GNU bug report logs - #64861
[PATCH] guix: Add lint check for guix as propagated-input

Previous Next

To reply to this bug, email your comments to 64861 AT debbugs.gnu.org.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Karl Hallsby <karl <at> hallsby.com>
To: guix-patches <at> gnu.org
Cc: Karl Hallsby <karl <at> hallsby.com>
Subject: [PATCH] guix: Add lint check for guix as propagated-input
Date: Tue, 25 Jul 2023 13:04:43 -0500

* guix/lint.scm (new check): Run lint check warning user if the provided
package uses guix as a propagated-input.

Passing guix as a propagated-input is problematic when users install a package
into their profiles. This can cause the guix propagated by the package to be
used in preference of the real one in $HOME/.config/. It was first noticed on
IRC with https://logs.guix.gnu.org/guix/2023-07-22.log#044534, and reproduced
with a different package https://logs.guix.gnu.org/guix/2023-07-25.log#054737.
---
 guix/lint.scm | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/guix/lint.scm b/guix/lint.scm
index d173563e51..5fae34ca22 100644
--- a/guix/lint.scm
+++ b/guix/lint.scm
@@ -575,6 +575,16 @@ (define (check-input-labels package)
                 (inputs ,package-inputs)
                 (propagated-inputs ,package-propagated-inputs))))
 
+(define (check-guix-propagated-inputs package)
+  (if (and (not (null? (package-propagated-inputs package)))
+           (not (memq (@ (gnu packages package-management) guix)
+                      (package-propagated-inputs package))))
+      (list
+       (make-warning package
+                     (G_ "are you sure guix should be a propagated-input?")
+                     #:field 'propagated-inputs))
+      '()))
+
 (define (report-wrap-program-error package wrapper-name)
   "Warn that \"bash-minimal\" is missing from 'inputs', while WRAPPER-NAME
 requires it."
@@ -1884,6 +1894,10 @@ (define %local-checkers
      (name        'input-labels)
      (description "Identify input labels that do not match package names")
      (check       check-input-labels))
+   (lint-checker
+     (name        'warn-guix-propagated-inputs)
+     (description "Emit warning if guix package is propagated-input")
+     (check       check-guix-propagated-inputs))
    (lint-checker
      (name        'wrapper-inputs)
      (description "Make sure 'wrap-program' can finds its interpreter.")

base-commit: 9ff1e7652a407b88a3eeeab6a67261f6fee40807
-- 
2.40.1

Message #8 received at 64861 <at> debbugs.gnu.org (full text, mbox):

From: Christopher Baines <mail <at> cbaines.net>
To: Karl Hallsby <karl <at> hallsby.com>
Cc: 64861 <at> debbugs.gnu.org
Subject: Re: [bug#64861] [PATCH] guix: Add lint check for guix as
 propagated-input
Date: Tue, 25 Jul 2023 19:26:01 +0100

[Message part 1 (text/plain, inline)]

Karl Hallsby <karl <at> hallsby.com> writes:

> * guix/lint.scm (new check): Run lint check warning user if the provided
> package uses guix as a propagated-input.
>
> Passing guix as a propagated-input is problematic when users install a package
> into their profiles. This can cause the guix propagated by the package to be
> used in preference of the real one in $HOME/.config/. It was first noticed on
> IRC with https://logs.guix.gnu.org/guix/2023-07-22.log#044534, and reproduced
> with a different package https://logs.guix.gnu.org/guix/2023-07-25.log#054737.
> ---
>  guix/lint.scm | 14 ++++++++++++++
>  1 file changed, 14 insertions(+)

I'm not sure this lint warning is helpful as in both cases mentioned in
the IRC links you provided (guile-imanifest and guix-data-service),
these packages provide Guile modules that depend on Guix, and therefore
the guix package is expected in their propagated inputs.

