Re: Buster to Bullseye upgrade problem
On Fri 20 Aug 2021, at 04:45, David Wright <deblis@lionunicorn.co.uk> wrote:
> On Thu 19 Aug 2021 at 07:42:56 (+0100), Gareth Evans wrote:
> > On Thu 19 Aug 2021, at 05:50, David Wright <deblis@lionunicorn.co.uk> wrote:
> > > On Thu 19 Aug 2021 at 04:00:04 (+0100), Gareth Evans wrote:
> > > > On Wed 18 Aug 2021, at 23:33, piorunz <piorunz@gmx.com> wrote:
> > > > > On 18/08/2021 16:14, Gareth Evans wrote:
> > > > > > Unpacking gir1.2-gst-plugins-bad-1.0:amd64 (1.18.4-3) ...
> > > > > > [1mdpkg:[0m error processing archive /tmp/apt-dpkg-install-Un4rDW/28-gir1.2-gst-plugins-bad-1.0_1.18.4-3_amd64.deb (--unpack):
> > > > > > trying to overwrite '/usr/lib/x86_64-linux-gnu/girepository-1.0/GstTranscoder-1.0.typelib', which is also in package pitivi 0.999-1+b1
> > > > > > [1mdpkg-deb:[0m [1;31merror:[0m paste subprocess was killed by signal (Broken pipe)
> > > > >
> > > > > IMO that's the source of your problem.
> > > > > You have two packages fighting to overwrite the same file. You should
> > > > > inspect them.
> > > > > Are you sure they come from buster, not from foreign repository?
> > > >
> > > > Apparently they are both from Buster.
> > >
> > > I'm afraid not. You need to check the version numbers more carefully.
> >
Hello again,
Thanks David, for your clear explanation of the problem.
> > I meant the packages (gir1.2-gst-plugins-bad-1.0, pitivi) being upgraded, rather than the versions replacing them, were both from Buster at the point of the upgrade (weren't they?)
>
> One might assume so, but only you can check that ...
I was mistaken in thinking that gir1.2-gst-plugins-bad-1.0 was installed in Buster in the first place.
[Buster]
$ apt policy gir*bad*
gir1.2-gst-plugins-bad-1.0:
Installed: (none)
Candidate: 1.14.4-1+deb10u2
Version table:
1.14.4-1+deb10u2 500
500 http://deb.debian.org/debian buster/main amd64 Packages
500 http://security.debian.org/debian-security buster/updates/main amd64 Packages
$ aptitude why gir1.2-gst-plugins-bad-1.0
Not currently installed
The candidate version 1.14.4-1+deb10u2 has priority optional
No dependencies require to install gir1.2-gst-plugins-bad-1.0
$ apt policy pitivi
pitivi:
Installed: 0.999-1+b1
Candidate: 0.999-1+b1
Version table:
*** 0.999-1+b1 500
500 http://deb.debian.org/debian buster/main amd64 Packages
100 /var/lib/dpkg/status
So pitivi 0.999 as installed is a Buster package, and gir* is installed during the upgrade as a dependency of Bullseye's newer pitivi version.
[Bullseye]
$ aptitude why gir1.2-gst-plugins-bad-1.0
i pitivi Depends gir1.2-gst-plugins-bad-1.0 (>= 1.18.0)
The first upgrade interruption issue (repeated here for clarity):
--
Unpacking gir1.2-gst-plugins-bad-1.0:amd64 (1.18.4-3) ...
dpkg: error processing archive /tmp/apt-dpkg-install-YeCJ7K/28-gir1.2-gst-plugins-bad-1.0_1.18.4-3_amd64.deb (--unpack):
trying to overwrite '/usr/lib/x86_64-linux-gnu/girepository-1.0/GstTranscoder-1.0.typelib', which is also in package pitivi 0.999-1+b1
dpkg-deb: error: paste subprocess was killed by signal (Broken pipe)
--
appears to be a file conflict, per
https://www.debian.org/releases/stable/amd64/release-notes/ch-upgrading.en.html#file-conflicts
which includes that "File conflicts should not occur if you upgrade from a “pure” buster system..."
