Merge branch 'pp/maint-doc-pager-config'
[git/raj.git] / gettext.c
blob71e954563d7068500bd41269dd57f01b097f7c3d
1 /*
2 * Copyright (c) 2010 Ævar Arnfjörð Bjarmason
3 */
5 #include "git-compat-util.h"
6 #include "gettext.h"
7 #include "strbuf.h"
8 #include "utf8.h"
10 #ifndef NO_GETTEXT
11 # include <locale.h>
12 # include <libintl.h>
13 # ifdef HAVE_LIBCHARSET_H
14 # include <libcharset.h>
15 # else
16 # include <langinfo.h>
17 # define locale_charset() nl_langinfo(CODESET)
18 # endif
19 #endif
21 #ifdef GETTEXT_POISON
22 int use_gettext_poison(void)
24 static int poison_requested = -1;
25 if (poison_requested == -1)
26 poison_requested = getenv("GIT_GETTEXT_POISON") ? 1 : 0;
27 return poison_requested;
29 #endif
31 #ifndef NO_GETTEXT
32 static const char *charset;
33 static void init_gettext_charset(const char *domain)
36 This trick arranges for messages to be emitted in the user's
37 requested encoding, but avoids setting LC_CTYPE from the
38 environment for the whole program.
40 This primarily done to avoid a bug in vsnprintf in the GNU C
41 Library [1]. which triggered a "your vsnprintf is broken" error
42 on Git's own repository when inspecting v0.99.6~1 under a UTF-8
43 locale.
45 That commit contains a ISO-8859-1 encoded author name, which
46 the locale aware vsnprintf(3) won't interpolate in the format
47 argument, due to mismatch between the data encoding and the
48 locale.
50 Even if it wasn't for that bug we wouldn't want to use LC_CTYPE at
51 this point, because it'd require auditing all the code that uses C
52 functions whose semantics are modified by LC_CTYPE.
54 But only setting LC_MESSAGES as we do creates a problem, since
55 we declare the encoding of our PO files[2] the gettext
56 implementation will try to recode it to the user's locale, but
57 without LC_CTYPE it'll emit something like this on 'git init'
58 under the Icelandic locale:
60 Bj? til t?ma Git lind ? /hlagh/.git/
62 Gettext knows about the encoding of our PO file, but we haven't
63 told it about the user's encoding, so all the non-US-ASCII
64 characters get encoded to question marks.
66 But we're in luck! We can set LC_CTYPE from the environment
67 only while we call nl_langinfo and
68 bind_textdomain_codeset. That suffices to tell gettext what
69 encoding it should emit in, so it'll now say:
71 Bjó til tóma Git lind í /hlagh/.git/
73 And the equivalent ISO-8859-1 string will be emitted under a
74 ISO-8859-1 locale.
76 With this change way we get the advantages of setting LC_CTYPE
77 (talk to the user in his language/encoding), without the major
78 drawbacks (changed semantics for C functions we rely on).
80 However foreign functions using other message catalogs that
81 aren't using our neat trick will still have a problem, e.g. if
82 we have to call perror(3):
84 #include <stdio.h>
85 #include <locale.h>
86 #include <errno.h>
88 int main(void)
90 setlocale(LC_MESSAGES, "");
91 setlocale(LC_CTYPE, "C");
92 errno = ENODEV;
93 perror("test");
94 return 0;
97 Running that will give you a message with question marks:
99 $ LANGUAGE= LANG=de_DE.utf8 ./test
100 test: Kein passendes Ger?t gefunden
102 In the long term we should probably see about getting that
103 vsnprintf bug in glibc fixed, and audit our code so it won't
104 fall apart under a non-C locale.
106 Then we could simply set LC_CTYPE from the environment, which would
107 make things like the external perror(3) messages work.
109 See t/t0203-gettext-setlocale-sanity.sh's "gettext.c" tests for
110 regression tests.
112 1. http://sourceware.org/bugzilla/show_bug.cgi?id=6530
113 2. E.g. "Content-Type: text/plain; charset=UTF-8\n" in po/is.po
115 setlocale(LC_CTYPE, "");
116 charset = locale_charset();
117 bind_textdomain_codeset(domain, charset);
118 setlocale(LC_CTYPE, "C");
121 void git_setup_gettext(void)
123 const char *podir = getenv("GIT_TEXTDOMAINDIR");
125 if (!podir)
126 podir = GIT_LOCALE_PATH;
127 bindtextdomain("git", podir);
128 setlocale(LC_MESSAGES, "");
129 init_gettext_charset("git");
130 textdomain("git");
133 /* return the number of columns of string 's' in current locale */
134 int gettext_width(const char *s)
136 static int is_utf8 = -1;
137 if (is_utf8 == -1)
138 is_utf8 = !strcmp(charset, "UTF-8");
140 return is_utf8 ? utf8_strwidth(s) : strlen(s);
142 #endif