Dependence on an old version of a package.

Dependence on an old version of a package.

[Philippe Veber]

Hi all and sorry if it's a FAQ !

I would like to write a package definition for a program that expects
openmpi < 4.0.0, which is present in earlier versions of guix. If I'm not
mistaken I have to use inferiors in that case, and I tried this :

(use-modules (gnu packages mpi)
    (guix packages)
             (guix download)
             (guix build-system gnu)
             (guix licenses)
    (guix build utils)
    (guix gexp)
    (guix inferior)
    (guix channels))

(define mychannels
  (list (channel
         (name 'guix)
         (url "https://git.savannah.gnu.org/git/guix.git")
         ;; Last commit that still has openmpi < 4.0.0
         (commit
          "1482531cd7d0f1f7e565207240affb68dd822e17"))))

(define myinferior
  (inferior-for-channels mychannels))

(define myopenmpi
  (first (lookup-inferior-packages myinferior "openmpi")))

(package
 (name "coevol")
 (version "1.5b")
 (source (origin
 (method url-fetch)
 (uri (string-append "
https://github.com/bayesiancook/coevol/archive/refs/tags/v" version
".tar.gz"))
 (sha256 (base32 "112kjjxwlm9m3f3wks3q86gp561l8s5gsw17zrffrhljixd8qyyp"))))
 (build-system gnu-build-system)
 (arguments
  (list
   #:phases #~(modify-phases %standard-phases
    (delete 'configure)
    (add-after 'unpack 'chdir (lambda _ (chdir "sources") #t)))))
 (inputs (list myopenmpi))
 (synopsis "Correlated evolution of substitution rates and quantitative
traits")
 (description "...")
 (home-page "
https://megasun.bch.umontreal.ca/People/lartillot/www/downloadcoevol.html")
 (license gpl3+))

My problem is that when invoking guix I get this:

$ guix package --install-from-file=coevol.scm
Mise à jour du canal « guix » depuis le dépôt Git «
https://git.savannah.gnu.org/git/guix.git »...
Computing Guix derivation for 'x86_64-linux'... \fetching path
`/gnu/store/ir3092v7657h6g4g2vlsw3zrli3rndb3-zlib-1.2.11.tar.gz'...
substitute: updating substitutes from 'https://ci.guix.gnu.org'... 100.0%
substitute: updating substitutes from 'https://ci.guix.gnu.org'... 100.0%

and a spinning character for more than 10 hours. I can see a 'guix
substitute' process working and using quite some memory after a while (3-4
GB). My question is this: should I just be patient and wait for it to
complete or I'm doing this wrong?

Thanks for any hint!
  Philippe.

Re: Dependence on an old version of a package.

[Simon Tournier]

Hi,

On Sun, 10 Dec 2023 at 09:18, Philippe Veber <philippe.veber@gmail.com> wrote:

> (define mychannels
>   (list (channel
>          (name 'guix)
>          (url "https://git.savannah.gnu.org/git/guix.git")
>          ;; Last commit that still has openmpi < 4.0.0
>          (commit
>           "1482531cd7d0f1f7e565207240affb68dd822e17"))))

--8<---------------cut here---------------start------------->8---
$ guix time-machine --commit=1482531cd7d0f1f7e565207240affb68dd822e17 -- help
guix time-machine: error: cannot travel past commit `6298c3ffd965' from May 1st, 2019
--8<---------------cut here---------------end--------------->8---


>      My question is this: should I just be patient and wait for it to
> complete or I'm doing this wrong?

You are trying to time-travel to something before the introduction of
time-travel.  Hence the unexpected behaviour, I guess.

Thanks for reporting that something is missing.  The inferior should
stop and error out.  It does not because ’cached-channel-instance’ is
currently called with the argument ’#:validate-channels’ set to the
default (const #t).

Currently, the verification is only done in
guix/scripts/time-machine.scm.  Well, this issue seems being part of
bug#65788 [1].


