User-Agent: Mozilla/5.0 (X11; U; Linux i686; de; rv:1.8.1.1) Gecko/20061217 SUSE/2.0.0.1-25.1 Firefox/2.0.0.1 Build Identifier: Mozilla/5.0 (X11; U; Linux i686; de; rv:1.8.1.1) Gecko/20061217 SUSE/2.0.0.1-25.1 Firefox/2.0.0.1 Firefox-qs is a program that keeps several instances of firefox running invisible in the background so that the user can open a new window quickly. Reproducible: Always Steps to Reproduce: 1.install firefox-qs (guru,...) 2.make it start on boot 3.reboot Actual Results: On every bootup the user is asked whether he wants to restore some previous firefox session. Expected Results: 1.) Firefox-QS should be able to start firefox in plain mode without any content and without prompting the user for anything 2.) Termination of Firefox instances still running in plain mode (no site opened) should not trigger session restoration It would be possible to do totally without FirefoxQS if the main /usr/bin/firefox executable was a dummy program just checking if any instance of Firefox is already running. If yes it should make the running instance fork: Just as if the user selecet File->NewWindow If not it should start firefox-bin.

You're hitting one of the Linux versions of bug 342885 (see e.g. bug 342885 comment #26) to one of which this probably should be DUPEd. This particular request should be fixed by bug 354894 though, so leaving as an enhancement request which can be WFMd (or whatever) once bug 354894 has been fixed.

I don't really know what this enhancement is asking for, please re-file with something that is concrete and well-specified.

Although it is already aeons ago that I have used firefox-qs I woold appreciate a general support for preloading utilities(like the warp-wmiface preloader). Windows not changed on from startup (about:blank/home) must not take part in session restoration*. Besides this firefox should start immediately by a fork of a running process if another firefox instance is already started. It should be possible to make a running firefox instance fork a new window via Firefoxes dbus interface (instead of having to select Document->New Window). I think those issues are already not too bad described in comment #0. (* all new windows not yet used for browsing should be in plain mode at first)

to put it short: Perhaps we do not need an explicit plain mode to start new instances in; just put all windows not yet used for browsing in plain mode and exclude these windows from session restoration. Expose a function like File->New Window via DBUS. That`s it.

Preloaders will work fine once bug 354894 has been fixed. There's really no need for any other change in behavior (as in: preloaders can keep a hidden non-browser window open and Session Restore will continue to work as expected).

I do not really see what the resolution of bug 354894 should do for preloaders. It says that Firefox should invoke session restoration even if just the download window has stayed open and Firefox is still running. The problem with Firefox-qs was that even instances held invisibly in the background were restored by session restoration. These windows must be excluded from session restoration ('plain windows'). Otherwise Firefox will end up restoring thousands of empty windows.

(In reply to comment #6) > The problem with Firefox-qs was that even instances held invisibly in the > background were restored by session restoration. That bit of the issue is actually firefox-qs's problem: it shouldn't keep a browser window open (and force us to introduce the notion of a "plain", unused browser window) but rather a non-browser window (which Session Restore doesn't track in the first place). Then the desired behavior would already be achievable, if it weren't for bug 354894. BTW: You get a non-browser window e.g. with: firefox -chrome about:blank As for the request to make it possible to keep Firefox running without any windows open, that would have to be filed as a new Toolkit:Startup bug (if it hasn't been filed already).

Thanks; so it becomes possible to prelaunch firefox without creating a window that takes place in session restoration. Nevertheless not restoring empty windows would be a cleaner solution, because then preloaders like warp would not need to take care of firefox as a special application. Warp just launches an app; hides its window and unhides it as soon as the user needs it. Very simple and straight forward. Besides this I do not see what sense it should make to restore an empty window upon session restoration.