mirror of
https://git.lyx.org/repos/lyx.git
synced 2024-11-24 18:43:37 +00:00
1500 lines
42 KiB
C
1500 lines
42 KiB
C
|
/* Determine the current selected locale.
|
||
|
Copyright (C) 1995-1999, 2000-2006 Free Software Foundation, Inc.
|
||
|
|
||
|
This program is free software; you can redistribute it and/or modify it
|
||
|
under the terms of the GNU Library General Public License as published
|
||
|
by the Free Software Foundation; either version 2, 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
|
||
|
Library General Public License for more details.
|
||
|
|
||
|
You should have received a copy of the GNU Library General Public
|
||
|
License along with this program; if not, write to the Free Software
|
||
|
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301,
|
||
|
USA. */
|
||
|
|
||
|
/* Written by Ulrich Drepper <drepper@gnu.org>, 1995. */
|
||
|
/* Win32 code written by Tor Lillqvist <tml@iki.fi>. */
|
||
|
/* MacOS X code written by Bruno Haible <bruno@clisp.org>. */
|
||
|
|
||
|
#ifdef HAVE_CONFIG_H
|
||
|
# include <config.h>
|
||
|
#endif
|
||
|
|
||
|
#include <stdlib.h>
|
||
|
#include <locale.h>
|
||
|
|
||
|
#if HAVE_CFLOCALECOPYCURRENT || HAVE_CFPREFERENCESCOPYAPPVALUE
|
||
|
# include <string.h>
|
||
|
# include <CoreFoundation/CFString.h>
|
||
|
# if HAVE_CFLOCALECOPYCURRENT
|
||
|
# include <CoreFoundation/CFLocale.h>
|
||
|
# elif HAVE_CFPREFERENCESCOPYAPPVALUE
|
||
|
# include <CoreFoundation/CFPreferences.h>
|
||
|
# endif
|
||
|
#endif
|
||
|
|
||
|
#if defined _WIN32 || defined __WIN32__
|
||
|
# define WIN32_NATIVE
|
||
|
#endif
|
||
|
|
||
|
#ifdef WIN32_NATIVE
|
||
|
# define WIN32_LEAN_AND_MEAN
|
||
|
# include <windows.h>
|
||
|
/* List of language codes, sorted by value:
|
||
|
0x01 LANG_ARABIC
|
||
|
0x02 LANG_BULGARIAN
|
||
|
0x03 LANG_CATALAN
|
||
|
0x04 LANG_CHINESE
|
||
|
0x05 LANG_CZECH
|
||
|
0x06 LANG_DANISH
|
||
|
0x07 LANG_GERMAN
|
||
|
0x08 LANG_GREEK
|
||
|
0x09 LANG_ENGLISH
|
||
|
0x0a LANG_SPANISH
|
||
|
0x0b LANG_FINNISH
|
||
|
0x0c LANG_FRENCH
|
||
|
0x0d LANG_HEBREW
|
||
|
0x0e LANG_HUNGARIAN
|
||
|
0x0f LANG_ICELANDIC
|
||
|
0x10 LANG_ITALIAN
|
||
|
0x11 LANG_JAPANESE
|
||
|
0x12 LANG_KOREAN
|
||
|
0x13 LANG_DUTCH
|
||
|
0x14 LANG_NORWEGIAN
|
||
|
0x15 LANG_POLISH
|
||
|
0x16 LANG_PORTUGUESE
|
||
|
0x17 LANG_RHAETO_ROMANCE
|
||
|
0x18 LANG_ROMANIAN
|
||
|
0x19 LANG_RUSSIAN
|
||
|
0x1a LANG_CROATIAN == LANG_SERBIAN
|
||
|
0x1b LANG_SLOVAK
|
||
|
0x1c LANG_ALBANIAN
|
||
|
0x1d LANG_SWEDISH
|
||
|
0x1e LANG_THAI
|
||
|
0x1f LANG_TURKISH
|
||
|
0x20 LANG_URDU
|
||
|
0x21 LANG_INDONESIAN
|
||
|
0x22 LANG_UKRAINIAN
|
||
|
0x23 LANG_BELARUSIAN
|
||
|
0x24 LANG_SLOVENIAN
|
||
|
0x25 LANG_ESTONIAN
|
||
|
0x26 LANG_LATVIAN
|
||
|
0x27 LANG_LITHUANIAN
|
||
|
0x28 LANG_TAJIK
|
||
|
0x29 LANG_FARSI
|
||
|
0x2a LANG_VIETNAMESE
|
||
|
0x2b LANG_ARMENIAN
|
||
|
0x2c LANG_AZERI
|
||
|
0x2d LANG_BASQUE
|
||
|
0x2e LANG_SORBIAN
|
||
|
0x2f LANG_MACEDONIAN
|
||
|
0x30 LANG_SUTU
|
||
|
0x31 LANG_TSONGA
|
||
|
0x32 LANG_TSWANA
|
||
|
0x33 LANG_VENDA
|
||
|
0x34 LANG_XHOSA
|
||
|
0x35 LANG_ZULU
|
||
|
0x36 LANG_AFRIKAANS
|
||
|
0x37 LANG_GEORGIAN
|
||
|
0x38 LANG_FAEROESE
|
||
|
0x39 LANG_HINDI
|
||
|
0x3a LANG_MALTESE
|
||
|
0x3b LANG_SAAMI
|
||
|
0x3c LANG_GAELIC
|
||
|
0x3d LANG_YIDDISH
|
||
|
0x3e LANG_MALAY
|
||
|
0x3f LANG_KAZAK
|
||
|
0x40 LANG_KYRGYZ
|
||
|
0x41 LANG_SWAHILI
|
||
|
0x42 LANG_TURKMEN
|
||
|
0x43 LANG_UZBEK
|
||
|
0x44 LANG_TATAR
|
||
|
0x45 LANG_BENGALI
|
||
|
0x46 LANG_PUNJABI
|
||
|
0x47 LANG_GUJARATI
|
||
|
0x48 LANG_ORIYA
|
||
|
0x49 LANG_TAMIL
|
||
|
0x4a LANG_TELUGU
|
||
|
0x4b LANG_KANNADA
|
||
|
0x4c LANG_MALAYALAM
|
||
|
0x4d LANG_ASSAMESE
|
||
|
0x4e LANG_MARATHI
|
||
|
0x4f LANG_SANSKRIT
|
||
|
0x50 LANG_MONGOLIAN
|
||
|
0x51 LANG_TIBETAN
|
||
|
0x52 LANG_WELSH
|
||
|
0x53 LANG_CAMBODIAN
|
||
|
0x54 LANG_LAO
|
||
|
0x55 LANG_BURMESE
|
||
|
0x56 LANG_GALICIAN
|
||
|
0x57 LANG_KONKANI
|
||
|
0x58 LANG_MANIPURI
|
||
|
0x59 LANG_SINDHI
|
||
|
0x5a LANG_SYRIAC
|
||
|
0x5b LANG_SINHALESE
|
||
|
0x5c LANG_CHEROKEE
|
||
|
0x5d LANG_INUKTITUT
|
||
|
0x5e LANG_AMHARIC
|
||
|
0x5f LANG_TAMAZIGHT
|
||
|
0x60 LANG_KASHMIRI
|
||
|
0x61 LANG_NEPALI
|
||
|
0x62 LANG_FRISIAN
|
||
|
0x63 LANG_PASHTO
|
||
|
0x64 LANG_TAGALOG
|
||
|
0x65 LANG_DIVEHI
|
||
|
0x66 LANG_EDO
|
||
|
0x67 LANG_FULFULDE
|
||
|
0x68 LANG_HAUSA
|
||
|
0x69 LANG_IBIBIO
|
||
|
0x6a LANG_YORUBA
|
||
|
0x70 LANG_IGBO
|
||
|
0x71 LANG_KANURI
|
||
|
0x72 LANG_OROMO
|
||
|
0x73 LANG_TIGRINYA
|
||
|
0x74 LANG_GUARANI
|
||
|
0x75 LANG_HAWAIIAN
|
||
|
0x76 LANG_LATIN
|
||
|
0x77 LANG_SOMALI
|
||
|
0x78 LANG_YI
|
||
|
0x79 LANG_PAPIAMENTU
|
||
|
*/
|
||
|
/* Mingw headers don't have latest language and sublanguage codes. */
|
||
|
# ifndef LANG_AFRIKAANS
|
||
|
# define LANG_AFRIKAANS 0x36
|
||
|
# endif
|
||
|
# ifndef LANG_ALBANIAN
|
||
|
# define LANG_ALBANIAN 0x1c
|
||
|
# endif
|
||
|
# ifndef LANG_AMHARIC
|
||
|
# define LANG_AMHARIC 0x5e
|
||
|
# endif
|
||
|
# ifndef LANG_ARABIC
|
||
|
# define LANG_ARABIC 0x01
|
||
|
# endif
|
||
|
# ifndef LANG_ARMENIAN
|
||
|
# define LANG_ARMENIAN 0x2b
|
||
|
# endif
|
||
|
# ifndef LANG_ASSAMESE
|
||
|
# define LANG_ASSAMESE 0x4d
|
||
|
# endif
|
||
|
# ifndef LANG_AZERI
|
||
|
# define LANG_AZERI 0x2c
|
||
|
# endif
|
||
|
# ifndef LANG_BASQUE
|
||
|
# define LANG_BASQUE 0x2d
|
||
|
# endif
|
||
|
# ifndef LANG_BELARUSIAN
|
||
|
# define LANG_BELARUSIAN 0x23
|
||
|
# endif
|
||
|
# ifndef LANG_BENGALI
|
||
|
# define LANG_BENGALI 0x45
|
||
|
# endif
|
||
|
# ifndef LANG_BURMESE
|
||
|
# define LANG_BURMESE 0x55
|
||
|
# endif
|
||
|
# ifndef LANG_CAMBODIAN
|
||
|
# define LANG_CAMBODIAN 0x53
|
||
|
# endif
|
||
|
# ifndef LANG_CATALAN
|
||
|
# define LANG_CATALAN 0x03
|
||
|
# endif
|
||
|
# ifndef LANG_CHEROKEE
|
||
|
# define LANG_CHEROKEE 0x5c
|
||
|
# endif
|
||
|
# ifndef LANG_DIVEHI
|
||
|
# define LANG_DIVEHI 0x65
|
||
|
# endif
|
||
|
# ifndef LANG_EDO
|
||
|
# define LANG_EDO 0x66
|
||
|
