Firefox window moves to current workspace

Running Ubuntu Gutsy Tribe 5, GNOME 2.19.7, Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a8pre) Gecko/2007090304 Minefield/3.0a8pre

This has been happening to me every time I click a link from the gnome-terminal.

A bit of clarification:

Gutsy Tribe 5 is the 5th development (alpha) release of Ubuntu 7.10 Gutsy Gibbon.

New pages are set to open in a new tab, and I am not using MOZ_NO_REMOTE/-no-remote

This is a real pain. I find myself clicking links and quickly switching to the desktop with Firefox on it, hoping I can outrace the request so Firefox comes to the foreground on its proper desktop and I don't have to move it back into place (again). Requesting blocking status.

Sounds like something that could be a regression from bug 295447 (which introduced gtk_window_present calls).

Description of problem:

When I open a new tab (via right click on a terminal, clicking on something in
liferea, whatever), firefox immediately jumps onto the current workspace.

This is irritating.

Version-Release number of selected component (if applicable):

firefox-2.0.0.6-11.fc8

Additional info:

Seems related to the startup notification patch, as it worked ok in 2.0.0.6-8.fc8.

I will second this. Very irritating!

Owen says this is a metacity bug. The following program also hops windows.

#!/bin/env python

import gtk
import gobject
w = gtk.Window()
w.set_title("Foo")
w.show()

def on_timeout():
        w.present()
        return False

gobject.timeout_add(5000, on_timeout)
gtk.main()

er, hops workspaces

To give some background on this issue, Firefox now supports startup notification:
https://bugzilla.mozilla.org/show_bug.cgi?id=223492

The issue in this bug only occurs when there are multiple workspaces.

Removing this Firefox patch would be a regression in the single workspace case
because it would require clicking on a flashing browser entry in the task list
after activating an external link.

From IRC:

<owen> walters: metacity could look at the window role and alter the behavior
based on that

FWIW, commenting out the call to gdk_window_focus() in the patch makes it DTRT
for me.

Actually, what I meant was that based on the "type" of the window application
vs. dialog, you could do different things.

A application never, under any circumstances, should be moved between desktops.

Bill, what effect does that have in the single-workspace case?

Also, I've filed this bug upstream here:

http://bugzilla.gnome.org/show_bug.cgi?id=482354

Since there seems to be agreement this is an upstream Metacity issue, please
continue discussion there.

Colin: when it's on the same workspace, I get the throbbing entry in the panel.
When it's on the other workspace, I don't.

We need to fix this for F8...

This is also happening in compiz... so more than just metacity at work here.

Jeremy: not quite...

in metacity, the _window_ switches to the current workspace.

In compiz, the _active workspace_ switches.

This is about the metacity issue, which everyone agrees is clearly wrong. The
compiz behavior seems more or less sane. While different, i've already gotten
used to it.

The long discussion leading to the current metacity behaviour can be found
here: http://bugzilla.gnome.org/show_bug.cgi?id=166379

So, going back to the beginning. The original patch (for firefox) has the
following comments usptream from Elijah
(https://bugzilla.mozilla.org/show_bug.cgi?id=223492#c31):

************
>+ if (desktopStartupID.IsEmpty()) {
>+ PRUint32 timestamp = GTKToolkit->GetFocusTimestamp();
>+ if (timestamp) {
>+ gdk_window_focus(aWindow->window, timestamp);
>+ GTKToolkit->SetFocusTimestamp(0);
>+ }
>+ return;
>+ }

This block of SetUserTimeAndStartupIDForActivatedWindow() is a little
confusing. It doesn't set the user time or the startup id for the activated
window. Instead, it sends an "activate me, please" message to the window
manager. I understand it's just a workaround when no startup id is available
(because either firefox wasn't launched with startup-notification or wasn't
built with MOZ_ENABLE_STARTUP_NOTIFICATION), but it'd be nice to have it
documented as such. It'd also be good to verify in testing that this block of
code isn't being executed (i.e. that startup id is non-empty), since it is just
a workaround.
*****************

I find it odd that the supposed 'workaround' part of the code is what's causing
the problem in my testing.

