[bug#55227,core-updates,2/8] commencement: Add stage0-posix.

Thiago Jung Bauermann writes:

Hello Thiago,

> This is great work!

Thanks!

> This is not an area I have experience in, so I didn't read it closely,
> at least as of yet. I have only one minor comment:

>> diff --git a/gnu/packages/commencement.scm b/gnu/packages/commencement.scm
>> index 663ffacd56..24e1e6ec8c 100644
>> --- a/gnu/packages/commencement.scm
>> +++ b/gnu/packages/commencement.scm
[..]
>> +                  (stage0-cpu
>> +                   (cond
>> +                    ((or (string-prefix? "x86_64-linux" target)
>
> IMHO the code is clearer and less error prone if the (guix utils)
> convenience functions are used to test the target rather than directly
> doing string pattern matching.
>
> In this case, ‘target-x86-64?’ is the one. There's no need to even pass
> an argument to it since the default is what you need.
>
> The check for Linux could be done separately in the body of the builder
> with ‘target-linux?’. What do you think?
>
>> +                         (string-prefix? "i686-linux" target))
>
> Here it would be ‘target-x86-32?’.
>
>> +                     "x86")
>> +                    ((or (string-prefix? "aarch64-linux" target)
>> +                         (string-prefix? "armhf-linux" target))
>
> This is a good case in point. There's a bug above: “armhf-linux” matches
> 32-bit ARM when building natively because that's what's in
> %current-target, but when cross-building %current-target-system will
> contain “arm-unknown-linux-gnueabihf” and this branch won't be taken.
>
> Using ‘target-arm32?’ avoids the bug.

Nice catch and good suggestion.  Find a new version attached.

Greetings
Janneke

Jan Nieuwenhuizen <janneke@gnu.org> skribis:

>>From f013e88563f73a7513faa330c79eef2f653daf28 Mon Sep 17 00:00:00 2001
> From: "Jan (janneke) Nieuwenhuizen" <janneke@gnu.org>
> Date: Thu, 31 Dec 2020 18:25:21 +0100
> Subject: [PATCH core-updates v2 2/8] commencement: Add stage0-posix.
>
> Stage0-posix is a skeleton for bootstrapping all of Stage0 for POSIX systems.
> On x86-linux, from the 357-byte hex0-seed binary from the bootstrap-seeds, it
> builds hex0, kaem, hex1, catm, hex2, M0, cc_x86, M1, M2,
> get_machine, (mescc-tools), and M2-Planet.
>
> * gnu/packages/commencement.scm (stage0-posix): New variable.

[...]

> +      (name "stage0-posix")
> +      (version "1.4")
> +      (source (bootstrap-origin
> +               (origin
> +                 (method url-fetch)
> +                 (uri (string-append
> +                       "https://lilypond.org/janneke/guix/20220502/"
> +                       "stage0-posix-" version ".tar.gz"))

Same comment as earlier regarding ‘bootstrap-origin’ and the URL.

> +      (supported-systems '("i686-linux" "x86_64-linux"
> +                           "arm-linux" "aarch64-linux"
> +                           "riscv64-linux"))

Woow.  :-)

s/arm-linux/armhf-linux/

> +      (native-inputs
> +       `(("bootstrap-seeds" ,bootstrap-seeds)
> +         ("mescc-tools"
> +          ,(bootstrap-origin
> +            (origin
> +              (method url-fetch)
> +              (uri (string-append
> +                    "https://lilypond.org/janneke/guix/20220502/"
> +                    "mescc-tools-" mescc-tools-version ".tar.gz"))
> +              (sha256
> +               (base32
> +                "1xi6f48pf5bhajhfis189gpizxij7nbp1vzvsb1aafhz4skkiqvg")))))
> +         ("m2-planet"
> +          ,(bootstrap-origin
> +            (origin
> +              (method url-fetch)
> +              (uri (string-append
> +                    "https://lilypond.org/janneke/guix/20220502/"
> +                    "M2-Planet-" m2-planet-version ".tar.gz"))
> +              (sha256
> +               (base32
> +                "1xrn69sc5nz4hwaishqyrcidp1ncxwib9zswl45x378ddz3mmk7g")))))
> +         ,@(%boot-gash-inputs)))

Here you can avoid ‘bootstrap-origin’ too and add mirror://gnu URLs.

The “new style” (with gexps) doesn’t work well with non-package inputs
in the sense that you cannot use ‘this-package-input’ to access them.