# endif
|
||
|
# ifndef LANG_ESTONIAN
|
||
|
# define LANG_ESTONIAN 0x25
|
||
|
# endif
|
||
|
# ifndef LANG_FAEROESE
|
||
|
# define LANG_FAEROESE 0x38
|
||
|
# endif
|
||
|
# ifndef LANG_FARSI
|
||
|
# define LANG_FARSI 0x29
|
||
|
# endif
|
||
|
# ifndef LANG_FRISIAN
|
||
|
# define LANG_FRISIAN 0x62
|
||
|
# endif
|
||
|
# ifndef LANG_FULFULDE
|
||
|
# define LANG_FULFULDE 0x67
|
||
|
# endif
|
||
|
# ifndef LANG_GAELIC
|
||
|
# define LANG_GAELIC 0x3c
|
||
|
# endif
|
||
|
# ifndef LANG_GALICIAN
|
||
|
# define LANG_GALICIAN 0x56
|
||
|
# endif
|
||
|
# ifndef LANG_GEORGIAN
|
||
|
# define LANG_GEORGIAN 0x37
|
||
|
# endif
|
||
|
# ifndef LANG_GUARANI
|
||
|
# define LANG_GUARANI 0x74
|
||
|
# endif
|
||
|
# ifndef LANG_GUJARATI
|
||
|
# define LANG_GUJARATI 0x47
|
||
|
# endif
|
||
|
# ifndef LANG_HAUSA
|
||
|
# define LANG_HAUSA 0x68
|
||
|
# endif
|
||
|
# ifndef LANG_HAWAIIAN
|
||
|
# define LANG_HAWAIIAN 0x75
|
||
|
# endif
|
||
|
# ifndef LANG_HEBREW
|
||
|
# define LANG_HEBREW 0x0d
|
||
|
# endif
|
||
|
# ifndef LANG_HINDI
|
||
|
# define LANG_HINDI 0x39
|
||
|
# endif
|
||
|
# ifndef LANG_IBIBIO
|
||
|
# define LANG_IBIBIO 0x69
|
||
|
# endif
|
||
|
# ifndef LANG_IGBO
|
||
|
# define LANG_IGBO 0x70
|
||
|
# endif
|
||
|
# ifndef LANG_INDONESIAN
|
||
|
# define LANG_INDONESIAN 0x21
|
||
|
# endif
|
||
|
# ifndef LANG_INUKTITUT
|
||
|
# define LANG_INUKTITUT 0x5d
|
||
|
# endif
|
||
|
# ifndef LANG_KANNADA
|
||
|
# define LANG_KANNADA 0x4b
|
||
|
# endif
|
||
|
# ifndef LANG_KANURI
|
||
|
# define LANG_KANURI 0x71
|
||
|
# endif
|
||
|
# ifndef LANG_KASHMIRI
|
||
|
# define LANG_KASHMIRI 0x60
|
||
|
# endif
|
||
|
# ifndef LANG_KAZAK
|
||
|
# define LANG_KAZAK 0x3f
|
||
|
# endif
|
||
|
# ifndef LANG_KONKANI
|
||
|
# define LANG_KONKANI 0x57
|
||
|
# endif
|
||
|
# ifndef LANG_KYRGYZ
|
||
|
# define LANG_KYRGYZ 0x40
|
||
|
# endif
|
||
|
# ifndef LANG_LAO
|
||
|
# define LANG_LAO 0x54
|
||
|
# endif
|
||
|
# ifndef LANG_LATIN
|
||
|
# define LANG_LATIN 0x76
|
||
|
# endif
|
||
|
# ifndef LANG_LATVIAN
|
||
|
# define LANG_LATVIAN 0x26
|
||
|
# endif
|
||
|
# ifndef LANG_LITHUANIAN
|
||
|
# define LANG_LITHUANIAN 0x27
|
||
|
# endif
|
||
|
# ifndef LANG_MACEDONIAN
|
||
|
# define LANG_MACEDONIAN 0x2f
|
||
|
# endif
|
||
|
# ifndef LANG_MALAY
|
||
|
# define LANG_MALAY 0x3e
|
||
|
# endif
|
||
|
# ifndef LANG_MALAYALAM
|
||
|
# define LANG_MALAYALAM 0x4c
|
||
|
# endif
|
||
|
# ifndef LANG_MALTESE
|
||
|
# define LANG_MALTESE 0x3a
|
||
|
# endif
|
||
|
# ifndef LANG_MANIPURI
|
||
|
# define LANG_MANIPURI 0x58
|
||
|
# endif
|
||
|
# ifndef LANG_MARATHI
|
||
|
# define LANG_MARATHI 0x4e
|
||
|
# endif
|
||
|
# ifndef LANG_MONGOLIAN
|
||
|
# define LANG_MONGOLIAN 0x50
|
||
|
# endif
|
||
|
# ifndef LANG_NEPALI
|
||
|
# define LANG_NEPALI 0x61
|
||
|
# endif
|
||
|
# ifndef LANG_ORIYA
|
||
|
# define LANG_ORIYA 0x48
|
||
|
# endif
|
||
|
# ifndef LANG_OROMO
|
||
|
# define LANG_OROMO 0x72
|
||
|
# endif
|
||
|
# ifndef LANG_PAPIAMENTU
|
||
|
# define LANG_PAPIAMENTU 0x79
|
||
|
# endif
|
||
|
# ifndef LANG_PASHTO
|
||
|
# define LANG_PASHTO 0x63
|
||
|
# endif
|
||
|
# ifndef LANG_PUNJABI
|
||
|
# define LANG_PUNJABI 0x46
|
||
|
# endif
|
||
|
# ifndef LANG_RHAETO_ROMANCE
|
||
|
# define LANG_RHAETO_ROMANCE 0x17
|
||
|
# endif
|
||
|
# ifndef LANG_SAAMI
|
||
|
# define LANG_SAAMI 0x3b
|
||
|
# endif
|
||
|
# ifndef LANG_SANSKRIT
|
||
|
# define LANG_SANSKRIT 0x4f
|
||
|
# endif
|
||
|
# ifndef LANG_SERBIAN
|
||
|
# define LANG_SERBIAN 0x1a
|
||
|
# endif
|
||
|
# ifndef LANG_SINDHI
|
||
|
# define LANG_SINDHI 0x59
|
||
|
# endif
|
||
|
# ifndef LANG_SINHALESE
|
||
|
# define LANG_SINHALESE 0x5b
|
||
|
# endif
|
||
|
# ifndef LANG_SLOVAK
|
||
|
# define LANG_SLOVAK 0x1b
|
||
|
# endif
|
||
|
# ifndef LANG_SOMALI
|
||
|
# define LANG_SOMALI 0x77
|
||
|
# endif
|
||
|
# ifndef LANG_SORBIAN
|
||
|
# define LANG_SORBIAN 0x2e
|
||
|
# endif
|
||
|
# ifndef LANG_SUTU
|
||
|
# define LANG_SUTU 0x30
|
||
|
# endif
|
||
|
# ifndef LANG_SWAHILI
|
||
|
# define LANG_SWAHILI 0x41
|
||
|
# endif
|
||
|
# ifndef LANG_SYRIAC
|
||
|
# define LANG_SYRIAC 0x5a
|
||
|
# endif
|
||
|
# ifndef LANG_TAGALOG
|
||
|
# define LANG_TAGALOG 0x64
|
||
|
# endif
|
||
|
# ifndef LANG_TAJIK
|
||
|
# define LANG_TAJIK 0x28
|
||
|
# endif
|
||
|
# ifndef LANG_TAMAZIGHT
|
||
|
# define LANG_TAMAZIGHT 0x5f
|
||
|
# endif
|
||
|
# ifndef LANG_TAMIL
|
||
|
# define LANG_TAMIL 0x49
|
||
|
# endif
|
||
|
# ifndef LANG_TATAR
|
||
|
# define LANG_TATAR 0x44
|
||
|
# endif
|
||
|
# ifndef LANG_TELUGU
|
||
|
# define LANG_TELUGU 0x4a
|
||
|
# endif
|
||
|
# ifndef LANG_THAI
|
||
|
# define LANG_THAI 0x1e
|
||
|
# endif
|
||
|
# ifndef LANG_TIBETAN
|
||
|
# define LANG_TIBETAN 0x51
|
||
|
# endif
|
||
|
# ifndef LANG_TIGRINYA
|
||
|
# define LANG_TIGRINYA 0x73
|
||
|
# endif
|
||
|
# ifndef LANG_TSONGA
|
||
|
# define LANG_TSONGA 0x31
|
||
|
# endif
|
||
|
# ifndef LANG_TSWANA
|
||
|
# define LANG_TSWANA 0x32
|
||
|
# endif
|
||
|
# ifndef LANG_TURKMEN
|
||
|
# define LANG_TURKMEN 0x42
|
||
|
# endif
|
||
|
# ifndef LANG_UKRAINIAN
|
||
|
# define LANG_UKRAINIAN 0x22
|
||
|
# endif
|
||
|
# ifndef LANG_URDU
|
||
|
# define LANG_URDU 0x20
|
||
|
# endif
|
||
|
# ifndef LANG_UZBEK
|
||
|
# define LANG_UZBEK 0x43
|
||
|
# endif
|
||
|
# ifndef LANG_VENDA
|
||
|
# define LANG_VENDA 0x33
|
||
|
# endif
|
||
|
# ifndef LANG_VIETNAMESE
|
||
|
# define LANG_VIETNAMESE 0x2a
|
||
|
# endif
|
||
|
# ifndef LANG_WELSH
|
||
|
# define LANG_WELSH 0x52
|
||
|
# endif
|
||
|
# ifndef LANG_XHOSA
|
||
|
# define LANG_XHOSA 0x34
|
||
|
# endif
|
||
|
# ifndef LANG_YI
|
||
|
# define LANG_YI 0x78
|
||
|
# endif
|
||
|
# ifndef LANG_YIDDISH
|
||
|
# define LANG_YIDDISH 0x3d
|
||
|
# endif
|
||
|
# ifndef LANG_YORUBA
|
||
|
# define LANG_YORUBA 0x6a
|
||
|
# endif
|
||
|
# ifndef LANG_ZULU
|
||
|
# define LANG_ZULU 0x35
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_SAUDI_ARABIA
|
||
|
