Re: Trying to replicate guix packages list on another machine

[zimoun <zimon.toutoune@gmail.com>]

Hi,

On Tue, 28 Jun 2022 at 08:30, Emmanuel Medernach <Emmanuel.Medernach@iphc.cnrs.fr> wrote:

> But glibc is 2.31 instead of 2.33

On ’Machine-A’, if you run,

      machine-A$ guix package --list-generations

then I guess ’glibc’ is not installed in the same transaction
(generation) as all the others.

You can open the internals. :-)  For instance, on ’Machine-A’, give a
look at the internal file ’~/.guix-profile/manifest’.  You should see
’glibc’ and its provenance (commit) and I bet it is different from the
other provenance.

Other said, the profile on ’Machine-A’ is populated by various states
(revision).  I guess glibc is at 2.31 at one state and then become 2.33
at another state.  The command “guix package --export-channels” makes
approximation to pick only one state when the profile is composed by
several ones.

Then it becomes hard, nor impossible, to replicate this multi-state
profile on another machine.  “Nor impossible” because it is possible to
use inferiors [1] although I would not recommend going this way.


1: <https://guix.gnu.org/manual/devel/en/guix.html#Inferiors>




> Thank you, yes packages installed are always evolving. I will now 
> promote to use manifests.

Well, from my point of view, the best practise is:

 1. track channels.scm from “guix pull -f channels”
 2. track manifest.scm

then

  guix time-machine -C channels.scm -- <cmd> -m manifest.scm

works independently of the time and of the machine; or it should be. ;-)


Cheers,
simon