04b, 0xf04c, 0xf04d, 0xf04e, 0xf04f,
0xf050, 0xf051, 0xf052, 0xf053, 0xf054, 0xf055, 0xf056, 0xf057,
0xf058, 0xf059, 0xf05a, 0xf05b, 0xf05c, 0xf05d, 0xf05e, 0xf05f,
0xf060, 0xf061, 0xf062, 0xf063, 0xf064, 0xf065, 0xf066, 0xf067,
0xf068, 0xf069, 0xf06a, 0xf06b, 0xf06c, 0xf06d, 0xf06e, 0xf06f,
0xf070, 0xf071, 0xf072, 0xf073, 0xf074, 0xf075, 0xf076, 0xf077,
0xf078, 0xf079, 0xf07a, 0xf07b, 0xf07c, 0xf07d, 0xf07e, 0xf07f,
0xf080, 0xf081, 0xf082, 0xf083, 0xf084, 0xf085, 0xf086, 0xf087,
0xf088, 0xf089, 0xf08a, 0xf08b, 0xf08c, 0xf08d, 0xf08e, 0xf08f,
0xf090, 0xf091, 0xf092, 0xf093, 0xf094, 0xf095, 0xf096, 0xf097,
0xf098, 0xf099, 0xf09a, 0xf09b, 0xf09c, 0xf09d, 0xf09e, 0xf09f,
0xf0a0, 0xf0a1, 0xf0a2, 0xf0a3, 0xf0a4, 0xf0a5, 0xf0a6, 0xf0a7,
0xf0a8, 0xf0a9, 0xf0aa, 0xf0ab, 0xf0ac, 0xf0ad, 0xf0ae, 0xf0af,
0xf0b0, 0xf0b1, 0xf0b2, 0xf0b3, 0xf0b4, 0xf0b5, 0xf0b6, 0xf0b7,
0xf0b8, 0xf0b9, 0xf0ba, 0xf0bb, 0xf0bc, 0xf0bd, 0xf0be, 0xf0bf,
0xf0c0, 0xf0c1, 0xf0c2, 0xf0c3, 0xf0c4, 0xf0c5, 0xf0c6, 0xf0c7,
0xf0c8, 0xf0c9, 0xf0ca, 0xf0cb, 0xf0cc, 0xf0cd, 0xf0ce, 0xf0cf,
0xf0d0, 0xf0d1, 0xf0d2, 0xf0d3, 0xf0d4, 0xf0d5, 0xf0d6, 0xf0d7,
0xf0d8, 0xf0d9, 0xf0da, 0xf0db, 0xf0dc, 0xf0dd, 0xf0de, 0xf0df,
0xf0e0, 0xf0e1, 0xf0e2, 0xf0e3, 0xf0e4, 0xf0e5, 0xf0e6, 0xf0e7,
0xf0e8, 0xf0e9, 0xf0ea, 0xf0eb, 0xf0ec, 0xf0ed, 0xf0ee, 0xf0ef,
0xf0f0, 0xf0f1, 0xf0f2, 0xf0f3, 0xf0f4, 0xf0f5, 0xf0f6, 0xf0f7,
0xf0f8, 0xf0f9, 0xf0fa, 0xf0fb, 0xf0fc, 0xf0fd, 0xf0fe, 0xf0ff
}
};
PRIVATE u16 *UC_translate = NULL;
PRIVATE struct UC_charset UCInfo[MAXCHARSETS];
/*
* The standard kernel character-to-font mappings are not invertible
* -- this is just a best effort.
*/
#define MAX_GLYPH 512 /* Max possible glyph value */
PRIVATE unsigned char * inv_translate = NULL;
PRIVATE unsigned char inv_norm_transl[MAX_GLYPH];
PRIVATE unsigned char * inverse_translations[4] = { NULL, NULL, NULL, NULL };
PRIVATE void set_inverse_transl PARAMS((
int i));
PRIVATE u16 *set_translate PARAMS((
int m));
#ifdef NOTDEFINED
PRIVATE unsigned char inverse_translate PARAMS((int glyph));
PRIVATE int con_set_trans_old PARAMS((unsigned char *arg));
PRIVATE int con_get_trans_old PARAMS((unsigned char *arg));
PRIVATE int con_set_trans_new PARAMS((u16 *arg));
PRIVATE int con_get_trans_new PARAMS((u16 *arg));
#endif /* NOTDEFINED */
PRIVATE int UC_valid_UC_charset PARAMS((
int UC_charset_hndl));
PRIVATE void UC_con_set_trans PARAMS((
int UC_charset_in_hndl,
int Gn,
int update_flag));
PRIVATE int con_insert_unipair PARAMS((
u16 unicode,
u16 fontpos,
int fordefault));
PRIVATE int con_insert_unipair_str PARAMS((
u16 unicode,
CONST char * replace_str,
int fordefault));
PRIVATE void con_clear_unimap PARAMS((
int fordefault));
PRIVATE void con_clear_unimap_str PARAMS((
int fordefault));
#ifdef NOTDEFINED
PRIVATE int con_set_unimap PARAMS((
u16 ct,
struct unipair * list));
#endif /* NOTDEFINED */
PRIVATE void con_set_default_unimap NOPARAMS;
PRIVATE int UC_con_set_unimap PARAMS((
int UC_charset_out_hndl,
int update_flag));
PRIVATE int UC_con_set_unimap_str PARAMS((
u16 ct,
struct unipair_str * list,
int fordefault));
#ifdef NOTDEFINED
PRIVATE int con_get_unimap PARAMS((
u16 ct,
u16 * uct,
struct unipair * list));
#endif /* NOTDEFINED */
PRIVATE int conv_uni_to_pc PARAMS((
long ucs,
int usedefault));
PRIVATE int conv_uni_to_str PARAMS((
char* outbuf,
int buflen,
long ucs,
int usedefault));
PRIVATE void UCconsole_map_init NOPARAMS;
PRIVATE int UC_MapGN PARAMS((
int UChndl,
int update_flag));
PRIVATE int UC_FindGN_byMIME PARAMS((
CONST char * UC_MIMEcharset));
PRIVATE void UCreset_allocated_LYCharSets NOPARAMS;
PRIVATE CONST char ** UC_setup_LYCharSets_repl PARAMS((
int UC_charset_in_hndl,
unsigned lowest8));
PRIVATE int UC_Register_with_LYCharSets PARAMS((
int s,
CONST char * UC_MIMEcharset,
CONST char * UC_LYNXcharset,
int lowest_eightbit));
#ifdef LY_FIND_LEAKS
PRIVATE void UCfree_allocated_LYCharSets NOPARAMS;
PRIVATE void UCcleanup_mem NOPARAMS;
#endif
PRIVATE int default_UChndl = -1;
PRIVATE void set_inverse_transl ARGS1(
int, i)
{
int j, glyph;
u16 *p = translations[i];
unsigned char *q = inverse_translations[i];
if (!q) {
/*
* Slightly messy to avoid calling kmalloc too early.
*/
q = inverse_translations[i] = ((i == LAT1_MAP) ?
inv_norm_transl :
(unsigned char *)malloc(MAX_GLYPH));
if (!q)
return;
}
for (j = 0; j < MAX_GLYPH; j++)
q[j] = 0;
for (j = 0; j < E_TABSZ; j++) {
glyph = conv_uni_to_pc(p[j], 0);
if (glyph >= 0 && glyph < MAX_GLYPH && q[glyph] < 32) {
/*
* Prefer '-' above SHY etc.
*/
q[glyph] = j;
}
}
}
PRIVATE u16 *set_translate ARGS1(
int, m)
{
if (!inverse_translations[m])
set_inverse_transl(m);
inv_translate = inverse_translations[m];
return translations[m];
}
#ifdef NOTDEFINED
/*
* Inverse translation is impossible for several reasons:
* 1. The font<->character maps are not 1-1.