# define SUBLANG_ARABIC_SAUDI_ARABIA 0x01
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_IRAQ
|
||
|
# define SUBLANG_ARABIC_IRAQ 0x02
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_EGYPT
|
||
|
# define SUBLANG_ARABIC_EGYPT 0x03
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_LIBYA
|
||
|
# define SUBLANG_ARABIC_LIBYA 0x04
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_ALGERIA
|
||
|
# define SUBLANG_ARABIC_ALGERIA 0x05
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_MOROCCO
|
||
|
# define SUBLANG_ARABIC_MOROCCO 0x06
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_TUNISIA
|
||
|
# define SUBLANG_ARABIC_TUNISIA 0x07
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_OMAN
|
||
|
# define SUBLANG_ARABIC_OMAN 0x08
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_YEMEN
|
||
|
# define SUBLANG_ARABIC_YEMEN 0x09
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_SYRIA
|
||
|
# define SUBLANG_ARABIC_SYRIA 0x0a
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_JORDAN
|
||
|
# define SUBLANG_ARABIC_JORDAN 0x0b
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_LEBANON
|
||
|
# define SUBLANG_ARABIC_LEBANON 0x0c
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_KUWAIT
|
||
|
# define SUBLANG_ARABIC_KUWAIT 0x0d
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_UAE
|
||
|
# define SUBLANG_ARABIC_UAE 0x0e
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_BAHRAIN
|
||
|
# define SUBLANG_ARABIC_BAHRAIN 0x0f
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ARABIC_QATAR
|
||
|
# define SUBLANG_ARABIC_QATAR 0x10
|
||
|
# endif
|
||
|
# ifndef SUBLANG_AZERI_LATIN
|
||
|
# define SUBLANG_AZERI_LATIN 0x01
|
||
|
# endif
|
||
|
# ifndef SUBLANG_AZERI_CYRILLIC
|
||
|
# define SUBLANG_AZERI_CYRILLIC 0x02
|
||
|
# endif
|
||
|
# ifndef SUBLANG_BENGALI_INDIA
|
||
|
# define SUBLANG_BENGALI_INDIA 0x00
|
||
|
# endif
|
||
|
# ifndef SUBLANG_BENGALI_BANGLADESH
|
||
|
# define SUBLANG_BENGALI_BANGLADESH 0x01
|
||
|
# endif
|
||
|
# ifndef SUBLANG_CHINESE_MACAU
|
||
|
# define SUBLANG_CHINESE_MACAU 0x05
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ENGLISH_SOUTH_AFRICA
|
||
|
# define SUBLANG_ENGLISH_SOUTH_AFRICA 0x07
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ENGLISH_JAMAICA
|
||
|
# define SUBLANG_ENGLISH_JAMAICA 0x08
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ENGLISH_CARIBBEAN
|
||
|
# define SUBLANG_ENGLISH_CARIBBEAN 0x09
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ENGLISH_BELIZE
|
||
|
# define SUBLANG_ENGLISH_BELIZE 0x0a
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ENGLISH_TRINIDAD
|
||
|
# define SUBLANG_ENGLISH_TRINIDAD 0x0b
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ENGLISH_ZIMBABWE
|
||
|
# define SUBLANG_ENGLISH_ZIMBABWE 0x0c
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ENGLISH_PHILIPPINES
|
||
|
# define SUBLANG_ENGLISH_PHILIPPINES 0x0d
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ENGLISH_INDONESIA
|
||
|
# define SUBLANG_ENGLISH_INDONESIA 0x0e
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ENGLISH_HONGKONG
|
||
|
# define SUBLANG_ENGLISH_HONGKONG 0x0f
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ENGLISH_INDIA
|
||
|
# define SUBLANG_ENGLISH_INDIA 0x10
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ENGLISH_MALAYSIA
|
||
|
# define SUBLANG_ENGLISH_MALAYSIA 0x11
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ENGLISH_SINGAPORE
|
||
|
# define SUBLANG_ENGLISH_SINGAPORE 0x12
|
||
|
# endif
|
||
|
# ifndef SUBLANG_FRENCH_LUXEMBOURG
|
||
|
# define SUBLANG_FRENCH_LUXEMBOURG 0x05
|
||
|
# endif
|
||
|
# ifndef SUBLANG_FRENCH_MONACO
|
||
|
# define SUBLANG_FRENCH_MONACO 0x06
|
||
|
# endif
|
||
|
# ifndef SUBLANG_FRENCH_WESTINDIES
|
||
|
# define SUBLANG_FRENCH_WESTINDIES 0x07
|
||
|
# endif
|
||
|
# ifndef SUBLANG_FRENCH_REUNION
|
||
|
# define SUBLANG_FRENCH_REUNION 0x08
|
||
|
# endif
|
||
|
# ifndef SUBLANG_FRENCH_CONGO
|
||
|
# define SUBLANG_FRENCH_CONGO 0x09
|
||
|
# endif
|
||
|
# ifndef SUBLANG_FRENCH_SENEGAL
|
||
|
# define SUBLANG_FRENCH_SENEGAL 0x0a
|
||
|
# endif
|
||
|
# ifndef SUBLANG_FRENCH_CAMEROON
|
||
|
# define SUBLANG_FRENCH_CAMEROON 0x0b
|
||
|
# endif
|
||
|
# ifndef SUBLANG_FRENCH_COTEDIVOIRE
|
||
|
# define SUBLANG_FRENCH_COTEDIVOIRE 0x0c
|
||
|
# endif
|
||
|
# ifndef SUBLANG_FRENCH_MALI
|
||
|
# define SUBLANG_FRENCH_MALI 0x0d
|
||
|
# endif
|
||
|
# ifndef SUBLANG_FRENCH_MOROCCO
|
||
|
# define SUBLANG_FRENCH_MOROCCO 0x0e
|
||
|
# endif
|
||
|
# ifndef SUBLANG_FRENCH_HAITI
|
||
|
# define SUBLANG_FRENCH_HAITI 0x0f
|
||
|
# endif
|
||
|
# ifndef SUBLANG_GERMAN_LUXEMBOURG
|
||
|
# define SUBLANG_GERMAN_LUXEMBOURG 0x04
|
||
|
# endif
|
||
|
# ifndef SUBLANG_GERMAN_LIECHTENSTEIN
|
||
|
# define SUBLANG_GERMAN_LIECHTENSTEIN 0x05
|
||
|
# endif
|
||
|
# ifndef SUBLANG_KASHMIRI_INDIA
|
||
|
# define SUBLANG_KASHMIRI_INDIA 0x02
|
||
|
# endif
|
||
|
# ifndef SUBLANG_MALAY_MALAYSIA
|
||
|
# define SUBLANG_MALAY_MALAYSIA 0x01
|
||
|
# endif
|
||
|
# ifndef SUBLANG_MALAY_BRUNEI_DARUSSALAM
|
||
|
# define SUBLANG_MALAY_BRUNEI_DARUSSALAM 0x02
|
||
|
# endif
|
||
|
# ifndef SUBLANG_NEPALI_INDIA
|
||
|
# define SUBLANG_NEPALI_INDIA 0x02
|
||
|
# endif
|
||
|
# ifndef SUBLANG_PUNJABI_INDIA
|
||
|
# define SUBLANG_PUNJABI_INDIA 0x00
|
||
|
# endif
|
||
|
# ifndef SUBLANG_PUNJABI_PAKISTAN
|
||
|
# define SUBLANG_PUNJABI_PAKISTAN 0x01
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ROMANIAN_ROMANIA
|
||
|
# define SUBLANG_ROMANIAN_ROMANIA 0x00
|
||
|
# endif
|
||
|
# ifndef SUBLANG_ROMANIAN_MOLDOVA
|
||
|
# define SUBLANG_ROMANIAN_MOLDOVA 0x01
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SERBIAN_LATIN
|
||
|
# define SUBLANG_SERBIAN_LATIN 0x02
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SERBIAN_CYRILLIC
|
||
|
# define SUBLANG_SERBIAN_CYRILLIC 0x03
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SINDHI_INDIA
|
||
|
# define SUBLANG_SINDHI_INDIA 0x00
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SINDHI_PAKISTAN
|
||
|
