2 * Copyright (c) 2010 Ævar Arnfjörð Bjarmason
14 # ifdef HAVE_LIBCHARSET_H
15 # include <libcharset.h>
17 # include <langinfo.h>
18 # define locale_charset() nl_langinfo(CODESET)
22 static const char *charset;
25 * Guess the user's preferred languages from the value in LANGUAGE environment
26 * variable and LC_MESSAGES locale category if NO_GETTEXT is not defined.
28 * The result can be a colon-separated list like "ko:ja:en".
30 const char *get_preferred_languages(void)
34 retval = getenv("LANGUAGE");
35 if (retval && *retval)
39 retval = setlocale(LC_MESSAGES, NULL);
40 if (retval && *retval &&
41 strcmp(retval, "C") &&
42 strcmp(retval, "POSIX"))
50 int use_gettext_poison(void)
52 static int poison_requested = -1;
53 if (poison_requested == -1)
54 poison_requested = getenv("GIT_GETTEXT_POISON") ? 1 : 0;
55 return poison_requested;
60 static int test_vsnprintf(const char *fmt, ...)
66 ret = vsnprintf(buf, sizeof(buf), fmt, ap);
71 static void init_gettext_charset(const char *domain)
74 This trick arranges for messages to be emitted in the user's
75 requested encoding, but avoids setting LC_CTYPE from the
76 environment for the whole program.
78 This primarily done to avoid a bug in vsnprintf in the GNU C
79 Library [1]. which triggered a "your vsnprintf is broken" error
80 on Git's own repository when inspecting v0.99.6~1 under a UTF-8
83 That commit contains a ISO-8859-1 encoded author name, which
84 the locale aware vsnprintf(3) won't interpolate in the format
85 argument, due to mismatch between the data encoding and the
88 Even if it wasn't for that bug we wouldn't want to use LC_CTYPE at
89 this point, because it'd require auditing all the code that uses C
90 functions whose semantics are modified by LC_CTYPE.
92 But only setting LC_MESSAGES as we do creates a problem, since
93 we declare the encoding of our PO files[2] the gettext
94 implementation will try to recode it to the user's locale, but
95 without LC_CTYPE it'll emit something like this on 'git init'
96 under the Icelandic locale:
98 Bj? til t?ma Git lind ? /hlagh/.git/
100 Gettext knows about the encoding of our PO file, but we haven't
101 told it about the user's encoding, so all the non-US-ASCII
102 characters get encoded to question marks.
104 But we're in luck! We can set LC_CTYPE from the environment
105 only while we call nl_langinfo and
106 bind_textdomain_codeset. That suffices to tell gettext what
107 encoding it should emit in, so it'll now say:
109 Bjó til tóma Git lind í /hlagh/.git/
111 And the equivalent ISO-8859-1 string will be emitted under a
114 With this change way we get the advantages of setting LC_CTYPE
115 (talk to the user in his language/encoding), without the major
116 drawbacks (changed semantics for C functions we rely on).
118 However foreign functions using other message catalogs that
119 aren't using our neat trick will still have a problem, e.g. if
120 we have to call perror(3):
128 setlocale(LC_MESSAGES, "");
129 setlocale(LC_CTYPE, "C");
135 Running that will give you a message with question marks:
137 $ LANGUAGE= LANG=de_DE.utf8 ./test
138 test: Kein passendes Ger?t gefunden
140 The vsnprintf bug has been fixed since glibc 2.17.
142 Then we could simply set LC_CTYPE from the environment, which would
143 make things like the external perror(3) messages work.
145 See t/t0203-gettext-setlocale-sanity.sh's "gettext.c" tests for
148 1. http://sourceware.org/bugzilla/show_bug.cgi?id=6530
149 2. E.g. "Content-Type: text/plain; charset=UTF-8\n" in po/is.po
151 setlocale(LC_CTYPE, "");
152 charset = locale_charset();
153 bind_textdomain_codeset(domain, charset);
154 /* the string is taken from v0.99.6~1 */
155 if (test_vsnprintf("%.*s", 13, "David_K\345gedal") < 0)
156 setlocale(LC_CTYPE, "C");
159 void git_setup_gettext(void)
161 const char *podir = getenv(GIT_TEXT_DOMAIN_DIR_ENVIRONMENT);
165 podir = p = system_path(GIT_LOCALE_PATH);
167 if (!is_directory(podir)) {
172 bindtextdomain("git", podir);
173 setlocale(LC_MESSAGES, "");
174 setlocale(LC_TIME, "");
175 init_gettext_charset("git");
181 /* return the number of columns of string 's' in current locale */
182 int gettext_width(const char *s)
184 static int is_utf8 = -1;
186 is_utf8 = is_utf8_locale();
188 return is_utf8 ? utf8_strwidth(s) : strlen(s);
192 int is_utf8_locale(void)
196 const char *env = getenv("LC_ALL");
198 env = getenv("LC_CTYPE");
200 env = getenv("LANG");
203 if (strchr(env, '.'))
204 env = strchr(env, '.') + 1;
205 charset = xstrdup(env);
208 return is_encoding_utf8(charset);