* 2. The text may have been written while a different translation map
* was active, or using Unicode.
* Still, it is now possible to a certain extent to cut and paste non-ASCII.
*/
PRIVATE unsigned char inverse_translate ARGS1(
int, glyph)
{
if (glyph < 0 || glyph >= MAX_GLYPH) {
return 0;
} else {
return ((inv_translate && inv_translate[glyph]) ?
inv_translate[glyph] :
(unsigned char)(glyph & 0xff));
}
}
/*
* Load customizable translation table.
* 'arg' points to a 256 byte translation table.
*
* The "old" variants are for translation directly to font (using the
* 0xf000-0xf0ff "transparent" Unicodes) whereas the "new" variants set
* Unicodes explicitly.
*/
PRIVATE int con_set_trans_old ARGS1(
unsigned char *, arg)
{
int i;
u16 *p = translations[USER_MAP];
#if(0)
i = verify_area(VERIFY_READ, (void *)arg, E_TABSZ);
if (i)
return i;
#endif
for (i = 0; i < E_TABSZ; i++)
p[i] = UNI_DIRECT_BASE | (u16)arg[i];
set_inverse_transl(USER_MAP);
return 0;
}
PRIVATE int con_get_trans_old ARGS1(
unsigned char *, arg)
{
int i, ch;
u16 *p = translations[USER_MAP];
#if(0)
i = verify_area(VERIFY_WRITE, (void *)arg, E_TABSZ);
if (i)
return i;
#endif
for (i = 0; i < E_TABSZ; i++) {
ch = conv_uni_to_pc(p[i]);
#ifdef NOTDEFINED
put_user((ch & ~0xff) ? 0 : ch, arg+i);
#endif /* NOTDEFINED */
arg[i] = (unsigned char)((ch & ~0xff) ? 0 : ch);
}
return 0;
}
PRIVATE int con_set_trans_new ARGS1(
u16 *, arg)
{
int i;
u16 *p = translations[USER_MAP];
#if(0)
i = verify_area(VERIFY_READ, (void *)arg, E_TABSZ*sizeof(u16));
if (i)
return i;
#endif
for (i = 0; i < E_TABSZ; i++)
p[i] = arg[i];
set_inverse_transl(USER_MAP);
return 0;
}
PRIVATE int con_get_trans_new ARGS1(
u16 * arg)
{
int i;
u16 *p = translations[USER_MAP];
#if(0)
i = verify_area(VERIFY_WRITE, (void *)arg, E_TABSZ*sizeof(u16));
if (i)
return i;
#endif
for (i = 0; i < E_TABSZ; i++)
arg[i] = p[i];
return 0;
}
#endif /* NOTDEFINED */
PRIVATE int UC_valid_UC_charset ARGS1(
int, UC_charset_hndl)
{
return (UC_charset_hndl >= 0 && UC_charset_hndl < UCNumCharsets);
}
PRIVATE void UC_con_set_trans ARGS3(
int, UC_charset_in_hndl,
int, Gn,
int, update_flag)
{
int i, j;
CONST u16 *p;
u16 *ptrans;
if (!UC_valid_UC_charset(UC_charset_in_hndl)) {
CTRACE(tfp, "UC_con_set_trans: Invalid charset handle %d.\n",
UC_charset_in_hndl);
return;
}
ptrans = translations[Gn];
p = UCInfo[UC_charset_in_hndl].unitable;
#if(0)
if (p == UC_current_unitable) { /* test whether pointers are equal */
return; /* nothing to be done */
}
/*
* The font is always 256 characters - so far.
* (this function preserved by num_uni==0 so unicount=NULL for built-in
* charsets like CJK or x-transparent should not be a problem?)
*/
con_clear_unimap();
#endif
for (i = 0; i < 256; i++) {
if ((j = UCInfo[UC_charset_in_hndl].unicount[i])) {
ptrans[i] = *p;
for (; j; j--) {
p++;
}
} else {
ptrans[i] = 0xfffd;
}
}
if (update_flag) {
set_inverse_transl(Gn); /* Update inverse translation for this one */
}
}
/*
* Unicode -> current font conversion
*
* A font has at most 512 chars, usually 256.
* But one font position may represent several Unicode chars.
* A hashtable is somewhat of a pain to deal with, so use a
* "paged table" instead. Simulation has shown the memory cost of
* this 3-level paged table scheme to be comparable to a hash table.
*/
PRIVATE int hashtable_contents_valid = 0; /* Use ASCII-only mode for bootup*/
PRIVATE int hashtable_str_contents_valid = 0;
PRIVATE u16 **uni_pagedir[32] =
{
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL
};
PRIVATE char* **uni_pagedir_str[32] =
{
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL
};
PRIVATE CONST u16 * UC_current_unitable = NULL;
PRIVATE struct unimapdesc_str *UC_current_unitable_str = NULL;
/*
* Keep a second set of structures for the translation designated
* as "default" - kw
*/
PRIVATE int unidefault_contents_valid = 0; /* Use ASCII-only mode for bootup*/
PRIVATE int unidefault_str_contents_valid = 0;
static u16 **unidefault_pagedir[32] =
{
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL
};
static char* **unidefault_pagedir_str[32] =
{
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL
};
PRIVATE CONST u16 * UC_default_unitable = 0;
PRIVATE CONST struct unimapdesc_str *UC_default_unitable_str = 0;
PRIVATE int con_insert_unipair ARGS3(
u16, unicode,
u16, fontpos,
int, fordefault)
{
int i, n;
u16 **p1, *p2;
if(fordefault)
p1 = unidefault_pagedir[n = unicode >> 11];
else
p1 = uni_pagedir[n = unicode >> 11];
if (!p1) {
p1 = (u16* *)malloc(32*sizeof(u16 *));
if (fordefault)
unidefault_pagedir[n] = p1;
else
uni_pagedir[n] = p1;
if ( !p1 )
return -ENOMEM;
for (i = 0; i < 32; i++) {
p1[i] = NULL;
}
}
if (!(p2 = p1[n = (unicode >> 6) & 0x1f])) {
p2 = p1[n] = (u16 *)malloc(64*sizeof(u16));
if (!p2)
return -ENOMEM;
for (i = 0; i < 64; i++) {
p2[i] = 0xffff; /* No glyph for this character (yet) */
}
}
p2[unicode & 0x3f] = fontpos;
return 0;
}
PRIVATE int con_insert_unipair_str ARGS3(
u16, unicode,
CONST char *, replace_str,
int, fordefault)
{
int i, n;
char ***p1;
CONST char **p2;
if(fordefault)
p1 = unidefault_pagedir_str[n = unicode >> 11];
else
p1 = uni_pagedir_str[n = unicode >> 11];
if (!p1) {
p1 = (char** *)malloc(32*sizeof(char **));
if (fordefault)
unidefault_pagedir_str[n] = p1;
else
uni_pagedir_str[n] = p1;
if ( !p1 )
return -ENOMEM;
for (i = 0; i < 32; i++) {
p1[i] = NULL;
}
}
n = ((unicode >> 6) & 0x1f);
if (!p1[n]) {
p1[n] = (char **)malloc(64*sizeof(char *));
if (!p1[n])
return -ENOMEM;
p2 = (CONST char **)p1[n];
for (i = 0; i < 64; i++) {
p2[i] = NULL; /* No replace string this character (yet) */
}
}
p2 = (CONST char **)p1[n];
p2[unicode & 0x3f] = replace_str;
return 0;
}
/*
* ui arg was a leftover, deleted. - KW
*/
PRIVATE void con_clear_unimap ARGS1(
int, fordefault)
{
int i, j;
u16 **p1;
if (fordefault) {
for (i = 0; i < 32; i++) {
if ((p1 = unidefault_pagedir[i]) != NULL) {
for (j = 0; j < 32; j++) {
FREE(p1[j]);
}
FREE(p1);
}
unidefault_pagedir[i] = NULL;
}
unidefault_contents_valid = 1;
} else {
for (i = 0; i < 32; i++) {
if ((p1 = uni_pagedir[i]) != NULL) {
for (j = 0; j < 32; j++) {
FREE(p1[j]);
}
FREE(p1);
}
uni_pagedir[i] = NULL;
}
hashtable_contents_valid = 1;
}
}
PRIVATE void con_clear_unimap_str ARGS1(int, fordefault)
{
int i, j;
char ***p1;
if (fordefault) {
for (i = 0; i < 32; i++) {
if ((p1 = unidefault_pagedir_str[i]) != NULL) {
for (j = 0; j < 32; j++) {
FREE(p1[j]);
}
FREE(p1);
}
unidefault_pagedir_str[i] = NULL;
}
unidefault_str_contents_valid = 1; /* ??? probably no use... */
} else {
for (i = 0; i < 32; i++) {
if ((p1 = uni_pagedir_str[i]) != NULL) {
for (j = 0; j < 32; j++) {
FREE(p1[j]);
}
FREE(p1);
}
uni_pagedir_str[i] = NULL;
}
hashtable_str_contents_valid = 1; /* ??? probably no use... */
}
}
#ifdef NOTDEFINED
PRIVATE int con_set_unimap ARGS2(
u16, ct,
struct unipair *, list)
{
int err = 0, err1, i;
while (ct--) {
if ((err1 = con_insert_unipair(list->unicode, list->fontpos)) != 0) {
err = err1;
}
list++;
}
for (i = 0; i <= 3; i++) {
set_inverse_transl(i); /* Update all inverse translations */
}
return err;
}
#endif /* NOTDEFINED */
/*
* Loads the unimap for the hardware font, as defined in uni_hash.tbl.
