Created attachment 10976 [details]
journalctl -b > journal.txt output from system after install halted

possible issue from journal.txt:

May 04 12:28:58 localhost.localdomain kernel: traps: urpmi[27196] trap divide error ip:7f1ff0f2b876 sp:7ffdf3c6df00 error:0 in libdb-5.3.so (deleted)[7f1ff0e57000+1b1000]

the issue does not occur in the i586 install,

however it fails with a different error.

this can be resolved with the following:
# rm -f /var/lib/rpm/__*
# rpm --rebuilddb

(In reply to ben mcmonagle from comment #1)
> Created attachment 10976 [details]
> journalctl -b > journal.txt output from system after install halted
> 
> possible issue from journal.txt:
> 
> May 04 12:28:58 localhost.localdomain kernel: traps: urpmi[27196] trap
> divide error ip:7f1ff0f2b876 sp:7ffdf3c6df00 error:0 in libdb-5.3.so
> (deleted)[7f1ff0e57000+1b1000]

That does indeed look fishy.

It does look a bit like https://bugzilla.redhat.com/show_bug.cgi?id=468437
( Summary: rpmdb environment corruption on ext4 )

and you mentioned 

(In reply to ben mcmonagle from comment #2)
> the issue does not occur in the i586 install,
> 
> however it fails with a different error.
> 
> this can be resolved with the following:
> # rm -f /var/lib/rpm/__*
> # rpm --rebuilddb

Does doing the same help to finish the 64bit upgrade?

CC'ing tv and the kernel maintainers, because I'm not sure whom to assign to.
CC'ing the rpm stack maintainers, too.

Summary: during on-line upgrade of all DE install, floating point exception occurs(core dumped) => kernel: traps: urpmi[...] trap divide error ip:.. sp:.. error:0 in libdb-5.3.so (deleted)[..] -- during on-line upgrade of all DE install, floating point exception occurs(core dumped)
Source RPM: (none) => db53?, kernel?
CC: (none) => kernel, marja11, rpmstack, thierry.vignaud

oh, is that indeed an ext4 partition?

(In reply to Marja Van Waes from comment #4)
> oh, is that indeed an ext4 partition?

It is, it's in your logs, sorry for asking :-)

(In reply to ben mcmonagle from comment #1)

> May 04 12:28:58 localhost.localdomain kernel: traps: urpmi[27196] trap
> divide error ip:7f1ff0f2b876 sp:7ffdf3c6df00 error:0 in libdb-5.3.so
> (deleted)[7f1ff0e57000+1b1000]

I'm wondering why it says "deleted" after libdb-5.3.so

What is the output of

   ls -al /usr/lib64/libdb-5.3.so

?

Hm,
Thierry, do we need to to:

diff --git a/urpm/select.pm b/urpm/select.pm
index 9e16b0b8..37cc3d0b 100644
--- a/urpm/select.pm
+++ b/urpm/select.pm
@@ -7,7 +7,7 @@ use urpm::util qw(any formatList intersection member min partition uniq);
 use urpm::sys;
 use URPM;
 
-my $default_priority_list = 'rpm,perl-base,perl-URPM,perl-MDV-Distribconf,urpmi,meta-task,glibc,aria2';
+my $default_priority_list = 'rpm,perl-base,perl-URPM,perl-MDV-Distribconf,urpmi,meta-task,glibc,aria2,libdb5.3,lib64db5.3';
 my @priority_list = split(',', $default_priority_list);


?

CC: (none) => tmb

(or enforce deps with rpm)

(In reply to Marja Van Waes from comment #6)

> 
> I'm wondering why it says "deleted" after libdb-5.3.so
> 
> What is the output of
> 
>    ls -al /usr/lib64/libdb-5.3.so
> 
> ?

ls -al /usr/lib64/libdb-5.3.so
-rwxr-xr-x 1 root root 2306848 Sep 13  2018 /usr/lib64/libdb-5.3.so*

(In reply to Marja Van Waes from comment #3)

> 
> Does doing the same help to finish the 64bit upgrade?
> 
yes

*** Bug 24731 has been marked as a duplicate of this bug. ***

CC: (none) => remco

(In reply to Thomas Backlund from comment #7)
> Hm,
> Thierry, do we need to to:
> 
> diff --git a/urpm/select.pm b/urpm/select.pm
> index 9e16b0b8..37cc3d0b 100644
> --- a/urpm/select.pm
> +++ b/urpm/select.pm
> @@ -7,7 +7,7 @@ use urpm::util qw(any formatList intersection member min
> partition uniq);
>  use urpm::sys;
>  use URPM;
>  
> -my $default_priority_list =
> 'rpm,perl-base,perl-URPM,perl-MDV-Distribconf,urpmi,meta-task,glibc,aria2';
> +my $default_priority_list =
> 'rpm,perl-base,perl-URPM,perl-MDV-Distribconf,urpmi,meta-task,glibc,aria2,
> libdb5.3,lib64db5.3';
>  my @priority_list = split(',', $default_priority_list);
> 
> 
> ?

@ Thierry

See tmb's question above :-)

(All my noise may have made you miss it)

Source RPM: db53?, kernel? => urpmi
Assignee: bugsquad => thierry.vignaud

I tried to reproduce the initial bug but my upgrade completed without error. My starting point was an all-DE, default package install from the Mageia-6-x86_64-DVD ISO, fully updated. I ran the upgrade from a terminal in Xfce.

I did notice that, as in Ben's log, the initial priority update list included all the libreoffice packages. That seems wrong.

CC: (none) => mageia

ran the upgrade all DE (i586) today.
(princeton mirror)

needed to run:
# rpm --rebuilddb

to complete upgrade.

(In reply to ben mcmonagle from comment #14)
> ran the upgrade all DE (i586) today.
> (princeton mirror)
> 
> needed to run:
> # rpm --rebuilddb
> 
> to complete upgrade.

Tried to reproduce that (using my local mirror), but failed. Either we are starting from a different baseline, or it's sensitive to the order in which packages are upgraded.

I had the same problem upgrading from MGA6.1 Live with latest updates :

segfault .....error 4 in libdb-5.3.so (deleted)...

I will attach the result of "journalctl | grep RPM" so that rpm update order can be seen.

CC: (none) => lists.jjorge

Created attachment 11033 [details]
An upgrade RPM list which failed

Sorry this bug was hidden the the bug stream :-(

(In reply to Thomas Backlund from comment #8)
> (or enforce deps with rpm)


Yeah I would rather versionnate the requires in rpm

(In reply to Martin Whitaker from comment #13)
> I tried to reproduce the initial bug but my upgrade completed without error.
> My starting point was an all-DE, default package install from the
> Mageia-6-x86_64-DVD ISO, fully updated. I ran the upgrade from a terminal in
> Xfce.
> 
> I did notice that, as in Ben's log, the initial priority update list
> included all the libreoffice packages. That seems wrong.

uh? LO in priority list?

If you can reproduce that, could you use:
urpmi --bug bug24767 <other_args>
tar cfa bug24767{.tar.xz,}

And send me the tarball?

(In reply to Thierry Vignaud from comment #18)
> Sorry this bug was hidden the the bug stream :-(
> 
> (In reply to Thomas Backlund from comment #8)
> > (or enforce deps with rpm)
> 
> 
> Yeah I would rather versionnate the requires in rpm

Versioned deps added in rpm-4.14.2.1-12.mga7 currently building

> Versioned deps added in rpm-4.14.2.1-12.mga7 currently building

So is this fixed, or what's left to do?

I will have a look at this tomorrow - need to dig out a Mga6 system image and re-install it.
then see how it goes.

i586 upgrade no longer reports the issue

x86_64 upgrade no longer displays the issue.

if there are no objections,
closing as fixed

Status: NEW => RESOLVED
Resolution: (none) => FIXED