Created attachment 225081
partial fix for firefox

Note that due to missing changes to the configury, we aren't even *building*
the SN support at the moment - this patch fixes that.

Doesn't change the behavior, though.

"This is not a blocker"

<notting> halfline: why not just take the patch out of ff, since we're not even
applying it in a way that works
<juhp> notting: exactly
<halfline> notting: that's not a horrible idea
 notting: might be worth talking to stransky about
 or caillon if he pops on

*** Bug 337721 has been marked as a duplicate of this bug. ***

Note there is now a patch for metacity upstream at:
http://bugzilla.gnome.org/show_bug.cgi?id=482354#c2

Also in response to:

<notting> halfline: why not just take the patch out of ff, since we're not even
applying it in a way that works

The patch does function as expected in the single workspace case, even if it may
be entering into a fallback section of the code (will investigate).

I have built a new version of metacity with the patch:
http://koji.fedoraproject.org/koji/buildinfo?buildID=21629

Please test, if it works well then we can reable to re-enable the Firefox fix.

(In reply to comment #20)
> Also in response to:
>
> <notting> halfline: why not just take the patch out of ff, since we're not even
> applying it in a way that works
>
> The patch does function as expected in the single workspace case, even if it may
> be entering into a fallback section of the code (will investigate).

The patch adds configury for using libstartupnotification, and ifdef'd code to
use libsn. We then build moz without sn support. If it's functioning, it's
completely by accident.

(In reply to comment #21)
> I have built a new version of metacity with the patch:
> http://koji.fedoraproject.org/koji/buildinfo?buildID=21629
>
> Please test, if it works well then we can reable to re-enable the Firefox fix.

Thanks, Colin! Yes, that works well for me so far. :)

(In reply to comment #22)
>
> The patch adds configury for using libstartupnotification, and ifdef'd code to
> use libsn. We then build moz without sn support. If it's functioning, it's
> completely by accident.

Ok, I'm at home so I can't actually test this with my rawhide computer, but from
looking at the patch again, the section you identified earlier (which calls
gdk_window_focus which ultimately causes the focus event) is the source of the
behavior under discussion.

The startup notification specification if supported correctly would make Firefox
display a progress status on first launch, but this (to me) is not the most
useful part of the patch. The most useful part is the behavior that when one
clicks on a link in an external program, the web browser appears and displays
the link.

Some people feel that behavior wasn't right for multiple desktops (and the
metacity patch addresses this), but please when testing things, try the single
workspace case. The expected behavior is that the browser appears; it should
not pulse as in comment #9. This is the way clicking on links works on other
platforms, and is the most sensible thing I believe for most situations in the
single-workspace case.

There might be some more special cases such as feed readers (liferea?) where one
wants to click a number of links to "queue" urls; I am open to discussion around
those. Probably the right solution for them will be to have the ability at
link-click time whether or not you want to see it.

I meant to mention above; the reason the patch still has effect is because the
gdk_window_focus call is outside the #ifdef MOZ_ENABLE_STARTUP_NOTIFICATION.

Also, bear in mind this patch is already upstreamed and will be in the Firefox 3
release; I am not yet sure whether it will be in the 2.0.0.7 stream.

We'll probably want this metacity fix for FC7 too before Firefox's support makes
it there.

Fixed in F8, moving off blocker

Its fixed for me after last update in F8
Cheers

http://bugzilla.gnome.org/show_bug.cgi?id=482354 thanks to Mart Raudsepp for the link

 1. new-window : makes sense to open on current desktop
 2. new-tab: moving all tabs to current desktop is annoying; but opening new tab on other desktop is probably too confusing.

Thus: if new-tab is selected as open mode, use new-window if there is no firefox window on current desktop. optionally, we could add a pref to disable this auto-new-window feature for power users that want the ffox 2 behaviour.

(fwiw gedit behaves in a similar fashion)

*** Bug 414879 has been marked as a duplicate of this bug. ***

*** Bug 420186 has been marked as a duplicate of this bug. ***

(In reply to comment #6)
> 1. new-window : makes sense to open on current desktop
> 2. new-tab: moving all tabs to current desktop is annoying; but opening new
> tab on other desktop is probably too confusing.

What about opening new tab on other desktop and flashing the window title button?

