1 files changed, 29 insertions, 9 deletions

diff --git a/libbb/unicode.c b/libbb/unicode.c
index 99dc1dfa6..9c4da50d3 100644
--- a/libbb/unicode.c
+++ b/libbb/unicode.c
@@ -28,19 +28,37 @@ void FAST_FUNC reinit_unicode(const char *LANG)
         static const char unicode_0x394[] = { 0xce, 0x94, 0 };
         size_t width;
 
+        /* We pass "" instead of "C" because some libc's have
+         * non-ASCII default locale for setlocale("") call
+         * (this allows users of such libc to have Unicoded
+         * system without having to mess with env).
+         *
+         * We set LC_CTYPE because (a) we may be called with $LC_CTYPE
+         * value in LANG, not with $LC_ALL, (b) internationalized
+         * LC_NUMERIC and LC_TIME are more PITA than benefit
+         * (for one, some utilities have hard time with comma
+         * used as a fractional separator).
+         */
 //TODO: avoid repeated calls by caching last string?
-        setlocale(LC_ALL, (LANG && LANG[0]) ? LANG : "C");
+        setlocale(LC_CTYPE, LANG ? LANG : "");
 
         /* In unicode, this is a one character string */
-// can use unicode_strlen(string) too, but otherwise unicode_strlen() is unused
-        width = mbstowcs(NULL, unicode_0x394, INT_MAX);
+        width = unicode_strlen(unicode_0x394);
         unicode_status = (width == 1 ? UNICODE_ON : UNICODE_OFF);
 }
 
 void FAST_FUNC init_unicode(void)
 {
-        if (unicode_status == UNICODE_UNKNOWN)
-                reinit_unicode(getenv("LANG"));
+        /* Some people set only $LC_CTYPE, not $LC_ALL, because they want
+         * only Unicode to be activated on their system, not the whole
+         * shebang of wrong decimal points, strange date formats and so on.
+         */
+        if (unicode_status == UNICODE_UNKNOWN) {
+                char *s = getenv("LC_ALL");
+                if (!s) s = getenv("LC_CTYPE");
+                if (!s) s = getenv("LANG");
+                reinit_unicode(s);
+        }
 }
 
 #else
@@ -58,8 +76,12 @@ void FAST_FUNC reinit_unicode(const char *LANG)
 
 void FAST_FUNC init_unicode(void)
 {
-        if (unicode_status == UNICODE_UNKNOWN)
-                reinit_unicode(getenv("LANG"));
+        if (unicode_status == UNICODE_UNKNOWN) {
+                char *s = getenv("LC_ALL");
+                if (!s) s = getenv("LC_CTYPE");
+                if (!s) s = getenv("LANG");
+                reinit_unicode(s);
+        }
 }
 # endif
 
@@ -963,7 +985,6 @@ int FAST_FUNC unicode_bidi_is_neutral_wchar(wint_t wc)
 
 /* The rest is mostly same for libc and for "homegrown" support */
 
-#if 0 // UNUSED
 size_t FAST_FUNC unicode_strlen(const char *string)
 {
         size_t width = mbstowcs(NULL, string, INT_MAX);
@@ -971,7 +992,6 @@ size_t FAST_FUNC unicode_strlen(const char *string)
                 return strlen(string);
         return width;
 }
-#endif
 
 size_t FAST_FUNC unicode_strwidth(const char *string)
 {