* The representation used was the most compact I could come up
* with. This routine is executed at sys_setup time, and when the
* PIO_FONTRESET ioctl is called.
*/
PRIVATE void con_set_default_unimap NOARGS
{
int i, j;
CONST u16 *p;
/*
* The default font is always 256 characters.
* (default font can not be a fake one, so unicout!=NULL for sure.)
*/
con_clear_unimap(1);
p = dfont_unitable;
for (i = 0; i < 256; i++) {
for (j = dfont_unicount[i]; j; j--) {
con_insert_unipair(*(p++), i, 1);
}
}
#if 0
for (i = 0; i <= 3; i++) {
set_inverse_transl(i); /* Update all inverse translations */
}
#endif
UC_default_unitable = dfont_unitable;
con_clear_unimap_str(1);
UC_con_set_unimap_str(dfont_replacedesc.entry_ct, repl_map, 1);
UC_default_unitable_str = &dfont_replacedesc;
}
PUBLIC int UCNumCharsets = 0;
PUBLIC int UCLYhndl_HTFile_for_unspec = -1;
PUBLIC int UCLYhndl_HTFile_for_unrec = -1;
PUBLIC int UCLYhndl_for_unspec = -1;
PUBLIC int UCLYhndl_for_unrec = -1;
/* easy to type, will initialize later */
PUBLIC int LATIN1 = -1; /* UCGetLYhndl_byMIME("iso-8859-1") */
PUBLIC int US_ASCII = -1; /* UCGetLYhndl_byMIME("us-ascii") */
PUBLIC int UTF8 = -1; /* UCGetLYhndl_byMIME("utf-8") */
PRIVATE int UC_con_set_unimap ARGS2(
int, UC_charset_out_hndl,
int, update_flag)
{
int i, j;
CONST u16 *p;
if (!UC_valid_UC_charset(UC_charset_out_hndl)) {
CTRACE(tfp, "UC_con_set_unimap: Invalid charset handle %d.\n",
UC_charset_out_hndl);
return -1;
}
p = UCInfo[UC_charset_out_hndl].unitable;
if (p == UC_current_unitable) { /* test whether pointers are equal */
return update_flag; /* nothing to be done */
}
UC_current_unitable = p;
/*
* The font is always 256 characters - so far.
* (fake 0 for built-in charsets like CJK or x-transparent, add a check)
*/
con_clear_unimap(0);
for (i = 0; i < 256 && UCInfo[UC_charset_out_hndl].unicount != NULL; i++) {
for (j = UCInfo[UC_charset_out_hndl].unicount[i]; j; j--) {
con_insert_unipair(*(p++), i, 0);
}
}
if (update_flag) {
for (i = 0; i <= 3; i++) {
set_inverse_transl(i); /* Update all inverse translations */
}
}
return 0;
}
PRIVATE int UC_con_set_unimap_str ARGS3(
u16, ct,
struct unipair_str *, list,
int, fordefault)
{
int err = 0, err1;
while (ct--) {
if ((err1 = con_insert_unipair_str(list->unicode,
list->replace_str,
fordefault)) != 0) {
err = err1;
}
list++;
}
/*
* No inverse translations for replacement strings!
*/
if (!err) {
if (fordefault)
unidefault_str_contents_valid = 1;
else
hashtable_str_contents_valid = 1;
}
return err;
}
#ifdef NOTDEFINED
PRIVATE int con_get_unimap ARGS3(
u16, ct,
u16 *, uct,
struct unipair *, list)
{
int i, j, k, ect;
u16 **p1, *p2;
ect = 0;
if (hashtable_contents_valid) {
for (i = 0; i < 32; i++) {
if ((p1 = uni_pagedir[i]) != NULL) {
for (j = 0; j < 32; j++) {
if ((p2 = *(p1++)) != NULL) {
for (k = 0; k < 64; k++) {
if (*p2 < MAX_GLYPH && ect++ < ct) {
list->unicode = (u16) ((i<<11)+(j<<6)+k);
list->fontpos = (u16) *p2;
list++;
}
p2++;
}
}
}
}
}
}
*uct = ect;
return ((ect <= ct) ? 0 : -ENOMEM);
}
#endif /* NOTDEFINED */
PRIVATE int conv_uni_to_pc ARGS2(
long, ucs,
int, usedefault)
{
int h;
u16 **p1, *p2;
/*
* Only 16-bit codes supported at this time.
*/
if (ucs > 0xffff) {
/*
* U+FFFD: REPLACEMENT CHARACTER.
*/
ucs = 0xfffd;
} else if (ucs < 0x20 || ucs >= 0xfffe) {
/*
* Not a printable character.
*/
return -1;
} else if (ucs == 0xfeff || (ucs >= 0x200b && ucs <= 0x200f)) {
/*
* Zero-width space.
*/
return -2;
} else if ((ucs & ~UNI_DIRECT_MASK) == UNI_DIRECT_BASE) {
/*
* UNI_DIRECT_BASE indicates the start of the region in the
* User Zone which always has a 1:1 mapping to the currently
* loaded font. The UNI_DIRECT_MASK indicates the bit span
* of the region.
*/
return (ucs & UNI_DIRECT_MASK);
}
if (usedefault) {
if (!unidefault_contents_valid)
return -3;
p1 = unidefault_pagedir[ucs >> 11];
} else {
if (!hashtable_contents_valid)
return -3;
p1 = uni_pagedir[ucs >> 11];
}
if (p1 &&
(p2 = p1[(ucs >> 6) & 0x1f]) &&
(h = p2[ucs & 0x3f]) < MAX_GLYPH) {
return h;
}
/*
* Not found.
