Kerala Cyber
Warriors
KCW Uploader V1.1
/* setlocale() function that respects the locale chosen by the user.
Copyright (C) 2009, 2015-2016 Free Software Foundation, Inc.
Written by Bruno Haible <bruno@clisp.org>, 2009.
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Lesser General Public License as published by
the Free Software Foundation; either version 2.1 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>. */
#ifdef HAVE_CONFIG_H
# include <config.h>
#endif
/* Override setlocale() and newlocale() so that when the default locale is
requested (locale = "") and no relevant environment variable is set, the
locale chosen by the user is used.
This matters on MacOS X 10 and Windows.
See the comments in localename.c, function gl_locale_name_default. */
#include <locale.h>
#include <stdlib.h>
#include <string.h>
/* When building a DLL, we must export some functions. Note that because
the functions are only defined for binary backward compatibility, we
don't need to use __declspec(dllimport) in any case. */
#if HAVE_VISIBILITY && BUILDING_DLL
# define DLL_EXPORTED __attribute__((__visibility__("default")))
#elif defined _MSC_VER && BUILDING_DLL
# define DLL_EXPORTED __declspec(dllexport)
#else
# define DLL_EXPORTED
#endif
#include "gettextP.h"
#if (defined __APPLE__ && defined __MACH__) || defined _WIN32 || defined __WIN32__ || defined __CYGWIN__
# undef setlocale
# undef newlocale
/* Return string representation of locale category CATEGORY. */
static const char *
category_to_name (int category)
{
const char *retval;
switch (category)
{
case LC_COLLATE:
retval = "LC_COLLATE";
break;
case LC_CTYPE:
retval = "LC_CTYPE";
break;
case LC_MONETARY:
retval = "LC_MONETARY";
break;
case LC_NUMERIC:
retval = "LC_NUMERIC";
break;
case LC_TIME:
retval = "LC_TIME";
break;
case LC_MESSAGES:
retval = "LC_MESSAGES";
break;
default:
/* If you have a better idea for a default value let me know. */
retval = "LC_XXX";
}
return retval;
}
# if (defined _WIN32 || defined __WIN32__) && ! defined __CYGWIN__
/* The native Win32 setlocale() function expects locale names of the form
"German" or "German_Germany" or "DEU", but not "de" or "de_DE". We need
to convert the names from the form with ISO 639 language code and ISO 3166
country code to the form with English names or with three-letter identifier.
The three-letter identifiers known by a Windows XP SP2 or SP3 are:
AFK Afrikaans_South Africa.1252
ARA Arabic_Saudi Arabia.1256
ARB Arabic_Lebanon.1256
ARE Arabic_Egypt.1256
ARG Arabic_Algeria.1256
ARH Arabic_Bahrain.1256
ARI Arabic_Iraq.1256
ARJ Arabic_Jordan.1256
ARK Arabic_Kuwait.1256
ARL Arabic_Libya.1256
ARM Arabic_Morocco.1256
ARO Arabic_Oman.1256
ARQ Arabic_Qatar.1256
ARS Arabic_Syria.1256
ART Arabic_Tunisia.1256
ARU Arabic_U.A.E..1256
ARY Arabic_Yemen.1256
AZE Azeri (Latin)_Azerbaijan.1254
BEL Belarusian_Belarus.1251
BGR Bulgarian_Bulgaria.1251
BSB Bosnian_Bosnia and Herzegovina.1250
BSC Bosnian (Cyrillic)_Bosnia and Herzegovina.1250 (wrong encoding!)
