The code here: http://mxr.mozilla.org/comm-central/source/mailnews/base/prefs/content/accountcreation/verifyConfig.js#116 Is incorrect. There is an errorCallback with (e), but we're not in a try catch at this point. So if we get to the end of this function, there will be an error about e not being defined. Checked in with bug 1155491

112 catch (e) { 113 gEmailWizardLogger.error("ERROR: verify logon shouldn't have failed"); 114 } 115 // Avoid pref pollution, clear out server prefs. 116 MailServices.accounts.removeIncomingServer(inServer, true); 117 errorCallback(e); 118 } From the code above this, it looks like these two lines should also be inside the catch block?

> From the code above this, it looks like these two lines should also be inside the catch block? That's my thought. I'll put together a patch and have rkent review.

It would be great if you could test and explain why the current code works, as I am not aware of any reported symptoms of this problem.

(In reply to Kent James (:rkent) from comment #3) > It would be great if you could test and explain why the current code works, > as I am not aware of any reported symptoms of this problem. It won't cause a problem in todays code unless something fails. I was porting this code to another mail client and I had a bug in my code which exposed this code. Nothing will break today because this is the tail end of this function. You'll just see a message on the console and wonder why it is there. It's just a correctness issue.

Yes currently the only time you'd get there is when e is defined.

This isn't worth fixing. I have no idea why I saw the problem in Postbox code.