When parsing HTML attachment, ignore UTF-16 charset label because the function is not supposed to parse it,

also avoid the parse when the attachment already has a charset assigned,
bug 160729, r=ducarroz, sr=bienvenu.

mailnews/base/util/nsMsgI18N.cpp

@@ -532,6 +532,16 @@ nsMsgI18NParseMetaCharset(nsFileSpec* fileSpec)
       { 
         PL_strncpy(charset, token, sizeof(charset));
         charset[sizeof(charset)-1] = '\0';
+
+        // this function cannot parse a file if it is really
+        // encoded by one of the following charsets
+        // so we can say that the charset label must be incorrect for
+        // the .html if we actually see those charsets parsed
+        // and we should ignore them
+        if (!nsCRT::strncasecmp("UTF-16", charset, sizeof("UTF-16")-1) || 
+            !nsCRT::strncasecmp("UTF-32", charset, sizeof("UTF-32")-1))
+          charset[0] = '\0';
+
         break;
       } 
     } 

mailnews/compose/src/nsMsgAttachmentHandler.cpp

@@ -953,8 +953,11 @@ nsMsgAttachmentHandler::UrlExit(nsresult status, const PRUnichar* aMsg)
   // First things first, we are now going to see if this is an HTML
   // Doc and if it is, we need to see if we can determine the charset 
   // for this part by sniffing the HTML file.
+  // This is needed only when the charset is not set already.
+  // (e.g. a charset may be specified in HTTP header)
   //
-  if ( (m_type) &&  (*m_type) ) 
+  if ( (m_type) &&  (*m_type) &&
+       (!m_charset || !(*m_charset)) ) 
   {
     if (PL_strcasecmp(m_type, TEXT_HTML) == 0)
     {