tweaking Unicode support

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

3 files changed, 57 insertions, 79 deletions

diff --git a/coreutils/ls.c b/coreutils/ls.c
index 20b979d..827b350 100644
--- a/coreutils/ls.c
+++ b/coreutils/ls.c
@@ -289,20 +289,6 @@ enum {
 } while (0)
 
 
-#if ENABLE_FEATURE_ASSUME_UNICODE
-/* libbb candidate */
-static size_t mbstrlen(const char *string)
-{
-	size_t width = mbstowcs(NULL, string, INT_MAX);
-	if (width == (size_t)-1L)
-		return strlen(string);
-	return width;
-}
-#else
-#define mbstrlen(string) strlen(string)
-#endif
-
-
 static struct dnode *my_stat(const char *fullname, const char *name, int force_follow)
 {
 	struct stat dstat;
@@ -570,7 +556,7 @@ static void showfiles(struct dnode **dn, int nfiles)
 	} else {
 		/* find the longest file name, use that as the column width */
 		for (i = 0; i < nfiles; i++) {
-			int len = mbstrlen(dn[i]->name);
+			int len = bb_mbstrlen(dn[i]->name);
 			if (column_width < len)
 				column_width = len;
 		}
@@ -717,7 +703,7 @@ static int print_name(const char *name)
 {
 	if (option_mask32 & OPT_Q) {
 #if ENABLE_FEATURE_ASSUME_UNICODE
-		int len = 2 + mbstrlen(name);
+		int len = 2 + bb_mbstrlen(name);
 #else
 		int len = 2;
 #endif
@@ -737,7 +723,7 @@ static int print_name(const char *name)
 	/* No -Q: */
 #if ENABLE_FEATURE_ASSUME_UNICODE
 	fputs(name, stdout);
-	return mbstrlen(name);
+	return bb_mbstrlen(name);
 #else
 	return printf("%s", name);
 #endif
diff --git a/include/unicode.h b/include/unicode.h
index be64a50..e006147 100644
--- a/include/unicode.h
+++ b/include/unicode.h
@@ -7,10 +7,13 @@
 
 #if !ENABLE_FEATURE_ASSUME_UNICODE
 
+# define bb_mbstrlen(string) strlen(string)
 # define check_unicode_in_env() ((void)0)
 
 #else
 
+size_t bb_mbstrlen(const char *string) FAST_FUNC;
+
 # if ENABLE_LOCALE_SUPPORT
 
 #  include <wchar.h>
@@ -19,6 +22,8 @@
 
 # else
 
+/* Crude "locale support" which knows only C and Unicode locales */
+
 #  if !ENABLE_FEATURE_CHECK_UNICODE_IN_ENV
 #   define check_unicode_in_env() ((void)0)
 #  else
@@ -50,8 +55,8 @@ int iswspace(wint_t wc) FAST_FUNC;
 int iswalnum(wint_t wc) FAST_FUNC;
 int iswpunct(wint_t wc) FAST_FUNC;
 
-# endif
+# endif /* !LOCALE_SUPPORT */
 
-#endif
+#endif /* FEATURE_ASSUME_UNICODE */
 
 #endif
diff --git a/libbb/unicode.c b/libbb/unicode.c
index a99f5ed..b977437 100644
--- a/libbb/unicode.c
+++ b/libbb/unicode.c
@@ -7,13 +7,22 @@
  * Licensed under GPL version 2, see file LICENSE in this tarball for details.
  */
 #include "libbb.h"
+# include "unicode.h"
 
-/* if LOCALE_SUPPORT, libc locale stuff takes care of it, else: */
+size_t FAST_FUNC bb_mbstrlen(const char *string)
+{
+	size_t width = mbstowcs(NULL, string, INT_MAX);
+	if (width == (size_t)-1L)
+		return strlen(string);
+	return width;
+}
 
 #if !ENABLE_LOCALE_SUPPORT
-#include "unicode.h"
 
-/* 0: not known yet,
+/* Crude "locale support" which knows only C and Unicode locales */
+
+/* unicode_is_enabled:
+ * 0: not known yet,
  * 1: not unicode (IOW: assuming one char == one byte)
  * 2: unicode
  */
@@ -39,6 +48,7 @@ void FAST_FUNC check_unicode_in_env(void)
 
 static size_t wcrtomb_internal(char *s, wchar_t wc)
 {
+	int n;
 	uint32_t v = wc;
 
 	if (v <= 0x7f) {
@@ -46,65 +56,38 @@ static size_t wcrtomb_internal(char *s, wchar_t wc)
 		return 1;
 	}
 