If you remove guix or make it a input instead of a propagated input,
then this breaks using the package, for example, with this change:

  modified   gnu/packages/guile-xyz.scm
  @@ -2098,8 +2098,10 @@ (define-public guile-imanifest
         (build-system guile-build-system)
         (native-inputs
          (list guile-3.0))
  +      (inputs
  +       (list guix))
         (propagated-inputs
  -       (list guile-readline guile-colorized guix))
  +       (list guile-readline guile-colorized))
         (home-page "https://sr.ht/~brown121407/guile-imanifest")
         (synopsis "Interactive Guix manifests")
         (description "This package provides functions to generate Guix manifests

Using guile-imanifest breaks:

  → ./pre-inst-env guix environment --pure --ad-hoc guile guile-imanifest -- guile -c "(use-modules (imanifest))"
  Backtrace:
    ...

  ice-9/boot-9.scm:3330:6: In procedure resolve-interface:
  no code for module (guix profiles)


I think some other approach is needed to avoid people having problems
with the guix package appearing in their users profile.

[signature.asc (application/pgp-signature, inline)]

Message #11 received at 64861 <at> debbugs.gnu.org (full text, mbox):

From: "Karl G. Hallsby" <karl <at> hallsby.com>
To: Christopher Baines <mail <at> cbaines.net>
Cc: 64861 <at> debbugs.gnu.org
Subject: Re: [bug#64861] [PATCH] guix: Add lint check for guix as
 propagated-input
Date: Tue, 25 Jul 2023 13:41:55 -0500

Christopher Baines <mail <at> cbaines.net> writes:

> [[PGP Signed Part:Undecided]]
>
> Karl Hallsby <karl <at> hallsby.com> writes:
>
>> * guix/lint.scm (new check): Run lint check warning user if the provided
>> package uses guix as a propagated-input.
>>
>> Passing guix as a propagated-input is problematic when users install a package
>> into their profiles. This can cause the guix propagated by the package to be
>> used in preference of the real one in $HOME/.config/. It was first noticed on
>> IRC with https://logs.guix.gnu.org/guix/2023-07-22.log#044534, and reproduced
>> with a different package https://logs.guix.gnu.org/guix/2023-07-25.log#054737.
>> ---
>>  guix/lint.scm | 14 ++++++++++++++
>>  1 file changed, 14 insertions(+)
>
> I'm not sure this lint warning is helpful as in both cases mentioned in
> the IRC links you provided (guile-imanifest and guix-data-service),
> these packages provide Guile modules that depend on Guix, and therefore
> the guix package is expected in their propagated inputs.
>
> If you remove guix or make it a input instead of a propagated input,
> then this breaks using the package, for example, with this change:
>
>   modified   gnu/packages/guile-xyz.scm
>   @@ -2098,8 +2098,10 @@ (define-public guile-imanifest
>          (build-system guile-build-system)
>          (native-inputs
>           (list guile-3.0))
>   +      (inputs
>   +       (list guix))
>          (propagated-inputs
>   -       (list guile-readline guile-colorized guix))
>   +       (list guile-readline guile-colorized))
>          (home-page "https://sr.ht/~brown121407/guile-imanifest")
>          (synopsis "Interactive Guix manifests")
>          (description "This package provides functions to generate Guix manifests
>
> Using guile-imanifest breaks:
>
>   → ./pre-inst-env guix environment --pure --ad-hoc guile guile-imanifest -- guile -c "(use-modules (imanifest))"
>   Backtrace:
>     ...
>
>   ice-9/boot-9.scm:3330:6: In procedure resolve-interface:
>   no code for module (guix profiles)
>
>
> I think some other approach is needed to avoid people having problems
> with the guix package appearing in their users profile.
>
> [[End of PGP Signed Part]]

That makes sense. Though, this lint is just a warning, so if nothing
else, it should make people think if guix really needs to be propagated.
If this lint is not the solution, then a way to mark packages that are
not recommended to be installed like this would be nice.

Message #14 received at 64861 <at> debbugs.gnu.org (full text, mbox):

From: Ludovic Courtès <ludo <at> gnu.org>
To: Christopher Baines <mail <at> cbaines.net>
Cc: Karl Hallsby <karl <at> hallsby.com>, 64861 <at> debbugs.gnu.org
Subject: Re: bug#64861: [PATCH] guix: Add lint check for guix as
 propagated-input
Date: Sun, 20 Aug 2023 22:58:10 +0200

Hi,

Christopher Baines <mail <at> cbaines.net> skribis:

> Using guile-imanifest breaks:
>
>   → ./pre-inst-env guix environment --pure --ad-hoc guile guile-imanifest -- guile -c "(use-modules (imanifest))"
>   Backtrace:
>     ...
>
>   ice-9/boot-9.scm:3330:6: In procedure resolve-interface:
>   no code for module (guix profiles)

Maybe ‘guile-imanifest’ should be made a Guix extension, which Guix
searches for in $GUIX_EXTENSIONS_PATH?