CAT Catalan_Spain.1252
CHH Chinese_Hong Kong S.A.R..950
CHI Chinese_Singapore.936
CHS Chinese_People's Republic of China.936
CHT Chinese_Taiwan.950
CSY Czech_Czech Republic.1250
CYM Welsh_United Kingdom.1252
DAN Danish_Denmark.1252
DEA German_Austria.1252
DEC German_Liechtenstein.1252
DEL German_Luxembourg.1252
DES German_Switzerland.1252
DEU German_Germany.1252
ELL Greek_Greece.1253
ENA English_Australia.1252
ENB English_Caribbean.1252
ENC English_Canada.1252
ENG English_United Kingdom.1252
ENI English_Ireland.1252
ENJ English_Jamaica.1252
ENL English_Belize.1252
ENP English_Republic of the Philippines.1252
ENS English_South Africa.1252
ENT English_Trinidad and Tobago.1252
ENU English_United States.1252
ENW English_Zimbabwe.1252
ENZ English_New Zealand.1252
ESA Spanish_Panama.1252
ESB Spanish_Bolivia.1252
ESC Spanish_Costa Rica.1252
ESD Spanish_Dominican Republic.1252
ESE Spanish_El Salvador.1252
ESF Spanish_Ecuador.1252
ESG Spanish_Guatemala.1252
ESH Spanish_Honduras.1252
ESI Spanish_Nicaragua.1252
ESL Spanish_Chile.1252
ESM Spanish_Mexico.1252
ESN Spanish_Spain.1252
ESO Spanish_Colombia.1252
ESP Spanish_Spain.1252
ESR Spanish_Peru.1252
ESS Spanish_Argentina.1252
ESU Spanish_Puerto Rico.1252
ESV Spanish_Venezuela.1252
ESY Spanish_Uruguay.1252
ESZ Spanish_Paraguay.1252
ETI Estonian_Estonia.1257
EUQ Basque_Spain.1252
FAR Farsi_Iran.1256
FIN Finnish_Finland.1252
FOS Faroese_Faroe Islands.1252
FPO Filipino_Philippines.1252
FRA French_France.1252
FRB French_Belgium.1252
FRC French_Canada.1252
FRL French_Luxembourg.1252
FRM French_Principality of Monaco.1252
FRS French_Switzerland.1252
FYN Frisian_Netherlands.1252
GLC Galician_Spain.1252
HEB Hebrew_Israel.1255
HRB Croatian_Bosnia and Herzegovina.1250
HRV Croatian_Croatia.1250
HUN Hungarian_Hungary.1250
IND Indonesian_Indonesia.1252
IRE Irish_Ireland.1252
ISL Icelandic_Iceland.1252
ITA Italian_Italy.1252
ITS Italian_Switzerland.1252
IUK Inuktitut (Latin)_Canada.1252
JPN Japanese_Japan.932
KKZ Kazakh_Kazakhstan.1251
KOR Korean_Korea.949
KYR Kyrgyz_Kyrgyzstan.1251
LBX Luxembourgish_Luxembourg.1252
LTH Lithuanian_Lithuania.1257
LVI Latvian_Latvia.1257
MKI FYRO Macedonian_Former Yugoslav Republic of Macedonia.1251
MON Mongolian_Mongolia.1251
MPD Mapudungun_Chile.1252
MSB Malay_Brunei Darussalam.1252
MSL Malay_Malaysia.1252
MWK Mohawk_Canada.1252
NLB Dutch_Belgium.1252
NLD Dutch_Netherlands.1252
NON Norwegian-Nynorsk_Norway.1252
NOR Norwegian (Bokmål)_Norway.1252
NSO Northern Sotho_South Africa.1252
PLK Polish_Poland.1250
PTB Portuguese_Brazil.1252
PTG Portuguese_Portugal.1252
QUB Quechua_Bolivia.1252
QUE Quechua_Ecuador.1252
QUP Quechua_Peru.1252
RMC Romansh_Switzerland.1252
ROM Romanian_Romania.1250
RUS Russian_Russia.1251
SKY Slovak_Slovakia.1250
SLV Slovenian_Slovenia.1250
SMA Sami (Southern)_Norway.1252
SMB Sami (Southern)_Sweden.1252
SME Sami (Northern)_Norway.1252
SMF Sami (Northern)_Sweden.1252
SMG Sami (Northern)_Finland.1252
SMJ Sami (Lule)_Norway.1252
SMK Sami (Lule)_Sweden.1252
SMN Sami (Inari)_Finland.1252
SMS Sami (Skolt)_Finland.1252
SQI Albanian_Albania.1250
SRB Serbian (Cyrillic)_Serbia and Montenegro.1251
SRL Serbian (Latin)_Serbia and Montenegro.1250
SRN Serbian (Cyrillic)_Bosnia and Herzegovina.1251
SRS Serbian (Latin)_Bosnia and Herzegovina.1250
SVE Swedish_Sweden.1252
SVF Swedish_Finland.1252
SWK Swahili_Kenya.1252
THA Thai_Thailand.874
TRK Turkish_Turkey.1254
TSN Tswana_South Africa.1252
TTT Tatar_Russia.1251
UKR Ukrainian_Ukraine.1251
URD Urdu_Islamic Republic of Pakistan.1256
USA English_United States.1252
UZB Uzbek (Latin)_Uzbekistan.1254
VIT Vietnamese_Viet Nam.1258
XHO Xhosa_South Africa.1252
ZHH Chinese_Hong Kong S.A.R..950
ZHI Chinese_Singapore.936
ZHM Chinese_Macau S.A.R..950
ZUL Zulu_South Africa.1252
*/
/* Table from ISO 639 language code, optionally with country or script suffix,
to English name.
