Update to libdb-5.3.28-21.fc25 in F25 will break the rpm database. Repairable by rebuilding rpm database with rpm --rebuilddb

[stan]

There is an update (or test update) of libdb for F25 that, when
installed, breaks the rpm database because it is incompatible.  The dnf
update completes, and then hangs (at least it did on my system).  The
fix is to remove the old rpm databases (this might be optional), and
rebuild the rpm databases.

rm /var/lib/rpm/__db.00?  # optional?
rpm --rebuilddb

I did a reboot to sync everything that uses libdb (anything that uses
the Berkeley db).  It might be possible just to restart affected
applications if that isn't an option.  That worked for my mail client
before I rebooted.

[Samuel Sieb]

On 06/09/2017 06:12 PM, stan wrote:
>  There is an update (or test update) of libdb for F25 that, when
>  installed, breaks the rpm database because it is incompatible.  The dnf
>  update completes, and then hangs (at least it did on my system).  The
>  fix is to remove the old rpm databases (this might be optional), and
>  rebuild the rpm databases.
>  
>  rm /var/lib/rpm/__db.00?  # optional?
>  rpm --rebuilddb 
That's not deleting the RPM database, just removing temporary files that 
rpm uses.

>  I did a reboot to sync everything that uses libdb (anything that
> uses
>  the Berkeley db).  It might be possible just to restart affected
>  applications if that isn't an option.  That worked for my mail client
>  before I rebooted. 
A reboot would be recommended, but restarting any affected applications 
should be sufficient.

[stan]

On Fri, 9 Jun 2017 18:20:39 -0700
Samuel Sieb <samuel(a)sieb.net&gt; wrote:

>  On 06/09/2017 06:12 PM, stan wrote: 
>  > rm /var/lib/rpm/__db.00?  # optional? 
>  That's not deleting the RPM database, just removing temporary
> files
>  that rpm uses. 
Thanks for the clarification.

[Ed Greshko]

On 06/10/17 09:12, stan wrote:
>  There is an update (or test update) of libdb for F25 that, when
>  installed, breaks the rpm database because it is incompatible.  The dnf
>  update completes, and then hangs (at least it did on my system).  The
>  fix is to remove the old rpm databases (this might be optional), and
>  rebuild the rpm databases.
>
>  rm /var/lib/rpm/__db.00?  # optional?
>  rpm --rebuilddb
>
>  I did a reboot to sync everything that uses libdb (anything that uses
>  the Berkeley db).  It might be possible just to restart affected
>  applications if that isn't an option.  That worked for my mail client
>  before I rebooted. 

FWIW, so far I've seen no ill effects from updating to
libdb-5.3.28-21.fc25.  However, I should note that I had previously
updated to libdb-5.3.28-16 from testing also with no ill effects.

-- 
Fedora Users List - The place to go to speculate endlessly

[Ed Greshko]

On 06/10/17 10:22, Ed Greshko wrote:
>  On 06/10/17 09:12, stan wrote:
> > There is an update (or test update) of libdb for F25 that, when
> > installed, breaks the rpm database because it is incompatible.  The dnf
> > update completes, and then hangs (at least it did on my system).  The
> > fix is to remove the old rpm databases (this might be optional), and
> > rebuild the rpm databases.
> >
> > rm /var/lib/rpm/__db.00?  # optional?
> > rpm --rebuilddb
> >
> > I did a reboot to sync everything that uses libdb (anything that uses
> > the Berkeley db).  It might be possible just to restart affected
> > applications if that isn't an option.  That worked for my mail client
> > before I rebooted.
>
>  FWIW, so far I've seen no ill effects from updating to
>  libdb-5.3.28-21.fc25.  However, I should note that I had previously
>  updated to libdb-5.3.28-16 from testing also with no ill effects.
>  
Oh, I just realized I updated a second system to libdb-5.3.28-21.fc25
but it hadn't been updated from the testing repository.  No problems for
me with that update either.


