[OPEN-1331] Extjs.Ajax (3.2.2) Leak

[jay@moduscreate.com]

FWIW, the *leanest* code (does not recreate strings, objects, etc);

var success_handler = function(response, opts){
        makeRequest();
    },
    failure_handler =  function(){
        makeRequest();
    },
    ajaxCfg = {
       url: 'router.php',
       success: success_handler,
       failure: failure_handler,
       params:  { 
            foo: 'bar' 
        }
    };


function makeRequest(){
    Ext.Ajax.request(ajaxCfg);
}

makeRequest();

[jay@moduscreate.com]

He only sets value - I would expect that that cannot/shouldn't cause the leak.

DOM manipulation is one of the leading causes of memory leaks, IMHO. each browser acts differently.

[meroy]

This has been entered into Trac. Thanks!

[wsi]

Thank you, what is Trac?

[Condor]

Trac is the issue tracker used internally by the Sencha team.

[Reimius]

Sorry to respond to such an old thread, but this leak is still apparent in Ext 4.0.7 even. I have tested it on IE9 and it seems that the test case will leak more if you return a larger dataset from the server. The odd thing is that the current version of Google Chrome does not leak. With my test case I have also noticed that the memory does hit a ceiling and then stop (but I'm not sure why). I have been using Drip to test the explorer leak after reports of our actually application crashing browsers.memory_leak.png
The only difference between my test cases with 4.0.7 and 3.2.1 is that 4.0.7 takes about 3x longer to get to the same ceiling. I can provide my test cases if that will help.

And just so you know, the ceiling it hits is not reasonable for our actual user base, it's about 1.7 Gb

[Reimius]

Hello again,

According to my research and testing, the issue is with this code in the Ext Base in 3.2.1:

function handleReadyState(o, callback){    callback = callback || {};
        var conn = o.conn,
            tId = o.tId,
            poll = pub.poll,
            cbTimeout = callback.timeout || null;


        if (cbTimeout) {
            pub.conn[tId] = conn;
            pub.timeout[tId] = setTimeout(function() {
                pub.abort(o, callback, true);
            }, cbTimeout);
        }


        poll[tId] = setInterval(
            function() {
                if (conn && conn.readyState == 4) {
                    clearInterval(poll[tId]);
                    poll[tId] = null;


                    if (cbTimeout) {
                        clearTimeout(pub.timeout[tId]);
                        pub.timeout[tId] = null;
                    }


                    handleTransactionResponse(o, callback);
                }
            },
            pub.pollInterval);
    }

the variable conn should not be referenced within the onreadystatechange, and 'this' should be used instead. There are also other variables being used from the outer scope and I'm not sure how to handle those, but according to the section: XmlHttpRequest memory management and leaks on this page: "http://javascript.info/tutorial/memo...js-memory-leak" this code will cause a memory leak. And I have tested from within my app to verify that it happens.

Since 4.0.7 also leaks in the same manner, I assume this code has been carried over, or is still similar. I'm also surprised that more people haven't had this problem.

[Reimius]

Actually, after more review of the code, I'm not sure if my last post makes sense, but I'm going to leave it there because I believe the issue is still around that code area.

[Reimius]

I saw this thread again today... fyi, the source of the problem (or one of the sources, or part of the problem) is the way Extjs decodes json (which is used in the ajax call). It uses the eval function which is known to cause memory leaks when used. To work around this I used jQuery's ajax call (which is the only framework I could find that didn't leak in all browsers) and json_parse from https://github.com/douglascrockford/.../json_parse.js which does not use eval to avoid memory leaks. Just posting this to help others in the same situation as I, and maybe to help the Ext with their own solution to this problem.

[dracofyre]

Any more information on this topic?