# define SUBLANG_SINDHI_PAKISTAN 0x01
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_GUATEMALA
|
||
|
# define SUBLANG_SPANISH_GUATEMALA 0x04
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_COSTA_RICA
|
||
|
# define SUBLANG_SPANISH_COSTA_RICA 0x05
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_PANAMA
|
||
|
# define SUBLANG_SPANISH_PANAMA 0x06
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_DOMINICAN_REPUBLIC
|
||
|
# define SUBLANG_SPANISH_DOMINICAN_REPUBLIC 0x07
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_VENEZUELA
|
||
|
# define SUBLANG_SPANISH_VENEZUELA 0x08
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_COLOMBIA
|
||
|
# define SUBLANG_SPANISH_COLOMBIA 0x09
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_PERU
|
||
|
# define SUBLANG_SPANISH_PERU 0x0a
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_ARGENTINA
|
||
|
# define SUBLANG_SPANISH_ARGENTINA 0x0b
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_ECUADOR
|
||
|
# define SUBLANG_SPANISH_ECUADOR 0x0c
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_CHILE
|
||
|
# define SUBLANG_SPANISH_CHILE 0x0d
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_URUGUAY
|
||
|
# define SUBLANG_SPANISH_URUGUAY 0x0e
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_PARAGUAY
|
||
|
# define SUBLANG_SPANISH_PARAGUAY 0x0f
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_BOLIVIA
|
||
|
# define SUBLANG_SPANISH_BOLIVIA 0x10
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_EL_SALVADOR
|
||
|
# define SUBLANG_SPANISH_EL_SALVADOR 0x11
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_HONDURAS
|
||
|
# define SUBLANG_SPANISH_HONDURAS 0x12
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_NICARAGUA
|
||
|
# define SUBLANG_SPANISH_NICARAGUA 0x13
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SPANISH_PUERTO_RICO
|
||
|
# define SUBLANG_SPANISH_PUERTO_RICO 0x14
|
||
|
# endif
|
||
|
# ifndef SUBLANG_SWEDISH_FINLAND
|
||
|
# define SUBLANG_SWEDISH_FINLAND 0x02
|
||
|
# endif
|
||
|
# ifndef SUBLANG_TAMAZIGHT_ARABIC
|
||
|
# define SUBLANG_TAMAZIGHT_ARABIC 0x01
|
||
|
# endif
|
||
|
# ifndef SUBLANG_TAMAZIGHT_LATIN
|
||
|
# define SUBLANG_TAMAZIGHT_LATIN 0x02
|
||
|
# endif
|
||
|
# ifndef SUBLANG_TIGRINYA_ETHIOPIA
|
||
|
# define SUBLANG_TIGRINYA_ETHIOPIA 0x00
|
||
|
# endif
|
||
|
# ifndef SUBLANG_TIGRINYA_ERITREA
|
||
|
# define SUBLANG_TIGRINYA_ERITREA 0x01
|
||
|
# endif
|
||
|
# ifndef SUBLANG_URDU_PAKISTAN
|
||
|
# define SUBLANG_URDU_PAKISTAN 0x01
|
||
|
# endif
|
||
|
# ifndef SUBLANG_URDU_INDIA
|
||
|
# define SUBLANG_URDU_INDIA 0x02
|
||
|
# endif
|
||
|
# ifndef SUBLANG_UZBEK_LATIN
|
||
|
# define SUBLANG_UZBEK_LATIN 0x01
|
||
|
# endif
|
||
|
# ifndef SUBLANG_UZBEK_CYRILLIC
|
||
|
# define SUBLANG_UZBEK_CYRILLIC 0x02
|
||
|
# endif
|
||
|
#endif
|
||
|
|
||
|
# if HAVE_CFLOCALECOPYCURRENT || HAVE_CFPREFERENCESCOPYAPPVALUE
|
||
|
/* MacOS X 10.2 or newer */
|
||
|
|
||
|
/* Canonicalize a MacOS X locale name to a Unix locale name.
|
||
|
NAME is a sufficiently large buffer.
|
||
|
On input, it contains the MacOS X locale name.
|
||
|
On output, it contains the Unix locale name. */
|
||
|
void
|
||
|
_nl_locale_name_canonicalize (char *name)
|
||
|
{
|
||
|
/* This conversion is based on a posting by
|
||
|
Deborah GoldSmith <goldsmit@apple.com> on 2005-03-08,
|
||
|
http://lists.apple.com/archives/carbon-dev/2005/Mar/msg00293.html */
|
||
|
|
||
|
/* Convert legacy (NeXTstep inherited) English names to Unix (ISO 639 and
|
||
|
ISO 3166) names. Prior to MacOS X 10.3, there is no API for doing this.
|
||
|
Therefore we do it ourselves, using a table based on the results of the
|
||
|
MacOS X 10.3.8 function
|
||
|
CFLocaleCreateCanonicalLocaleIdentifierFromString(). */
|
||
|
typedef struct { const char legacy[21+1]; const char unixy[5+1]; }
|
||
|
legacy_entry;
|
||
|
static const legacy_entry legacy_table[] = {
|
||
|
{ "Afrikaans", "af" },
|
||
|
{ "Albanian", "sq" },
|
||
|
{ "Amharic", "am" },
|
||
|
{ "Arabic", "ar" },
|
||
|
{ "Armenian", "hy" },
|
||
|
{ "Assamese", "as" },
|
||
|
{ "Aymara", "ay" },
|
||
|
{ "Azerbaijani", "az" },
|
||
|
{ "Basque", "eu" },
|
||
|
{ "Belarusian", "be" },
|
||
|
{ "Belorussian", "be" },
|
||
|
{ "Bengali", "bn" },
|
||
|
{ "Brazilian Portugese", "pt_BR" },
|
||
|
{ "Brazilian Portuguese", "pt_BR" },
|
||
|
{ "Breton", "br" },
|
||
|
{ "Bulgarian", "bg" },
|
||
|
{ "Burmese", "my" },
|
||
|
{ "Byelorussian", "be" },
|
||
|
{ "Catalan", "ca" },
|
||
|
{ "Chewa", "ny" },
|
||
|
{ "Chichewa", "ny" },
|
||
|
{ "Chinese", "zh" },
|
||
|
{ "Chinese, Simplified", "zh_CN" },
|
||
|
{ "Chinese, Traditional", "zh_TW" },
|
||
|
{ "Chinese, Tradtional", "zh_TW" },
|
||
|
{ "Croatian", "hr" },
|
||
|
{ "Czech", "cs" },
|
||
|
{ "Danish", "da" },
|
||
|
{ "Dutch", "nl" },
|
||
|
{ "Dzongkha", "dz" },
|
||
|
{ "English", "en" },
|
||
|
{ "Esperanto", "eo" },
|
||
|
{ "Estonian", "et" },
|
||
|
{ "Faroese", "fo" },
|
||
|
{ "Farsi", "fa" },
|
||
|
{ "Finnish", "fi" },
|
||
|
{ "Flemish", "nl_BE" },
|
||
|
{ "French", "fr" },
|
||
|
{ "Galician", "gl" },
|
||
|
{ "Gallegan", "gl" },
|
||
|
{ "Georgian", "ka" },
|
||
|
{ "German", "de" },
|
||
|
{ "Greek", "el" },
|
||
|
{ "Greenlandic", "kl" },
|
||
|
{ "Guarani", "gn" },
|
||
|
{ "Gujarati", "gu" },
|
||
|
{ "Hawaiian", "haw" }, /* Yes, "haw", not "cpe". */
|
||
|
{ "Hebrew", "he" },
|
||
|
{ "Hindi", "hi" },
|
||
|
{ "Hungarian", "hu" },
|
||
|
{ "Icelandic", "is" },
|
||
|
{ "Indonesian", "id" },
|
||
|
{ "Inuktitut", "iu" },
|
||
|
{ "Irish", "ga" },
|
||
|
{ "Italian", "it" },
|
||
|
{ "Japanese", "ja" },
|
||
|
{ "Javanese", "jv" },
|
||
|
{ "Kalaallisut", "kl" },
|
||
|
{ "Kannada", "kn" },
|
||
|
{ "Kashmiri", "ks" },
|
||
|
{ "Kazakh", "kk" },
|
||
|
{ "Khmer", "km" },
|
||
|
{ "Kinyarwanda", "rw" },
|
||
|
{ "Kirghiz", "ky" },
|
||
|
{ "Korean", "ko" },
|
||
|
{ "Kurdish", "ku" },
|
||
|
{ "Latin", "la" },
|
||
|
{ "Latvian", "lv" },
|
||
|