1: bug#65788: poor information when updating using “guix time-machine”
Simon Tournier <zimon.toutoune@gmail.com>
Wed, 06 Sep 2023 18:57:38 +0200
id:87pm2vme7x.fsf@gmail.com
https://issues.guix.gnu.org/65788
https://issues.guix.gnu.org/msgid/87pm2vme7x.fsf@gmail.com
https://yhetil.org/guix/87pm2vme7x.fsf@gmail.com

Cheers,
simon

Re: Dependence on an old version of a package.

[Philippe Veber]

Thanks a lot for having a look at it!

So I'm guessing that even when the bug is corrected, it won't be possible
to depend on a commit prior to May 1st 2019, is that right?

Le mar. 19 déc. 2023 à 15:42, Simon Tournier <zimon.toutoune@gmail.com> a
écrit :

> Hi,
>
> On Sun, 10 Dec 2023 at 09:18, Philippe Veber <philippe.veber@gmail.com>
> wrote:
>
> > (define mychannels
> >   (list (channel
> >          (name 'guix)
> >          (url "https://git.savannah.gnu.org/git/guix.git")
> >          ;; Last commit that still has openmpi < 4.0.0
> >          (commit
> >           "1482531cd7d0f1f7e565207240affb68dd822e17"))))
>
> --8<---------------cut here---------------start------------->8---
> $ guix time-machine --commit=1482531cd7d0f1f7e565207240affb68dd822e17 --
> help
> guix time-machine: error: cannot travel past commit `6298c3ffd965' from
> May 1st, 2019
> --8<---------------cut here---------------end--------------->8---
>
>
> >      My question is this: should I just be patient and wait for it to
> > complete or I'm doing this wrong?
>
> You are trying to time-travel to something before the introduction of
> time-travel.  Hence the unexpected behaviour, I guess.
>
> Thanks for reporting that something is missing.  The inferior should
> stop and error out.  It does not because ’cached-channel-instance’ is
> currently called with the argument ’#:validate-channels’ set to the
> default (const #t).
>
> Currently, the verification is only done in
> guix/scripts/time-machine.scm.  Well, this issue seems being part of
> bug#65788 [1].
>
>
> 1: bug#65788: poor information when updating using “guix time-machine”
> Simon Tournier <zimon.toutoune@gmail.com>
> Wed, 06 Sep 2023 18:57:38 +0200
> id:87pm2vme7x.fsf@gmail.com
> https://issues.guix.gnu.org/65788
> https://issues.guix.gnu.org/msgid/87pm2vme7x.fsf@gmail.com
> https://yhetil.org/guix/87pm2vme7x.fsf@gmail.com
>
> Cheers,
> simon
>

Re: Dependence on an old version of a package.

[Simon Tournier]

Hi Philippe,

Sorry for the late reply.

On Thu, 21 Dec 2023 at 11:31, Philippe Veber <philippe.veber@gmail.com> wrote:

> So I'm guessing that even when the bug is corrected, it won't be possible
> to depend on a commit prior to May 1st 2019, is that right?

Yes, it is not possible to travel back before the invention of the
DeLorean. ;-)  It is not possible to depend on a commit older the
implementation of inferiors; roughly May 2019.

Cheers,
simon

Re: Dependence on an old version of a package.

[Philippe Veber]

No worries, and thanks for your answer!

Le ven. 12 janv. 2024 à 13:51, Simon Tournier <zimon.toutoune@gmail.com> a
écrit :

> Hi Philippe,
>
> Sorry for the late reply.
>
> On Thu, 21 Dec 2023 at 11:31, Philippe Veber <philippe.veber@gmail.com>
> wrote:
>
> > So I'm guessing that even when the bug is corrected, it won't be possible
> > to depend on a commit prior to May 1st 2019, is that right?
>
> Yes, it is not possible to travel back before the invention of the
> DeLorean. ;-)  It is not possible to depend on a commit older the
> implementation of inferiors; roughly May 2019.
>
> Cheers,
> simon
>