-- 
Fedora Users List - The place to go to speculate endlessly

[Samuel Sieb]

On 06/09/2017 07:29 PM, Ed Greshko wrote:
>  On 06/10/17 10:22, Ed Greshko wrote:
> > FWIW, so far I've seen no ill effects from updating to
> > libdb-5.3.28-21.fc25.  However, I should note that I had previously
> > updated to libdb-5.3.28-16 from testing also with no ill effects.
>  
>  Oh, I just realized I updated a second system to libdb-5.3.28-21.fc25
>  but it hadn't been updated from the testing repository.  No problems for
>  me with that update either. 
My observations of the discussions indicate that it depends on the 
packages that are getting updated.  Some packages try to run rpm during 
the installation process and that is where the problem happens.

[stan]

On Sat, 10 Jun 2017 10:22:17 +0800
Ed Greshko <ed.greshko(a)greshko.com&gt; wrote:


>  FWIW, so far I've seen no ill effects from updating to
>  libdb-5.3.28-21.fc25.  However, I should note that I had previously
>  updated to libdb-5.3.28-16 from testing also with no ill effects. 
I suppose the subject should be *might* break the rpm database.  I
wonder why it worked for you and not for me?  I have test-updates
enabled, so I should have followed the same path you did.

[Patrick O'Callaghan]

On Sat, 2017-06-10 at 10:22 +0800, Ed Greshko wrote:
>  On 06/10/17 09:12, stan wrote:
>  > There is an update (or test update) of libdb for F25 that, when
>  > installed, breaks the rpm database because it is incompatible.  The dnf
>  > update completes, and then hangs (at least it did on my system).  The
>  > fix is to remove the old rpm databases (this might be optional), and
>  > rebuild the rpm databases.
>  > 
>  > rm /var/lib/rpm/__db.00?  # optional?
>  > rpm --rebuilddb
>  > 
>  > I did a reboot to sync everything that uses libdb (anything that uses
>  > the Berkeley db).  It might be possible just to restart affected
>  > applications if that isn't an option.  That worked for my mail client
>  > before I rebooted.
>  
>  
>  FWIW, so far I've seen no ill effects from updating to
>  libdb-5.3.28-21.fc25.  However, I should note that I had previously
>  updated to libdb-5.3.28-16 from testing also with no ill effects. 
On running 'dnf update' this morning I got a segfault from
/usr/lib/libdb-5.3.so, which is part of libdb-5.3.28-21.fc25, so the
problem is more than just an incompatible database.

However removing the temp files and rebuilding the db seems to have
cleared it.

poc

[Samuel Sieb]

On 06/10/2017 02:42 AM, Patrick O'Callaghan wrote:
>  On running 'dnf update' this morning I got a segfault from
>  /usr/lib/libdb-5.3.so, which is part of libdb-5.3.28-21.fc25, so the
>  problem is more than just an incompatible database.
>  
>  However removing the temp files and rebuilding the db seems to have
>  cleared it. 
Don't those two paragraphs contradict each other?  You rebuilt the 
database and there's no more segfault.  The problem is that a certain 
structure from glibc that is stored in the database has changed it's 
format.  Clearing the temp files removes that structure so after that 
there's no problem.

[Patrick O'Callaghan]

On Sat, 2017-06-10 at 10:07 -0700, Samuel Sieb wrote:
>  On 06/10/2017 02:42 AM, Patrick O'Callaghan wrote:
>  > On running 'dnf update' this morning I got a segfault from
>  > /usr/lib/libdb-5.3.so, which is part of libdb-5.3.28-21.fc25, so the
>  > problem is more than just an incompatible database.
>  > 
>  > However removing the temp files and rebuilding the db seems to have
>  > cleared it.
>  
>  Don't those two paragraphs contradict each other?  You rebuilt the 
>  database and there's no more segfault.  The problem is that a certain 
>  structure from glibc that is stored in the database has changed it's 
>  format.  Clearing the temp files removes that structure so after that 
>  there's no problem. 
My point is that an incompatibility in a database should never cause a
segfault under any circumstances. If it happens it means the db library
is not adequately testing its input. This kind of thing is what leads
to security holes as well as broken software.

poc