gnome-session storing broken since intrepid

thank you for your bug report, those are known upstream issues

Sebatien,
Would you be kind enough to point us to the mailing list or WIKI that discusses these upstream issues?

Per the August 8th changelog entry within the gnome-session source indicates this was fixed.
Traffic on the desktop-devel-list is somewhat sparse after the initial proposal to revamp gnome-session to dbus instead of xsmp. Traffic to gnome-list concerning the problem seems to be 2.22 oriented.
Nothing is mentioned in the Gnome 2.24 showstopper review.

A quick review of the desktop files generated for gnome-terminal indicate that gnome-terminal does not currently have AutostartCondition nor X-GNOME-Autostart-*.

Reading through the proposal thread http://mail.gnome.org/archives/desktop-devel-list/2008-June/msg00210.html indicates that xsmp was completely dropped in favor of the new session protocol with no concern over legacy apps, including core gnome apps such as gnome-terminal
This is not "known upstream issues." This is "This will never work for legacy applications." A warning of some sort in /usr/share/doc/gnome-session directing annoyed users to the guilty parties would have been appreciated.

the upstream bug reference is listed several times on the launchpad bug page, in the bugs table and in the sidebar

Is this really "low" priority? I find it extremely irritating that none of my applications are restarted when I log back in. It's even more irritating that Gnome claims to be able to do this!

Also, is gnome-terminal really a "legacy" application?

For years, I've had applications automatically started upon login -- I would consider this bug a serious regression. (Imagine the press after Intrepid is released: "It's 2008 and Ubuntu can't ship a distro that remembers what applications to start when logging in!")

There seems to be no activity on the upstream bug; is there any work within the Ubuntu community to do something about this?

This isn't something that Ubuntu can really fix anyhow. Only Gnome's developers should be willing to spend the time to fix, not Ubuntu's.

On Fri, Oct 03, 2008 at 04:06:03AM -0000, Johne wrote:
> This isn't something that Ubuntu can really fix anyhow. Only Gnome's
> developers should be willing to spend the time to fix, not Ubuntu's.

Well, there's a fix, but at this point it would be a mess.
Revert to 2.22.x GNOME. Upgrade to 2.24.x GNOME when either
1) gnome-session is able to properly handle the legacy session protocol.
(Preferable, breaks the fewest apps).
2) The new protocol is adopted by KDE as well and a significant number of
apps support the new protocol.
(Acceptable, but can lead to a few annoyances).
3) A significant portion of gnome-core 2.24.x supports the new protocol.

Then there is politeness. Include in /usr/share/doc/gnome-session
a file indicating that GNOME upstream broke backwards compatibility
with the old session management protocol and why it was considered a good
idea for ubuntu to upgrade anyways.

Being able to have applications automatically start up when I log in
where I want them has been a standard part of my computing experience
for 18 years now. Being able to do that without having to drag out
'vi' has been part of my experience for 7 or 8 years now.

--
Chris Dukes
"In cynicism she's about 35" -- Terry Pratchett's "Hog Father"

pakraticus: As far as I know, GNOME *didn't* switch to a new session protocol. There are plans to eventually do that, but, for the time being, the session protocol continues to be XSMP. You may be confused by the fact that the GNOME 2.24 release notes mention DBus for session management, but, as far as I understand it, this is just an interface to control the session manager itself (start the shutdown sequence, for example).

First, correction: GNOME is indeed switching to a new D-Bus-based protocol for session management, but the new session manager *should have legacy support for XSMP*. The latest documentation seem to be http://live.gnome.org/SessionManagement/GnomeSession and XSMP is mentioned there prominently. Also on my machine, the session manager is indeed creating the SESSION_MANAGER environment variable needed by XSMP clients, and the variable points to a UNIX domain socket as expected.

Some further observation gives me the impression that, in intrepid, the session is not being saved at all, and that applications aren't being shut down cleanly. Try, for example, starting firefox, loading a page, logging out from the GNOME session, logging in again, and starting firefox again (it won't be restarted automatically, anyway.) Firefox will then report that the previous session was ended forcefully, and ask if its state should be restored. I tried with other applications, such as OpenOffice and Lyx (just open a new document, put some text in it, and log out from GNOME) and this leads to data loss. So I don't think this problem should be considered low priority, it isn't just anoying, it is rather dangerous.

Also, Sebastien, the linked upstream bug doesn't really seem to be related to this problem. As far as I understand it, it has to do with the new session manager not being able to restore old session data, saved by the older session manager.

