all messages for Guix-related lists mirrored at yhetil.org
 help / color / mirror / code / Atom feed
From: "Ludovic Courtès" <ludovic.courtes@inria.fr>
To: 50335@debbugs.gnu.org
Cc: philippe.swartvagher@inria.fr
Subject: bug#50335: Inheritance vs. ‘this-package’
Date: Thu, 02 Sep 2021 15:48:27 +0200	[thread overview]
Message-ID: <87r1e73x78.fsf@inria.fr> (raw)

Hi!

We stumbled upon an idiom for package arguments that, combined with
package inheritance, would not work as expected.  It goes like this
(taken from <https://gitlab.inria.fr/guix-hpc/guix-hpc/>):

--8<---------------cut here---------------start------------->8---
(define-public starpu
  (package
    ;; …
    (arguments
      (substitute-keyword-arguments (package-arguments starpu-1.2)
        ((#:configure-flags flags '())
         (match (assoc "mpi" (package-propagated-inputs this-package))
           (("mpi" mpi)
            (if (string=? (package-name mpi) "nmad")
                `(cons "--enable-nmad" ,flags)
                flags))))))
    (propagated-inputs `(("mpi" ,openmpi)))))

(define-public starpu+fxt
  (package
    (inherit starpu)
    ;; …
    (arguments
     (substitute-keyword-arguments (package-arguments starpu)
       ((#:configure-flags flags '())
        `(cons "--with-fxt" ,flags))))))
--8<---------------cut here---------------end--------------->8---

The first package, ‘starpu’, has its arguments depend on:

  (package-propagated-inputs this-package)

This is smart because if you inherit from it without changing the
‘arguments’ field (for instance by doing ‘--with-input=openmpi=nmad’),
it’ll adjust its configure flags based on the MPI package it actually
depends on.

However, this trick doesn’t work with the second package, ‘starpu-fxt’:
if you do:

  guix build starpu-fxt --with-input=openmpi=nmad

the ‘--enable-nmad’ configure flag will not be passed.  This is because
‘starpu-fxt’ calls:

  (package-arguments starpu)

which returns the original arguments of ‘starpu’.


Still here?  :-)

This is admittedly a pretty far-fetched example and there are ways to
work around it¹.  Still, I wonder if we could up with a linguistic
device (ah ha!) to address that.

We don’t want to (package-arguments this-package) here: that would
naturally lead to an infinite loop.  But we want something close to it,
in spirit: (package-arguments parent), where ‘parent’ is the package we
inherit from.

(package-arguments parent) could call the ‘arguments’ thunk of the
parent, but it would pass it the child as its argument.

Or something like that.

Thoughts?

Ludo’.

¹ https://gitlab.inria.fr/guix-hpc/guix-hpc/-/commit/761cfe97d5bb42dbbde538593cef93e803086e3b




             reply	other threads:[~2021-09-02 13:55 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-09-02 13:48 Ludovic Courtès [this message]
2023-03-04 15:15 ` bug#50335: Inheritance vs. ‘this-package’ Ludovic Courtès

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=87r1e73x78.fsf@inria.fr \
    --to=ludovic.courtes@inria.fr \
    --cc=50335@debbugs.gnu.org \
    --cc=philippe.swartvagher@inria.fr \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
Code repositories for project(s) associated with this external index

	https://git.savannah.gnu.org/cgit/guix.git

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.