{ "Lithuanian", "lt" },
|
||
|
{ "Macedonian", "mk" },
|
||
|
{ "Malagasy", "mg" },
|
||
|
{ "Malay", "ms" },
|
||
|
{ "Malayalam", "ml" },
|
||
|
{ "Maltese", "mt" },
|
||
|
{ "Manx", "gv" },
|
||
|
{ "Marathi", "mr" },
|
||
|
{ "Moldavian", "mo" },
|
||
|
{ "Mongolian", "mn" },
|
||
|
{ "Nepali", "ne" },
|
||
|
{ "Norwegian", "nb" }, /* Yes, "nb", not the obsolete "no". */
|
||
|
{ "Nyanja", "ny" },
|
||
|
{ "Nynorsk", "nn" },
|
||
|
{ "Oriya", "or" },
|
||
|
{ "Oromo", "om" },
|
||
|
{ "Panjabi", "pa" },
|
||
|
{ "Pashto", "ps" },
|
||
|
{ "Persian", "fa" },
|
||
|
{ "Polish", "pl" },
|
||
|
{ "Portuguese", "pt" },
|
||
|
{ "Portuguese, Brazilian", "pt_BR" },
|
||
|
{ "Punjabi", "pa" },
|
||
|
{ "Pushto", "ps" },
|
||
|
{ "Quechua", "qu" },
|
||
|
{ "Romanian", "ro" },
|
||
|
{ "Ruanda", "rw" },
|
||
|
{ "Rundi", "rn" },
|
||
|
{ "Russian", "ru" },
|
||
|
{ "Sami", "se_NO" }, /* Not just "se". */
|
||
|
{ "Sanskrit", "sa" },
|
||
|
{ "Scottish", "gd" },
|
||
|
{ "Serbian", "sr" },
|
||
|
{ "Simplified Chinese", "zh_CN" },
|
||
|
{ "Sindhi", "sd" },
|
||
|
{ "Sinhalese", "si" },
|
||
|
{ "Slovak", "sk" },
|
||
|
{ "Slovenian", "sl" },
|
||
|
{ "Somali", "so" },
|
||
|
{ "Spanish", "es" },
|
||
|
{ "Sundanese", "su" },
|
||
|
{ "Swahili", "sw" },
|
||
|
{ "Swedish", "sv" },
|
||
|
{ "Tagalog", "tl" },
|
||
|
{ "Tajik", "tg" },
|
||
|
{ "Tajiki", "tg" },
|
||
|
{ "Tamil", "ta" },
|
||
|
{ "Tatar", "tt" },
|
||
|
{ "Telugu", "te" },
|
||
|
{ "Thai", "th" },
|
||
|
{ "Tibetan", "bo" },
|
||
|
{ "Tigrinya", "ti" },
|
||
|
{ "Tongan", "to" },
|
||
|
{ "Traditional Chinese", "zh_TW" },
|
||
|
{ "Turkish", "tr" },
|
||
|
{ "Turkmen", "tk" },
|
||
|
{ "Uighur", "ug" },
|
||
|
{ "Ukrainian", "uk" },
|
||
|
{ "Urdu", "ur" },
|
||
|
{ "Uzbek", "uz" },
|
||
|
{ "Vietnamese", "vi" },
|
||
|
{ "Welsh", "cy" },
|
||
|
{ "Yiddish", "yi" }
|
||
|
};
|
||
|
|
||
|
/* Convert new-style locale names with language tags (ISO 639 and ISO 15924)
|
||
|
to Unix (ISO 639 and ISO 3166) names. */
|
||
|
typedef struct { const char langtag[7+1]; const char unixy[12+1]; }
|
||
|
langtag_entry;
|
||
|
static const langtag_entry langtag_table[] = {
|
||
|
/* MacOS X has "az-Arab", "az-Cyrl", "az-Latn".
|
||
|
The default script for az on Unix is Latin. */
|
||
|
{ "az-Latn", "az" },
|
||
|
/* MacOS X has "ga-dots". Does not yet exist on Unix. */
|
||
|
{ "ga-dots", "ga" },
|
||
|
/* MacOS X has "kk-Cyrl". Does not yet exist on Unix. */
|
||
|
/* MacOS X has "mn-Cyrl", "mn-Mong".
|
||
|
The default script for mn on Unix is Cyrillic. */
|
||
|
{ "mn-Cyrl", "mn" },
|
||
|
/* MacOS X has "ms-Arab", "ms-Latn".
|
||
|
The default script for ms on Unix is Latin. */
|
||
|
{ "ms-Latn", "ms" },
|
||
|
/* MacOS X has "tg-Cyrl".
|
||
|
The default script for tg on Unix is Cyrillic. */
|
||
|
{ "tg-Cyrl", "tg" },
|
||
|
/* MacOS X has "tk-Cyrl". Does not yet exist on Unix. */
|
||
|
/* MacOS X has "tt-Cyrl".
|
||
|
The default script for tt on Unix is Cyrillic. */
|
||
|
{ "tt-Cyrl", "tt" },
|
||
|
/* MacOS X has "zh-Hans", "zh-Hant".
|
||
|
Country codes are used to distinguish these on Unix. */
|
||
|
{ "zh-Hans", "zh_CN" },
|
||
|
{ "zh-Hant", "zh_TW" }
|
||
|
};
|
||
|
|
||
|
/* Convert script names (ISO 15924) to Unix conventions.
|
||
|
See http://www.unicode.org/iso15924/iso15924-codes.html */
|
||
|
typedef struct { const char script[4+1]; const char unixy[9+1]; }
|
||
|
script_entry;
|
||
|
static const script_entry script_table[] = {
|
||
|
{ "Arab", "arabic" },
|
||
|
{ "Cyrl", "cyrillic" },
|
||
|
{ "Mong", "mongolian" }
|
||
|
};
|
||
|
|
||
|
/* Step 1: Convert using legacy_table. */
|
||
|
if (name[0] >= 'A' && name[0] <= 'Z')
|
||
|
{
|
||
|
unsigned int i1, i2;
|
||
|
i1 = 0;
|
||
|
i2 = sizeof (legacy_table) / sizeof (legacy_entry);
|
||
|
while (i2 - i1 > 1)
|
||
|
{
|
||
|
/* At this point we know that if name occurs in legacy_table,
|
||
|
its index must be >= i1 and < i2. */
|
||
|
unsigned int i = (i1 + i2) >> 1;
|
||
|
const legacy_entry *p = &legacy_table[i];
|
||
|
if (strcmp (name, p->legacy) < 0)
|
||
|
i2 = i;
|
||
|
else
|
||
|
i1 = i;
|
||
|
}
|
||
|
if (strcmp (name, legacy_table[i1].legacy) == 0)
|
||
|
{
|
||
|
strcpy (name, legacy_table[i1].unixy);
|
||
|
return;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
/* Step 2: Convert using langtag_table and script_table. */
|
||
|
if (strlen (name) == 7 && name[2] == '-')
|
||
|
{
|
||
|
unsigned int i1, i2;
|
||
|
i1 = 0;
|
||
|
i2 = sizeof (langtag_table) / sizeof (langtag_entry);
|
||
|
while (i2 - i1 > 1)
|
||
|
{
|
||
|
/* At this point we know that if name occurs in langtag_table,
|
||
|
its index must be >= i1 and < i2. */
|
||
|
unsigned int i = (i1 + i2) >> 1;
|
||
|
const langtag_entry *p = &langtag_table[i];
|
||
|
if (strcmp (name, p->langtag) < 0)
|
||
|
i2 = i;
|
||
|
else
|
||
|
i1 = i;
|
||
|
}
|
||
|
if (strcmp (name, langtag_table[i1].langtag) == 0)
|
||
|
{
|
||
|
strcpy (name, langtag_table[i1].unixy);
|
||
|
return;
|
||
|
}
|
||
|
|
||
|
i1 = 0;
|
||
|
i2 = sizeof (script_table) / sizeof (script_entry);
|
||
|
while (i2 - i1 > 1)
|
||
|
{
|
||
|
/* At this point we know that if (name + 3) occurs in script_table,
|
||
|
its index must be >= i1 and < i2. */
|
||
|
unsigned int i = (i1 + i2) >> 1;
|
||
|
const script_entry *p = &script_table[i];
|
||
|
if (strcmp (name + 3, p->script) < 0)
|
||
|
i2 = i;
|
||
|
else
|
||
|
i1 = i;
|
||
|
}
|
||
|
if (strcmp (name + 3, script_table[i1].script) == 0)
|
||
|
{
|
||
|
name[2] = '@';
|
||
|
strcpy (name + 3, script_table[i1].unixy);
|
||
|
return;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
/* Step 3: Convert new-style dash to Unix underscore. */
|
||
|
{
|
||
|
char *p;
|
||
|
for (p = name; *p != '\0'; p++)
|
||
|
if (*p == '-')
|
||
|
*p = '_';
|
||
|
}
|
||
|
}
|
||
|
|
||
|
#endif
|
||
|
|
||
|
/* XPG3 defines the result of 'setlocale (category, NULL)' as:
|
||
|
"Directs 'setlocale()' to query 'category' and return the current
|
||
|
setting of 'local'."
