[Pkg-xfce-devel] Debian testing installs with XDM as dependency of Xfce fail (eventually) when switched to LightDM

Wed Sep 7 22:03:00 UTC 2011

On 09/07/2011 11:16 AM, Yves-Alexis Perez wrote:
> On mer., 2011-09-07 at 11:02 -0400, Gilbert Sullivan wrote:
>
>>
>> I'll attach /var/log/syslog and the present output of dmesg. I hope the
>> files aren't too large. If so, please suggest an alternative way for me
>> to provide the information.
>
> That's fine, I've received them directly.
>
>>
>>>>
>>>> I am NOT seeing the Init ID "co" respawning too fast errors on any
>>>> systems other than the AMD64 installations which had XDM as a dependency
>>>> of Xfce4. I still have no idea whether or not those errors had anything
>>>> to do with my Xfce session launch failures.
>>>
>>> Ok.
>
> Could you check on those system the content of /etc/inittab too? (join
> them if possible). Did you change anything there? Do you use some kind
> of serial login? (I know this is weird on a laptop but...)
>
> But in any case, I don't think those are related to the login failures.
>

I have not changed /etc/inittab manually on any of the systems, but I'm 
not quite sure what you mean by serial login. Do you mean via serial 
port connection? If so, no. The laptop doesn't even have a serial port, 
and I wish it did, because I'd like to use minicom with it for 
configuring some Cisco switches. (I guess I might be able to use a 
USB-to-serial converter for this, but haven't looked into it.)

I have more systems available to me now. I'll send the /etc/inittab file 
from the old desktop (PCV-rx462d) and also from an i386 notebook (CF-R3) 
and from the AMD64 system (T520i). Neither the PCV-rx462d nor the CF-R3 
have ever had the login problem. The interesting point is that the CF-R3 
is an i386 installation from the same daily build version as the AMD64 
system that was failing.

>>>>
>>>> I did gather the dpkg -l information on one symptomatic AMD64 and one
>>>> asymptomatic i386 system -- and I gathered lightdm.log examples, too.
>>>> I'm doubtful that there's anything of interest here, since it looks very
>>>> much as though the systems aren't going to fail again.
>>>
>>> Please share them.
>>
>> Done. Again, I hope this doesn't make the mail too large. If the normal
>> approach is to upload them to a hosting service somewhere, please let me
>> know which one you'd prefer.
>
> It's fine :)
> There are quite some difference between the systems installs (in dpkg
> -l) so it's hard to say for sure.

Yes, the PCV-rx462d system is an ancient desktop on which Debian testing 
was installed over a year ago, and is a poor basis for comparison. But 
the i386 installation on a little subnotebook (CF-R3) was made from the 
same daily build as the AMD64 installation on the Thinkpad (T520i). I 
can now send the result of dpkg-l on the CF-R3 for comparison to that of 
the T520i. I went through the OS setups on those two systems 
side-by-side and installed the same packages via aptitude on both 
systems, so the only differences between them should be due to the 
differences in installations for those architectures. I'm sorry I 
couldn't send data from the CF-R3 earlier. It was being used heavily by 
a student who couldn't let me work with it while she was preparing a 
paper on it.

>
> For the lightdm log, could you explain which is which, and especially if
> there's a “failed” one there?
>

The lightdm logs are all from the system that was having trouble, but 
not *while* it was having trouble. Those logs were all gathered after 
this most recent lightdm installation -- which is no longer (apparently) 
inclined to cause failure to log in. The only sign of any real trouble 
still showing on that system are the "respawning" errors, and I guess 
those may not be related to the login failure problem. Anyway, the 
lightdm log files are named for the machine, the date, and the order in 
which they were recorded. I doubt they contain anything useful.

It is unfortunate that I have been unable to get this T520i system to 
fail again since the most recent installation of lightdm. I had assumed 
that I would be able to re-create the problem easily, since I had seen 
it occur three times in a row -- each time within two days after 
installing lightdm on that system. Perhaps a software upgrade that has 
occurred in the interim between the most recent two lightdm 
installations has fixed some component which was causing the failures. I 
did see a changelog note that an update had fixed issues specific to the 
AMD64 architecture. But why said component would cause failures only 
when using lightdm is beyond me. I guess it's true that lightdm is 
rather much more "hooked" into the system than is xdm.

I am still noticing one thing on the T520i that seems suggestive of the 
previous login failure. If I have a long session such that I see a long 
list of the respawning error messages (so long that they take the entire 
screen) when logging out, then I don't get my chosen mouse cursor theme 
if I log back in. I appear to get the default cursor theme for Xfce 
instead. If I want to get the Adwaita cursor theme to reappear, I have 
to reboot the system, and then log on. I have noticed no other 
degradation of the Xfce session than this one.

I'm afraid all of the lightdm logs that might have been of use are in 
the bit bucket. It was stupid of me not to have gathered those logs at 
the time of failure. That last time around with the login failure, it 
was one of those cases where I just needed the system to work, and I was 
not feeling well -- so I just did the easy, brainless thing and went 
back to XDM. I purged the system following removal of lightdm because I 
was certain that I was going to try lightdm again and didn't want 
possible problems with old configurations screwing up the new 
installation attempt. I should have saved the evidence before purging.

I hope this message makes sense, and that I sent everything you 
requested. I'm still not operating on all cylinders, and I'm not exactly 
a Ferrari (12 cylinders) to begin with.

;-)

Best regards,
Gilbert
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: CF-R3_dpkg-l.txt
URL: <http://lists.alioth.debian.org/pipermail/pkg-xfce-devel/attachments/20110907/8bfe9fe6/attachment-0004.txt>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: CF-R3_inittab.txt
URL: <http://lists.alioth.debian.org/pipermail/pkg-xfce-devel/attachments/20110907/8bfe9fe6/attachment-0005.txt>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: T520i_inittab.txt
URL: <http://lists.alioth.debian.org/pipermail/pkg-xfce-devel/attachments/20110907/8bfe9fe6/attachment-0006.txt>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: PCV_rx_462d_inittab.txt
URL: <http://lists.alioth.debian.org/pipermail/pkg-xfce-devel/attachments/20110907/8bfe9fe6/attachment-0007.txt>