[bug#40084] gnu: Add v86d.

* gnu/packages/xorg.scm (v86d): New variable.
---
 gnu/packages/xorg.scm | 109 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 109 insertions(+)

On Sun, Mar 15, 2020 at 10:48:51PM +0100, Florian Pelz wrote:
> +         ;; Replace the bundled x86emu with its upstream copy from Xorg-server:
> +         (add-after 'unpack 'unpack-x86emu-sources
> +           (lambda* (#:key inputs #:allow-other-keys)
> +             (begin
> +               (format #t "decompressing x86emu source code~%")
> +               (with-directory-excursion "libs"
> +                 (call-with-output-file "xorg-server.tar"
> +                   (lambda (out)
> +                     (let* ((xz (assoc-ref inputs "xz"))
> +                            (srcs (assoc-ref inputs "xorg-server-sources"))
> +                            (pipe (open-pipe* OPEN_READ
> +                                              (string-append xz "/bin/xz")
> +                                              "-cd" srcs)))
> +                       (dump-port pipe out)
> +                       (unless (= (status:exit-val (close-pipe pipe)) 0)
> +                         (error "xz decompress failed")))))
> +                 (invoke (string-append (assoc-ref inputs "tar") "/bin/tar")
> +                         "xvf" "xorg-server.tar" "--strip-components=3"
> +                         "--wildcards" "*/hw/xfree86/x86emu/")
> +                 ;; extract license:
> +                 (with-directory-excursion "x86emu"
> +                   (invoke (string-append (assoc-ref inputs "tar") "/bin/tar")
> +                           "xvf" "../xorg-server.tar" "--strip-components=1"
> +                           "--wildcards" "*/COPYING"))
> +                 (delete-file "xorg-server.tar")
> +                 #t))))

I don't see why you're decompressing the tarball manually, this seems to
work just as well:

 (add-after 'unpack 'unpack-x86emu-sources
   (lambda* (#:key inputs #:allow-other-keys)
     (begin
       (format #t "decompressing x86emu source code~%")
       (with-directory-excursion "libs"
         (invoke (string-append (assoc-ref inputs "tar") "/bin/tar")
                 "xvf" (assoc-ref inputs "xorg-server-sources")
                 "--strip-components=3"
                 "--wildcards" "*/hw/xfree86/x86emu/")
         ;; extract license:
         (with-directory-excursion "x86emu"
           (invoke (string-append (assoc-ref inputs "tar") "/bin/tar")
                   "xvf" (assoc-ref inputs "xorg-server-sources")
                   "--strip-components=1"
                   "--wildcards" "*/COPYING"))
         #t))))

If it's about efficiency or robustness in some way, I'd suggest leaving
a comment detailing the reason. Either way, I'd bind the path to tar in
a `let', as well as the path to "xorg-server-sources" if we decide to
not decompress separately.

> +         (replace 'configure
> +           (lambda* (#:key outputs #:allow-other-keys)
> +             (let ((out (assoc-ref outputs "out")))
> +               (setenv "CC" (which "gcc"))
> +               (setenv "DESTDIR" out)
> +               (invoke "./configure" "--with-x86emu"))))
> +         (add-after 'build 'build-testvbe
> +           (lambda _
> +             (invoke "make" "testvbe")))
> +         (add-after 'install 'install-testvbe
> +           (lambda* (#:key outputs #:allow-other-keys)
> +             (let ((testvbe (assoc-ref outputs "testvbe"))
> +                   (olddest (getenv "DESTDIR")))
> +               (setenv "DESTDIR" testvbe)
> +               (invoke "make" "install_testvbe")
> +               (setenv "DESTDIR" olddest))))

These phases are missing the trailing #t.

I don't know how to test the package, but it passes lint & build,
including --rounds=2 to check for reproducibility.

Regards,
Jakub Kądziołka

On Thu, Mar 19, 2020 at 03:49:08PM +0100, Jakub Kądziołka wrote:
> I don't see why you're decompressing the tarball manually, this seems to
> work just as well:
> […]
> If it's about efficiency or robustness in some way, I'd suggest leaving
> a comment detailing the reason. Either way, I'd bind the path to tar in
> a `let', as well as the path to "xorg-server-sources" if we decide to
> not decompress separately.
> 

Your review is much appreciated.  Having tar do the decompression is
much better.  Before I had copied the decompression without checking
from my earlier patch for nginx-accept-language-module.  I think at
the time tar for some reason did not do auto-decompression when
building.  Or maybe manual decompression already was unnecessary back
then.

I have tested on multiple computers, sufficiently I believe, so I
pushed with your changes as e2303e8e375ed2e07c1fd760c86a204eb51fbc6e.
I also pushed similar changes to the decompression in
nginx-accept-language-module as e84490346d8dac3720a57a331f533ce67ff0da1c.


> > +         (replace 'configure
> > +           (lambda* (#:key outputs #:allow-other-keys)
> > +             (let ((out (assoc-ref outputs "out")))
> > +               (setenv "CC" (which "gcc"))
> > +               (setenv "DESTDIR" out)
> > +               (invoke "./configure" "--with-x86emu"))))
> > +         (add-after 'build 'build-testvbe
> > +           (lambda _
> > +             (invoke "make" "testvbe")))
> > +         (add-after 'install 'install-testvbe
> > +           (lambda* (#:key outputs #:allow-other-keys)
> > +             (let ((testvbe (assoc-ref outputs "testvbe"))
> > +                   (olddest (getenv "DESTDIR")))
> > +               (setenv "DESTDIR" testvbe)
> > +               (invoke "make" "install_testvbe")
> > +               (setenv "DESTDIR" olddest))))
> 
> These phases are missing the trailing #t.
>

I missed the #t for the install-testvbe phase.  Thank you!  According
to the docstring of

(define (invoke program . args)
  "Invoke PROGRAM with the given ARGS.  Raise an exception
if the exit code is non-zero; otherwise return #t."

in guix/build/utils.scm, invoke always returns #t, so I did not change
the phases ending in invoke.


> I don't know how to test the package, but it passes lint & build,
> including --rounds=2 to check for reproducibility.
> 
> Regards,
> Jakub Kądziołka

The important user of v86d is uvesafb.  uvesafb allows me to make Xorg
work on multiple computers of mine with

#!/bin/sh
modprobe uvesafb mode_option=1280x800-32 \
         v86d=/run/current-system/profile/sbin/v86d
chmod o+rw /dev/fb0
sleep 1
herd restart xorg-server


On many computers, the behavior can be tested when having a service

(set-xorg-configuration
 (xorg-configuration
  (modules
   (list xf86-video-fbdev
         xf86-input-libinput))))


and passing the kernel parameter 'nomodeset' e.g. in GRUB, because
then the framebuffer /dev/fb0 is only present when using uvesafb.

But I believe I have tested sufficiently.

Thank you!

Florian