However, it might be simpler to not have them in ‘native-inputs’ and to
instead refer to them in the code, as in:

  #~(begin
      …
      (invoke "tar" "xvf" #$mescc-tools)
      …)

where:

  (define mescc-tools
    (origin …))

> +      (synopsis "The initial bootstrap package, builds stage0 up to M2-Planet")
> +      (description
> +       "Starting from an 357-byte hex0 provided by the bootstrap-seeds,
> +stage0-posix builds hex0, kaem, hex1, catm, hex2, M0, cc_x86, M1,
> +M2, get_machine, (mescc-tools), and M2-Planet.")

Bonus points if you can make it a full sentence.  :-)

(I feel ridiculous commenting on this on a patch series that’s this
important, but hey!)

Ludo’.

Ludovic Courtès writes:

Hi,

> Jan Nieuwenhuizen <janneke@gnu.org> skribis:
>
>>>From f013e88563f73a7513faa330c79eef2f653daf28 Mon Sep 17 00:00:00 2001
>> * gnu/packages/commencement.scm (stage0-posix): New variable.

> Same comment as earlier regarding ‘bootstrap-origin’ and the URL.

bootstrap-origin removed and URL changed to:

         (uri (list
               (string-append
               "mirror://gnu/guix/mirror/"
               "stage0-posix-" version ".tar.gz")
               (string-append
                "https://lilypond.org/janneke/guix/20220502/"
                "stage0-posix-" version ".tar.gz")))
>
>> +      (supported-systems '("i686-linux" "x86_64-linux"
>> +                           "arm-linux" "aarch64-linux"
>> +                           "riscv64-linux"))
>
> Woow.  :-)

:-)  Yeah, that will mean some work for us to come...

> s/arm-linux/armhf-linux/

Oops.

>> +      (native-inputs
>> +       `(("bootstrap-seeds" ,bootstrap-seeds)
>> +         ("mescc-tools"
>> +          ,(bootstrap-origin
>> +            (origin
>> +              (method url-fetch)
>> +              (uri (string-append
>> +                    "https://lilypond.org/janneke/guix/20220502/"
>> +                    "mescc-tools-" mescc-tools-version ".tar.gz"))
>> +              (sha256
>> +               (base32
>> +                "1xi6f48pf5bhajhfis189gpizxij7nbp1vzvsb1aafhz4skkiqvg")))))
>> +         ("m2-planet"
>> +          ,(bootstrap-origin
>> +            (origin
>> +              (method url-fetch)
>> +              (uri (string-append
>> +                    "https://lilypond.org/janneke/guix/20220502/"
>> +                    "M2-Planet-" m2-planet-version ".tar.gz"))
>> +              (sha256
>> +               (base32
>> +                "1xrn69sc5nz4hwaishqyrcidp1ncxwib9zswl45x378ddz3mmk7g")))))
>> +         ,@(%boot-gash-inputs)))
>
> Here you can avoid ‘bootstrap-origin’ too and add mirror://gnu URLs.

Ok.

> The “new style” (with gexps) doesn’t work well with non-package inputs
> in the sense that you cannot use ‘this-package-input’ to access them.
>
> However, it might be simpler to not have them in ‘native-inputs’ and to
> instead refer to them in the code, as in:
>
>   #~(begin
>       …
>       (invoke "tar" "xvf" #$mescc-tools)
>       …)
>
> where:
>
>   (define mescc-tools
>     (origin …))

Ok.  I put the mescc-tools and m2-planet in the initial let.

>> +      (synopsis "The initial bootstrap package, builds stage0 up to M2-Planet")
>> +      (description
>> +       "Starting from an 357-byte hex0 provided by the bootstrap-seeds,
>> +stage0-posix builds hex0, kaem, hex1, catm, hex2, M0, cc_x86, M1,
>> +M2, get_machine, (mescc-tools), and M2-Planet.")
>
> Bonus points if you can make it a full sentence.  :-)

I tried:

       "Starting from the 357-byte hex0-seed binary provided by the
bootstrap-seeds, the stage0-posix package first builds hex0 and then all
the way up: hex1, catm, hex2, M0, cc_x86, M1, M2, get_machine (that's
all of MesCC-Tools), and finally M2-Planet."

> (I feel ridiculous commenting on this on a patch series that’s this
> important, but hey!)

(otoh, after putting in all this work, some polishing doesn't hurt, so:
thanks!)

Greetings,
Janneke