So I would like to know if apt is not handling this properly, or if the scenario of a file changing packages (see David's previous email) is an expected exception to the (sort of) rule.
Shouldn't pitivi 0.999 be disregarded anyway as it's being upgraded? It's not a conflict involving two Bullseye packages, nor with one from Bullseye and one held/pinned etc, so I don't see why it should happen.
FWIW I followed s4.2 of the release notes ("start from pure Debian") meticulously
https://www.debian.org/releases/stable/amd64/release-notes/ch-upgrading.en.html#system-status
as well as running the commands kindly suggested by piorunz:
> On Wed 18 Aug 2021, at 23:33, piorunz <piorunz@gmx.com> wrote:
> [...]
> check if any packages are newer than repository (possibly foreign)
> apt-show-versions | grep newer
>
> check if any packages are foreign (not from Debian)
> aptitude search '~i(!~ODebian)'
There is also no explanation in term.log, syslog or dpkg.log for the second interruption:
--
Processing triggers for libapache2-mod-php7.4 (7.4.21-1+deb11u1) ...
[upgrade interrupted...]
W: APT had planned for dpkg to do more than it reported back (5014 vs 5047).
Affected packages: texlive-fonts-recommended:amd64 texlive-lang-greek:amd64 texlive-latex-base:amd64 texlive-latex-extra:amd64 texlive-latex-recommended:amd64 texlive-pictures:amd64 texlive-plain-generic:amd64 texlive-science:amd64
---
which occurs even if pitivi is removed before upgrading, and the warning doesn't appear in term.log either.
If anyone can shed further light, I would be interested, but it's not ultimately a roadblock to upgrading so possibly not worth worrying about.
Thanks,
Gareth
> ... There are two logs
> of the upgrade. /var/log/apt/history.log (and its predecessors) shows
> the command issued, followed by the packages affected, with the old
> and new version numbers in parentheses. /var/log/apt/term.log (and its
> predecessors) shows the various packages being unpacked and then set
> up. Again, there are old and new version numbers. And it's all timestamped.
>
> BTW I can't immediately see why you would have
> gir1.2-gst-plugins-bad-1.0 installed in buster (as a dependency),
> because the only package that depends on it is
> libgstreamer-plugins-bad1.0-dev, and I can't find anything that
> depends on that.
>
> > On Bullseye now:
> >
> > $ apt policy *gir*bad*
> > gir1.2-gst-plugins-bad-1.0:
> > Installed: 1.18.4-3
> > Candidate: 1.18.4-3
> > Version table:
> > *** 1.18.4-3 500
> > 500 http://deb.debian.org/debian bullseye/main amd64 Packages
> > 100 /var/lib/dpkg/status
> >
> > $ aptitude why gir1.2-gst-plugins-bad-1.0
> > i pitivi Depends gir1.2-gst-plugins-bad-1.0 (>= 1.18.0)
> >
> > (having installed pitivi again after removing it before upgrading to see what difference it made)
> >
> > I don't understand what the problem is/was -
>
> To summarise,
>
> /usr/lib/x86_64-linux-gnu/girepository-1.0/GstTranscoder-1.0.typelib included
> ↓↙
> ↓
>
> yes buster 0.999-1+b1 pitivi
> no bullseye 2020.09.2-3 pitivi
>
> no buster 1.14.4-1+deb10u2 gir1.2-gst-plugins-bad-1.0
> yes bullseye 1.18.4-3 gir1.2-gst-plugins-bad-1.0
>
> pitivi/bullseye depends on gir1.2-gst-plugins-bad-1.0 (>= 1.18.0) so
> it can't be installed while buster's gir1.2-gst-plugins-bad-1.0 is there.
>
> pitivi/buster must be removed before gir1.2-gst-plugins-bad-1.0/bullseye
> is installed, to avoid having two versions of that file.
>
> > shouldn't apt just have handled this?
>
> AFAIK, assuming that buster was up-to-date at the start of the
> dist-upgrade, and there were no "foreign" packages etc.
> (I don't know anything about apt-forktracer.)
>
> Cheers,
> David.
>
>
Reply to: