Canonical-packages restoration.

Canonical-packages restoration.

[Mathieu Othacehe]

Hello,

With f30d84d32db0f4f6cb84e139868e1727a7dc0a51 and
dfc8ccbf5da96a67eb1cade499f0def21e7fdb02, I did remove most of the
"canonical-package" calls because they were breaking system
cross-compilation.

Now, I'd like to somehow restore them, using the new "let-system". My
idea is to define something like:

--8<---------------cut here---------------start------------->8---
(define (canonical-package* package)
  (let ((canonical
         (module-ref (resolve-interface '(gnu packages base))
                     'canonical-package)))
    (let-system (system target)
      (if target
          package
          (canonical package)))))
--8<---------------cut here---------------end--------------->8---

And then use for instance this way:

--8<---------------cut here---------------start------------->8---
diff --git a/gnu/system/locale.scm b/gnu/system/locale.scm
index 689d238d1a..fd50538e9a 100644
--- a/gnu/system/locale.scm
+++ b/gnu/system/locale.scm
@@ -106,7 +106,8 @@ of LIBC."
 
          ;; 'localedef' executes 'gzip' to access compressed locale sources.
          (setenv "PATH"
-                 (string-append #+gzip "/bin:" #+libc "/bin"))
+                 (string-append #+gzip "/bin:"
+                                #+(canonical-package* libc) "/bin"))
--8<---------------cut here---------------end--------------->8---

However, it seems that nesting a "let-system" inside "file-append" does
not work:

--8<---------------cut here---------------start------------->8---
(use-modules (guix))
(use-modules (gnu))

(run-with-store (open-connection)
  (mlet* %store-monad
      ((drv (lower-object
             (computed-file "computed"
                            #~(begin
                                (mkdir #$output)
                                (symlink #$(file-append
                                            (let-system (s t) glibc) "/bin")
                                         (string-append #$output "/ref"))))))
       (output -> (derivation->output-path drv)))
    (mbegin %store-monad
      (built-derivations (list drv))
      (return (format #t "~a~%" output)))))
--8<---------------cut here---------------end--------------->8---

and it prevents me from going further. Does this stuff make any sense?
If yes I can try to fix it, but I just want to be sure first :)

Thanks,

Mathieu

Re: Canonical-packages restoration.

[Jan Nieuwenhuizen]

Mathieu Othacehe writes:

> With f30d84d32db0f4f6cb84e139868e1727a7dc0a51 and
> dfc8ccbf5da96a67eb1cade499f0def21e7fdb02, I did remove most of the
> "canonical-package" calls because they were breaking system
> cross-compilation.
>
> Now, I'd like to somehow restore them

Just a quick question: why?; would that reduce a system's closure size?

Janneke

-- 
Jan Nieuwenhuizen <janneke@gnu.org> | GNU LilyPond http://lilypond.org
Freelance IT http://JoyofSource.com | Avatar® http://AvatarAcademy.com

Re: Canonical-packages restoration.

[Mathieu Othacehe]

Hey Jan,

> Just a quick question: why?; would that reduce a system's closure size?

Yes mostly, even if the gains are not huge (~100MiB). However, I feel
like its easier the tackle the system closure size issue if we get rid
of the "noise".

Thanks,

Mathieu

Re: Canonical-packages restoration.

[Ludovic Courtès]

Hi,

Mathieu Othacehe <othacehe@gnu.org> skribis:

> With f30d84d32db0f4f6cb84e139868e1727a7dc0a51 and
> dfc8ccbf5da96a67eb1cade499f0def21e7fdb02, I did remove most of the
> "canonical-package" calls because they were breaking system
> cross-compilation.
>
> Now, I'd like to somehow restore them, using the new "let-system". 

What if, instead, we removed those “canonical” packages entirely from
the reference graph?  Do you think that’s an option?

> My idea is to define something like:
>
> (define (canonical-package* package)
>   (let ((canonical
>          (module-ref (resolve-interface '(gnu packages base))
>                      'canonical-package)))
>     (let-system (system target)
>       (if target
>           package
>           (canonical package)))))

Aren’t there pieces of code that expect these things to be <package>
records?

> However, it seems that nesting a "let-system" inside "file-append" does
> not work:
>
> (use-modules (guix))
> (use-modules (gnu))
>
> (run-with-store (open-connection)
>   (mlet* %store-monad
>       ((drv (lower-object
>              (computed-file "computed"
>                             #~(begin
>                                 (mkdir #$output)
>                                 (symlink #$(file-append
>                                             (let-system (s t) glibc) "/bin")
>                                          (string-append #$output "/ref"))))))
>        (output -> (derivation->output-path drv)))
>     (mbegin %store-monad
>       (built-derivations (list drv))
>       (return (format #t "~a~%" output)))))
>
> and it prevents me from going further. Does this stuff make any sense?
> If yes I can try to fix it, but I just want to be sure first :)

Bah yes, that’s why initially I didn’t push ‘let-system’:

  https://issues.guix.gnu.org/29296#4

Perhaps we could avoid the expander = #f special case.

