Fix for Bug #6298

[csnover]

This bug has about over 9000 dupes in the bug tracker (at least 3), so here is a patch. Passes QUnit and the test case provided in duplicate #6586. I hate the JSLint assignment fix, but am not sure how to keep it from complaining about the fact that, yes, we really do want to call new XMLHttpRequest() for its side-effects.

[jeresig]

Oh - I landed a fix in commit 873c284 while I was on the plane earlier - also closed #6298 at the same time. Thanks for the patch, though!

I notice that you did things a little bit differently in your implementation - I don't think we have to be quite so aggressive on making sure that we fall back to a solution (or throwing an error) - just having the one extra try/catch (as I did it) is probably fine.

[csnover]

Indeed, yes. My aims were to 1. avoid the overhead of an extra try/catch on every request, 2. keep XHR initialization all in one place, 3. improve minification slightly by using references to window.ActiveXObject and window.XMLHttpRequest, and 4. avoid any silent failures.

I agree that silent failure is not such a big deal right now since every browser that jQuery supports also supports XHR, but it might be more of an issue when thinking ahead to mobile development and the inevitable “why is ajax not working on my [whatever]” (you would be the expert on that though, I don’t know what the XHR looks like on mobiles :)).