Keep in sync with the gl_locale_name_from_win32_LANGID function in
localename.c! */
struct table_entry
{
const char *code;
const char *english;
};
static const struct table_entry language_table[] =
{
{ "af", "Afrikaans" },
{ "am", "Amharic" },
{ "ar", "Arabic" },
{ "arn", "Mapudungun" },
{ "as", "Assamese" },
{ "az@cyrillic", "Azeri (Cyrillic)" },
{ "az@latin", "Azeri (Latin)" },
{ "ba", "Bashkir" },
{ "be", "Belarusian" },
{ "ber", "Tamazight" },
{ "ber@arabic", "Tamazight (Arabic)" },
{ "ber@latin", "Tamazight (Latin)" },
{ "bg", "Bulgarian" },
{ "bin", "Edo" },
{ "bn", "Bengali" },
{ "bn_BD", "Bengali (Bangladesh)" },
{ "bn_IN", "Bengali (India)" },
{ "bnt", "Sutu" },
{ "bo", "Tibetan" },
{ "br", "Breton" },
{ "bs", "BSB" }, /* "Bosnian (Latin)" */
{ "bs@cyrillic", "BSC" }, /* Bosnian (Cyrillic) */
{ "ca", "Catalan" },
{ "chr", "Cherokee" },
{ "co", "Corsican" },
{ "cpe", "Hawaiian" },
{ "cs", "Czech" },
{ "cy", "Welsh" },
{ "da", "Danish" },
{ "de", "German" },
{ "dsb", "Lower Sorbian" },
{ "dv", "Divehi" },
{ "el", "Greek" },
{ "en", "English" },
{ "es", "Spanish" },
{ "et", "Estonian" },
{ "eu", "Basque" },
{ "fa", "Farsi" },
{ "ff", "Fulfulde" },
{ "fi", "Finnish" },
{ "fo", "Faroese" }, /* "Faeroese" does not work */
{ "fr", "French" },
{ "fy", "Frisian" },
{ "ga", "IRE" }, /* Gaelic (Ireland) */
{ "gd", "Gaelic (Scotland)" },
{ "gd", "Scottish Gaelic" },
{ "gl", "Galician" },
{ "gn", "Guarani" },
{ "gsw", "Alsatian" },
{ "gu", "Gujarati" },
{ "ha", "Hausa" },
{ "he", "Hebrew" },
{ "hi", "Hindi" },
{ "hr", "Croatian" },
{ "hsb", "Upper Sorbian" },
{ "hu", "Hungarian" },
{ "hy", "Armenian" },
{ "id", "Indonesian" },
{ "ig", "Igbo" },
{ "ii", "Yi" },
{ "is", "Icelandic" },
{ "it", "Italian" },
{ "iu", "IUK" }, /* Inuktitut */
{ "ja", "Japanese" },
{ "ka", "Georgian" },
{ "kk", "Kazakh" },
{ "kl", "Greenlandic" },
{ "km", "Cambodian" },
{ "km", "Khmer" },
{ "kn", "Kannada" },
{ "ko", "Korean" },
{ "kok", "Konkani" },
{ "kr", "Kanuri" },
{ "ks", "Kashmiri" },
{ "ks_IN", "Kashmiri_India" },
{ "ks_PK", "Kashmiri (Arabic)_Pakistan" },
