XmlHttpRequest.onload not called

问题:

I'm playing around with this XmlHttpRequest thing. In some tutorials and books, it is the onload function the one that is called when the request is done. In my little experiment, this function is never called. Here's my code:

    window.onload = function() {
        var url = "http://www.google.com";
        var request = new XMLHttpRequest();


        request.onload = function() {
            var state = this.readyState;
            var responseCode = request.status;
            console.log("request.onload called. readyState: " + state + "; status: " + responseCode);

            if (state == this.DONE && responseCode == 200) {
                var responseData = this.responseText;
                alert("Success: " + responseData.length  + " chars received.");
            }
        };

        request.error = function(e) {
            console.log("request.error called. Error: " + e);
        };

        request.onreadystatechange = function(){
            console.log("request.onreadystatechange called. readyState: " + this.readyState);
        };

        request.open("GET", url);
        request.send(null);
    };

I'm testing this on the last Firefox release (just updated today). The log line in onload is never printed, and the breakpoint I set in the first line is never hit. However, the onreadystatechange function is called twice, and the http request is actually made. This is what firebug's console shows:

    request.onreadystatechange called. readyState: 1
    GET http://www.google.com/    302 Found    174ms
    request.onreadystatechange called. readyState: 4
    NS_ERROR_FAILURE: Failure
        request.send(null);

There's an error in the send line. I've tried changing it to request.send() with identical result.

At first I thought this might be the browser trying to prevent XSS, so I moved my html driver page to a Tomcat instance in my dev machine, but the result is the same.

Is this function guaranteed to be called? As I've said above, it's common to be seen in most tutorials, but on the other hand in the W3C spec page, the hello world snippet uses onreadystatechange instead:

    function processData(data) {
      // taking care of data
    }

    function handler() {
      if(this.readyState == this.DONE) {
        if(this.status == 200 &&
           this.responseXML != null &&
           this.responseXML.getElementById('test').textContent) {
          // success!
          processData(this.responseXML.getElementById('test').textContent);
          return;
        }
        // something went wrong
        processData(null);
      }
    }

    var client = new XMLHttpRequest();
    client.onreadystatechange = handler;
    client.open("GET", "unicorn.xml");
    client.send();

It checks readyState == this.DONE. DONE has the value 4, which is what I can see in my log. So if this were a XSS related issue, and the browser were preventing me to make the connection to a different domain, then why the actual connection is made and the DONE status is received???

PS: Yes, I know there are powerful libraries to do this easily, but I'm still a JavaScript noob so I'd like to understand the low level first.

---

UPDATE:
I've changed the URL to one inside my domain (localhost), and the error is gone but the onload function is still not being called. Tested in IE8 and does not work. Tested in Chrome and works. How's that?

---

UPDATE 2:
Tested again in Firefox, and now it works. Probably the old page was still cached so that's why I couldn't notice it immediatly. Still failing in IE8, I'll try to test it in a newer version.

回答1:

It looks like it was indeed a XSS issue and Firefox was blocking the onload call. I can't still understand why the http network request was actually being done and the onreadystatechange was being called with the DONE readyState.

I changed the URL to another one in the same domain, and now it works in Firefox (after some cache-related false attempts) and in Chrome. It still does not work in IE8, despite the official docs say it is supported. I've found this SO answer which states otherwise. It looks like the onload function is a more modern convenience method and the old way of checking the result is using onreadystatechange instead.

I guess I'll accept this answer as the solution unless a more detailed answer is provided.

回答2:

IE has different method to create xmlhttprequest.

function createCORSRequest(method, url) {
  var xhr = new XMLHttpRequest();
  if ("withCredentials" in xhr) {

    // Check if the XMLHttpRequest object has a "withCredentials" property.
    // "withCredentials" only exists on XMLHTTPRequest2 objects.
    xhr.open(method, url, true);

  } else if (typeof XDomainRequest != "undefined") {

    // Otherwise, check if XDomainRequest.
    // XDomainRequest only exists in IE, and is IE's way of making CORS requests.
    xhr = new XDomainRequest();
    xhr.open(method, url);

  } else {

    // Otherwise, CORS is not supported by the browser.
    xhr = null;

  }
  return xhr;
};

same this article:https://www.html5rocks.com/en/tutorials/cors/

回答3:

The onload handler won't be called for yet another reason, I'm adding it here just so it can be helpful to someone else referencing this page.

If the HTTP response is malformed, the onload handler will not be called either. For example, a plaintext response of 10 bytes that advertises a length of 14 in Content-Length header will not invoke the onload handler. I wasted hours on client code before I start to replace back-end units with test stubs.