*/
return -4;
}
/*
* Note: contents of outbuf is not changes for negative return value!
*/
PRIVATE int conv_uni_to_str ARGS4(
char*, outbuf,
int, buflen,
long, ucs,
int, usedefault)
{
char *h;
char ***p1, **p2;
/*
* Only 16-bit codes supported at this time.
*/
if (ucs > 0xffff) {
/*
* U+FFFD: REPLACEMENT CHARACTER.
*/
ucs = 0xfffd;
/*
* Maybe the following two cases should be allowed here?? - KW
*/
} else if (ucs < 0x20 || ucs >= 0xfffe) {
/*
* Not a printable character.
*/
return -1;
} else if (ucs == 0xfeff || (ucs >= 0x200b && ucs <= 0x200f)) {
/*
* Zero-width space.
*/
return -2;
#ifdef NOTDEFINED /* We don't handle the following here: */
} else if ((ucs & ~UNI_DIRECT_MASK) == UNI_DIRECT_BASE) {
/*
* UNI_DIRECT_BASE indicates the start of the region in the
* User Zone which always has a 1:1 mapping to the currently
* loaded font. The UNI_DIRECT_MASK indicates the bit span
* of the region.
*/
return ucs & UNI_DIRECT_MASK;
#endif /* NOTDEFINED */
}
if (usedefault) {
if (!unidefault_str_contents_valid)
return -3;
p1 = unidefault_pagedir_str[ucs >> 11];
} else {
if (!hashtable_str_contents_valid)
return -3;
p1 = uni_pagedir_str[ucs >> 11];
}
if (p1 &&
(p2 = p1[(ucs >> 6) & 0x1f]) &&
(h = p2[ucs & 0x3f])) {
strncpy (outbuf,h,(size_t) (buflen-1));
return 1; /* ok ! */
}
/*
* Not found.
*/
return -4;
}
PUBLIC int UCInitialized = 0;
/*
* [ original comment: - KW ]
* This is called at sys_setup time, after memory and the console are
* initialized. It must be possible to call kmalloc(..., GFP_KERNEL)
* from this function, hence the call from sys_setup.
*/
PRIVATE void UCconsole_map_init NOARGS
{
con_set_default_unimap();
UCInitialized = 1;
}
/*
* OK now, finally, some stuff that is more specifically for Lynx: - KW
*/
PUBLIC int UCTransUniChar ARGS2(
long, unicode,
int, charset_out)
{
int rc = 0;
int UChndl_out;
int isdefault, trydefault = 0;
CONST u16 * ut;
if ((UChndl_out = LYCharSet_UC[charset_out].UChndl) < 0) {
if ((UChndl_out = default_UChndl) < 0)
return -12;
isdefault = 1;
} else {
isdefault = UCInfo[UChndl_out].replacedesc.isdefault;
trydefault = UCInfo[UChndl_out].replacedesc.trydefault;
}
if (!isdefault) {
ut = UCInfo[UChndl_out].unitable;
if (ut != UC_current_unitable) {
rc = UC_con_set_unimap(UChndl_out, 1);
if (rc < 0) {
return rc;
}
}
rc = conv_uni_to_pc(unicode, 0);
if (rc >= 0)
return rc;
}
if (isdefault || trydefault) {
rc = conv_uni_to_pc(unicode, 1);
if (rc >= 0)
return rc;
}
if (!isdefault && (rc == -4)) {
rc = conv_uni_to_pc(0xfffd, 0);
}
if ((isdefault || trydefault) && (rc == -4)) {
rc = conv_uni_to_pc(0xfffd, 1);
}
return rc;
}
/*
* Returns string length, or negative value for error.
*/
PUBLIC int UCTransUniCharStr ARGS5(
char *, outbuf,
int, buflen,
long, unicode,
int, charset_out,
int, chk_single_flag)
{
int rc = -14, src = 0, ignore_err;
int UChndl_out;
int isdefault, trydefault = 0;
struct unimapdesc_str * repl;
CONST u16 * ut;
if (buflen < 2)
return -13;
if ((UChndl_out = LYCharSet_UC[charset_out].UChndl) < 0) {
if ((UChndl_out = default_UChndl) < 0)
return -12;
isdefault = 1;
} else {
isdefault = UCInfo[UChndl_out].replacedesc.isdefault;
trydefault = UCInfo[UChndl_out].replacedesc.trydefault;
}
if (chk_single_flag) {
if (!isdefault) {
ut = UCInfo[UChndl_out].unitable;
if (ut != UC_current_unitable) {
src = UC_con_set_unimap(UChndl_out, 1);
if (src < 0) {
return src;
}
}
}
src = conv_uni_to_pc(unicode, isdefault);
if (src >= 32) {
outbuf[0] = src; outbuf[1] = '\0';
return 1;
}
}
repl = &(UCInfo[UChndl_out].replacedesc);
if (!isdefault) {
if (repl != UC_current_unitable_str) {
con_clear_unimap_str(0);
ignore_err = UC_con_set_unimap_str(repl->entry_ct, repl->entries, 0);
UC_current_unitable_str = repl;
}
rc = conv_uni_to_str(outbuf, buflen, unicode, 0);
if (rc >= 0)
return (strlen(outbuf));
}
if (trydefault && chk_single_flag) {
src = conv_uni_to_pc(unicode, 1);
if (src >= 32) {
outbuf[0] = src; outbuf[1] = '\0';
return 1;
}
}
if (isdefault || trydefault) {
rc = conv_uni_to_str(outbuf, buflen, unicode, 1);
if (rc >= 0)
return (strlen(outbuf));
}
if (rc == -4) {
if (!isdefault)
rc = conv_uni_to_str(outbuf, buflen, 0xfffd, 0);
if ((rc == -4) && (isdefault || trydefault))
rc = conv_uni_to_str(outbuf, buflen, 0xfffd, 1);
if (rc >= 0)
return (strlen(outbuf));
}
if (chk_single_flag && src == -4) {
if (!isdefault)
rc = conv_uni_to_pc(0xfffd, 0);
if ((rc == -4) && (isdefault || trydefault))
rc = conv_uni_to_pc(0xfffd, 1);
if (rc >= 32) {
outbuf[0] = rc; outbuf[1] = '\0';
return 1;
}
return rc;
}
return -4;
}
PRIVATE int UC_lastautoGN = 0;
PRIVATE int UC_MapGN ARGS2(
int, UChndl,
int, update_flag)
{
int i, Gn, found, lasthndl;
found = 0;
Gn = -1;
for (i = 0; i < 4 && Gn < 0; i++) {
if (UC_GNhandles[i] < 0) {
Gn = i;
} else if (UC_GNhandles[i] == UChndl) {
Gn = i;
found = 1;
}
}
if (found)
return Gn;
if (Gn >= 0) {
UCInfo[UChndl].GN = Gn;
UC_GNhandles[Gn] = UChndl;
} else {
if (UC_lastautoGN == GRAF_MAP) {
Gn = IBMPC_MAP;
} else {
Gn = GRAF_MAP;
}
UC_lastautoGN = Gn;
lasthndl = UC_GNhandles[Gn];
UCInfo[lasthndl].GN = -1;
UCInfo[UChndl].GN = Gn;
UC_GNhandles[Gn] = UChndl;
}
CTRACE(tfp, "UC_MapGN: Using %d <- %d (%s)\n",
Gn, UChndl, UCInfo[UChndl].MIMEname);
UC_con_set_trans(UChndl,Gn,update_flag);
return Gn;
}
PUBLIC int UCTransChar ARGS3(
char, ch_in,
int, charset_in,
int, charset_out)
{
int unicode, Gn;
int rc = -4;
int UChndl_in, UChndl_out;
int isdefault, trydefault = 0;
CONST u16 * ut;
int upd = 0;
#ifndef UC_NO_SHORTCUTS
if (charset_in == charset_out)
return (unsigned char)ch_in;
#endif /* UC_NO_SHORTCUTS */
if (charset_in < 0)
return -11;
if ((UChndl_in = LYCharSet_UC[charset_in].UChndl) < 0)
return -11;
if ((UChndl_out = LYCharSet_UC[charset_out].UChndl) < 0) {
if ((UChndl_out = default_UChndl) < 0)
return -12;
isdefault = 1;
} else {
isdefault = UCInfo[UChndl_out].replacedesc.isdefault;
trydefault = UCInfo[UChndl_out].replacedesc.trydefault;