{ "ky", "Kyrgyz" },
{ "la", "Latin" },
{ "lb", "Luxembourgish" },
{ "lo", "Lao" },
{ "lt", "Lithuanian" },
{ "lv", "Latvian" },
{ "mi", "Maori" },
{ "mk", "FYRO Macedonian" },
{ "mk", "Macedonian" },
{ "ml", "Malayalam" },
{ "mn", "Mongolian" },
{ "mni", "Manipuri" },
{ "moh", "Mohawk" },
{ "mr", "Marathi" },
{ "ms", "Malay" },
{ "mt", "Maltese" },
{ "my", "Burmese" },
{ "nb", "NOR" }, /* Norwegian Bokmål */
{ "ne", "Nepali" },
{ "nic", "Ibibio" },
{ "nl", "Dutch" },
{ "nn", "NON" }, /* Norwegian Nynorsk */
{ "no", "Norwegian" },
{ "nso", "Northern Sotho" },
{ "nso", "Sepedi" },
{ "oc", "Occitan" },
{ "om", "Oromo" },
{ "or", "Oriya" },
{ "pa", "Punjabi" },
{ "pap", "Papiamentu" },
{ "pl", "Polish" },
{ "prs", "Dari" },
{ "ps", "Pashto" },
{ "pt", "Portuguese" },
{ "qu", "Quechua" },
{ "qut", "K'iche'" },
{ "rm", "Romansh" },
{ "ro", "Romanian" },
{ "ru", "Russian" },
{ "rw", "Kinyarwanda" },
{ "sa", "Sanskrit" },
{ "sah", "Yakut" },
{ "sd", "Sindhi" },
{ "se", "Sami (Northern)" },
{ "se", "Northern Sami" },
{ "si", "Sinhalese" },
{ "sk", "Slovak" },
{ "sl", "Slovenian" },
{ "sma", "Sami (Southern)" },
{ "sma", "Southern Sami" },
{ "smj", "Sami (Lule)" },
{ "smj", "Lule Sami" },
{ "smn", "Sami (Inari)" },
{ "smn", "Inari Sami" },
{ "sms", "Sami (Skolt)" },
{ "sms", "Skolt Sami" },
{ "so", "Somali" },
{ "sq", "Albanian" },
{ "sr", "Serbian (Latin)" },
{ "sr@cyrillic", "SRB" }, /* Serbian (Cyrillic) */
{ "sw", "Swahili" },
{ "syr", "Syriac" },
{ "ta", "Tamil" },
{ "te", "Telugu" },
{ "tg", "Tajik" },
{ "th", "Thai" },
{ "ti", "Tigrinya" },
{ "tk", "Turkmen" },
{ "tl", "Filipino" },
{ "tn", "Tswana" },
{ "tr", "Turkish" },
{ "ts", "Tsonga" },
{ "tt", "Tatar" },
{ "ug", "Uighur" },
{ "uk", "Ukrainian" },
{ "ur", "Urdu" },
{ "uz", "Uzbek" },
{ "uz", "Uzbek (Latin)" },
{ "uz@cyrillic", "Uzbek (Cyrillic)" },
{ "ve", "Venda" },
{ "vi", "Vietnamese" },
{ "wen", "Sorbian" },
{ "wo", "Wolof" },
{ "xh", "Xhosa" },
{ "yi", "Yiddish" },
{ "yo", "Yoruba" },
{ "zh", "Chinese" },
{ "zu", "Zulu" }
};
/* Table from ISO 3166 country code to English name.
