unofficial mirror of guix-devel@gnu.org 
 help / color / mirror / code / Atom feed
From: ludo@gnu.org (Ludovic Courtès)
To: David Thompson <dthompson2@worcester.edu>
Cc: Guix-devel <guix-devel@gnu.org>,
	bug-guix@gnu.org, Federico Beffa <beffa@ieee.org>
Subject: Re: guix environment
Date: Tue, 03 Feb 2015 21:29:16 +0100	[thread overview]
Message-ID: <87k2zypwmb.fsf@inria.fr> (raw)
In-Reply-To: 87lhkldwfs.fsf@izanagi.i-did-not-set--mail-host-address--so-tickle-me

David Thompson <dthompson2@worcester.edu> skribis:

> Ludovic Courtès <ludo@gnu.org> writes:
>
>> On closer inspection, I see two issues:
>>
>>   (define (packages->transitive-inputs packages)
>>     "Return a list of the transitive inputs for all PACKAGES."
>>     (define (transitive-inputs package)
>>       (filter-map (match-lambda
>>                    ((_ (? package? package)) package)
>>                    (_ #f))    ; <---- !
>>                   (bag-transitive-inputs
>>                    (package->bag package))))
>>     (delete-duplicates
>>      (append-map transitive-inputs packages)))
>>
>> Here only inputs of the form ("foo" PKG) are considered; things like
>> ("glib" ,glib "bin") are discarded.
>>
>>   (define (for-each-search-path proc inputs derivations pure?)
>>     (let ((paths (map derivation->output-path derivations))) ; <-- !
>>       [...]
>>
>> Above, ‘derivation->output-path’ considers only the “out” output,
>> ignoring others if they are needed.
>
> Here's a patch.  WDYT?
>
>
> From 9609806fb78557d74cf5b3fb47802898ef9d1ecf Mon Sep 17 00:00:00 2001
> From: David Thompson <dthompson2@worcester.edu>
> Date: Thu, 29 Jan 2015 17:53:17 -0500
> Subject: [PATCH] guix: environment: Consider all package outputs.
>
> * guix/scripts/environment.scm (for-each-search-path): Iterate over all
>   derivation output paths.
>   (packages->transitive-inputs): Process inputs that specify an output, too.

[...]

> --- a/guix/scripts/environment.scm
> +++ b/guix/scripts/environment.scm
> @@ -40,7 +40,12 @@
>  Use the output paths of DERIVATIONS to build each search path.  When PURE? is
>  #t, the existing search path value is ignored.  Otherwise, the existing search
>  path value is appended."
> -  (let ((paths (map derivation->output-path derivations)))
> +  (let ((paths (append-map (lambda (drv)
> +                             (map (match-lambda
> +                                   ((_ . output)
> +                                    (derivation-output-path output)))
> +                                  (derivation-outputs drv)))
> +                           derivations)))
>      (for-each (match-lambda
>                 (($ <search-path-specification>
>                     variable directories separator)
> @@ -177,7 +182,9 @@ packages."
>    "Return a list of the transitive inputs for all PACKAGES."
>    (define (transitive-inputs package)
>      (filter-map (match-lambda
> -                 ((_ (? package? package)) package)
> +                 ((or (_ (? package? package))
> +                      (_ (? package? package) _))
> +                  package)
>                   (_ #f))

LGTM, please push!

There’s another problem, though.  When a dependency is a multiple-output
package, all its outputs are added to the environment, because
‘package->transitive-inputs’ discards the information of which output is
needed.

So for instance, both the ‘out’ and the ‘debug’ output of Coreutils end
up being downloaded and added to the environment, even though only ‘out’
is an input.

Now, the problem is that ‘build-derivations’ can only build *all* the
outputs of the given derivation.  This could be worked around either:

  1. by creating a “sink” derivation, for instance with
     ‘profile-derivation’, that could refer precisely to the output(s)
     needed; not ideal.

  2. by using (build-things (list "/the/output/path")) and resorting to
     ‘build-derivations’ only if the ‘build-things’ call did nothing
     (when passed a non-.drv store item, ‘build-things’ tries to
     substitute and does nothing if that fails.)

Thoughts?

Ludo’.

  reply	other threads:[~2015-02-03 20:29 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-01-20 17:21 guix environment Federico Beffa
2015-01-20 21:43 ` Ludovic Courtès
2015-01-20 21:56   ` David Thompson
2015-01-29 23:00   ` David Thompson
2015-02-03 20:29     ` Ludovic Courtès [this message]
2015-02-03 21:15       ` David Thompson
2015-02-08 18:22       ` bug#19641: " Ludovic Courtès
  -- strict thread matches above, loose matches on Subject: below --
2015-06-15  6:55 Pjotr Prins
2015-06-15  7:14 ` Ricardo Wurmus
2015-06-15 13:10 ` Thompson, David
2015-06-15 19:10   ` Pjotr Prins
2015-06-15 19:22     ` Thompson, David
2015-06-16  5:52       ` Pjotr Prins
2015-06-16  7:59         ` Ludovic Courtès
2015-06-16 10:04           ` Pjotr Prins
2015-06-16 11:23             ` Ludovic Courtès
2015-06-16 14:39               ` Pjotr Prins
2015-06-15 13:27 ` Claes Wallin (韋嘉誠)
2015-06-15 20:09   ` Ludovic Courtès
2015-06-16  5:06     ` Claes Wallin (韋嘉誠)

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

  List information: https://guix.gnu.org/

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

  git send-email \
    --in-reply-to=87k2zypwmb.fsf@inria.fr \
    --to=ludo@gnu.org \
    --cc=beffa@ieee.org \
    --cc=bug-guix@gnu.org \
    --cc=dthompson2@worcester.edu \
    --cc=guix-devel@gnu.org \
    /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 public inbox

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

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).