An example of that is ‘guix-modules’.

Ludo’.

Message #17 received at 64861 <at> debbugs.gnu.org (full text, mbox):

From: Simon Tournier <zimon.toutoune <at> gmail.com>
To: Ludovic Courtès <ludo <at> gnu.org>, Christopher Baines
 <mail <at> cbaines.net>
Cc: Karl Hallsby <karl <at> hallsby.com>, 64861 <at> debbugs.gnu.org
Subject: Re: [bug#64861] [PATCH] guix: Add lint check for guix as
 propagated-input
Date: Thu, 07 Sep 2023 16:22:14 +0200

Hi,

On Sun, 20 Aug 2023 at 22:58, Ludovic Courtès <ludo <at> gnu.org> wrote:

> Maybe ‘guile-imanifest’ should be made a Guix extension, which Guix
> searches for in $GUIX_EXTENSIONS_PATH?

And probably renamed ’guix-imanifest’?

Back to the submission, I think that the propagation of the package guix
means something is wrong.  From my point of view, there is two cases:

 1. The package uses the stable library API and thus it makes sense to
    rely on the package ’guix’.  That’s the case for ’gwl’,
    ’guix-data-service’ for example.

 2. The aim of package is to collaborate with the current Guix and thus
    there is no point to have the package ’guix’ as inputs.  Instead,
    the package must rely on GUIX_EXTENSIONS_PATH.  That’s the case for
    ’guix-modules’ or ’guile-imanifest’ (so that needs a fix ;-))

Therefore, I think this new checker makes sense.  WDYT?


About #1, IMHO, this is expected:

--8<---------------cut here---------------start------------->8---
$ guix shell -C gwl -- guix --version
guix shell: error: guix: command not found

        (define-public gwl
            (inputs
        [...]
               (list guix
--8<---------------cut here---------------end--------------->8---

and this is not expected:

--8<---------------cut here---------------start------------->8---
$ guix shell -C guix-data-service -- guix --version
guix (GNU Guix) 1.4.0-10.4dfdd82
Copyright (C) 2023 the Guix authors
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.

        (define-public guix-data-service
        [...]
              (propagated-inputs
               (list guix
--8<---------------cut here---------------end--------------->8---


Cheers,
simon

Message #20 received at 64861 <at> debbugs.gnu.org (full text, mbox):

From: "Karl G. Hallsby" <karl <at> hallsby.com>
To: 64861 <at> debbugs.gnu.org
Cc: Ludovic Courtès <ludo <at> gnu.org>,
 Christopher Baines <mail <at> cbaines.net>, Karl Hallsby <karl <at> hallsby.com>,
 Simon Tournier <zimon.toutoune <at> gmail.com>
Subject: Re: [bug#64861] [PATCH] guix: Add lint check for guix as
 propagated-input
Date: Sat, 11 Jan 2025 15:54:59 -0600

Hi all,

What is the state of this patch's proposal? Is this warning check for
guix in propagated-inputs something we still want to warn about?


I know we have not talked about it in a while. I am going through all of
my contributions to Guix & Guix-based projects and
closing/merging/resolving them.

--
Raven Hallsby

Message #23 received at 64861 <at> debbugs.gnu.org (full text, mbox):

From: Ludovic Courtès <ludo <at> gnu.org>
To: "Karl G. Hallsby" <karl <at> hallsby.com>
Cc: Christopher Baines <mail <at> cbaines.net>, 64861 <at> debbugs.gnu.org,
 Simon Tournier <zimon.toutoune <at> gmail.com>
Subject: Re: [bug#64861] [PATCH] guix: Add lint check for guix as
 propagated-input
Date: Thu, 16 Jan 2025 09:07:37 +0100

Hi,

"Karl G. Hallsby" <karl <at> hallsby.com> skribis:

> What is the state of this patch's proposal? Is this warning check for
> guix in propagated-inputs something we still want to warn about?

I second Christopher here: there are legitimate cases for propagating
Guix (Scheme libraries that themselves use Guix modules) so I would lean
towards not including this lint check.

Closing?

Thanks,
Ludo’.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.