Keep in sync with the gl_locale_name_from_win32_LANGID function in
localename.c! */
static const struct table_entry country_table[] =
{
{ "AE", "U.A.E." },
{ "AF", "Afghanistan" },
{ "AL", "Albania" },
{ "AM", "Armenia" },
{ "AN", "Netherlands Antilles" },
{ "AR", "Argentina" },
{ "AT", "Austria" },
{ "AU", "Australia" },
{ "AZ", "Azerbaijan" },
{ "BA", "Bosnia and Herzegovina" },
{ "BD", "Bangladesh" },
{ "BE", "Belgium" },
{ "BG", "Bulgaria" },
{ "BH", "Bahrain" },
{ "BN", "Brunei Darussalam" },
{ "BO", "Bolivia" },
{ "BR", "Brazil" },
{ "BT", "Bhutan" },
{ "BY", "Belarus" },
{ "BZ", "Belize" },
{ "CA", "Canada" },
{ "CG", "Congo" },
{ "CH", "Switzerland" },
{ "CI", "Cote d'Ivoire" },
{ "CL", "Chile" },
{ "CM", "Cameroon" },
{ "CN", "People's Republic of China" },
{ "CO", "Colombia" },
{ "CR", "Costa Rica" },
{ "CS", "Serbia and Montenegro" },
{ "CZ", "Czech Republic" },
{ "DE", "Germany" },
{ "DK", "Denmark" },
{ "DO", "Dominican Republic" },
{ "DZ", "Algeria" },
{ "EC", "Ecuador" },
{ "EE", "Estonia" },
{ "EG", "Egypt" },
{ "ER", "Eritrea" },
{ "ES", "Spain" },
{ "ET", "Ethiopia" },
{ "FI", "Finland" },
{ "FO", "Faroe Islands" },
{ "FR", "France" },
{ "GB", "United Kingdom" },
{ "GD", "Caribbean" },
{ "GE", "Georgia" },
{ "GL", "Greenland" },
{ "GR", "Greece" },
{ "GT", "Guatemala" },
{ "HK", "Hong Kong" },
{ "HK", "Hong Kong S.A.R." },
{ "HN", "Honduras" },
{ "HR", "Croatia" },
{ "HT", "Haiti" },
{ "HU", "Hungary" },
{ "ID", "Indonesia" },
{ "IE", "Ireland" },
{ "IL", "Israel" },
{ "IN", "India" },
{ "IQ", "Iraq" },
{ "IR", "Iran" },
{ "IS", "Iceland" },
{ "IT", "Italy" },
{ "JM", "Jamaica" },
{ "JO", "Jordan" },
{ "JP", "Japan" },
{ "KE", "Kenya" },
{ "KG", "Kyrgyzstan" },
{ "KH", "Cambodia" },
{ "KR", "South Korea" },
{ "KW", "Kuwait" },
{ "KZ", "Kazakhstan" },
{ "LA", "Laos" },
{ "LB", "Lebanon" },
{ "LI", "Liechtenstein" },
{ "LK", "Sri Lanka" },
{ "LT", "Lithuania" },
{ "LU", "Luxembourg" },
{ "LV", "Latvia" },
{ "LY", "Libya" },
{ "MA", "Morocco" },
{ "MC", "Principality of Monaco" },
{ "MD", "Moldava" },
{ "MD", "Moldova" },
{ "ME", "Montenegro" },
{ "MK", "Former Yugoslav Republic of Macedonia" },
{ "ML", "Mali" },
{ "MM", "Myanmar" },
{ "MN", "Mongolia" },
{ "MO", "Macau S.A.R." },
{ "MT", "Malta" },
{ "MV", "Maldives" },
{ "MX", "Mexico" },
{ "MY", "Malaysia" },
{ "NG", "Nigeria" },
{ "NI", "Nicaragua" },
{ "NL", "Netherlands" },
{ "NO", "Norway" },
{ "NP", "Nepal" },
{ "NZ", "New Zealand" },
{ "OM", "Oman" },
{ "PA", "Panama" },
{ "PE", "Peru" },
{ "PH", "Philippines" },
{ "PK", "Islamic Republic of Pakistan" },
{ "PL", "Poland" },
{ "PR", "Puerto Rico" },
{ "PT", "Portugal" },
{ "PY", "Paraguay" },
{ "QA", "Qatar" },
{ "RE", "Reunion" },
{ "RO", "Romania" },
{ "RS", "Serbia" },
{ "RU", "Russia" },
{ "RW", "Rwanda" },
{ "SA", "Saudi Arabia" },
{ "SE", "Sweden" },
{ "SG", "Singapore" },
{ "SI", "Slovenia" },
{ "SK", "Slovak" },
{ "SN", "Senegal" },
{ "SO", "Somalia" },
{ "SR", "Suriname" },
{ "SV", "El Salvador" },
{ "SY", "Syria" },
{ "TH", "Thailand" },
{ "TJ", "Tajikistan" },
{ "TM", "Turkmenistan" },
{ "TN", "Tunisia" },
{ "TR", "Turkey" },
{ "TT", "Trinidad and Tobago" },
{ "TW", "Taiwan" },
{ "TZ", "Tanzania" },
{ "UA", "Ukraine" },
{ "US", "United States" },
{ "UY", "Uruguay" },
{ "VA", "Vatican" },
{ "VE", "Venezuela" },
{ "VN", "Viet Nam" },
{ "YE", "Yemen" },
{ "ZA", "South Africa" },
{ "ZW", "Zimbabwe" }
};
/* Given a string STRING, find the set of indices i such that TABLE[i].code is
the given STRING. It is a range [lo,hi-1]. */
typedef struct { size_t lo; size_t hi; } range_t;
static void
search (const struct table_entry *table, size_t table_size, const char *string,
range_t *result)
{
/* The table is sorted. Perform a binary search. */
size_t hi = table_size;
size_t lo = 0;
while (lo < hi)
{
/* Invariant:
for i < lo, strcmp (table[i].code, string) < 0,
for i >= hi, strcmp (table[i].code, string) > 0. */
size_t mid = (hi + lo) >> 1; /* >= lo, < hi */
int cmp = strcmp (table[mid].code, string);
if (cmp < 0)
lo = mid + 1;
else if (cmp > 0)
hi = mid;
else
{
/* Found an i with
strcmp (language_table[i].code, string) == 0.