}
if (!UCInfo[UChndl_in].num_uni)
return -11;
if ((Gn = UCInfo[UChndl_in].GN) < 0) {
Gn = UC_MapGN(UChndl_in,0);
upd = 1;
}
ut = UCInfo[UChndl_out].unitable;
if (!isdefault) {
if (ut == UC_current_unitable) {
if (upd) {
set_inverse_transl(Gn);
}
} else {
rc = UC_con_set_unimap(UChndl_out, 1);
if (rc > 0) {
set_inverse_transl(Gn);
} else if (rc < 0) {
return rc;
}
}
}
UC_translate = set_translate(Gn);
unicode = UC_translate[(unsigned char)ch_in];
if (!isdefault) {
rc = conv_uni_to_pc(unicode, 0);
if (rc >= 0)
return rc;
}
if ((rc == -4) && (isdefault || trydefault)) {
rc = conv_uni_to_pc(unicode, 1);
}
if ((rc == -4) && !isdefault) {
rc = conv_uni_to_pc(0xfffd, 0);
}
if ((rc == -4) && (isdefault || trydefault)) {
rc = conv_uni_to_pc(0xfffd, 1);
}
return rc;
}
PUBLIC long int UCTransToUni ARGS2(
char, ch_in,
int, charset_in)
{
int unicode, Gn;
unsigned char ch_iu;
int UChndl_in;
ch_iu = (unsigned char)ch_in;
#ifndef UC_NO_SHORTCUTS
if (charset_in == LATIN1)
return ch_iu;
if ((unsigned char)ch_in < 128 && (unsigned char)ch_in >= 32)
return ch_iu;
#endif /* UC_NO_SHORTCUTS */
if (charset_in < 0)
return -11;
if ((unsigned char)ch_in < 32 &&
LYCharSet_UC[charset_in].enc != UCT_ENC_8BIT_C0)
/*
* Don't translate C0 chars except for specific charsets.
*/
return ch_iu;
if ((UChndl_in = LYCharSet_UC[charset_in].UChndl) < 0)
return -11;
if (!UCInfo[UChndl_in].num_uni)
return -11;
if ((Gn = UCInfo[UChndl_in].GN) < 0) {
Gn = UC_MapGN(UChndl_in,1);
}
UC_translate = set_translate(Gn);
unicode = UC_translate[(unsigned char)ch_in];
return unicode;
}
PUBLIC int UCReverseTransChar ARGS3(
char, ch_out,
int, charset_in,
int, charset_out)
{
int Gn;
int rc;
int UChndl_in, UChndl_out;
int isdefault;
int i_ch = (unsigned char)ch_out;
CONST u16 * ut;
#ifndef UC_NO_SHORTCUTS
if (charset_in == charset_out)
return ch_out;
#endif /* UC_NO_SHORTCUTS */
if (charset_in < 0)
return -11;
if ((UChndl_in = LYCharSet_UC[charset_in].UChndl) < 0)
return -11;
if (!UCInfo[UChndl_in].num_uni)
return -11;
if (charset_out < 0)
return -12;
if ((UChndl_out = LYCharSet_UC[charset_out].UChndl) < 0) {
if ((UChndl_out = default_UChndl) < 0)
return -12;
isdefault = 1;
} else {
isdefault = UCInfo[UChndl_out].replacedesc.isdefault;
}
if (!isdefault) {
/*
* Try to use the inverse table if charset_out is not equivalent
* to using just the default table. If it is, it should have
* just ASCII chars and trying to back-translate those should
* not give anything but themselves. - kw
*/
ut = UCInfo[UChndl_out].unitable;
if (ut == UC_current_unitable) {
if ((Gn = UCInfo[UChndl_in].GN) >= 0) {
UC_translate = set_translate(Gn);
rc = inv_translate[i_ch];
if (rc >= 32) {
return rc;
}
} else {
Gn = UC_MapGN(UChndl_in,1);
UC_translate = set_translate(Gn);
rc = inv_translate[i_ch];
if (rc >= 32) {
return rc;
}
}
}
}
return UCTransChar(ch_out, charset_out, charset_in);
}
/*
* Returns string length, or negative value for error.
*/
PUBLIC int UCTransCharStr ARGS6(
char *, outbuf,
int, buflen,
char, ch_in,
int, charset_in,
int, charset_out,
int, chk_single_flag)
{
int unicode, Gn;
int rc = -14, src = 0, ignore_err;
int UChndl_in, UChndl_out;
int isdefault, trydefault = 0;
struct unimapdesc_str * repl;
CONST u16 * ut;
int upd = 0;
if (buflen < 2)
return -13;
#ifndef UC_NO_SHORTCUTS
if (chk_single_flag && charset_in == charset_out) {
outbuf[0] = ch_in;
outbuf[1] = '\0';
return 1;
}
#endif /* UC_NO_SHORTCUTS */
if (charset_in < 0)
return -11;
if ((UChndl_in = LYCharSet_UC[charset_in].UChndl) < 0)
return -11;
if (!UCInfo[UChndl_in].num_uni)
return -11;
if ((UChndl_out = LYCharSet_UC[charset_out].UChndl) < 0) {
if ((UChndl_out = default_UChndl) < 0)
return -12;
isdefault = 1;
} else {
isdefault = UCInfo[UChndl_out].replacedesc.isdefault;
trydefault = UCInfo[UChndl_out].replacedesc.trydefault;
}
if ((Gn = UCInfo[UChndl_in].GN) < 0) {
Gn = UC_MapGN(UChndl_in, !chk_single_flag);
upd = chk_single_flag;
}
UC_translate = set_translate(Gn);
unicode = UC_translate[(unsigned char)ch_in];
if (chk_single_flag) {
if (!isdefault) {
ut = UCInfo[UChndl_out].unitable;
if (ut == UC_current_unitable) {
if (upd) set_inverse_transl(Gn);
} else {
src = UC_con_set_unimap(UChndl_out, 1);
if (src > 0) {
set_inverse_transl(Gn);
} else if (src < 0) {
return src;
}
}
}
src = conv_uni_to_pc(unicode, isdefault);
if (src >= 32) {
outbuf[0] = src; outbuf[1] = '\0';
return 1;
}
}
repl = &(UCInfo[UChndl_out].replacedesc);
if (!isdefault) {
if (repl != UC_current_unitable_str) {
con_clear_unimap_str(0);
ignore_err = UC_con_set_unimap_str(repl->entry_ct, repl->entries, 0);
UC_current_unitable_str = repl;
}
rc = conv_uni_to_str(outbuf, buflen, unicode, 0);
if (rc >= 0)
return (strlen(outbuf));
}
if (trydefault && chk_single_flag) {
src = conv_uni_to_pc(unicode, 1);
if (src >= 32) {
outbuf[0] = src; outbuf[1] = '\0';
return 1;
}
}
if (isdefault || trydefault) {
rc = conv_uni_to_str(outbuf, buflen, unicode, 1);
if (rc >= 0)
return (strlen(outbuf));
}
if (rc == -4) {
if (!isdefault)
rc = conv_uni_to_str(outbuf, buflen, 0xfffd, 0);
if ((rc == -4) && (isdefault || trydefault))
rc = conv_uni_to_str(outbuf, buflen, 0xfffd, 1);
if (rc >= 0)
return (strlen(outbuf));
}
if (chk_single_flag && src == -4) {
if (!isdefault)
rc = conv_uni_to_pc(0xfffd, 0);
if ((rc == -4) && (isdefault || trydefault))
rc = conv_uni_to_pc(0xfffd, 1);
if (rc >= 32) {
outbuf[0] = rc; outbuf[1] = '\0';
return 1;
} else if (rc <= 0) {
outbuf[0] = '\0';
return rc;
}
return rc;
}
return -4;
}
PRIVATE int UC_FindGN_byMIME ARGS1(
CONST char *, UC_MIMEcharset)
{
int i;
for (i = 0; i < 4; i++) {
if (!strcmp(UC_MIMEcharset,UC_GNsetMIMEnames[i])) {
return i;
}
}
return -1;
}
PUBLIC int UCGetRawUniMode_byLYhndl ARGS1(
int, i)
{
if (i < 0)
return 0;
return LYCharSet_UC[i].enc;
}
/*
* Get Lynx internal charset handler from MIME name,
* return -1 if we got NULL or did not recognize value.