Can people please check if they can repeat my results with Firefox and OpenOffice?

Yes, the firefox is just killed. Anyway, gnome-session-properties complains that saving session is not implemented yet :) And I don't see any session file under ~/.gnome2, ~/.gnome2_private, ~/.config

On Wed, Oct 08, 2008 at 06:02:10PM -0000, Lukas Hejtmanek wrote:
> Yes, the firefox is just killed. Anyway, gnome-session-properties
> complains that saving session is not implemented yet :) And I don't see
> any session file under ~/.gnome2, ~/.gnome2_private, ~/.config

the new session protocol has each app that supports it
(Which as far as I can tell are the apps in gnome-session and nothing else)
directs each application to store session information in gconf2.

Backwards compatibility still remains on the todo list.
>
> --
> gnome session does not start window manager nor restores the previous session
> https://bugs.launchpad.net/bugs/249373
> You received this bug notification because you are a direct subscriber
> of the bug.
>
> Status in The GNOME 2 Session Manager: New
> Status in “gnome-session” source package in Ubuntu: Triaged
>
> Bug description:
> Binary package hint: gnome-session
>
> The latest gnome-session in Ubuntu/intrepid does not restore saved session at all. It does not start window manager (I'm not using craps like compiz or metacity, I'm using openbox, but /usr/bin/gnome-wm correctly starts the openbox).
>
> I have gnome-session: 2.23.4.1-0ubuntu1
>
> I did gnome-session-save before quiting the session but nothing is restored. Moreover, I use save session on exit option.

--
Chris Dukes
"In cynicism she's about 35" -- Terry Pratchett's "Hog Father"

> the new session protocol has each app that supports it
> (Which as far as I can tell are the apps in gnome-session and nothing else)
> directs each application to store session information in gconf2.

I wonder who starts all the applications that have their session saved in gconf2. I don't care whether application remembers its last state, but I care that applications do not start again as they were.

>Is this really "low" priority? I find it extremely irritating that none of my applications are restarted when I log back >in.
>It's even more irritating that Gnome claims to be able to do this!

I think the same. Low priority?.
This makes my system almost unusable. With low characteristics of the 90 systems.
This should not happen even in a Beta.

i have to second this opinion! this is a serious regression! 'low' is way to low ;).

I can't believe that this so upset and seriously bug still hapens in a RC!.
Are we supposed to be that every time you turn on the PC into intrepid run all the applications?, skype, pidgim, evolution, terminal, etc.. every time.!?
This marked low and it would be one of the most annoying bugs so far.
Amazing and worrying.

you should comment upstream the ubuntu desktop team doesn't have the ressources to work on such changes right now, you can add softwares to the session start though

I agree that this bug should be a high priority fix as many users of previous releases are used to this option and expect it to work in the next release. I have this problem and it is _very_ annoying!

On Tue, Oct 28, 2008 at 03:01:56PM -0000, Mozg wrote:
> I agree that this bug should be a high priority fix as many users of
> previous releases are used to this option and expect it to work in the
> next release. I have this problem and it is _very_ annoying!

Complaining to the ubuntu bug tracker about a serious defect in
upstream gnome won't get anything fixed.

However, you might want to complain that there is no mention of
the gnome 2.24.x design problem mentioned in
https://wiki.ubuntu.com/IntrepidReleaseNotes

As for the bug with upstream...
Start following
http://bugzilla.gnome.org/show_bug.cgi?id=552387

I strongly suspect that Sebastien is already tracking this upstream
bug and will apply the fixes from upstream as soon as they are available.

Hopefully Gnome upstream will have a fix shortly after the 30th.
Otherwise, hopefully Gnome upstream will fix it by April :-).

--
Chris Dukes
"In cynicism she's about 35" -- Terry Pratchett's "Hog Father"

I have this same issue after upgrading, session is not saved, I had removed ~/.gnome2/session for test and it is not created now.

Session Save is a very important feature, I agree that this bug should be a high priority fix.

Please include this so important and annoying bug in:
https://wiki.ubuntu.com/IntrepidReleaseNotes

Becoming more and more people that are being upgraded is discovering it.
I think it is so important and annoying to be appointed.

I just upgraded to Intrepid and have this bug.

It is very very annoying. Please mark this bug has high priority.

Is there a workaround? a quick-dirty fix?

Hi Kees,

Someone has a workaround, but it requires a bit of effort, so it seems: http://blogs.sun.com/mattman/entry/gnome_2_24_session_save1.