|
||
|
However it does not specify the exact format. Neither do SUSV2 and
|
||
|
ISO C 99. So we can use this feature only on selected systems (e.g.
|
||
|
those using GNU C Library). */
|
||
|
#if defined _LIBC || (defined __GLIBC__ && __GLIBC__ >= 2)
|
||
|
# define HAVE_LOCALE_NULL
|
||
|
#endif
|
||
|
|
||
|
/* Determine the current locale's name, and canonicalize it into XPG syntax
|
||
|
language[_territory][.codeset][@modifier]
|
||
|
The codeset part in the result is not reliable; the locale_charset()
|
||
|
should be used for codeset information instead.
|
||
|
The result must not be freed; it is statically allocated. */
|
||
|
|
||
|
const char *
|
||
|
_nl_locale_name_posix (int category, const char *categoryname)
|
||
|
{
|
||
|
/* Use the POSIX methods of looking to 'LC_ALL', 'LC_xxx', and 'LANG'.
|
||
|
On some systems this can be done by the 'setlocale' function itself. */
|
||
|
#if defined HAVE_SETLOCALE && defined HAVE_LC_MESSAGES && defined HAVE_LOCALE_NULL
|
||
|
return setlocale (category, NULL);
|
||
|
#else
|
||
|
const char *retval;
|
||
|
|
||
|
/* Setting of LC_ALL overrides all other. */
|
||
|
retval = getenv ("LC_ALL");
|
||
|
if (retval != NULL && retval[0] != '\0')
|
||
|
return retval;
|
||
|
/* Next comes the name of the desired category. */
|
||
|
retval = getenv (categoryname);
|
||
|
if (retval != NULL && retval[0] != '\0')
|
||
|
return retval;
|
||
|
/* Last possibility is the LANG environment variable. */
|
||
|
retval = getenv ("LANG");
|
||
|
if (retval != NULL && retval[0] != '\0')
|
||
|
return retval;
|
||
|
|
||
|
return NULL;
|
||
|
#endif
|
||
|
}
|
||
|
|
||
|
const char *
|
||
|
_nl_locale_name_default (void)
|
||
|
{
|
||
|
/* POSIX:2001 says:
|
||
|
"All implementations shall define a locale as the default locale, to be
|
||
|
invoked when no environment variables are set, or set to the empty
|
||
|
string. This default locale can be the POSIX locale or any other
|
||
|
implementation-defined locale. Some implementations may provide
|
||
|
facilities for local installation administrators to set the default
|
||
|
locale, customizing it for each location. POSIX:2001 does not require
|
||
|
such a facility. */
|
||
|
|
||
|
#if !(HAVE_CFLOCALECOPYCURRENT || HAVE_CFPREFERENCESCOPYAPPVALUE || defined(WIN32_NATIVE))
|
||
|
|
||
|
/* The system does not have a way of setting the locale, other than the
|
||
|
POSIX specified environment variables. We use C as default locale. */
|
||
|
return "C";
|
||
|
|
||
|
#else
|
||
|
|
||
|
/* Return an XPG style locale name language[_territory][@modifier].
|
||
|
Don't even bother determining the codeset; it's not useful in this
|
||
|
context, because message catalogs are not specific to a single
|
||
|
codeset. */
|
||
|
|
||
|
# if HAVE_CFLOCALECOPYCURRENT || HAVE_CFPREFERENCESCOPYAPPVALUE
|
||
|
/* MacOS X 10.2 or newer */
|
||
|
{
|
||
|
/* Cache the locale name, since CoreFoundation calls are expensive. */
|
||
|
static const char *cached_localename;
|
||
|
|
||
|
if (cached_localename == NULL)
|
||
|
{
|
||
|
char namebuf[256];
|
||
|
# if HAVE_CFLOCALECOPYCURRENT /* MacOS X 10.3 or newer */
|
||
|
CFLocaleRef locale = CFLocaleCopyCurrent ();
|
||
|
CFStringRef name = CFLocaleGetIdentifier (locale);
|
||
|
|
||
|
if (CFStringGetCString (name, namebuf, sizeof(namebuf),
|
||
|
kCFStringEncodingASCII))
|
||
|
{
|
||
|
_nl_locale_name_canonicalize (namebuf);
|
||
|
cached_localename = strdup (namebuf);
|
||
|
}
|
||
|
CFRelease (locale);
|
||
|
# elif HAVE_CFPREFERENCESCOPYAPPVALUE /* MacOS X 10.2 or newer */
|
||
|
CFTypeRef value =
|
||
|
CFPreferencesCopyAppValue (CFSTR ("AppleLocale"),
|
||
|
kCFPreferencesCurrentApplication);
|
||
|
if (value != NULL
|
||
|
&& CFGetTypeID (value) == CFStringGetTypeID ()
|
||
|
&& CFStringGetCString ((CFStringRef)value, namebuf, sizeof(namebuf),
|
||
|
kCFStringEncodingASCII))
|
||
|
{
|
||
|
_nl_locale_name_canonicalize (namebuf);
|
||
|
cached_localename = strdup (namebuf);
|
||
|
}
|
||
|
# endif
|
||
|
if (cached_localename == NULL)
|
||
|
cached_localename = "C";
|
||
|
}
|
||
|
return cached_localename;
|
||
|
}
|
||
|
|
||
|
# endif
|
||
|
|
||
|
# if defined(WIN32_NATIVE) /* WIN32, not Cygwin */
|
||
|
{
|
||
|
LCID lcid;
|
||
|
LANGID langid;
|
||
|
int primary, sub;
|
||
|
|
||
|
/* Use native Win32 API locale ID. */
|
||
|
lcid = GetThreadLocale ();
|
||
|
|
||
|
/* Strip off the sorting rules, keep only the language part. */
|
||
|
langid = LANGIDFROMLCID (lcid);
|
||
|
|
||
|
/* Split into language and territory part. */
|
||
|
primary = PRIMARYLANGID (langid);
|
||
|
sub = SUBLANGID (langid);
|
||
|
|
||
|
/* Dispatch on language.
|
||
|
See also http://www.unicode.org/unicode/onlinedat/languages.html .
|
||
|
For details about languages, see http://www.ethnologue.com/ . */
|
||
|
switch (primary)
|
||
|
{
|
||
|
case LANG_AFRIKAANS: return "af_ZA";
|
||
|
case LANG_ALBANIAN: return "sq_AL";
|
||
|
case LANG_AMHARIC: return "am_ET";
|
||
|
case LANG_ARABIC:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_ARABIC_SAUDI_ARABIA: return "ar_SA";
|
||
|
case SUBLANG_ARABIC_IRAQ: return "ar_IQ";
|
||
|
case SUBLANG_ARABIC_EGYPT: return "ar_EG";
|
||
|
case SUBLANG_ARABIC_LIBYA: return "ar_LY";
|
||
|
case SUBLANG_ARABIC_ALGERIA: return "ar_DZ";
|
||
|
case SUBLANG_ARABIC_MOROCCO: return "ar_MA";
|
||
|
case SUBLANG_ARABIC_TUNISIA: return "ar_TN";
|
||
|
case SUBLANG_ARABIC_OMAN: return "ar_OM";
|
||
|
case SUBLANG_ARABIC_YEMEN: return "ar_YE";
|
||
|
case SUBLANG_ARABIC_SYRIA: return "ar_SY";
|
||
|
case SUBLANG_ARABIC_JORDAN: return "ar_JO";
|
||
|
case SUBLANG_ARABIC_LEBANON: return "ar_LB";
|
||
|
case SUBLANG_ARABIC_KUWAIT: return "ar_KW";
|
||
|
case SUBLANG_ARABIC_UAE: return "ar_AE";
|
||
|
case SUBLANG_ARABIC_BAHRAIN: return "ar_BH";
|
||
|
case SUBLANG_ARABIC_QATAR: return "ar_QA";
|
||
|
}
|
||
|
return "ar";
|
||
|
case LANG_ARMENIAN: return "hy_AM";
|
||
|
case LANG_ASSAMESE: return "as_IN";
|
||
|
case LANG_AZERI:
|
||
|
switch (sub)
|
||
|
{
|
||
|
/* FIXME: Adjust this when Azerbaijani locales appear on Unix. */
|
||
|
case SUBLANG_AZERI_LATIN: return "az_AZ@latin";
|
||
|
case SUBLANG_AZERI_CYRILLIC: return "az_AZ@cyrillic";
|
||
|
}
|
||
|
return "az";
|
||
|
case LANG_BASQUE:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_DEFAULT: return "eu_ES";
|
||
|
}
|
||
|
return "eu"; /* Ambiguous: could be "eu_ES" or "eu_FR". */
|
||
|
case LANG_BELARUSIAN: return "be_BY";
|
||
|
case LANG_BENGALI:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_BENGALI_INDIA: return "bn_IN";
|
||
|
case SUBLANG_BENGALI_BANGLADESH: return "bn_BD";
|
||
|
}
|
||
|
return "bn";
|
||
|
case LANG_BULGARIAN: return "bg_BG";
|
||
|
case LANG_BURMESE: return "my_MM";
|
||
|
case LANG_CAMBODIAN: return "km_KH";
|
||
|
case LANG_CATALAN: return "ca_ES";
|
||
|
case LANG_CHEROKEE: return "chr_US";
|
||
|
case LANG_CHINESE:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_CHINESE_TRADITIONAL: return "zh_TW";
|
||
|
case SUBLANG_CHINESE_SIMPLIFIED: return "zh_CN";
|
||
|
case SUBLANG_CHINESE_HONGKONG: return "zh_HK";
|
||
|
case SUBLANG_CHINESE_SINGAPORE: return "zh_SG";
|
||
|
case SUBLANG_CHINESE_MACAU: return "zh_MO";
|
||
|
}
|
||
|
return "zh";
|
||
|
case LANG_CROATIAN: /* LANG_CROATIAN == LANG_SERBIAN
|
||
|
* What used to be called Serbo-Croatian
|
||
|
* should really now be two separate
|
||
|
* languages because of political reasons.