-	/* 80-7FF -> 110yyyxx 10xxxxxx */
-	if (v <= 0x7ff) {
-		s[1] = (v & 0x3f) | 0x80;
-		v >>= 6;
-		s[0] = v | 0xc0;
-		return 2;
-	}
+	/* RFC 3629 says that Unicode ends at 10FFFF,
+	 * but we cover entire 32 bits */
 
-	/* 800-FFFF -> 1110yyyy 10yyyyxx 10xxxxxx */
-	if (v <= 0xffff) {
-		s[2] = (v & 0x3f) | 0x80;
-		v >>= 6;
-		s[1] = (v & 0x3f) | 0x80;
-		v >>= 6;
-		s[0] = v | 0xe0;
-		return 3;
+	n = 2;
+	/* 4000000-FFFFFFFF -> 111111tt 10tttttt 10zzzzzz 10zzyyyy 10yyyyxx 10xxxxxx */
+	if (v >= 0x4000000) {
+		s[5] = (wc & 0x3f) | 0x80;
+		wc = (uint32_t)wc >> 6; /* ensuring that high bits are 0 */
+		n++;
+	}
+	/* 200000-3FFFFFF -> 111110tt 10zzzzzz 10zzyyyy 10yyyyxx 10xxxxxx */
+	if (v >= 0x200000) {
+		s[4] = (wc & 0x3f) | 0x80;
+		wc >>= 6;
+		n++;
 	}
-
-	/* RFC 3629 says that Unicode ends at 10FFFF */
-
 	/* 10000-1FFFFF -> 11110zzz 10zzyyyy 10yyyyxx 10xxxxxx */
-	if (v <= 0x1fffff) {
-		s[3] = (v & 0x3f) | 0x80;
-		v >>= 6;
-		s[2] = (v & 0x3f) | 0x80;
-		v >>= 6;
-		s[1] = (v & 0x3f) | 0x80;
-		v >>= 6;
-		s[0] = v | 0xf0;
-		return 4;
+	if (v >= 0x10000) {
+		s[3] = (wc & 0x3f) | 0x80;
+		wc >>= 6;
+		n++;
 	}
-
-	/* 200000-3FFFFFF -> 111110tt 10zzzzzz 10zzyyyy 10yyyyxx 10xxxxxx */
-	if (v <= 0x3ffffff) {
-		s[4] = (v & 0x3f) | 0x80;
-		v >>= 6;
-		s[3] = (v & 0x3f) | 0x80;
-		v >>= 6;
-		s[2] = (v & 0x3f) | 0x80;
-		v >>= 6;
-		s[1] = (v & 0x3f) | 0x80;
-		v >>= 6;
-		s[0] = v | 0xf8;
-		return 5;
+	/* 800-FFFF -> 1110yyyy 10yyyyxx 10xxxxxx */
+	if (v >= 0x800) {
+		s[2] = (wc & 0x3f) | 0x80;
+		wc >>= 6;
+		n++;
 	}
-
-	/* 4000000-FFFFFFFF -> 111111tt 10tttttt 10zzzzzz 10zzyyyy 10yyyyxx 10xxxxxx */
-	s[5] = (v & 0x3f) | 0x80;
-	v >>= 6;
-	s[4] = (v & 0x3f) | 0x80;
-	v >>= 6;
-	s[3] = (v & 0x3f) | 0x80;
-	v >>= 6;
-	s[2] = (v & 0x3f) | 0x80;
-	v >>= 6;
-	s[1] = (v & 0x3f) | 0x80;
-	v >>= 6;
-	s[0] = v | 0xfc;
-	return 6;
+	s[1] = (wc & 0x3f) | 0x80;
+	wc >>= 6;
+	s[0] = wc | (uint8_t)(0x3f00 >> n);
+	return n;
 }
 
 size_t FAST_FUNC wcrtomb(char *s, wchar_t wc, mbstate_t *ps UNUSED_PARAM)
@@ -164,7 +147,9 @@ size_t FAST_FUNC mbstowcs(wchar_t *dest, const char *src, size_t n)
 	if (unicode_is_enabled != 2) {
 		while (n) {
 			unsigned char c = *src++;
-			*dest++ = c;
+
+			if (dest)
+				*dest++ = c;
 			if (c == 0)
 				break;
 			n--;
@@ -177,7 +162,8 @@ size_t FAST_FUNC mbstowcs(wchar_t *dest, const char *src, size_t n)
 		unsigned c = (unsigned char) *src++;
 
 		if (c <= 0x7f) {
-			*dest++ = c;
+			if (dest)
+				*dest++ = c;
 			if (c == '\0')
 				break;
 			n--;
@@ -216,7 +202,8 @@ size_t FAST_FUNC mbstowcs(wchar_t *dest, const char *src, size_t n)
 			//or maybe: c = 0xfffd; /* replacement character */
 		}
 
-		*dest++ = c;
+		if (dest)
+			*dest++ = c;
 		n--;
 	}