[PATCH] gettext: use libcharset when available

["Ævar Arnfjörð Bjarmason" <avarab@gmail.com>]

From: Erik Faye-Lund <kusmabite@gmail.com>

Change the git_setup_gettext function to use libcharset to query the
character set of the current locale if it's available. I.e. use it
instead of nl_langinfo if HAVE_LIBCHARSET_H is set.

The GNU gettext manual recommends using langinfo.h's
nl_langinfo(CODESET) to acquire the current character set, but on
systems that have libcharset.h's locale_charset() using the latter is
either saner, or the only option on those systems.

GNU and Solaris have a nl_langinfo(CODESET), FreeBSD can use either,
but MingW and some others need to use libcharset.h's locale_charset()
instead.

Since locale_charset returns a const char* instead of char* as
nl_langinfo does the type of the variable we're using to store the
charset in git_setup_gettext has been changed.

Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com>
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
---

Junio, this goes on top of ab/i18n.

On Wed, Sep 29, 2010 at 11:41, Erik Faye-Lund <kusmabite@gmail.com> wrote:
> On Wed, Sep 29, 2010 at 12:00 PM, Ævar Arnfjörð Bjarmason
> <avarab@gmail.com> wrote:
>> On Tue, Sep 28, 2010 at 21:47, Erik Faye-Lund <kusmabite@gmail.com> wrote:
>>> On Tue, Sep 28, 2010 at 8:29 PM, Ævar Arnfjörð Bjarmason
>>>>  * Added defaults for NO_LIBCHARSET to the default, I only changed the
>>>>   defaults for the MINGW entry, maybe it should be changed on Cygwin
>>>>   and Windows too? And probably on OpenBSD and NetBSD too.
>>>>
>>>
>>> I don't think NO_LIBCHARSET should be the default. libcharset is
>>> reported to be a bit better than nl_langinfo at normalizing the
>>> encoding, and GNU gettext depends on libcharset (through libiconv,
>>> which libcharset is distributed with). So in the case of a GNU
>>> gettext, libcharset should really be present.
>>
>> I can't find any package (with apt-file) on Debian or Ubuntu that
>> provides libcharset.h, but I have langinfo.h on those systems.
>>
>
> Strange. A 'make install' on libiconv installed libcharset.h to
> $prefix/include on my system. But looking a bit deeper, it seems that
> glibc supplies it's own iconv implementation (perhaps based on
> libiconv, I don't know). So yes, I tend to agree with you. GNU
> platforms should not be expected to have libcharset.

I asked around and none of Debian/Ubuntu/Fedora and other Linux
systems have libcharset.h, and using the gettext branch without
libcharset.h on FreeBSD works fine.

>> The GNU gettext manual also reccomends the use of nl_langinfo in
>> "11.2.4 How to specify the output character set `gettext' uses", so it
>> seems that using that and not libiconv is the default way of doing
>> things on GNU gettext + GNU libc systems.
>>
>
> OK, fair enough. I based my comment on some comment by the GNU gettext
> maintainer (who is also the libcharset maintainer - libcharset does in
> fact use nl_langinfo if present), but since this is in the manual I
> fully withdraw my comment.
>
> Then again, if this is an opt-in rather than an opt-out, perhaps we
> should change the switch to HAVE_LIBCHARSET? I don't mean to go in
> circles here, but it sounds more self-documenting to me.

Agreed, changed in this version. I also added a bit to config.mak.in
to make the configure.ac change actually do something, and changed the
docs & commit message.

 Makefile      |   17 +++++++++++++++++
 config.mak.in |    1 +
 configure.ac  |    6 ++++++
 gettext.c     |   10 +++++++++-
 4 files changed, 33 insertions(+), 1 deletions(-)

diff --git a/Makefile b/Makefile
index 680e578..a05396b 100644
--- a/Makefile
+++ b/Makefile
@@ -43,6 +43,12 @@ all::
 # on platforms where we don't expect glibc (Linux, Hurd,
 # GNU/kFreeBSD), which includes libintl.
 #
+# Define HAVE_LIBCHARSET_H if you haven't set NO_GETTEXT and you can't
+# trust the langinfo.h's nl_langinfo(CODESET) function to return the
+# current character set. GNU and Solaris have a nl_langinfo(CODESET),
+# FreeBSD can use either, but MingW and some others need to use
+# libcharset.h's locale_charset() instead.
+#
 # Define GNU_GETTEXT if you're using the GNU implementation of
 # libintl. We define this everywhere except on Solaris, which has its
 # own gettext implementation. If GNU_GETTEXT is set we'll use GNU
@@ -792,6 +798,10 @@ ifndef NO_GETTEXT
 	# Systems that don't use GNU gettext are the exception. Only
 	# Solaris has a mature non-GNU gettext implementation.
 	GNU_GETTEXT = YesPlease
+
+	# Since we assume a GNU gettext by default we also assume a
+	# GNU-like langinfo.h by default
+	HAVE_LIBCHARSET_H =
 endif
 
