jQuery Forum

or better yet, why not just delegate to 'console' if globally available:
QUnit.log = (console&&console.log)?
    function(result, msg){console.log('['+!!result+'] '+msg);}:
    function(result, msg){};
then we still have a global hook to logging while keeping it inside the QUnit namespace internally.

I'm confused - why are you attaching the QUnit.done/log handlers inside window.onload? Ideally they should be the very first thing done (after loading the test suite itself).
<br clear="all">--John

hi john, your right that is where it should be done.  the problem arises from the fact that we load the scripts for the test from the html itself, as part of the browser emulation process.  this means that we can only affect the global scope before the html is loaded, onload, and after load.  otherwise we would need to copy the script loading process outside of the native browser loading process and either duplicate the process (at best) and usually only approach something similar.
for example our old script for kick off the tests was:
// Init
load("build/runtest/env.js");
var isLocal;
window.onload = function(){
    isLocal  = !!(window.location.protocol == 'file:');
   
    // Load the test runner
    load("dist/jquery.js",
        "build/runtest/testrunner.js");
   
    // Load the tests
    load(
        "test/unit/core.js",
        "test/unit/selector.js",
        "test/unit/event.js",
        "test/unit/fx.js",
        "test/unit/dimensions.js",
        "test/unit/data.js",
       
        // offset relies on window.open, which is currently unimplemented in env.js
        //"test/unit/offset.js",
       
        // these tests require hitting a server, so we will need some clever env.js
        // way of testing them
        "test/unit/ajax.js"
    );
   
    // Display the results
    results();
};
and now its just:
(function($env){
   
    //let it load the script from the html
    $env.scriptTypes = {
        "text/javascript"   :true
    };
   
    var count = 0;
    window.console = {
        log: function(result, message){
            $env.log('(' + (count++) + ')[' + ((!!result) ? 'PASS' : 'FAIL') + '] ' + message);
        }
    };
   
    window.onload = function(){
        $env.warn('Defining QUnit.done');
        QUnit.done = function(pass, fail){
            $env.warn('Writing Results to File');
            jQuery('script').each(function(){
                this.type = 'text/envjs';
            });
            $env.writeToFile(
                document.documentElement.xml,
                $env.location('jqenv.html')
            );
        };
       
    }
   
    window.location = "test/index.html";
   
})(Envjs);
where i did modify the qunit to define QUnit.log as in the last email.
does this make sense?  the latter approach lets the html essentially define everything we need to know and just load and go, setting up our hooks prior to setting window.location
thatcher

cut and paste error after the old example, should also include the line
window.location = "test/index.html";

i think there is a reasonable possibility i am mistaken or misunderstanding what i'm intending to do.  let me try to create a canonical example, and assuming i can't it's on my plate.

PS after writing this i realize that it's possible to still load the script from the html directly as part of the loading process, while also providing a hook in envjs to register a callback to execute before/after/around a particular script is loaded.  this means you could 'expect' qunit and after loading it apply a registered callback to modify it in some way so that the environment is harmonized.  this is the correct solution in my opinion.
<div id=":2nb" class="ii gt">
apologies for asking qunit to change before i fully explored envjs' ability to adapt.
thanks
thatcher
</div>