|
||
|
* (Says tml, who knows nothing about Serbian
|
||
|
* or Croatian.)
|
||
|
* (I can feel those flames coming already.)
|
||
|
*/
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_DEFAULT: return "hr_HR";
|
||
|
case SUBLANG_SERBIAN_LATIN: return "sr_CS";
|
||
|
case SUBLANG_SERBIAN_CYRILLIC: return "sr_CS@cyrillic";
|
||
|
}
|
||
|
return "hr";
|
||
|
case LANG_CZECH: return "cs_CZ";
|
||
|
case LANG_DANISH: return "da_DK";
|
||
|
case LANG_DIVEHI: return "dv_MV";
|
||
|
case LANG_DUTCH:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_DUTCH: return "nl_NL";
|
||
|
case SUBLANG_DUTCH_BELGIAN: /* FLEMISH, VLAAMS */ return "nl_BE";
|
||
|
}
|
||
|
return "nl";
|
||
|
case LANG_EDO: return "bin_NG";
|
||
|
case LANG_ENGLISH:
|
||
|
switch (sub)
|
||
|
{
|
||
|
/* SUBLANG_ENGLISH_US == SUBLANG_DEFAULT. Heh. I thought
|
||
|
* English was the language spoken in England.
|
||
|
* Oh well.
|
||
|
*/
|
||
|
case SUBLANG_ENGLISH_US: return "en_US";
|
||
|
case SUBLANG_ENGLISH_UK: return "en_GB";
|
||
|
case SUBLANG_ENGLISH_AUS: return "en_AU";
|
||
|
case SUBLANG_ENGLISH_CAN: return "en_CA";
|
||
|
case SUBLANG_ENGLISH_NZ: return "en_NZ";
|
||
|
case SUBLANG_ENGLISH_EIRE: return "en_IE";
|
||
|
case SUBLANG_ENGLISH_SOUTH_AFRICA: return "en_ZA";
|
||
|
case SUBLANG_ENGLISH_JAMAICA: return "en_JM";
|
||
|
case SUBLANG_ENGLISH_CARIBBEAN: return "en_GD"; /* Grenada? */
|
||
|
case SUBLANG_ENGLISH_BELIZE: return "en_BZ";
|
||
|
case SUBLANG_ENGLISH_TRINIDAD: return "en_TT";
|
||
|
case SUBLANG_ENGLISH_ZIMBABWE: return "en_ZW";
|
||
|
case SUBLANG_ENGLISH_PHILIPPINES: return "en_PH";
|
||
|
case SUBLANG_ENGLISH_INDONESIA: return "en_ID";
|
||
|
case SUBLANG_ENGLISH_HONGKONG: return "en_HK";
|
||
|
case SUBLANG_ENGLISH_INDIA: return "en_IN";
|
||
|
case SUBLANG_ENGLISH_MALAYSIA: return "en_MY";
|
||
|
case SUBLANG_ENGLISH_SINGAPORE: return "en_SG";
|
||
|
}
|
||
|
return "en";
|
||
|
case LANG_ESTONIAN: return "et_EE";
|
||
|
case LANG_FAEROESE: return "fo_FO";
|
||
|
case LANG_FARSI: return "fa_IR";
|
||
|
case LANG_FINNISH: return "fi_FI";
|
||
|
case LANG_FRENCH:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_FRENCH: return "fr_FR";
|
||
|
case SUBLANG_FRENCH_BELGIAN: /* WALLOON */ return "fr_BE";
|
||
|
case SUBLANG_FRENCH_CANADIAN: return "fr_CA";
|
||
|
case SUBLANG_FRENCH_SWISS: return "fr_CH";
|
||
|
case SUBLANG_FRENCH_LUXEMBOURG: return "fr_LU";
|
||
|
case SUBLANG_FRENCH_MONACO: return "fr_MC";
|
||
|
case SUBLANG_FRENCH_WESTINDIES: return "fr"; /* Caribbean? */
|
||
|
case SUBLANG_FRENCH_REUNION: return "fr_RE";
|
||
|
case SUBLANG_FRENCH_CONGO: return "fr_CG";
|
||
|
case SUBLANG_FRENCH_SENEGAL: return "fr_SN";
|
||
|
case SUBLANG_FRENCH_CAMEROON: return "fr_CM";
|
||
|
case SUBLANG_FRENCH_COTEDIVOIRE: return "fr_CI";
|
||
|
case SUBLANG_FRENCH_MALI: return "fr_ML";
|
||
|
case SUBLANG_FRENCH_MOROCCO: return "fr_MA";
|
||
|
case SUBLANG_FRENCH_HAITI: return "fr_HT";
|
||
|
}
|
||
|
return "fr";
|
||
|
case LANG_FRISIAN: return "fy_NL";
|
||
|
case LANG_FULFULDE:
|
||
|
/* Spoken in Nigeria, Guinea, Senegal, Mali, Niger, Cameroon, Benin. */
|
||
|
return "ff_NG";
|
||
|
case LANG_GAELIC:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case 0x01: /* SCOTTISH */ return "gd_GB";
|
||
|
case 0x02: /* IRISH */ return "ga_IE";
|
||
|
}
|
||
|
return "C";
|
||
|
case LANG_GALICIAN: return "gl_ES";
|
||
|
case LANG_GEORGIAN: return "ka_GE";
|
||
|
case LANG_GERMAN:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_GERMAN: return "de_DE";
|
||
|
case SUBLANG_GERMAN_SWISS: return "de_CH";
|
||
|
case SUBLANG_GERMAN_AUSTRIAN: return "de_AT";
|
||
|
case SUBLANG_GERMAN_LUXEMBOURG: return "de_LU";
|
||
|
case SUBLANG_GERMAN_LIECHTENSTEIN: return "de_LI";
|
||
|
}
|
||
|
return "de";
|
||
|
case LANG_GREEK: return "el_GR";
|
||
|
case LANG_GUARANI: return "gn_PY";
|
||
|
case LANG_GUJARATI: return "gu_IN";
|
||
|
case LANG_HAUSA: return "ha_NG";
|
||
|
case LANG_HAWAIIAN:
|
||
|
/* FIXME: Do they mean Hawaiian ("haw_US", 1000 speakers)
|
||
|
or Hawaii Creole English ("cpe_US", 600000 speakers)? */
|
||
|
return "cpe_US";
|
||
|
case LANG_HEBREW: return "he_IL";
|
||
|
case LANG_HINDI: return "hi_IN";
|
||
|
case LANG_HUNGARIAN: return "hu_HU";
|
||
|
case LANG_IBIBIO: return "nic_NG";
|
||
|
case LANG_ICELANDIC: return "is_IS";
|
||
|
case LANG_IGBO: return "ig_NG";
|
||
|
case LANG_INDONESIAN: return "id_ID";
|
||
|
case LANG_INUKTITUT: return "iu_CA";
|
||
|
case LANG_ITALIAN:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_ITALIAN: return "it_IT";
|
||
|
case SUBLANG_ITALIAN_SWISS: return "it_CH";
|
||
|
}
|
||
|
return "it";
|
||
|
case LANG_JAPANESE: return "ja_JP";
|
||
|
case LANG_KANNADA: return "kn_IN";
|
||
|
case LANG_KANURI: return "kr_NG";
|
||
|
case LANG_KASHMIRI:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_DEFAULT: return "ks_PK";
|
||
|
case SUBLANG_KASHMIRI_INDIA: return "ks_IN";
|
||
|
}
|
||
|
return "ks";
|
||
|
case LANG_KAZAK: return "kk_KZ";
|
||
|
case LANG_KONKANI:
|
||
|
/* FIXME: Adjust this when such locales appear on Unix. */
|
||
|
return "kok_IN";
|
||
|
case LANG_KOREAN: return "ko_KR";
|
||
|
case LANG_KYRGYZ: return "ky_KG";
|
||
|
case LANG_LAO: return "lo_LA";
|
||
|
case LANG_LATIN: return "la_VA";
|
||
|
case LANG_LATVIAN: return "lv_LV";
|
||
|
case LANG_LITHUANIAN: return "lt_LT";
|
||
|
case LANG_MACEDONIAN: return "mk_MK";
|
||
|
case LANG_MALAY:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_MALAY_MALAYSIA: return "ms_MY";
|
||
|
case SUBLANG_MALAY_BRUNEI_DARUSSALAM: return "ms_BN";
|
||
|
}
|
||
|
return "ms";
|
||
|
case LANG_MALAYALAM: return "ml_IN";
|
||
|
case LANG_MALTESE: return "mt_MT";
|
||
|
case LANG_MANIPURI:
|
||
|
/* FIXME: Adjust this when such locales appear on Unix. */
|
||
|
return "mni_IN";
|
||
|
case LANG_MARATHI: return "mr_IN";
|
||
|
case LANG_MONGOLIAN:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_DEFAULT: return "mn_MN";
|
||
|
}
|
||
|
return "mn"; /* Ambiguous: could be "mn_CN" or "mn_MN". */
|
||
|
case LANG_NEPALI:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_DEFAULT: return "ne_NP";
|
||
|
case SUBLANG_NEPALI_INDIA: return "ne_IN";
|
||
|
}
|
||
|
return "ne";
|
||
|
case LANG_NORWEGIAN:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_NORWEGIAN_BOKMAL: return "nb_NO";
|
||
|
case SUBLANG_NORWEGIAN_NYNORSK: return "nn_NO";
|
||
|
}
|
||
|
return "no";
|
||
|
case LANG_ORIYA: return "or_IN";
|
||
|
case LANG_OROMO: return "om_ET";
|
||
|
case LANG_PAPIAMENTU: return "pap_AN";
|
||
|
case LANG_PASHTO:
|
||
|
return "ps"; /* Ambiguous: could be "ps_PK" or "ps_AF". */
|
||
|
case LANG_POLISH: return "pl_PL";
|
||
|
case LANG_PORTUGUESE:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_PORTUGUESE: return "pt_PT";
|
||
|
/* Hmm. SUBLANG_PORTUGUESE_BRAZILIAN == SUBLANG_DEFAULT.