 # We choose to avoid "if .. else if .. else .. endif endif"
@@ -1180,6 +1190,9 @@ ifneq (,$(wildcard ../THIS_IS_MSYSGIT))
 	EXTLIBS += /mingw/lib/libz.a
 	NO_R_TO_GCC_LINKER = YesPlease
 	INTERNAL_QSORT = YesPlease
+ifndef NO_GETTEXT
+	HAVE_LIBCHARSET_H = YesPlease
+endif
 else
 	NO_CURL = YesPlease
 endif
@@ -1964,6 +1977,10 @@ config.s config.o: EXTRA_CPPFLAGS = -DETC_GITCONFIG='"$(ETC_GITCONFIG_SQ)"'
 
 http.s http.o: EXTRA_CPPFLAGS = -DGIT_HTTP_USER_AGENT='"git/$(GIT_VERSION)"'
 
+ifdef HAVE_LIBCHARSET_H
+gettext.s gettext.o: EXTRA_CPPFLAGS = -DHAVE_LIBCHARSET_H
+endif
+
 ifdef NO_EXPAT
 http-walker.s http-walker.o: EXTRA_CPPFLAGS = -DNO_EXPAT
 endif
diff --git a/config.mak.in b/config.mak.in
index 9f47aa5..969cbaa 100644
--- a/config.mak.in
+++ b/config.mak.in
@@ -34,6 +34,7 @@ NO_CURL=@NO_CURL@
 NO_EXPAT=@NO_EXPAT@
 NO_LIBGEN_H=@NO_LIBGEN_H@
 HAVE_PATHS_H=@HAVE_PATHS_H@
+HAVE_LIBCHARSET_H=@HAVE_LIBCHARSET_H@
 NO_GETTEXT=@NO_GETTEXT@
 NEEDS_LIBICONV=@NEEDS_LIBICONV@
 NEEDS_SOCKET=@NEEDS_SOCKET@
diff --git a/configure.ac b/configure.ac
index 1821d89..b06bad1 100644
--- a/configure.ac
+++ b/configure.ac
@@ -810,6 +810,12 @@ AC_CHECK_HEADER([libintl.h],
 [NO_GETTEXT=YesPlease])
 AC_SUBST(NO_GETTEXT)
 #
+# Define HAVE_LIBCHARSET_H if have libcharset.h
+AC_CHECK_HEADER([libcharset.h],
+[HAVE_LIBCHARSET_H=YesPlease],
+[HAVE_LIBCHARSET_H=])
+AC_SUBST(HAVE_LIBCHARSET_H)
+#
 # Define NO_STRCASESTR if you don't have strcasestr.
 GIT_CHECK_FUNC(strcasestr,
 [NO_STRCASESTR=],
diff --git a/gettext.c b/gettext.c
index 8644098..9bdac56 100644
--- a/gettext.c
+++ b/gettext.c
@@ -1,13 +1,17 @@
 #include "exec_cmd.h"
 #include <locale.h>
 #include <libintl.h>
+#ifdef HAVE_LIBCHARSET_H
+#include <libcharset.h>
+#else
 #include <langinfo.h>
+#endif
 #include <stdlib.h>
 
 extern void git_setup_gettext(void) {
 	char *podir;
 	char *envdir = getenv("GIT_TEXTDOMAINDIR");
-	char *charset;
+	const char *charset;
 
 	if (envdir) {
 		(void)bindtextdomain("git", envdir);
@@ -20,7 +24,11 @@ extern void git_setup_gettext(void) {
 
 	(void)setlocale(LC_MESSAGES, "");
 	(void)setlocale(LC_CTYPE, "");
+#ifdef HAVE_LIBCHARSET_H
+	charset = locale_charset();
+#else
 	charset = nl_langinfo(CODESET);
+#endif
 	(void)bind_textdomain_codeset("git", charset);
 	(void)setlocale(LC_CTYPE, "C");
 	(void)textdomain("git");
-- 
1.7.3.159.g610493