Find the entire interval of such i. */
{
size_t i;
for (i = mid; i > lo; )
{
i--;
if (strcmp (table[i].code, string) < 0)
{
lo = i + 1;
break;
}
}
}
{
size_t i;
for (i = mid; i < hi; i++)
{
if (strcmp (table[i].code, string) > 0)
{
hi = i;
break;
}
}
}
/* The set of i with
strcmp (language_table[i].code, string) == 0
is the interval [lo, hi-1]. */
break;
}
}
result->lo = lo;
result->hi = hi;
}
/* Like setlocale, but accept also locale names in the form ll or ll_CC,
where ll is an ISO 639 language code and CC is an ISO 3166 country code. */
static char *
setlocale_unixlike (int category, const char *locale)
{
char *result;
char llCC_buf[64];
char ll_buf[64];
char CC_buf[64];
/* First, try setlocale with the original argument unchanged. */
result = setlocale (category, locale);
if (result != NULL)
return result;
/* Otherwise, assume the argument is in the form
language[_territory][.codeset][@modifier]
and try to map it using the tables. */
if (strlen (locale) < sizeof (llCC_buf))
{
/* Second try: Remove the codeset part. */
{
const char *p = locale;
char *q = llCC_buf;
/* Copy the part before the dot. */
for (; *p != '\0' && *p != '.'; p++, q++)
*q = *p;
if (*p == '.')
/* Skip the part up to the '@', if any. */
for (; *p != '\0' && *p != '@'; p++)
;
/* Copy the part starting with '@', if any. */
for (; *p != '\0'; p++, q++)
*q = *p;
*q = '\0';
}
/* llCC_buf now contains
language[_territory][@modifier]
*/
if (strcmp (llCC_buf, locale) != 0)
{
result = setlocale (category, llCC_buf);
if (result != NULL)
return result;
}
/* Look it up in language_table. */
{
range_t range;
size_t i;
search (language_table,
sizeof (language_table) / sizeof (language_table[0]),
llCC_buf,
&range);
for (i = range.lo; i < range.hi; i++)
{
/* Try the replacement in language_table[i]. */
result = setlocale (category, language_table[i].english);
if (result != NULL)
return result;
}
}
/* Split language[_territory][@modifier]
into ll_buf = language[@modifier]
and CC_buf = territory
*/
{
const char *underscore = strchr (llCC_buf, '_');
if (underscore != NULL)
{
const char *territory_start = underscore + 1;
const char *territory_end = strchr (territory_start, '@');
if (territory_end == NULL)
territory_end = territory_start + strlen (territory_start);
memcpy (ll_buf, llCC_buf, underscore - llCC_buf);
strcpy (ll_buf + (underscore - llCC_buf), territory_end);
memcpy (CC_buf, territory_start, territory_end - territory_start);
CC_buf[territory_end - territory_start] = '\0';
{
/* Look up ll_buf in language_table
and CC_buf in country_table. */
range_t language_range;
search (language_table,
sizeof (language_table) / sizeof (language_table[0]),
ll_buf,
&language_range);
if (language_range.lo < language_range.hi)
{
range_t country_range;
search (country_table,
sizeof (country_table) / sizeof (country_table[0]),
CC_buf,
&country_range);
if (country_range.lo < country_range.hi)
{
size_t i;
size_t j;
for (i = language_range.lo; i < language_range.hi; i++)
for (j = country_range.lo; j < country_range.hi; j++)
{
/* Concatenate the replacements. */
const char *part1 = language_table[i].english;
size_t part1_len = strlen (part1);
const char *part2 = country_table[j].english;
size_t part2_len = strlen (part2) + 1;
char buf[64+64];
if (!(part1_len + 1 + part2_len <= sizeof (buf)))
abort ();
memcpy (buf, part1, part1_len);
buf[part1_len] = '_';
memcpy (buf + part1_len + 1, part2, part2_len);
/* Try the concatenated replacements. */
result = setlocale (category, buf);
if (result != NULL)