> Thus: if new-tab is selected as open mode, use new-window if there is no
> firefox window on current desktop. optionally, we could add a pref to disable
> this auto-new-window feature for power users that want the ffox 2 behaviour.

I would certainly make use of that pref. :)

This isn't a bug in Firefox, it is a bug in metacity which Colin Walters at Red Hat reported and supplied a patch for.

http://bugzilla.gnome.org/show_bug.cgi?id=482354

If you see this bug on Linux, you're probably not running Fedora Rawhide/9. But, I recommend you forward this to your distro and ask that they temporarily take the metacity patch until the upstream metacity developers decide this bug is important enough to warrant their attention.

Alternatively, switching window managers to something other than metacity is a workaround.

> Alternatively, switching window managers to something other than metacity is a
workaround.

I could say "lie". It's simply not true :P . I had this issue with Compiz running.

> If you see this bug on Linux, you're probably not running Fedora Rawhide/9.

Fedora is innovative ^^ .

Compiz in Fedora 8 certainly doesn't have that problem. If compiz regressed, that's unfortunate, and I'd suggest filing a bug against compiz.

Until I set an option in about:config I saw this bug. Now everything from other apps is opened in background tabs. That is great and doesn't require fixing anything.

(In reply to comment #14)
> Until I set an option in about:config I saw this bug. Now everything from other
> apps is opened in background tabs. That is great and doesn't require fixing
> anything.

Which option did you set, and what did you set it to?

browser.tabs.loadDivertedInBackground = true

I just tried every compiz version from F8 onward, and metacity from F8 onward. No compiz version features this bug. Metacity didn't get the fix until 2.21.5 and had it except for a day when it got accidentally removed around 2.21.8. It's been fine ever since.

(In reply to comment #16)
> browser.tabs.loadDivertedInBackground = true

Hmm, indeed, that prevents the window from moving, although it doesn't duplicate the behavior that has regressed, since it doesn't throb the browser's taskbar button nor open the link in a foreground tab.

*** Bug 406167 has been marked as a duplicate of this bug. ***

*** Bug 429901 has been marked as a duplicate of this bug. ***

This message is a reminder that Fedora 7 is nearing the end of life. Approximately 30 (thirty) days from now Fedora will stop maintaining and issuing updates for Fedora 7. It is Fedora's policy to close all bug reports from releases that are no longer maintained. At that time this bug will be closed as WONTFIX if it remains open with a Fedora 'version' of '7'.

Package Maintainer: If you wish for this bug to remain open because you plan to fix it in a currently maintained version, simply change the 'version' to a later Fedora version prior to Fedora 7's end of life.

Bug Reporter: Thank you for reporting this issue and we are sorry that we may not be able to fix it before Fedora 7 is end of life. If you would still like to see this bug fixed and are able to reproduce it against a later version of Fedora please change the 'version' of this bug. If you are unable to change the version, please add a comment here and someone will do it for you.

Although we aim to fix as many bugs as possible during every release's lifetime, sometimes those efforts are overtaken by events. Often a more recent Fedora release includes newer upstream software that fixes bugs or makes them obsolete. If possible, it is recommended that you try the newest available Fedora distribution to see if your bug still exists.

Please read the Release Notes for the newest Fedora distribution to make sure it will meet your needs:
http://docs.fedoraproject.org/release-notes/

The process we are following is described here: http://fedoraproject.org/wiki/BugZappers/HouseKeeping

Thank you for the bug report. Closing bug as per comment #27 and comment #28.

If you still experience this problem after updating to our latest Fedora
release, reopen this bug against that version if this bug exists there.

Hi, I have the same problem with stumpwm, although Firefox's window doesn't get moved to the current Group, I think that's stumpwm's term for a workspace, but I get switched to the group where firefox is.

(In reply to comment #6)
> 1. new-window : makes sense to open on current desktop
> 2. new-tab: moving all tabs to current desktop is annoying; but opening new
> tab on other desktop is probably too confusing.
>
> Thus: if new-tab is selected as open mode, use new-window if there is no
> firefox window on current desktop. optionally, we could add a pref to disable
> this auto-new-window feature for power users that want the ffox 2 behaviour.
>
> (fwiw gedit behaves in a similar fashion)

Maybe there need more options in multi-workspace environments?
E.g.:
1. new window unconditionally
2. new window if none on an active workspace, new tab otherwise [hidden in Windows]
3. new tab unconditionally

and also a checkbox to focus the window when link has been opened (having it checked would probably move the user to another workspace in case 3). Probably, the same preference mentioned in Comment #0 could be reused here.

