800167 - Per window private browsing observer dnw after using nsIPrivateBrowsingService

Status: NEW

(Firefox :: Private Browsing, defect)

Description

[Erik Vold [:erikvold] (please needinfo? me)]

If you try the code here https://gist.github.com/3868985 in scratchpad for a chrome window, then you will not get the 'change' alert, but that would be expected, so I think this is a bug.

Comment 1

[Erik Vold [:erikvold] (please needinfo? me)]

Also, in Nightly, https://gist.github.com/3869202 does not produce an alert when setting a window's pb mode to false, which I would expect.  The gist does not produce any alerts on FF 15.

Comment 2

[Erik Vold [:erikvold] (please needinfo? me)]

(In reply to Erik Vold [:erikvold] from comment #0)
> If you try the code here https://gist.github.com/3868985 in scratchpad for a
> chrome window, then you will not get the 'change' alert, but that would be
> expected, so I think this is a bug.

This was for FF15 only

Comment 3

[(no longer active)]

nsIPrivacyTransitionObserver dates back to when we didn't have a clear idea on how we will put windows in private browsing mode.  Back in those days, it seemed like what we would need to do is to open a window, then some time after it has loaded, set the usePrivateBrowsing property to true and get it do the right thing.

However, that has changed, and as of a few hours ago that I landed bug 798508, the right way to open a new PB window is to call:

OpenBrowserWindow({private: true})

or alternatively:

window.openDialog("chrome://browser/content/", "name", "chrome,private,...");

This will put the window in PB mode right when it's created and will make it unnecessary to set the usePrivateBrowsing property (in fact, we should consider making that property read-only.)

Comment 4

[(no longer active)]

Now this seems like a bug nevertheless.  Josh, any chance you can take a look into this please?

Comment 5

[Erik Vold [:erikvold] (please needinfo? me)]

I guess we can close this as wontfix now?

Comment 6

[(no longer active)]

I still think this is a bug, but I'm curious to know what Josh thinks too.

Comment 7

[Josh Matthews [:jdm]]

I believe there are two symptoms of the same problem in these snippets: in the first entering private browsing blows away all existing tabs, so I expect that window.gBrowser.docShell refers to two different objects before and after entering PB mode, therefore the observer won't be triggered. In the second one, I doubt that exiting private browsing mode brings back the same dcoshell objects, so window.gBrowser.docShell once against refers to a different docshell than the one that contains the observer.