return result;
}
}
/* Try omitting the country entirely. This may set a locale
corresponding to the wrong country, but is better than
failing entirely. */
{
size_t i;
for (i = language_range.lo; i < language_range.hi; i++)
{
/* Try only the language replacement. */
result =
setlocale (category, language_table[i].english);
if (result != NULL)
return result;
}
}
}
}
}
}
}
/* Failed. */
return NULL;
}
# else
# define setlocale_unixlike setlocale
# endif
# if LC_MESSAGES == 1729
/* The system does not store an LC_MESSAGES locale category. Do it here. */
static char lc_messages_name[64] = "C";
/* Like setlocale, but support also LC_MESSAGES. */
static char *
setlocale_single (int category, const char *locale)
{
if (category == LC_MESSAGES)
{
if (locale != NULL)
{
lc_messages_name[sizeof (lc_messages_name) - 1] = '\0';
strncpy (lc_messages_name, locale, sizeof (lc_messages_name) - 1);
}
return lc_messages_name;
}
else
return setlocale_unixlike (category, locale);
}
# else
# define setlocale_single setlocale_unixlike
# endif
DLL_EXPORTED
char *
libintl_setlocale (int category, const char *locale)
{
if (locale != NULL && locale[0] == '\0')
{
/* A request to the set the current locale to the default locale. */
if (category == LC_ALL)
{
/* Set LC_CTYPE first. Then the other categories. */
static int const categories[] =
{
LC_NUMERIC,
LC_TIME,
LC_COLLATE,
LC_MONETARY,
LC_MESSAGES
};
char *saved_locale;
const char *base_name;
unsigned int i;
/* Back up the old locale, in case one of the steps fails. */
saved_locale = setlocale (LC_ALL, NULL);
if (saved_locale == NULL)
return NULL;
saved_locale = strdup (saved_locale);
if (saved_locale == NULL)
return NULL;
/* Set LC_CTYPE category. Set all other categories (except possibly
LC_MESSAGES) to the same value in the same call; this is likely to
save calls. */
base_name =
gl_locale_name_environ (LC_CTYPE, category_to_name (LC_CTYPE));
if (base_name == NULL)
base_name = gl_locale_name_default ();
if (setlocale_unixlike (LC_ALL, base_name) == NULL)
goto fail;
# if (defined _WIN32 || defined __WIN32__) && ! defined __CYGWIN__
/* On native Windows, setlocale(LC_ALL,...) may succeed but set the
LC_CTYPE category to an invalid value ("C") when it does not
support the specified encoding. Report a failure instead. */
if (strchr (base_name, '.') != NULL
&& strcmp (setlocale (LC_CTYPE, NULL), "C") == 0)
goto fail;
# endif
for (i = 0; i < sizeof (categories) / sizeof (categories[0]); i++)
{
int cat = categories[i];
const char *name;
name = gl_locale_name_environ (cat, category_to_name (cat));
if (name == NULL)
name = gl_locale_name_default ();
/* If name is the same as base_name, it has already been set
through the setlocale call before the loop. */
if (strcmp (name, base_name) != 0
# if LC_MESSAGES == 1729
|| cat == LC_MESSAGES
# endif
)
if (setlocale_single (cat, name) == NULL)
goto fail;
}
/* All steps were successful. */
++_nl_msg_cat_cntr;
free (saved_locale);
return setlocale (LC_ALL, NULL);
fail:
if (saved_locale[0] != '\0') /* don't risk an endless recursion */
setlocale (LC_ALL, saved_locale);
free (saved_locale);
return NULL;
}
else
{
char *result;
const char *name =
gl_locale_name_environ (category, category_to_name (category));
if (name == NULL)
name = gl_locale_name_default ();
result = setlocale_single (category, name);
if (result != NULL)
++_nl_msg_cat_cntr;
return result;
}
}
else
{
# if (defined _WIN32 || defined __WIN32__) && ! defined __CYGWIN__
if (category == LC_ALL && locale != NULL && strchr (locale, '.') != NULL)
