Garbled characters happen when EAP 7/Undertow's FormEncodedDataDefinition parses raw multibyte characters in POST request data. However, the garbled character issue does not happen on EAP 6 (JBossWeb) and Tomcat.

example application:

...(snip)...
@WebServlet(name = "TestServlet", urlPatterns = {"/test"})
public class TestServlet extends HttpServlet {
    ...(snip)...

    protected void doPost(HttpServletRequest request, HttpServletResponse response)
            throws ServletException, IOException {
        request.setCharacterEncoding("UTF-8");
        response.setCharacterEncoding("UTF-8");
        response.setContentType("text/html;charset=UTF-8");
        try (PrintWriter out = response.getWriter()) {
           out.println("test = " + request.getParameter("test"));
        }
    }

result:

• EAP 7 (Undertow)

  $ curl http://localhost:8080/example/test -d "test=テスト"
  test = ￣ﾃﾆ￣ﾂﾹ￣ﾃﾈ 
  

• EAP 6 (JBossWeb) and JWS 5 Tomcat 9

  $ curl http://localhost:8080/example/test -d "test=テスト"
  test = テスト
  

—

IMO, as per HTTP specification, the client should use percent-encode (url-encode) when sending POST data as "Content-Type: application/x-www-form-urlencoded". Basically, web browsers (HTTP compliant clients) correctly uses percent-encode (url-encode) when sending POST data. So, I think the root cause of this issue is the client's wrong behavior. In fact, this garbled character issue does not happen when using the following curl command:

$ curl localhost:8080/test/ -d "test=%E3%83%86%E3%82%B9%E3%83%88"

However, I think we can improve Undertow's FormEncodedDataDefinition slightly to become more compatible with EAP 6/JBossWeb and not to break raw multibyte characters in some cases.

Here's an analysis of implementation differences between EAP 6 and EAP 7:

• EAP 6/JBossWeb (org.apache.tomcat.util.http.Parameters#processParameters) parses parameters as byte, then it finally converts byte array of key/value to String through ByteChunk.

• EAP 7/Undertow (io.undertow.server.handlers.form.FormEncodedDataDefinition#doParse) also parses parameters as byte. However, it directly converts each one byte to one char by using "StringBuilder#append((char) n)" where n is one byte. As multibyte characters (like Japanese characters) can not be correctly converted to one character from one byte, this processing results in a garbled character.

So, I think we can improve FormEncodedDataDefinition by using ByteArrayOutputStream instead of StringBuilder: https://github.com/undertow-io/undertow/compare/master...msfm:master_UNDERTOW-1802

Of course, if the parameter value contains both % and raw multibyte characters, the parameter still can not be parsed correctly because the value can not be decoded as a percent-encoded value correctly. However, I think this is not an issue because it's the same as EAP 6/JBossWeb.