* According to RFC, MIME headers should match case-insensitively.
*/
PUBLIC int UCGetLYhndl_byMIME ARGS1(
CONST char *, value)
{
int i;
int LYhndl = -1;
if (!value || !(*value)) {
CTRACE(tfp, "UCGetLYhndl_byMIME: NULL argument instead of MIME name.\n");
return -1;
}
for (i = 0;
(i < MAXCHARSETS && i < LYNumCharsets &&
LYchar_set_names[i]); i++) {
if (LYCharSet_UC[i].MIMEname &&
!strcasecomp(value, LYCharSet_UC[i].MIMEname)) {
return i;
}
}
/*
* Not yet found, try synonyms. - FM
*/
if (!strcasecomp(value, "unicode-1-1-utf-8") ||
!strcasecomp(value, "utf8")) {
/*
* Treat these as synonyms for the IANA registered name. - FM
*/
return UCGetLYhndl_byMIME("utf-8");
}
if (!strncasecomp(value, "iso-2022-jp", 11) ||
!strcasecomp(value, "x-euc-jp")) {
return UCGetLYhndl_byMIME("euc-jp");
}
if (!strcasecomp(value, "x-shift-jis")) {
return UCGetLYhndl_byMIME("shift_jis");
}
if (!strcasecomp(value, "iso-2022-kr")) {
return UCGetLYhndl_byMIME("euc-kr");
}
if (!strcasecomp(value, "gb2312") ||
!strncasecomp(value, "cn-gb", 5) ||
!strcasecomp(value, "iso-2022-cn")) {
return UCGetLYhndl_byMIME("euc-cn");
}
if (!strcasecomp(value, "cn-big5")) {
return UCGetLYhndl_byMIME("big5");
}
if (!strcasecomp(value, "x-mac-roman") ||
!strcasecomp(value, "mac-roman")) {
return UCGetLYhndl_byMIME("macintosh");
}
if (!strcasecomp(value, "x-next") ||
!strcasecomp(value, "nextstep") ||
!strcasecomp(value, "x-nextstep")) {
return UCGetLYhndl_byMIME("next");
}
if (!strcasecomp(value, "iso-8859-1-windows-3.1-latin-1") ||
!strcasecomp(value, "cp1252") ||
!strcasecomp(value, "cp-1252") ||
!strcasecomp(value, "ibm1252") ||
!strcasecomp(value, "iso-8859-1-windows-3.0-latin-1")) {
/*
* Treat these as synonyms for windows-1252, which is more
* commonly used than the IANA registered name. - FM
*/
return UCGetLYhndl_byMIME("windows-1252");
}
if (!strcasecomp(value, "iso-8859-2-windows-latin-2") ||
!strcasecomp(value, "cp1250") ||
!strcasecomp(value, "cp-1250") ||
!strcasecomp(value, "ibm1250")) {
/*
* Treat these as synonyms for windows-1250. - FM
*/
return UCGetLYhndl_byMIME("windows-1250");
}
if ((!strncasecomp(value, "ibm", 3) ||
!strncasecomp(value, "cp-", 3)) &&
isdigit((unsigned char)value[3]) &&
isdigit((unsigned char)value[4]) &&
isdigit((unsigned char)value[5])) {
/*
* For "ibmNNN<...>" or "cp-NNN", try "cpNNN<...>"
* if not yet found. - KW & FM
*/
char * cptmp = NULL;
StrAllocCopy(cptmp, (value + 1));
cptmp[0] = 'c';
cptmp[1] = 'p';
if ((LYhndl = UCGetLYhndl_byMIME(cptmp)) >= 0) {
FREE(cptmp);
return LYhndl;
}
/*
* Try windows-NNN<...> if not yet found. - FM
*/
StrAllocCopy(cptmp, "windows-");
StrAllocCat(cptmp, (value + 3));
LYhndl = UCGetLYhndl_byMIME(cptmp);
FREE(cptmp);
return LYhndl;
}
if (!strncasecomp(value, "windows-", 8) &&
isdigit((unsigned char)value[8]) &&
isdigit((unsigned char)value[9]) &&
isdigit((unsigned char)value[10])) {
/*
* For "windows-NNN<...>", try "cpNNN<...>" - FM
*/
char * cptmp = NULL;
StrAllocCopy(cptmp, (value + 6));
cptmp[0] = 'c';
cptmp[1] = 'p';
LYhndl = UCGetLYhndl_byMIME(cptmp);
FREE(cptmp);
return LYhndl;
}
if (!strcasecomp(value, "koi-8")) { /* accentsoft bugosity */
return UCGetLYhndl_byMIME("koi8-r");
}
/* no more synonyms if come here... */
CTRACE(tfp, "UCGetLYhndl_byMIME: unrecognized MIME name \"%s\"\n", value);
return -1; /* returns -1 if no charset found by that MIME name */
}
/*
* Function UC_setup_LYCharSets_repl() tries to set up a subtable in
* LYCharSets[] appropriate for this new charset, for compatibility
* with the "old method". Maybe not nice (maybe not even necessary
* any more), but it works (as far as it goes..).
*
* We try to be conservative and only allocate new memory for this
* if needed. If not needed, just point to SevenBitApproximations[i].
* [Could do the same for ISO_Latin1[] if it's identical to that, but
* would make it even *more* messy than it already is...]
* This the only function in this file that knows, or cares, about the
* HTMLDTD or details of LYCharSets[] subtables (and therefore somewhat
* violates the idea that this file should be independent of those).
* As in other places, we rely on ISO_Latin1 being the *first* table
* in LYCharSets. - KW
*/
/*
* We need to remember which ones were allocated and which are static.
