Update.
authorUlrich Drepper <drepper@redhat.com>
Tue, 12 May 1998 22:42:08 +0000 (22:42 +0000)
committerUlrich Drepper <drepper@redhat.com>
Tue, 12 May 1998 22:42:08 +0000 (22:42 +0000)
1998-05-12  Ulrich Drepper  <drepper@cygnus.com>

* wcsmbs/wcsmbsload.c (extract_charset_name): Pass correctly
formed charset names to getfct.  Let user override charset
selection by mention charset in locale name.

ChangeLog
wcsmbs/wcsmbsload.c

index ef8d9a1..9aa9479 100644 (file)
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,9 @@
+1998-05-12  Ulrich Drepper  <drepper@cygnus.com>
+
+       * wcsmbs/wcsmbsload.c (extract_charset_name): Pass correctly
+       formed charset names to getfct.  Let user override charset
+       selection by mention charset in locale name.
+
 1998-05-12 12:15  Ulrich Drepper  <drepper@cygnus.com>
 
        * posix/glob.c: Move to ...
index f4babc4..5373da9 100644 (file)
 
 #include <langinfo.h>
 #include <limits.h>
+#include <stdlib.h>
+#include <string.h>
 
+#include <locale/localeinfo.h>
 #include <wcsmbsload.h>
 #include <bits/libc-lock.h>
 #include <iconv/gconv_int.h>
@@ -85,6 +88,54 @@ getfct (const char *to, const char *from)
 }
 
 
+/* Extract from the given locale name the character set portion.  Since
+   only the XPG form of the name includes this information we don't have
+   to take care for the CEN form.  */
+#define extract_charset_name(str) \
+  ({                                                                         \
+    const char *cp = str;                                                    \
+    char *result = NULL;                                                     \
+                                                                             \
+    while (strchr ("@._+,", *cp) == NULL)                                    \
+      ++cp;                                                                  \
+    if (*cp == '.')                                                          \
+      {                                                                              \
+       const char *endp = cp;                                                \
+       while (*endp != '\0' && *endp != '@')                                 \
+         ++endp;                                                             \
+       if (endp != cp)                                                       \
+         result = strndupa (str, endp - cp);                                 \
+      }                                                                              \
+    result;                                                                  \
+  })
+
+
+/* The gconv functions expects the name to be complete, including the
+   trailing shashes if necessary.  */
+#define add_slashes(str) \
+  ({                                                                         \
+    const char *cp = str;                                                    \
+    char *result;                                                            \
+    char *tmp;                                                               \
+    size_t cnt = 0;                                                          \
+                                                                             \
+    while (*cp != '\0')                                                              \
+      if (*cp++ == '/')                                                              \
+       ++cnt;                                                                \
+                                                                             \
+    result = alloca (cp - str + 3);                                          \
+    tmp = __mempcpy (result, str, cp - str);                                 \
+    if (cnt < 2)                                                             \
+      {                                                                              \
+       *tmp++ = '/';                                                         \
+       if (cnt < 1)                                                          \
+         *tmp++ = '/';                                                       \
+      }                                                                              \
+    *tmp = '\0';                                                             \
+    result;                                                                  \
+  })
+
+
 /* Load conversion functions for the currently selected locale.  */
 void
 __wcsmbs_load_conv (const struct locale_data *new_category)
@@ -109,12 +160,22 @@ __wcsmbs_load_conv (const struct locale_data *new_category)
        {
          /* We must find the real functions.  */
          const char *charset_name;
-
-         /* Get name of charset of the locale.  */
-         charset_name = new_category->values[_NL_ITEM_INDEX(CODESET)].string;
-
-         __wcsmbs_gconv_fcts.tomb = getfct (charset_name, "INTERNAL");
-         __wcsmbs_gconv_fcts.towc = getfct ("INTERNAL", charset_name);
+         const char *complete_name;
+
+         /* Get name of charset of the locale.  We first examine
+            whether we have a character set mentioned in the locale
+            name.  If this isn't the case we use the information from
+            the locale files.  */
+         charset_name = extract_charset_name (setlocale (LC_CTYPE, NULL));
+         if (charset_name == NULL)
+           charset_name =
+             new_category->values[_NL_ITEM_INDEX(CODESET)].string;
+
+         /* Add the slashes necessary for a complete lookup.  */
+         complete_name = add_slashes (charset_name);
+
+         __wcsmbs_gconv_fcts.tomb = getfct (complete_name, "INTERNAL");
+         __wcsmbs_gconv_fcts.towc = getfct ("INTERNAL", complete_name);
 
          /* If any of the conversion functions is not available we don't
             use any since this would mean we cannot convert back and