Looks like MozXMLHttpRequestParameters to set {mozAnon: true} is only usable from a window context. http://mxr.mozilla.org/mozilla-central/source/dom/webidl/XMLHttpRequest.webidl#35 There isn't a good way to set mozAnon from a javascript module because there's no constructor and the mozAnon attribute is readonly. http://mxr.mozilla.org/mozilla-central/source/dom/base/nsIXMLHttpRequest.idl#322

Cu.importGlobalProperties("XMLHttpRequest") should work, I would think, and then you can just use the constructor. I wonder whether we should just expose that constructor to system scopes unconditionally...

Er, I meant Cu.importGlobalProperties(["XMLHttpRequest"])

Just making sure if I'm trying this correctly: Cu.importGlobalProperties(["XMLHttpRequest"]); let x = new XMLHttpRequest({mozAnon: true, mozSystem: false}); dump([x.mozAnon, x.mozSystem]); // "false,true" let x = new XMLHttpRequest({mozAnon: true}); dump([x.mozAnon, x.mozSystem]); // "false,true" let x = new XMLHttpRequest(); dump([x.mozAnon, x.mozSystem]); // "false,true"

Oh. That's dumb. The issue is that nsXMLHttpRequest::InitParameters tries to do stuff with windows. Fix coming up.

Oh, and mozSystem will continue to test true, since the XHR has a system principal, of course.

Comment on attachment 8604701 [details] [diff] [review] part 1. Give DOMEventTargetHelper a sane accessor for getting its nsIGlobalObject Shouldn't nsWindowRoot::GetOwnerGlobal() return the current inner window. nsIGlobalObject* GetParentObject() const { { goes to the next line.

> Shouldn't nsWindowRoot::GetOwnerGlobal() return the current inner window. Yes! Good catch. > { goes to the next line. I followed the existing style of GetParentObject. I guess I can change both.

https://hg.mozilla.org/integration/mozilla-inbound/rev/2db3491d59e4 https://hg.mozilla.org/integration/mozilla-inbound/rev/d9429069ec0e

This seems to have caused mochitest-1 failure: https://treeherder.mozilla.org/logviewer.html#?job_id=9790367&repo=mozilla-inbound 13:55:32 INFO - 1011 INFO TEST-START | dom/base/test/test_bug927196.html 13:55:32 INFO - 1012 INFO TEST-PASS | dom/base/test/test_bug927196.html | XMLHttpRequest should be mozAnon 13:55:32 INFO - 1013 INFO TEST-PASS | dom/base/test/test_bug927196.html | XMLHttpRequest should not be mozSystem 13:55:32 INFO - 1014 INFO TEST-UNEXPECTED-FAIL | dom/base/test/test_bug927196.html | XMLHttpRequest should be mozAnon - got true, expected false 13:55:32 INFO - 1015 INFO TEST-PASS | dom/base/test/test_bug927196.html | XMLHttpRequest should not be mozSystem 13:55:32 INFO - 1016 INFO TEST-PASS | dom/base/test/test_bug927196.html | XMLHttpRequest should be mozAnon 13:55:32 INFO - 1017 INFO TEST-PASS | dom/base/test/test_bug927196.html | XMLHttpRequest should be mozAnon 13:55:32 INFO - 1018 INFO TEST-PASS | dom/base/test/test_bug927196.html | XMLHttpRequest should not be mozSystem 13:55:32 INFO - 1019 INFO TEST-PASS | dom/base/test/test_bug927196.html | XMLHttpRequest should be mozAnon 13:55:32 INFO - 1020 INFO TEST-PASS | dom/base/test/test_bug927196.html | XMLHttpRequest should be mozSystem 13:55:32 INFO - 1021 INFO TEST-PASS | dom/base/test/test_bug927196.html | XMLHttpRequest should be mozAnon http://mxr.mozilla.org/mozilla-central/source/dom/base/test/test_bug927196.html?force=1#47 46 req = new XMLHttpRequest({mozAnon: true, mozSystem: true}); 47 is(req.mozAnon, false, "XMLHttpRequest should be mozAnon"); Not sure why "should be mozAnon" is checking false..

> Not sure why "should be mozAnon" is checking false.. Because the existing code was buggy and the test was written around the bug without even noticing it. The patches here fix that bug. Anyway, test fix pushed.

https://hg.mozilla.org/mozilla-central/rev/2db3491d59e4 https://hg.mozilla.org/mozilla-central/rev/d9429069ec0e https://hg.mozilla.org/mozilla-central/rev/6ffbd3150f6c