Created an attachment (id=343829)
Shell script to open a URL strictly in a new firefox tab on the current workspace

As a substitute for a fix of this bug, I've worked around it with a shell script that uses wmctrl to get the ideal behavior. It'll check to see if there's an existing firefox window open on the current workspace, and if so, (optionally) focus it, and load a url in a new tab there. If there's no such window, it'll simply open it in a new window on the current workspace.

First, how does this bug report relate to Mozilla bug #263553? Isn't it a duplicate?

I just posted the following to http://bugs.debian.org/486570, but I found no way to update this bug report.

After reading what James wrote[0], I went to investigate this a bit.

0. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=486570#15

First, I found the single call to gtk_window_present in the Firefox
source and disabled it:

  --- a/widget/src/gtk2/nsWindow.cpp
  +++ b/widget/src/gtk2/nsWindow.cpp
  @@ -1406,7 +1406,7 @@ nsWindow::SetFocus(PRBool aRaise)
          owningWindow->mContainerBlockFocus = PR_TRUE;

          // Set focus to the window
  - if (gRaiseWindows && aRaise && toplevelWidget &&
  + if (0 && gRaiseWindows && aRaise && toplevelWidget &&
              !GTK_WIDGET_HAS_FOCUS(toplevelWidget) &&
              owningWindow->mIsShown && GTK_IS_WINDOW(owningWindow->mShell))
            gtk_window_present(GTK_WINDOW(owningWindow->mShell));

and built new packages, but the problem remained. However, I found
a couple of curiousities. I have three heads one two graphics cards,
and I use Zaphod mode to put them all next to each other. head 0 is
on GPU 0, and heads 1&2 are on GPU 1.

  1. the focus is never stolen when the Firefox window is on heads
     1&2, not even if the link is clicked on another head.

  2. if I move the window from head 0 to head 1 and back, then it no
     longer steals focus, not even if the click is on a separate head.

  3. if I move the window from head 0 to head 1, click a link, and
     move it back to head 0, then the focus stealing happens again.

I then went back to investigate these three points with the original
Firefox packages, i.e. without the above patch, and the behaviour is
identical.

Then I investigated the urgency hint, using xprop -spy: I started it
in a terminal, then X-pasted a link into that terminal and clicked
it. I did this to avoid changing focus with the mouse while testing.
There was nothing in the output of xprop -spy mentioning an urgency
hint. I did verify that xprop -spy shows the urgency hint, using
URxvt.urgentOnBell and echo -e '\a'.

This is starting to feel a bit like a riddle. For me, (2.) above is
an okay workaround at the moment, but it really seems like the
Firefox code needs some cleaning.

(In reply to Jeremy Nickurak from comment #23)
> Created attachment 343829
> Shell script to open a URL strictly in a new firefox tab on the current
> workspace
>
> As a substitute for a fix of this bug, I've worked around it with a shell
> script that uses wmctrl to get the ideal behavior. It'll check to see if
> there's an existing firefox window open on the current workspace, and if so,
> (optionally) focus it, and load a url in a new tab there. If there's no such
> window, it'll simply open it in a new window on the current workspace.

Thanks for that script, it works very well.
This is the behavior I wish Firefox could have out of the box (Reuse-window-in-current-desktop otherwise create-new-in-current-desktop). Much better than jumping to another desktop or opening tabs hidden into another desktop. Any chance for this to happen?

The severity field for this bug is relatively low, S3. However, the bug has 4 duplicates.
:Amir, could you consider increasing the bug severity?

For more information, please visit [auto_nag documentation](https://wiki.mozilla.org/Release_Management/autonag#severity_underestimated.py).

The last needinfo from me was triggered in error by recent activity on the bug. I'm clearing the needinfo since this is a very old bug and I don't know if it's still relevant.