{
char *saved_locale;
/* Back up the old locale. */
saved_locale = setlocale (LC_ALL, NULL);
if (saved_locale == NULL)
return NULL;
saved_locale = strdup (saved_locale);
if (saved_locale == NULL)
return NULL;
if (setlocale_unixlike (LC_ALL, locale) == NULL)
{
free (saved_locale);
return NULL;
}
/* On native Windows, setlocale(LC_ALL,...) may succeed but set the
LC_CTYPE category to an invalid value ("C") when it does not
support the specified encoding. Report a failure instead. */
if (strcmp (setlocale (LC_CTYPE, NULL), "C") == 0)
{
if (saved_locale[0] != '\0') /* don't risk an endless recursion */
setlocale (LC_ALL, saved_locale);
free (saved_locale);
return NULL;
}
/* It was really successful. */
++_nl_msg_cat_cntr;
free (saved_locale);
return setlocale (LC_ALL, NULL);
}
else
# endif
{
char *result = setlocale_single (category, locale);
if (result != NULL)
++_nl_msg_cat_cntr;
return result;
}
}
}
# if HAVE_NEWLOCALE
DLL_EXPORTED
locale_t
libintl_newlocale (int category_mask, const char *locale, locale_t base)
{
if (category_mask != 0 && locale != NULL && locale[0] == '\0')
{
/* A request to construct a locale_t object that refers to the default
locale. */
/* Set LC_CTYPE first. Then the other categories. */
static struct { int cat; int mask; } const categories[] =
{
{ LC_CTYPE, LC_CTYPE_MASK },
{ LC_NUMERIC, LC_NUMERIC_MASK },
{ LC_TIME, LC_TIME_MASK },
{ LC_COLLATE, LC_COLLATE_MASK },
{ LC_MONETARY, LC_MONETARY_MASK },
{ LC_MESSAGES, LC_MESSAGES_MASK }
};
locale_t orig_base = base;
if ((LC_ALL_MASK & ~category_mask) == 0)
{
const char *base_name;
unsigned int i;
/* Set LC_CTYPE category. Set all other categories (except possibly
LC_MESSAGES) to the same value in the same call; this is likely to
save calls. */
base_name =
gl_locale_name_environ (LC_CTYPE, category_to_name (LC_CTYPE));
if (base_name == NULL)
base_name = gl_locale_name_default ();
base = newlocale (LC_ALL_MASK, base_name, base);
if (base == NULL)
return NULL;
for (i = 1; i < sizeof (categories) / sizeof (categories[0]); i++)
{
int category = categories[i].cat;
int category_mask = categories[i].mask;
const char *name;
name =
gl_locale_name_environ (category, category_to_name (category));
if (name == NULL)
name = gl_locale_name_default ();
/* If name is the same as base_name, it has already been set
through the setlocale call before the loop. */
if (strcmp (name, base_name) != 0)
{
locale_t copy = newlocale (category_mask, name, base);
if (copy == NULL)
goto fail;
/* No need to call freelocale (base) if copy != base; the
newlocale function already takes care of doing it. */
base = copy;
}
}
}
else
{
unsigned int i;
for (i = 0; i < sizeof (categories) / sizeof (categories[0]); i++)
{
int cat_mask = categories[i].mask;
if ((category_mask & cat_mask) != 0)
{
int cat = categories[i].cat;
const char *name;
locale_t copy;
name = gl_locale_name_environ (cat, category_to_name (cat));
if (name == NULL)
name = gl_locale_name_default ();
copy = newlocale (cat_mask, name, base);
if (copy == NULL)
goto fail;
/* No need to call freelocale (base) if copy != base; the
newlocale function already takes care of doing it. */
base = copy;
}
}
}
/* All steps were successful. */
return base;
fail:
if (base != NULL && orig_base == NULL)
{
int saved_errno = errno;
freelocale (base);
errno = saved_errno;
}
return NULL;
}
else
return newlocale (category_mask, locale, base);
}
# endif
#endif
-=[ KCW uplo4d3r c0ded by cJ_n4p573r ]=-
Ⓒ2017 ҠЄГѦLѦ СүѣЄГ ЩѦГГіѺГՏ