2 * Copyright (c) 2010 Ævar Arnfjörð Bjarmason
5 #include "git-compat-util.h"
13 # ifdef HAVE_LIBCHARSET_H
14 # include <libcharset.h>
16 # include <langinfo.h>
17 # define locale_charset() nl_langinfo(CODESET)
21 static const char *charset;
24 * Guess the user's preferred languages from the value in LANGUAGE environment
25 * variable and LC_MESSAGES locale category if NO_GETTEXT is not defined.
27 * The result can be a colon-separated list like "ko:ja:en".
29 const char *get_preferred_languages(void)
33 retval = getenv("LANGUAGE");
34 if (retval && *retval)
38 retval = setlocale(LC_MESSAGES, NULL);
39 if (retval && *retval &&
40 strcmp(retval, "C") &&
41 strcmp(retval, "POSIX"))
49 int use_gettext_poison(void)
51 static int poison_requested = -1;
52 if (poison_requested == -1)
53 poison_requested = getenv("GIT_GETTEXT_POISON") ? 1 : 0;
54 return poison_requested;
59 static int test_vsnprintf(const char *fmt, ...)
65 ret = vsnprintf(buf, sizeof(buf), fmt, ap);
70 static void init_gettext_charset(const char *domain)
73 This trick arranges for messages to be emitted in the user's
74 requested encoding, but avoids setting LC_CTYPE from the
75 environment for the whole program.
77 This primarily done to avoid a bug in vsnprintf in the GNU C
78 Library [1]. which triggered a "your vsnprintf is broken" error
79 on Git's own repository when inspecting v0.99.6~1 under a UTF-8
82 That commit contains a ISO-8859-1 encoded author name, which
83 the locale aware vsnprintf(3) won't interpolate in the format
84 argument, due to mismatch between the data encoding and the
87 Even if it wasn't for that bug we wouldn't want to use LC_CTYPE at
88 this point, because it'd require auditing all the code that uses C
89 functions whose semantics are modified by LC_CTYPE.
91 But only setting LC_MESSAGES as we do creates a problem, since
92 we declare the encoding of our PO files[2] the gettext
93 implementation will try to recode it to the user's locale, but
94 without LC_CTYPE it'll emit something like this on 'git init'
95 under the Icelandic locale:
97 Bj? til t?ma Git lind ? /hlagh/.git/
99 Gettext knows about the encoding of our PO file, but we haven't
100 told it about the user's encoding, so all the non-US-ASCII
101 characters get encoded to question marks.
103 But we're in luck! We can set LC_CTYPE from the environment
104 only while we call nl_langinfo and
105 bind_textdomain_codeset. That suffices to tell gettext what
106 encoding it should emit in, so it'll now say:
108 Bjó til tóma Git lind í /hlagh/.git/
110 And the equivalent ISO-8859-1 string will be emitted under a
113 With this change way we get the advantages of setting LC_CTYPE
114 (talk to the user in his language/encoding), without the major
115 drawbacks (changed semantics for C functions we rely on).
117 However foreign functions using other message catalogs that
118 aren't using our neat trick will still have a problem, e.g. if
119 we have to call perror(3):
127 setlocale(LC_MESSAGES, "");
128 setlocale(LC_CTYPE, "C");
134 Running that will give you a message with question marks:
136 $ LANGUAGE= LANG=de_DE.utf8 ./test
137 test: Kein passendes Ger?t gefunden
139 The vsnprintf bug has been fixed since glibc 2.17.
141 Then we could simply set LC_CTYPE from the environment, which would
142 make things like the external perror(3) messages work.
144 See t/t0203-gettext-setlocale-sanity.sh's "gettext.c" tests for
147 1. http://sourceware.org/bugzilla/show_bug.cgi?id=6530
148 2. E.g. "Content-Type: text/plain; charset=UTF-8\n" in po/is.po
150 setlocale(LC_CTYPE, "");
151 charset = locale_charset();
152 bind_textdomain_codeset(domain, charset);
153 /* the string is taken from v0.99.6~1 */
154 if (test_vsnprintf("%.*s", 13, "David_K\345gedal") < 0)
155 setlocale(LC_CTYPE, "C");
158 void git_setup_gettext(void)
160 const char *podir = getenv("GIT_TEXTDOMAINDIR");
163 podir = GIT_LOCALE_PATH;
164 bindtextdomain("git", podir);
165 setlocale(LC_MESSAGES, "");
166 setlocale(LC_TIME, "");
167 init_gettext_charset("git");
171 /* return the number of columns of string 's' in current locale */
172 int gettext_width(const char *s)
174 static int is_utf8 = -1;
176 is_utf8 = is_utf8_locale();
178 return is_utf8 ? utf8_strwidth(s) : strlen(s);
182 int is_utf8_locale(void)
186 const char *env = getenv("LC_ALL");
188 env = getenv("LC_CTYPE");
190 env = getenv("LANG");
193 if (strchr(env, '.'))
194 env = strchr(env, '.') + 1;
195 charset = xstrdup(env);
198 return is_encoding_utf8(charset);