The easiest way is have Gnome start your apps by setting them in your gnome-session-properties (System - Preferences - Sessions), but it's not the same as saving your session.

Another workaround is hibernate your computer instead of shutting it down.

At work I depend heavily on saving sessions, so I'm a bit crippled too.

Cheers, R.

Dammit, I should have waited longer before updating. This is *NOT* a low priority bug. It's the first one (there are more) that a user encounters. How could this go unnoticed in the release candidate?

How about something like this in the release notes, under "Other Known
issues":

User sessions are not currently saved on logout. The simple workaround
is to add the programs you need to start to the auto-start menu. We are
awaiting a fix from upstream, and expect to fix it in a post-release
update. https://bugs.launchpad.net/bugs/249373 .

discussing the setting will not change the fact that the bug is an upstream one and will require upstream work too

Of course, the fact that it's an upstream bug doesn't alter the fact that Ubuntu has made a serious error in judgement to move to a version of Gnome with such a serious regression.

At the top of the page I see a link to bugzilla of gnome:
http://bugzilla.gnome.org/show_bug.cgi?id=536685

Is this the description of the upstream bug?
If it is so then I think a wrong bug is reported!
Also I don't see any links back from that bug to this thread. So that upstream can see what the problems are.

ubuntu ships the current GNOME version every cycle that's not an error if you don't want it you can still use the lts version

Just because it's consistent with your policy doesn't mean the policy itself is not an error. I shouldn't have to stay with a previous version of the distro in order to avoid total and complete breakage of a useful feature (with no replacement in sight). Not to mention that downgrading to a previous version is vastly more work than the upgrade.

Ubuntu was an attractive distro precisely because it was slick, and had a high proportion of things Just Working out of the box. With several highly irritating regressions in each of the last few releases, I'm really beginning to think they've completely lost the plot on this front.

On Sat, Nov 01, 2008 at 12:33:56AM -0000, Kees Kuip wrote:
> At the top of the page I see a link to bugzilla of gnome:
> http://bugzilla.gnome.org/show_bug.cgi?id=536685

http://bugzilla.gnome.org/show_bug.cgi?id=552387
is probably the better bug to track.
536685 refers more to a bug in migrating to the new version.

>
> Is this the description of the upstream bug?
> If it is so then I think a wrong bug is reported!
> Also I don't see any links back from that bug to this thread. So that upstream can see what the problems are.

Upstream is well aware of it.
If you follow that bug you may even find some amusing bits.

--
Chris Dukes
"In cynicism she's about 35" -- Terry Pratchett's "Hog Father"

Same bug. Same annoying!

Very annoying bug for distro with such claim.

This is an extremely bad and annoying feature regression. It's a show blocker and should not have been released. It must have been known about before release. It's things like this that will drive users away from Ubuntu and Linux. Absolutely unacceptable!

To maintain a respectful atmosphere, please follow the code of conduct - http://www.ubuntu.com/community/conduct/ . Bug reports are handled by humans, the majority of whom are volunteers, such comments are not appropriate there

I agree that this is a serious bug, I have it since I upgraded from Hardy to Intrepid and is really annoying.
It should be mentioned in the release notes. Maybe I should have delayed the upgrade process if I knew about it.

Sorry if you found my tone was inappropriate. I certainly didn't mean to offend any of the developers or volunteers who work to make Ubuntu such a great product. I am very grateful for all the hard work they are doing and Ubuntu is by far my favorite OS!

However, I do believe it is unacceptable from a user's point of view that a release of Ubuntu should be made with this known feature broken/unimplemented when it worked perfectly well in all previous releases. Canonical has paid employees and releases a product that users pay for support for so there should be a level of quality control that would catch major problems like this before a new release is made. Even the most basic testing would have uncovered this.

My frustration is not with the developers, or the people who take care of bug tracking, but with the lack of quality control at Canonical that allowed something like this to make it into a new release of Ubuntu. This is a serious regression and I really can't believe Canonical wasn't aware of it before release.

Ubuntu is not writting GNOME but only contributing upstream and distributing it. The bug was known before intrepid but downgrading the version was not trivial, it was not clear that GNOME was not going to work on the issue before 2.24 and Ubuntu usually ships the current GNOME. If you are not happy about the choice you should rather complain directly to GNOME, if you want a stable Ubuntu version maybe you should consider staying on a lts version?