*/
PRIVATE CONST char ** remember_allocated_LYCharSets[MAXCHARSETS];
PRIVATE void UCreset_allocated_LYCharSets NOARGS
{
int i = 0;
for (; i < MAXCHARSETS; i++) {
remember_allocated_LYCharSets[i] = NULL;
}
}
#ifdef LY_FIND_LEAKS
PRIVATE void UCfree_allocated_LYCharSets NOARGS
{
int i = 0;
for (; i < MAXCHARSETS; i++) {
if (remember_allocated_LYCharSets[i] != NULL) {
FREE(remember_allocated_LYCharSets[i]);
}
}
}
#endif
PRIVATE CONST char ** UC_setup_LYCharSets_repl ARGS2(
int, UC_charset_in_hndl,
unsigned, lowest8)
{
CONST char **ISO_Latin1 = LYCharSets[0];
CONST char **p;
char **prepl;
CONST u16 *pp;
CONST char **tp;
CONST char *s7;
CONST char *s8;
size_t i;
int j, changed;
u16 k;
u8 *ti;
/*
* Create a temporary table for reverse lookup of latin1 codes:
*/
tp = (CONST char **)malloc(96 * sizeof(CONST char *));
if (!tp)
return NULL;
for (i = 0; i < 96; i++)
tp[i] = NULL;
ti = (u8 *)malloc(96 * sizeof(u8));
if (!ti) {
FREE(tp);
return NULL;
}
for (i = 0; i < 96; i++)
ti[i] = 0;
pp = UCInfo[UC_charset_in_hndl].unitable;
/*
* Determine if we have any mapping of a Unicode in the range 160-255
* to an allowed code point > 0x80 in our new charset...
* Store any mappings found in ti[].
*/
if (UCInfo[UC_charset_in_hndl].num_uni > 0) {
for (i = 0; i < 256; i++) {
if ((j = UCInfo[UC_charset_in_hndl].unicount[i])) {
if ((k = *pp) >= 160 && k < 256 && i >= lowest8) {
ti[k-160] = i;
}
for (; j; j--) {
pp++;
}
}
}
}
{
u16 ct;
struct unipair_str *list;
/*
* Determine if we have any mapping of a Unicode in the range
* 160-255 to a replacement string for our new charset...
* Store any mappings found in tp[].
*/
ct = UCInfo[UC_charset_in_hndl].replacedesc.entry_ct;
list = UCInfo[UC_charset_in_hndl].replacedesc.entries;
while (ct--) {
if ((k = list->unicode) >= 160 && k < 256) {
tp[k-160] = list->replace_str;
}
list++;
}
}
/*
* Now allocate a new table compatible with LYCharSets[]
* and with the HTMLDTD for entities.
* We don't know yet whether we'll keep it around. */
prepl = (char **)malloc(HTML_dtd.number_of_entities * sizeof(char *));
if (!prepl) {
FREE(tp);
FREE(ti);
return 0;
}
p = (CONST char **)prepl;
changed = 0;
for (i = 0; i < HTML_dtd.number_of_entities; i++, p++) {
/*
* For each of those entities, we check what the "old method"
* ISO_Latin1[] mapping does with them. If it is nothing we
* want to use, just point to the SevenBitApproximations[] string.
*/
s7 = SevenBitApproximations[i];
s8 = ISO_Latin1[i];
*p = s7;
if (s8 && (unsigned char)(*s8) >= 160 && strlen(s8) == 1) {
/*
* We have an entity that is mapped to
* one valid eightbit latin1 char.
*/
if (ti[(unsigned char)(*s8) - 160] >= lowest8 &&
!(s7[0] == ti[(unsigned char)(*s8) - 160] &&
s7[1] == '\0')) {
/*
* ...which in turn is mapped, by our "new method",
* to another valid eightbit char for this new
* charset: either to itself...
*/
if (ti[(unsigned char)(*s8) - 160] == (unsigned char)(*s8)) {
*p = s8;
} else {
/*
* ...or another byte...
*/
#ifdef NOTDEFINED
*p = (char *)malloc(2*sizeof(char));
if (!*p) {
FREE(tp);
FREE(ti);
FREE(prepl);
return NULL;
}
(*p)[0] = ti[(unsigned char)(*s8) - 160];
(*p)[1] = '\0';
#else
/*
* Use this instead... make those 1-char strings
* into HTAtoms, so they will be cleaned up
* at exit... all for the sake of preventing
* memory leaks, sigh.
*/
static char dummy[2]; /* one char dummy string */
dummy[0] = ti[(unsigned char)(*s8) - 160];
*p = HTAtom_name(HTAtom_for(dummy));
#endif /* NOTDEFINED */
}
changed = 1;
} else if (tp[(unsigned char)(*s8) - 160] &&
strcmp(s7, tp[(unsigned char)(*s8) - 160])) {
/*
* ...or which is mapped, by our "new method",
* to a replacement string for this new charset.
*/
*p = tp[(unsigned char)(*s8) - 160];
changed = 1;
}
}
}
FREE(tp);
FREE(ti);
if (!changed) {
FREE(prepl);
return NULL;
}
return (CONST char **)prepl;
}
/*
* "New method" meets "Old method" ...
*/
PRIVATE int UC_Register_with_LYCharSets ARGS4(
int, s,
CONST char *, UC_MIMEcharset,
CONST char *, UC_LYNXcharset,
int, lowest_eightbit)
{
int i, LYhndl, found;
CONST char **repl;
LYhndl = -1;
if (LYNumCharsets == 0) {
/*
* Initialize here; so whoever changes
* LYCharSets.c doesn't have to count...
*/
for (i = 0; (i < MAXCHARSETS) && LYchar_set_names[i]; i++) {
LYNumCharsets = i+1;
}
}
/*
* Search by MIME name, (LYchar_set_names may differ...)
*/
for (i = 0; i < MAXCHARSETS && LYchar_set_names[i] && LYhndl < 0; i++) {
if (LYCharSet_UC[i].MIMEname &&
!strcmp(UC_MIMEcharset, LYCharSet_UC[i].MIMEname)) {
LYhndl = i;
}
}
if (LYhndl < 0) { /* not found */
found = 0;
if (LYNumCharsets >= MAXCHARSETS) {
CTRACE(tfp, "UC_Register_with_LYCharSets: Too many. Ignoring %s/%s.",
UC_MIMEcharset, UC_LYNXcharset);
return -1;
}
/*
* Add to LYCharSets.c lists.
*/
LYhndl = LYNumCharsets;
LYNumCharsets ++;
LYlowest_eightbit[LYhndl] = 999;
LYCharSets[LYhndl] = SevenBitApproximations;
/*
* Hmm, try to be conservative here.
*/
LYchar_set_names[LYhndl] = UC_LYNXcharset;
LYchar_set_names[LYhndl+1] = NULL;
/*
* Terminating NULL may be looked for by Lynx code.
*/
} else {
found = 1;
}
LYCharSet_UC[LYhndl].UChndl = s;
/*
* Can we just copy the pointer? Hope so...
*/
LYCharSet_UC[LYhndl].MIMEname = UC_MIMEcharset;
LYCharSet_UC[LYhndl].enc = UCInfo[s].enc;
LYCharSet_UC[LYhndl].codepage = UCInfo[s].codepage;
/*
* @@@ We really SHOULD get more info from the table files,
* and set relevant flags in the LYCharSet_UC[] entry with
* that info... For now, let's try it without. - KW
*/
if (lowest_eightbit < LYlowest_eightbit[LYhndl]) {
LYlowest_eightbit[LYhndl] = lowest_eightbit;
} else if (lowest_eightbit > LYlowest_eightbit[LYhndl]) {
UCInfo[s].lowest_eight = LYlowest_eightbit[LYhndl];
}
if (!found && LYhndl > 0) {
repl = UC_setup_LYCharSets_repl(s,UCInfo[s].lowest_eight);
if (repl) {
LYCharSets[LYhndl] = repl;
/*
* Remember to FREE at exit.
*/
remember_allocated_LYCharSets[LYhndl] = repl;
}
}
return LYhndl;
}
/*
* This only sets up the structure - no initialization of the tables
* is done here yet.
*/
PUBLIC void UC_Charset_Setup ARGS9(
CONST char *, UC_MIMEcharset,
CONST char *, UC_LYNXcharset,
CONST u8 *, unicount,
CONST u16 *, unitable,
int, nnuni,
struct unimapdesc_str, replacedesc,
int, lowest_eight,
int, UC_rawuni,
int, codepage)
{
int s, Gn;
int i, status = 0, found;
/*
* Get (new?) slot.
