soup-headers: Update UTF-8 header param handling for RFC 5987

Update the handling of UTF-8 header parameters for the
changes/clarifications to RFC 2231 published in RFC 5987:

    * Decode iso-8859-1 params in addition to UTF-8 ones
    * An encoded UTF-8 param should override an unencoded param
      of the same name, regardless of which order they appear in

diff --git a/libsoup/soup-headers.c b/libsoup/soup-headers.c
index ecc3aec..33f6f6f 100644 (file)
--- a/libsoup/soup-headers.c
+++ b/libsoup/soup-headers.c
@@ -664,15 +664,21 @@ decode_quoted_string (char *quoted_string)
 }
 
 static gboolean
-decode_rfc2231 (char *encoded_string)
+decode_rfc5987 (char *encoded_string)
 {
        char *q, *decoded;
+       gboolean iso_8859_1 = FALSE;
 
        q = strchr (encoded_string, '\'');
        if (!q)
                return FALSE;
        if (g_ascii_strncasecmp (encoded_string, "UTF-8",
-                                q - encoded_string) != 0)
+                                q - encoded_string) == 0)
+               ;
+       else if (g_ascii_strncasecmp (encoded_string, "iso-8859-1",
+                                     q - encoded_string) == 0)
+               iso_8859_1 = TRUE;
+       else
                return FALSE;
 
        q = strchr (q + 1, '\'');
@@ -680,7 +686,23 @@ decode_rfc2231 (char *encoded_string)
                return FALSE;
 
        decoded = soup_uri_decode (q + 1);
-       /* strlen(decoded) <= strlen(q + 1) < strlen(encoded_string) */
+       if (iso_8859_1) {
+               char *utf8 =  g_convert_with_fallback (decoded, -1, "UTF-8",
+                                                      "iso-8859-1", "_",
+                                                      NULL, NULL, NULL);
+               g_free (decoded);
+               if (!utf8)
+                       return FALSE;
+               decoded = utf8;
+       }
+
+       /* If encoded_string was UTF-8, then each 3-character %-escape
+        * will be converted to a single byte, and so decoded is
+        * shorter than encoded_string. If encoded_string was
+        * iso-8859-1, then each 3-character %-escape will be
+        * converted into at most 2 bytes in UTF-8, and so it's still
+        * shorter.
+        */
        strcpy (encoded_string, decoded);
        g_free (decoded);
        return TRUE;
@@ -692,6 +714,7 @@ parse_param_list (const char *header, char delim)
        GHashTable *params;
        GSList *list, *iter;
        char *item, *eq, *name_end, *value;
+       gboolean override;
 
        list = parse_list (header, delim);
        if (!list)
@@ -703,6 +726,7 @@ parse_param_list (const char *header, char delim)
 
        for (iter = list; iter; iter = iter->next) {
                item = iter->data;
+               override = FALSE;
 
                eq = strchr (item, '=');
                if (eq) {
@@ -719,16 +743,18 @@ parse_param_list (const char *header, char delim)
 
                        if (name_end[-1] == '*' && name_end > item + 1) {
                                name_end[-1] = '\0';
-                               if (!decode_rfc2231 (value)) {
+                               if (!decode_rfc5987 (value)) {
                                        g_free (item);
                                        continue;
                                }
+                               override = TRUE;
                        } else if (*value == '"')
                                decode_quoted_string (value);
                } else
                        value = NULL;
 
-               g_hash_table_insert (params, item, value);
+               if (override || !g_hash_table_lookup (params, item))
+                       g_hash_table_replace (params, item, value);
        }
 
        g_slist_free (list);
@@ -745,7 +771,7 @@ parse_param_list (const char *header, char delim)
  * Tokens that don't have an associated value will still be added to
  * the resulting hash table, but with a %NULL value.
  * 
- * This also handles RFC2231 encoding (which in HTTP is mostly used
+ * This also handles RFC5987 encoding (which in HTTP is mostly used
  * for giving UTF8-encoded filenames in the Content-Disposition
  * header).
  *
@@ -771,7 +797,7 @@ soup_header_parse_param_list (const char *header)
  * Tokens that don't have an associated value will still be added to
  * the resulting hash table, but with a %NULL value.
  * 
- * This also handles RFC2231 encoding (which in HTTP is mostly used
+ * This also handles RFC5987 encoding (which in HTTP is mostly used
  * for giving UTF8-encoded filenames in the Content-Disposition
  * header).
  *
@@ -805,7 +831,7 @@ soup_header_free_param_list (GHashTable *param_list)
 }
 
 static void
-append_param_rfc2231 (GString    *string,
+append_param_rfc5987 (GString    *string,
                      const char *name,
                      const char *value)
 {
@@ -865,7 +891,7 @@ soup_header_g_string_append_param_quoted (GString    *string,
  * quotes or backslashes in @value.
  *
  * Alternatively, if @value is a non-ASCII UTF-8 string, it will be
- * appended using RFC2231 syntax. Although in theory this is supposed
+ * appended using RFC5987 syntax. Although in theory this is supposed
  * to work anywhere in HTTP that uses this style of parameter, in
  * reality, it can only be used portably with the Content-Disposition
  * "filename" parameter.