You can also note that the low priority focus for this bug upstream is due to the fact that the feature has never been working really correctly on many software and they don't consider that something stable and used by the majority of users, that might be wrong but should be discussed on bugzilla.gnome.org rather than here, intrepid ships now the current version which works mostly correctly and the way to go is to get the bug fixed now in the intrepid updates or jaunty

Is this regression already in: https://wiki.ubuntu.com/IntrepidReleaseNotes ? I think that People must be warned about this.

And I agree with ubuntu guys, If someone wants to complain, do not do it here, neither on gnome tracker, comments like 'me too' is completely useless, just put your name on CC and wait for someone to work on it or do it by yourself (use the force). I'm one of guys that used to use this feature, I don't have time to resolve it, so I will wait, but I agree that this issue must be warned, I was not warned about it.

At the risk of asking a really stupid question, does anyone have an idea whether there's any likelihood of a workaround such as installing and using the Gnome 2.22 session manager with 2.24? Or using a different session manager entirely? I don't mind hacking my system to get this to work, but if anyone can warn me that I'd be wasting my time, I'd appreciate that.

(I ran straight into this on the first machine I upgraded, and I don't want to upgrade any others unless I can find a fix. This regression in something I thought was one of the main functions of the desktop makes me feel extremely gloomy, but I won't vent any further here.)

Chris

On Tue, 2008-11-04 at 08:54 +0000, Sebastien Bacher wrote:
> You can also note that the low priority focus for this bug upstream is
> due to the fact that the feature has never been working really correctly
> on many software and they don't consider that something stable and used
> by the majority of users,

Well, as the comments here show, some people were relying on this
feature. This should be out of discussion now.

> that might be wrong but should be discussed on
> bugzilla.gnome.org rather than here, intrepid ships now the current
> version which works mostly correctly and the way to go is to get the bug
> fixed now in the intrepid updates or jaunty

With all due respect, I'm really surprised with this answer. Does this
imply that every time an upstream project delivers a new version with
regressions, Ubuntu will ship it just because it is the new upstream
version, no matter what damage it may cause? Does it imply that if
Ubuntu, for whatever reason, ships upstream code with a regression,
Ubuntu users will be directed to complain upstream as their only
resource?

I don't think this reflects Ubuntu's objectives (but please correct me
if I'm wrong.) As far as I can tell, Ubuntu has generally done a very
good job of selecting appropriate versions of upstream software: stable
and functional enough for most people, yet current. And most users now
rely on upgrades not containing serious regressions. So it can be argued
that the Ubuntu Desktop Team made the wrong decision when they let this
software in, because in fact, the problem was very well known to them
(https://lists.ubuntu.com/archives/ubuntu-devel/2008-September/026588.html look under "Release Status").

Now, my intention here is not to point fingers, but to ask for
reasonable damage control. As a minimum, I would expect this problem to
be acknowledged by the desktop team, and clearly reported in the Release
Notes, so that users can decide if they want to update. Of course, it
would also be good to try to fix gnome-session at least to the point of
not putting user data under risk (see my comment above in this bug
report) but I don't know if anyone can commit any resources to that.

Finally, regarding us complaining upstream, I'm not sure a lot of
disgruntled users commenting in GNOME Bugzilla will be of help. As far
as I can tell, this is not just a bug. What's going on here is that the
desired functionality is simply not implemented, or very immature in the
new gnome-session. So, it cannot be said that it "works mostly
correctly", because, in reality, it is able to start a session properly,
but far from being able to stop or restart it properly, which means it
currently does 30 to 50% of the intended functionality. This being the
actual state of affairs, I think we'll be very lucky if this is ready
for Jaunty.

Of course, GNOME should have also reported this clearly in their release
notes. They should also tell us what their policy regarding regressions
is, because they seem to be very picky regarding obscure ABI
regressions, but don't seem to mind large, user-visible regressions that
potentially put user data under risk of destruction. But this is
something we should try to bring t...

I don't want to clog up everyone's mailbox further but before you comment saying people are incompetent or that it is so easy to add such and such a feature, please try to make those changes yourself. Not only would you be doing the community a great service if you are successful, but you will also learn that these things are not as easy as they look and take a great deal of time to implement and then test. We should all be thankful that folks are donating their valuable time to work on this bug.

It's not a matter of adding such and such a feature. It's a matter of breaking something that already worked fine.

They didn't need to include the latest broken version of Gnome in Ubuntu, and they didn't need to release Ubuntu on a specific date with known bugs and regressions, but they did anyway.

(We're all clogging inboxes while saying that we shouldn't. Where *specifically* should this discussion take place instead of here?)