Thanks,
Ludo’.

Re: Canonical-packages restoration.

[Mathieu Othacehe]

[-- Attachment #1: Type: text/plain, Size: 1512 bytes --]


Hey Ludo!

> What if, instead, we removed those “canonical” packages entirely from
> the reference graph?  Do you think that’s an option?

It seems to be a better option! So, as I did remove most of the explicit
references to 'canonical-packages' the only references left are
implicit.

A good example is "isc-dhcp", that refers to implicit canonical inputs
"coreutils" and "sed" in the "dhclient-script". The attached patch fixes
it. I could do the same for a few other packages and it should fix the
issue.

The problem is that approach is not very long-term. Maybe we should take
a step back. Would it be an option to add all the canonical-packages as
disallowed-references? Or could we use implicit inputs that are not
rooted in the bootstrap chain?

>> My idea is to define something like:
>>
>> (define (canonical-package* package)
>>   (let ((canonical
>>          (module-ref (resolve-interface '(gnu packages base))
>>                      'canonical-package)))
>>     (let-system (system target)
>>       (if target
>>           package
>>           (canonical package)))))
>
> Aren’t there pieces of code that expect these things to be <package>
> records?

Yes, but I was planning to change that, not a good idea after all.

> Bah yes, that’s why initially I didn’t push ‘let-system’:
>
>   https://issues.guix.gnu.org/29296#4
>
> Perhaps we could avoid the expander = #f special case.

Ok, I will try to fix it then.

Thanks,

Mathieu

[-- Attachment #2: patch --]
[-- Type: application/octet-stream, Size: 1593 bytes --]

diff --git a/gnu/packages/admin.scm b/gnu/packages/admin.scm
index 77981e520a..537b162ba8 100644
--- a/gnu/packages/admin.scm
+++ b/gnu/packages/admin.scm
@@ -1004,10 +1004,10 @@ connection alive.")
                ;; if finds all the programs it needs.
                (let* ((out       (assoc-ref outputs "out"))
                       (libexec   (string-append out "/libexec"))
-                      (coreutils (assoc-ref inputs "coreutils"))
+                      (coreutils (assoc-ref inputs "coreutils*"))
                       (inetutils (assoc-ref inputs "inetutils"))
                       (net-tools (assoc-ref inputs "net-tools"))
-                      (sed       (assoc-ref inputs "sed")))
+                      (sed       (assoc-ref inputs "sed*")))
                  (substitute* "client/scripts/linux"
                    (("/sbin/ip")
                     (string-append (assoc-ref inputs "iproute")
@@ -1047,12 +1047,8 @@ connection alive.")
                      (base32
                       "0vws0zzb39mkphj4hhjrgfj9dzw951lc4pfa6pqg5ll5ma51mbsr"))))
 
-                ;; When cross-compiling, we need the cross Coreutils and sed.
-                ;; Otherwise just use those from %FINAL-INPUTS.
-                ,@(if (%current-target-system)
-                      `(("coreutils" ,coreutils)
-                        ("sed" ,sed))
-                      '())))
+                ("coreutils*" ,coreutils)
+                ("sed*" ,sed)))
 
       (home-page "https://www.isc.org/products/DHCP/")
       (synopsis "Dynamic Host Configuration Protocol (DHCP) tools")

Re: Canonical-packages restoration.

[Ludovic Courtès]

Hi,

Mathieu Othacehe <othacehe@gnu.org> skribis:

>> What if, instead, we removed those “canonical” packages entirely from
>> the reference graph?  Do you think that’s an option?
>
> It seems to be a better option! So, as I did remove most of the explicit
> references to 'canonical-packages' the only references left are
> implicit.
>
> A good example is "isc-dhcp", that refers to implicit canonical inputs
> "coreutils" and "sed" in the "dhclient-script". The attached patch fixes
> it. I could do the same for a few other packages and it should fix the
> issue.

Yes, sounds good to me!

> The problem is that approach is not very long-term. Maybe we should take
> a step back. Would it be an option to add all the canonical-packages as
> disallowed-references?

Hmm yes, we could try that, but there’ll probably be many changes to
make.

> Or could we use implicit inputs that are not rooted in the bootstrap
> chain?

The problem is not that they’re rooted in the bootstrap chain (they
necessary build-depend on the bootstrap chain, but they’re guaranteed
not to have references to it).  The problem is that we have two copies
of all these: coreutils-final vs. coreutils, guile-final vs. guile, etc.

As things are, it seems hard to avoid.

>> Bah yes, that’s why initially I didn’t push ‘let-system’:
>>
>>   https://issues.guix.gnu.org/29296#4
>>
>> Perhaps we could avoid the expander = #f special case.
>
> Ok, I will try to fix it then.

No rush though, we can do that anytime.

Thanks,
Ludo’.