Re: CR repo update disaster for my desktop.

Rob Kampen <rkampen@xxxxxxxxxxxxxxxxx> · Wed, 5 Dec 2018 21:30:32 +1300

On 5/12/18 4:17 AM, Simon Matter wrote:
On 4/12/18 9:06 AM, Rob Kampen wrote:
Sorry for top post, my android BlueMail will not let me insert at the
bottom.

I have found that tracker-extract seems to trigger a segfault. I note
this is not updated in CR, but comes from base. Removal of tracker seems
too harsh as it has dependant modules like brasero, evince, grilo,
nautilus and totem.
Maybe there is an obscure relationship in tracker that need attention.
Stumpped.

⁣Sent from BlueMail 

On 03 Dec 2018, 11:28 PM, at 11:28 PM, Fabian Arrotin
<arrfab@xxxxxxxxxx> wrote:
On 03/12/2018 11:14, John Hodrien wrote:
On Mon, 3 Dec 2018, Simon Matter wrote:

Le 03/12/2018 à 06:25, Rob Kampen a écrit :
I enabled the CR repo and did the yum update. Some 800+ rpms were
offered and all seemed to resolve depenancies OK, so yes it was
started. The updates completed and all looked good, until the
reboot.
I got a similar disaster here. I guess the lesson to be learned is
that
CR is nice to have on servers, but don't use it on desktops.
My question is what will change with the final release of 7.6? I
thought
the CR repo usually holds all updates with the exception of
centos-release, or are there more updates to come? I had the
impression
that in the past, the final release brought only cosmetic changes
with
the
centos-release being updated.
I've seen zero problems on Desktops I've installed CR on.

If CR has issues, as you say you'd expect 7.6 to have problems.  If
things are
failing with CR updates, you really want to investigate what's going
on.
jh
+1 as 7.6.1810 is exactly 7.5.1804+updates+CR so only missing packages
are centos-release/anaconda and install tree/media ..

So if there are issues with CR, using the list and bugs.centos.org
would
be good.
BTW : the goal of CR is exactly to find those issues earlier and then
write Release Notes with workarounds/warnings
So in an effort to narrow down the problem I also have an old Samsung
laptop - i5 with an nvidia card - all up-to-date 7.5 - I thought I would
try a more conservative upgrade approach.

first updated to the CR kernel with yum upgrade kernel*

then after successful reboot did an update to gdm* and gnome*

another reboot and all is well

then update of xorg* (only 20 files)

Now it will not boot - keyboard is non responsive so cannot even do a
Ctrl Alt F2 to access a shell. Old kernel is also non boot. Left this
machine for now and back to my main desktop workstation.

No idea if this xorg problem relates to my original workstation - so on
that machine I did a downgrade of xorg* - seemed to complete ok, but on
reboot and login - still the same problem.

As the only clue in the logs is the libc segfault I did a downgrade of
glibc - this too seemed to work but no change to the system - gnome
crashes after 5 - 10 seconds.

Wasted far too much time on this, no idea what to do now so I have done
a fresh 7.5 install and all works again - just need to install all the
additional stuff I use each day, but at least I have a desktop that
functions.

I will wait until 7.6 is officially released, watch the list for a week
or two and then with much fear and trepidation give it a go. <sad face>

As a long time user of CentOS (well over a decade - I started with 5.x
workstations, then had a couple of 6.x, and used two 7.x machines, not
counting my servers), I must observe that the core server stuff "just
works" e.g. mail, mysql / mariadb, postfix, dovecot, apache, php. Even
the migration to libreoffice was okay.

The major version upgrades each gave an all over better desktop
experience. What has also happened however is that some of the 7.x
upgrades have been difficult to deal with - one of the recent firefox
updates lost all my local password databases which I do not back up to
the cloud for obvious reasons. Unable to resolve the instability after
many hours of testing and trialing combinations, I have left firefox and
moved to vivaldi - we'll see how that works. As upstream seems to
include more bleeding edge versions of major desktop utilities such as
firefox, stability has been compromised.

Now with this latest CR update, I have no idea if it is gnome, gdm, xorg
or some other component, but to have two quite differently configured
machines both loose their desktops, I am unhappy. The use of roll back
of over 800 files just didn't pass the transaction test thus that too
was futile.

I thought the approach I used with my laptop today was sufficiently
cautious, yet it caught me and I am still trying to work out what to do
next.

As always, happy to entertain ideas and suggestions. My main workstation
however is now safely working at a clean install of 7.5, thus cannot do
further tests there, life must go on, four days of down time is far too
much.

Also let me give a heartfelt thanks to the hard working CentOS team, I
see the more and more difficult task you need to navigate with each
major update, you do an outstanding task, and I am most grateful!
The big question remains, what did you install which is not part of the
base CentOS distribution? You mentioned vivaldi browser, Nvidia stuff,
what else did you install from EPEL or other sources? I mean, is it
possible in your case that something causes problems now which is not part
of CentOS?
It is possible - I have now got a new clean install of 7.5 and am slowly 
making that workable, thus unable to do further testing until I do the 
7.6 upgrade, which I will hold off until Red Hat Bugzilla – Bug 1650634 
is resolved - this looks like the exact issue my laptop has stumbled in 
to. Will need to use a rescue USB to try the xorg downgrade on that as 
it also crashes the keyboard.

Regards,
Simon

_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
https://lists.centos.org/mailman/listinfo/centos
_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
https://lists.centos.org/mailman/listinfo/centos