Should this work in Jaunty Beta? I'm asking because it does not properly:
* There is NO prompt for saving open documents at restart/shutdown.
* On logout, it miraculously works, at least with Gedit and OpenOffice

- If someone here knows what the state /should/ be in Jaunty Beta, please post here, thanks!

the current version only works for logout indeed

So should someone post a new bug report for Jaunty? Or are the developers already aware?

My understanding from the GNOME bugtracker is that a fix exists, but is not included in 2.26.0 (hopefully in 2.26.1).

This presumably means the fix will not be in 9.04, but is likely to appear in a subsequent update -- is that right?

I see no mention of it in the known issues section of the notes at http://www.ubuntu.com/testing/jaunty/beta (not sure whether I would expect to see issues reported there that are not regressions new to this particular release), but this seems consistent with Christoph's comment above and Sebastien's change to the summary.

In the mean time, has anyone had a go at using gnome-session 2.22 with the rest of gnome 2.26? I currently have gnome-session and gnome-panel 2.22 packages on "hold" while using 2.24 for everything else, and I'd like to get an idea of whether it will be possible to leave them held while upgrading the rest of the distro from 8.10 to 9.04.

Session restore seems to be back with the latest updates for Jaunty.

but the position and the number of occurrences of an application are not restored correctly: e.g., if I had 3 terminals running, upon restoring only one terminal window is shown (and not in the right position: at the up left corner, with the window title under the panel so that it cannot be moved)

Just a short comment to confirm what user Christoph reported above: session shutdown and saving is sort of working when the "log out" option is chosen in the menu, but not when the "restart" or "shut down" options are selected.

gnome-session (2.26.0svn20090408-0ubuntu1) jaunty; urgency=low

  * Updated to the current svn version (lp: #257067, #272854)
  * debian/patches/91_upstream_change_fix_session_saving.patch:
    - update by Vincent Untz which should make session storing work
      on all the actions now (lp: #552387)
  * debian/patches/98_autotools.patch:
    - new version update
  * debian/rules:
    - update for manpage changes in the new version

This is still not working for me, unfortunately. I have the latest version of gnome-session available from Jaunty (2.26.0svn20090408-0ubuntu2) and the problem remains: "Log Out" from the FUSA applet works properly, whereas "Shut Down" and "Restart" don't. My test case is to leave Firefox running and terminate the session. After "Log Out", Firefox restarts automatically with the same tabs. When any one of the other two options is used, it complains later about having been terminated abnormally.

gnome-session is fixed but fusa needs updating still

This works somewhat better, but not entirely, for me in 9.04 (upgraded from 8.10).

For example: start a new session, then run up Firefox, Emacs, Inkscape, Scribus and gnome-terminal, open documents in all of them (where applicable), replace metacity with my preferred window manager (which is not metacity or compiz).

Then select "Restart", reboot, and log in again.

The restarted session contains Firefox with all of the correct tabs open (though none of them load, as network-manager has not yet reestablished the network connection). It also has Emacs and gnome-terminal. So that's a definite improvement.

But it does not restart Scribus or Inkscape at all, and it has the wrong window manager. Reading .xsession-errors it looks like gnome-session tries to run my preferred window manager, but it has already started metacity before that, so the "right" window manager then fails to start.

There are also some weird random errors -- once it started Emacs with a window only a few pixels large, and it often seems to open an additional window from some program that I did not have open at all before (a Nautilus window on some arbitrary directory, or a Tomboy index window).

It doesn't seem to make any difference whether I "Restart" or "Log Out", both work equally (i.e. partially).

Altogether, better than 8.10 but still a regression from 8.04. I haven't established yet whether the programs that were not restarted had been given a chance to save their documents before exit or not -- that's quite important.

Chris

 This bug persists with me, too: I exited Pidgin, run gnome-session-save, logged out logged in and Pidgin reappeared automatically. (Pidgin is not an entry in System > Preferences > Startup Applications.)
The same result when I rebooted instead of logging out.
uname -a; Linux MD97600 2.6.28-14-generic #47-Ubuntu SMP Sat Jul 25 00:28:35 UTC 2009 i686 GNU/Linux
This Ubuntu version is a dist-upgrade from 8.04.2 > 8.10 > 9.04.
GNOME 2.26.1

with Ubuntu 10.04:
1) the gnome-terminal and nautilus windows are restored not iconized, if they were iconized at shutdown time.
2) the application position inside task bar are not restored in the correct order (are always swapped)
3) Mozilla windows are all restored on the first desktop, even if they are in other desktops at shut down time.