*/
found = -1;
for (i = 0; i < UCNumCharsets && found < 0; i++) {
if (!strcmp(UCInfo[i].MIMEname, UC_MIMEcharset)) {
found = i;
}
}
if (found >= 0) {
s = found;
} else {
if (UCNumCharsets >= MAXCHARSETS) {
CTRACE(tfp, "UC_Charset_Setup: Too many. Ignoring %s/%s.",
UC_MIMEcharset, UC_LYNXcharset);
return;
}
s = UCNumCharsets;
UCInfo[s].MIMEname = UC_MIMEcharset;
}
UCInfo[s].LYNXname = UC_LYNXcharset;
UCInfo[s].unicount = unicount;
UCInfo[s].unitable = unitable;
UCInfo[s].num_uni = nnuni;
UCInfo[s].replacedesc = replacedesc;
if (replacedesc.isdefault) {
default_UChndl = s;
}
Gn = UC_FindGN_byMIME(UC_MIMEcharset);
if (Gn >= 0)
UC_GNhandles[Gn] = s;
UCInfo[s].GN = Gn;
if (UC_rawuni == UCT_ENC_UTF8)
lowest_eight = 128; /* cheat here */
UCInfo[s].lowest_eight = lowest_eight;
UCInfo[s].enc = UC_rawuni;
UCInfo[s].codepage = codepage;
UCInfo[s].LYhndl = UC_Register_with_LYCharSets(s,
UC_MIMEcharset,
UC_LYNXcharset,
lowest_eight);
UCInfo[s].uc_status = status;
if (found < 0)
UCNumCharsets++;
return;
}
#ifdef LY_FIND_LEAKS
PRIVATE void UCcleanup_mem NOARGS
{
int i;
UCfree_allocated_LYCharSets();
con_clear_unimap_str(0);
con_clear_unimap_str(1);
con_clear_unimap(0);
con_clear_unimap(1);
for (i = 1; i < 4; i++) { /* first one is static! */
FREE(inverse_translations[i]);
}
}
#endif /* LY_FIND_LEAKS */
PUBLIC void UCInit NOARGS
{
UCreset_allocated_LYCharSets();
#ifdef LY_FIND_LEAKS
atexit(UCcleanup_mem);
#endif
UCconsole_map_init();
/*
* The order of charset names visible in Lynx Options menu
* correspond to the order of lines below,
* except the first two described in LYCharSet.c
*
* Entries whose comment is marked with *** are declared in UCdomap.h,
* others are based on the included tables - UCdomap.c, near the top.
*/
UC_CHARSET_SETUP_iso_8859_1; /* ISO Latin 1 */
UC_CHARSET_SETUP_iso_8859_15; /* ISO 8859-15 (Latin 9)*/
UC_CHARSET_SETUP_cp850; /* DosLatin1 (cp850) */
UC_CHARSET_SETUP_windows_1252; /* WinLatin1 (cp1252) */
UC_CHARSET_SETUP_cp437; /* DosLatinUS (cp437) */
UC_CHARSET_SETUP_dec_mcs; /* DEC Multinational */
UC_CHARSET_SETUP_macintosh; /* Macintosh (8 bit) */
UC_CHARSET_SETUP_next; /* NeXT character set */
UC_CHARSET_SETUP_hp_roman8; /* HP Roman8 */
UC_CHARSET_SETUP_euc_cn; /*** Chinese */
UC_CHARSET_SETUP_euc_jp; /*** Japanese (EUC_JP) */
UC_CHARSET_SETUP_shift_jis; /*** Japanese (Shift_JIS) */
UC_CHARSET_SETUP_euc_kr; /*** Korean */
UC_CHARSET_SETUP_big5; /*** Taipei (Big5) */
UC_CHARSET_SETUP_viscii; /* Vietnamese (VISCII) */
UC_CHARSET_SETUP; /* us-ascii */ /* 7 bit approximations */
UC_CHARSET_SETUP_x_transparent; /*** Transparent */
UC_CHARSET_SETUP_iso_8859_2; /* ISO Latin 2 */
UC_CHARSET_SETUP_cp852; /* DosLatin2 (cp852) */
UC_CHARSET_SETUP_windows_1250; /* WinLatin2 (cp1250) */
UC_CHARSET_SETUP_iso_8859_3; /* ISO Latin 3 */
UC_CHARSET_SETUP_iso_8859_4; /* ISO Latin 4 */
UC_CHARSET_SETUP_cp775; /* DosBaltRim (cp775) */
UC_CHARSET_SETUP_windows_1257; /* WinBaltRim (cp1257) */
UC_CHARSET_SETUP_iso_8859_5; /* ISO 8859-5 Cyrillic */
UC_CHARSET_SETUP_cp866; /* DosCyrillic (cp866) */
UC_CHARSET_SETUP_windows_1251; /* WinCyrillic (cp1251) */
UC_CHARSET_SETUP_koi8_r; /* KOI8-R Cyrillic */
UC_CHARSET_SETUP_iso_8859_6; /* ISO 8869-6 Arabic */
UC_CHARSET_SETUP_cp864; /* DosArabic (cp864) */
UC_CHARSET_SETUP_windows_1256; /* WinArabic (cp1256) */
UC_CHARSET_SETUP_iso_8859_7; /* ISO 8859-7 Greek */
UC_CHARSET_SETUP_cp737; /* DosGreek (cp737) */
UC_CHARSET_SETUP_cp869; /* DosGreek2 (cp869) */
UC_CHARSET_SETUP_windows_1253; /* WinGreek (cp1253) */
UC_CHARSET_SETUP_iso_8859_8; /* ISO 8859-8 Hebrew */
UC_CHARSET_SETUP_cp862; /* DosHebrew (cp862) */
UC_CHARSET_SETUP_windows_1255; /* WinHebrew (cp1255) */
UC_CHARSET_SETUP_iso_8859_9; /* ISO 8859-9 (Latin 5) */
UC_CHARSET_SETUP_iso_8859_10; /* ISO 8859-10 */
UC_CHARSET_SETUP_utf_8; /*** UNICODE UTF-8 */
UC_CHARSET_SETUP_mnemonic_ascii_0; /* RFC 1345 w/o Intro */
UC_CHARSET_SETUP_mnemonic; /* RFC 1345 Mnemonic */
UC_CHARSET_SETUP_cp866u; /* Ukrainian Cyrillic (866) */
UC_CHARSET_SETUP_koi8_u; /* Ukrainian Cyrillic (koi8-u) */
#ifdef NOTDEFINED
UC_CHARSET_SETUP_mnem;
#endif /* NOTDEFINED */
/*
* To add synonyms for any charset name
* check function UCGetLYhndl_byMIME in this file.
*/
/* easy to type: */
LATIN1 = UCGetLYhndl_byMIME("iso-8859-1");
US_ASCII = UCGetLYhndl_byMIME("us-ascii");
UTF8 = UCGetLYhndl_byMIME("utf-8");
}
/*
* Safe variant of UCGetLYhndl_byMIME, with blind recovery from typo
* in user input: lynx.cfg, userdefs.h, switches from command line.
*/
PUBLIC int safeUCGetLYhndl_byMIME ARGS1 (CONST char *, value)
{
int i = UCGetLYhndl_byMIME(value);
if (i == -1) { /* was user's typo or not yet recognized value */
i = LATIN1; /* error recovery? */
CTRACE(tfp, "safeUCGetLYhndl_byMIME: ISO-8859-1 assumed.\n");
}
return(i);
}