You are not logged in.

#1 2016-03-28 10:42:09

osvein
Member
Registered: 2016-03-27
Posts: 3

[Review request] lib32-gimp (using i686 binaries from official repo)

I needed lib32-gimp for some printer drivers and I couldn't find it in AUR4, so I just grabbed it from the AUR archive and reuploaded it. I noticed the PKGBUILD merely retrieves the i686 gimp package from the official repo. I've never seen this approach to multilib before.

_pkgbase=gimp
pkgname=lib32-gimp
pkgver=2.8.16
pkgrel=1
pkgdesc="GNU Image Manipulation Program (32-bit)"
arch=('x86_64')
url="http://www.gimp.org/"
depends=( 'lib32-glibc' 'lib32-cairo' 'lib32-gdk-pixbuf2' 'lib32-glib2' 'lib32-pango' 'gimp')
license=('GPL' 'LGPL')
source=(https://www.archlinux.org/packages/extra/i686/gimp/download/#gimp-i686.pkg.tar.xz)
md5sums=('557c83fd5968c73e29163a7720528011')
pkgver() {
        cat $srcdir/.PKGINFO | grep -oP '(?<=pkgver = )\d+\.\d+\.\d+'
}
package() {
	mkdir -p ${pkgdir}/usr/lib32
	cp -rPf ${srcdir}/usr/lib/*.so* ${pkgdir}/usr/lib32
	install -dm755 "${pkgdir}"/usr/share/licenses
	ln -s ${_pkgbase} "${pkgdir}"/usr/share/licenses/${pkgname}
}

Is this really the right way to do it?

Offline

#2 2016-03-28 11:28:12

x33a
Forum Fellow
Registered: 2009-08-15
Posts: 4,587

Re: [Review request] lib32-gimp (using i686 binaries from official repo)

I am not sure about the repackaging bit, but I am curious about printer drivers needing GIMP. What sort of drivers are these?

Offline

#3 2016-03-28 11:37:31

Trilby
Inspector Parrot
Registered: 2011-11-29
Posts: 29,525
Website

Re: [Review request] lib32-gimp (using i686 binaries from official repo)

As written in this PKGBUILD the pkgname should have two suffixes.  It should have the '-bin' suffiix to identify it is a binary nor a source build.  But you also have it retrieving the latest version and determining that version at build time in the pkgver function.  This requires yet another suffix, but none of the standard ones apply (this isn't git, or hg, etc).  Perpaps -vcs would be good.

EDIT: oh, wait, you are getting this from the i686 repos.  That's odd.

EDIT 2: this package will not work for very long.  Get rid of the pkgver function, and use a specific version.  You definitely can't have a dynamic source and a fixed checksum.  As soon as the version in the i686 repo is bumped, this PKGBUILD will fail.  So just set the pkgver to the current version number, and add the '-bin' suffix.

It's also a bit odd that you requested a review of this PKGBUILD after you submitted it to the AUR.


"UNIX is simple and coherent..." - Dennis Ritchie, "GNU's Not UNIX" -  Richard Stallman

Offline

#4 2016-03-28 16:10:38

osvein
Member
Registered: 2016-03-27
Posts: 3

Re: [Review request] lib32-gimp (using i686 binaries from official repo)

x33a wrote:

What sort of drivers are these?

I should have said they were scanner drivers, not printer drivers. It's the ScanGear driver for the Canon MP-series printers. ScanGear was already in the AUR and listed lib32-gimp as a dependency, but lib32-gimp had yet to be uploaded to AUR4.

Trilby wrote:

this package will not work for very long.  Get rid of the pkgver function, and use a specific version.  You definitely can't have a dynamic source and a fixed checksum.  As soon as the version in the i686 repo is bumped, this PKGBUILD will fail.  So just set the pkgver to the current version number, and add the '-bin' suffix.

It's also a bit odd that you requested a review of this PKGBUILD after you submitted it to the AUR.

I merely grabbed it from the AUR3 archive and updated the checksum and pkgver, so I assumed it was fine.

Do you think that using binaries from the i686 repo is fine, as long as I fix the pkgver and add the -bin suffix? I've never seen other multilib packages do this.

Last edited by osvein (2016-03-28 16:20:32)

Offline

#5 2016-03-28 16:22:06

Trilby
Inspector Parrot
Registered: 2011-11-29
Posts: 29,525
Website

Re: [Review request] lib32-gimp (using i686 binaries from official repo)

Which package?  I just checked several of the scangear aur packages (not all, there are a lot) and none of the ones I checked needed lib32-gimp - some of them did depend on gimp.


"UNIX is simple and coherent..." - Dennis Ritchie, "GNU's Not UNIX" -  Richard Stallman

Offline

#6 2016-04-02 11:42:25

osvein
Member
Registered: 2016-03-27
Posts: 3

Re: [Review request] lib32-gimp (using i686 binaries from official repo)

They're all listed here. It's an x86_64-specific dependency, obviously.

Offline

#7 2016-04-02 11:47:55

Trilby
Inspector Parrot
Registered: 2011-11-29
Posts: 29,525
Website

Re: [Review request] lib32-gimp (using i686 binaries from official repo)

osvein wrote:

They're all listed here. It's an x86_64-specific dependency, obviously.

Thanks - sorry I had missed those somehow.

Back to your edit:

osvein wrote:

Do you think that using binaries from the i686 repo is fine, as long as I fix the pkgver and add the -bin suffix? I've never seen other multilib packages do this.

It struck me as a bit odd to repackage something from the repos in this way, but I don't see anything wrong with it given a fixed pkgver and -bin suffix.


"UNIX is simple and coherent..." - Dennis Ritchie, "GNU's Not UNIX" -  Richard Stallman

Offline

#8 2016-04-02 12:26:00

Lone_Wolf
Member
From: Netherlands, Europe
Registered: 2005-10-04
Posts: 11,919

Re: [Review request] lib32-gimp (using i686 binaries from official repo)

Trilby wrote:
osvein wrote:

Do you think that using binaries from the i686 repo is fine, as long as I fix the pkgver and add the -bin suffix? I've never seen other multilib packages do this.

It struck me as a bit odd to repackage something from the repos in this way, but I don't see anything wrong with it given a fixed pkgver and -bin suffix.

Canon has a habit of re-using i686 32-bit code in their x86-64 code as much as possible.
Usually they limit this to their own proprietary code,in this case i guess some of their proprietary 32-bit code needs gimp i686 version.


Disliking systemd intensely, but not satisfied with alternatives so focusing on taming systemd.


(A works at time B)  && (time C > time B ) ≠  (A works at time C)

Offline

Board footer

Powered by FluxBB