guix-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] gnu: make-bootstrap: Produce the correct %glibc-bootstrap-ta


From: Ludovic Courtès
Subject: Re: [PATCH] gnu: make-bootstrap: Produce the correct %glibc-bootstrap-tarball for Hurd systems.
Date: Tue, 30 Jun 2015 21:56:59 +0200
User-agent: Gnus/5.13 (Gnus v5.13) Emacs/24.5 (gnu/linux)

Γειά σας!

Manolis Ragkousis <address@hidden> skribis:

> I chose this approach in %glibc-stripped because later on, as more
> things will be needed to be Hurd-specific, it's going to get messy
> with conditional code. This way we can safely change/add things
> without worrying it will break the other. WDYT?

The problem is that this patch duplicates 54 lines that are mostly
identical.

I think it would be preferable to make it more concise.

> From fa84fdeb8a479f3b59b38f762f551d744def2e5f Mon Sep 17 00:00:00 2001
> From: Manolis Ragkousis <address@hidden>
> Date: Tue, 30 Jun 2015 12:53:24 +0300
> Subject: [PATCH] gnu: make-bootstrap: Produce the correct
>  %glibc-bootstrap-tarball for Hurd systems.
>
> * gnu/packages/make-bootstrap.scm (%glibc-bootstrap-tarball): Make it a 
> procedure.
>   (%glibc-stripped): Make it a procedure that produces the correct 
> %glibc-stripped
>   depending on the target system.

[...]

> +  (match target
> +    ("i586-pc-gnu"
> +     (let ((glibc (glibc-for-bootstrap)))
> +       (package (inherit glibc/hurd)
> +        (name "glibc-stripped")
> +        (build-system trivial-build-system)
> +        (arguments
> +         `(#:modules ((guix build utils))
> +           #:builder
> +           (begin
> +             (use-modules (guix build utils))
> +
> +             (setvbuf (current-output-port) _IOLBF)
> +             (let* ((out    (assoc-ref %outputs "out"))
> +                    (libdir (string-append out "/lib"))
> +                    (incdir (string-append out "/include"))
> +                    (libc   (assoc-ref %build-inputs "libc"))
> +                    (mach (assoc-ref %build-inputs "gnumach-headers")))
> +               (mkdir-p libdir)
> +               (for-each (lambda (file)
> +                           (let ((target (string-append libdir "/"
> +                                                        (basename file))))
> +                             (copy-file file target)
> +                             (remove-store-references target)))
> +                         (find-files (string-append libc "/lib")
> +                                     
> "^(crt.*|ld.*|lib(c|m|dl|rt|pthread|nsl|util).*\\.so(\\..*)?|libc_nonshared\\.a)$"))
> +
> +               (copy-recursively (string-append libc "/include") incdir)
> +
> +               (copy-recursively (string-append mach "/include/mach")
> +                                 (string-append incdir "/mach"))
> +               #t))))

What about introducing a (guix build make-bootstrap) module to hide all
this plumbing?

It could expose a procedure like:

  (define (make-stripped-libc output libc kernel-headers)
    "Copy to OUTPUT the subset of LIBC and KERNEL-HEADERS that is needed
  when producing a bootstrap libc."
    ;; ...
    (if (directory-exists? (string-append kernel-headers "/mach"))
        (copy-mach-headers output kernel-headers)
        (copy-linux-headers output kernel-headers)))

  ;; ...


That way, the #:builder above would become:

  #:builder (begin
              (use-modules (guix build make-bootstrap))
              (make-stripped-libc (assoc-ref %outputs "out")
                                  (assoc-ref %build-inputs "libc")
                                  (assoc-ref %build-inputs "kernel-headers")))

and so we’d be duplicating fewer lines.

The ‘match’ itself could be moved to the ‘inputs’ field, where we’d be
selecting the right headers:

  (define (%glibc-stripped)
    (package (inherit (glibc-for-bootstrap))
      ;; ...
      (inputs `(("kernel-headers"
                 ,(if (or (and (%current-target-system)
                               (hurd-triplet? (%current-target-system)))
                          (string-suffix? "-hurd" (%current-system)))
                      gnumach-headers
                      linux-libre-headers))
                ...))))

where:

  (define (hurd-triplet? triplet)
    (and (string-suffix? "-gnu" triplet)
         (not (string-contains triplet "linux"))))

Note that (%current-target-system) is a GNU triplet whereas
(%current-system) is ARCH-KERNEL.

Does that make sense?  Does that sound doable?

Thank you!

Ludo’.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]