[Pkg-xfce-devel] Bug#653965: Bug#653965: xfce4-session: Strange desktop behaviour after corruption of xfce4-session cache

Tue Jan 3 17:25:58 UTC 2012

Hi Alex,

Thank you for your reply.  It is much appreciated.

On 02/01/12 06:37, Yves-Alexis Perez wrote:ls -l
> On dim., 2012-01-01 at 22:53 +0000, Paul Bryan Roberts wrote:
>> Package: xfce4-session
>> Version: 4.6.2-3
>> Severity: normal
>>
>> *** Please type your report below this line ***
>>
>> Strange xfce desktop behaviour includes:
>>     - failure to log-out
>>       (reports "Failed to receive a reply from the session manager")
> looks like xfce4-session wasn't even running
I thought so too but not so:

debian    2163  2127  0 14:34 ?        00:00:00 xfce4-session
debian    2338  2163  0 14:35 ?        00:00:00 [xfce4-session] <defunct>

Looks more like sigchild confusion too me.  The error message also has 
text stating the
Session Manager must be in the idle state.
>>     - log-out panel icon offers to exit panel instead of logout
>>     - panel disappears
> same
>
>>     - windows have no min/max buttons etc. and cannot be moved
> xfwm4 not running either
Appears so.
>> The strange behaviour is not transient:  it persists over reboot.
>> Normal behaviour can be restored by deleting the xfce4-session file in
>> $HOME/.cache/sessions/.
> Indeed, you have a bad session cache.
>> Problems observed with xfce4-session 4.6.2-3 on both amd64 and i386
>> architectures.  Problems also observed with xfce4-session 4.8.1-3 and
>> amd64 architecture (Wheezy and LMDE).
> Yeah, looks like unrelated to the arch or version.
>
>> Cause of cache corruption unknown.
>> Corruption noticed after package upgrades but this may not be significant.
>> The xfce4-session package itself has not been upgraded.
> For me, it's just that you quit saving your session with no xfwm4 or
> something like that, but it might be something else. The fact that
> xfce4-session was missing from the session itself is weird though.
>> No idea how to reproduce (but corrupt cache files available on request).
> Please send them and diff with your current cache.
Attached - four bad ones and a good one for reference.

It seems these files list the 'applications' to start when restarting 
the desktop.

One (xfce4-session-squeeze32.bad) has no reference to xfce-panel so that 
could account for why
there was no panel on reboot.

One (xfce4-session-lmde.bad) has no reference to xfwm4 and that could 
account for why there
was no window manager on reboot.

The others have RestartCommands for xfwm4 that include a uuid.  The 
.cache/session directories
appear to contain files with these uuid in their name except on my 
wheezy installation - two cache
files but no uuid match.  Perhaps window manager barfs on start-up so no 
window manager.

Perhaps the software is doing what it was written to do and supports 
well the minority of users
who choose not to have a panel or a window manager but I suspect the 
majority might find the
the design sucks just a little because once something has gone wrong, it 
stays wrong.

The logout may be a separate issue but note ... with xfce-session 
4.6.2-3, log out doesn't happen
but with xfce4-session 4.8.1-3 it does eventually log out.  About a 
minute later, just when you've
given up.  I find this behaviour very disconcerting and undermines 
confidence in the whole desktop.
More to the point - I suspect that when this log out occurs, the window 
manager has exited and that
is why the session is saved with no reference to the window manger so 
there isn't one on reboot.
>> It took weeks (and a fresh installation) to make any the connection between
>> the diverse symptoms.
> Well, sorry to hear that, but should you have googled for the symptoms,
> it should have been pretty obvious what happened.
I was trying to suggest was there may be many apparently unrelated 
problem reports out there
that might be connected and seeing the connection might be the way to 
finding the root cause.

I looked on the net.  I found "Help, I've upgraded Debian Testing and my 
XFCE Desktop has gone
haywire" postings.  Long on hysterics, short on facts.  The most helpful 
suggestion was "uncheck
save session on exit".  I have now corresponded with one unhappy chap 
whose response was
along the lines of "Oh, that one again !  When will I learn ?"
> Note that you have a graphical frontend to the session cache in the Xfce
> settings manager, Session and Startup, Session tab.
Interesting.  From the looks of this, crashing say, wicd, is not serious 
as autostart will ensure
it is started next reboot but for the panel and window manger not so.
> Unfortunately, there's not so much we can do about this.
I take it by we you mean the Debian Project ?  Your own recommendations 
are that I report
problems to you rather than upstream even though I recognise that often 
all you can do
is refer the problem upstream
> Regards,
Regards too,

-- 
Paul Bryan Roberts
pbronline-debian at yahoo.co.uk

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: xfce4-session-lmde.bad
URL: <http://lists.alioth.debian.org/pipermail/pkg-xfce-devel/attachments/20120103/b42489bb/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: xfce4-session-squeeze32.bad
URL: <http://lists.alioth.debian.org/pipermail/pkg-xfce-devel/attachments/20120103/b42489bb/attachment-0001.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: xfce4-session-squeeze64.bad
URL: <http://lists.alioth.debian.org/pipermail/pkg-xfce-devel/attachments/20120103/b42489bb/attachment-0002.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: xfce4-session-squeeze64.good
URL: <http://lists.alioth.debian.org/pipermail/pkg-xfce-devel/attachments/20120103/b42489bb/attachment-0003.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: xfce4-session-wheezyxfce.bad
URL: <http://lists.alioth.debian.org/pipermail/pkg-xfce-devel/attachments/20120103/b42489bb/attachment-0004.ksh>