|
||
|
Same phenomenon as SUBLANG_ENGLISH_US == SUBLANG_DEFAULT. */
|
||
|
case SUBLANG_PORTUGUESE_BRAZILIAN: return "pt_BR";
|
||
|
}
|
||
|
return "pt";
|
||
|
case LANG_PUNJABI:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_PUNJABI_INDIA: return "pa_IN"; /* Gurmukhi script */
|
||
|
case SUBLANG_PUNJABI_PAKISTAN: return "pa_PK"; /* Arabic script */
|
||
|
}
|
||
|
return "pa";
|
||
|
case LANG_RHAETO_ROMANCE: return "rm_CH";
|
||
|
case LANG_ROMANIAN:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_ROMANIAN_ROMANIA: return "ro_RO";
|
||
|
case SUBLANG_ROMANIAN_MOLDOVA: return "ro_MD";
|
||
|
}
|
||
|
return "ro";
|
||
|
case LANG_RUSSIAN:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_DEFAULT: return "ru_RU";
|
||
|
}
|
||
|
return "ru"; /* Ambiguous: could be "ru_RU" or "ru_UA" or "ru_MD". */
|
||
|
case LANG_SAAMI: /* actually Northern Sami */ return "se_NO";
|
||
|
case LANG_SANSKRIT: return "sa_IN";
|
||
|
case LANG_SINDHI:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_SINDHI_INDIA: return "sd_IN";
|
||
|
case SUBLANG_SINDHI_PAKISTAN: return "sd_PK";
|
||
|
}
|
||
|
return "sd";
|
||
|
case LANG_SINHALESE: return "si_LK";
|
||
|
case LANG_SLOVAK: return "sk_SK";
|
||
|
case LANG_SLOVENIAN: return "sl_SI";
|
||
|
case LANG_SOMALI: return "so_SO";
|
||
|
case LANG_SORBIAN:
|
||
|
/* FIXME: Adjust this when such locales appear on Unix. */
|
||
|
return "wen_DE";
|
||
|
case LANG_SPANISH:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_SPANISH: return "es_ES";
|
||
|
case SUBLANG_SPANISH_MEXICAN: return "es_MX";
|
||
|
case SUBLANG_SPANISH_MODERN:
|
||
|
return "es_ES@modern"; /* not seen on Unix */
|
||
|
case SUBLANG_SPANISH_GUATEMALA: return "es_GT";
|
||
|
case SUBLANG_SPANISH_COSTA_RICA: return "es_CR";
|
||
|
case SUBLANG_SPANISH_PANAMA: return "es_PA";
|
||
|
case SUBLANG_SPANISH_DOMINICAN_REPUBLIC: return "es_DO";
|
||
|
case SUBLANG_SPANISH_VENEZUELA: return "es_VE";
|
||
|
case SUBLANG_SPANISH_COLOMBIA: return "es_CO";
|
||
|
case SUBLANG_SPANISH_PERU: return "es_PE";
|
||
|
case SUBLANG_SPANISH_ARGENTINA: return "es_AR";
|
||
|
case SUBLANG_SPANISH_ECUADOR: return "es_EC";
|
||
|
case SUBLANG_SPANISH_CHILE: return "es_CL";
|
||
|
case SUBLANG_SPANISH_URUGUAY: return "es_UY";
|
||
|
case SUBLANG_SPANISH_PARAGUAY: return "es_PY";
|
||
|
case SUBLANG_SPANISH_BOLIVIA: return "es_BO";
|
||
|
case SUBLANG_SPANISH_EL_SALVADOR: return "es_SV";
|
||
|
case SUBLANG_SPANISH_HONDURAS: return "es_HN";
|
||
|
case SUBLANG_SPANISH_NICARAGUA: return "es_NI";
|
||
|
case SUBLANG_SPANISH_PUERTO_RICO: return "es_PR";
|
||
|
}
|
||
|
return "es";
|
||
|
case LANG_SUTU: return "bnt_TZ"; /* or "st_LS" or "nso_ZA"? */
|
||
|
case LANG_SWAHILI: return "sw_KE";
|
||
|
case LANG_SWEDISH:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_DEFAULT: return "sv_SE";
|
||
|
case SUBLANG_SWEDISH_FINLAND: return "sv_FI";
|
||
|
}
|
||
|
return "sv";
|
||
|
case LANG_SYRIAC: return "syr_TR"; /* An extinct language. */
|
||
|
case LANG_TAGALOG: return "tl_PH";
|
||
|
case LANG_TAJIK: return "tg_TJ";
|
||
|
case LANG_TAMAZIGHT:
|
||
|
switch (sub)
|
||
|
{
|
||
|
/* FIXME: Adjust this when Tamazight locales appear on Unix. */
|
||
|
case SUBLANG_TAMAZIGHT_ARABIC: return "ber_MA@arabic";
|
||
|
case SUBLANG_TAMAZIGHT_LATIN: return "ber_MA@latin";
|
||
|
}
|
||
|
return "ber_MA";
|
||
|
case LANG_TAMIL:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_DEFAULT: return "ta_IN";
|
||
|
}
|
||
|
return "ta"; /* Ambiguous: could be "ta_IN" or "ta_LK" or "ta_SG". */
|
||
|
case LANG_TATAR: return "tt_RU";
|
||
|
case LANG_TELUGU: return "te_IN";
|
||
|
case LANG_THAI: return "th_TH";
|
||
|
case LANG_TIBETAN: return "bo_CN";
|
||
|
case LANG_TIGRINYA:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_TIGRINYA_ETHIOPIA: return "ti_ET";
|
||
|
case SUBLANG_TIGRINYA_ERITREA: return "ti_ER";
|
||
|
}
|
||
|
return "ti";
|
||
|
case LANG_TSONGA: return "ts_ZA";
|
||
|
case LANG_TSWANA: return "tn_BW";
|
||
|
case LANG_TURKISH: return "tr_TR";
|
||
|
case LANG_TURKMEN: return "tk_TM";
|
||
|
case LANG_UKRAINIAN: return "uk_UA";
|
||
|
case LANG_URDU:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_URDU_PAKISTAN: return "ur_PK";
|
||
|
case SUBLANG_URDU_INDIA: return "ur_IN";
|
||
|
}
|
||
|
return "ur";
|
||
|
case LANG_UZBEK:
|
||
|
switch (sub)
|
||
|
{
|
||
|
case SUBLANG_UZBEK_LATIN: return "uz_UZ";
|
||
|
case SUBLANG_UZBEK_CYRILLIC: return "uz_UZ@cyrillic";
|
||
|
}
|
||
|
return "uz";
|
||
|
case LANG_VENDA: return "ve_ZA";
|
||
|
case LANG_VIETNAMESE: return "vi_VN";
|
||
|
case LANG_WELSH: return "cy_GB";
|
||
|
case LANG_XHOSA: return "xh_ZA";
|
||
|
case LANG_YI: return "sit_CN";
|
||
|
case LANG_YIDDISH: return "yi_IL";
|
||
|
case LANG_YORUBA: return "yo_NG";
|
||
|
case LANG_ZULU: return "zu_ZA";
|
||
|
default: return "C";
|
||
|
}
|
||
|
}
|
||
|
# endif
|
||
|
#endif
|
||
|
}
|
||
|
|
||
|
const char *
|
||
|
_nl_locale_name (int category, const char *categoryname)
|
||
|
{
|
||
|
const char *retval;
|
||
|
|
||
|
retval = _nl_locale_name_posix (category, categoryname);
|
||
|
if (retval != NULL)
|
||
|
return retval;
|
||
|
|
||
|
return _nl_locale_name_default ();
|
||
|
}
|