← Back to team overview

cuneiform team mailing list archive

Re: Ответ: Arch Linux PKGBUILD

 

On 9/22/08, Jussi Pakkanen <jpakkane@xxxxxxxxx> wrote:

> Ok, this looks quite sensible. Some things I noticed:
>
> Why do have the version number in lib dir?

Firstly, just in case. Secondly... do you know how AUR works?
It is not some official repository. It is just a collection of users
contributed PKGBUILDs. For instance, to install cuneiform
you are to download PKGBUILD and run makepkg. It wil
download sources, build them and make .pkg.tar.gz--
Arch package format. Only then you can install it as an
ordinary package (:
If some PKGBUILD from AUR takes enough votes, it moves
to the official "community" repository. It makes able to install
some app just via # pacman -S appname
More info here: http://wiki.archlinux.org/index.php/AUR

So, I thought that it is better to keep libs in a versioned dir.
To avoid conflicts and make able to have several version
per same time (e.g. for debug).

If my arguments is weak--say, I'll fix (:
Anyway, that was my first PKGBUILD, guys.

> I looked at /usr/lib on a
> Debian machine and the subdirectories there only had the program name.
> Parallel installing two versions of the same program to the same
> prefix is probably just begging for trouble.

As I sad, I found such solution useful only for a case of AUR.

> Will there be problems with linking? In the future we may want to
> offer Cuneiform as a library. Does this mean that the main shared
> library (libpuma.so) has to be installed to /usr/lib? Or do we need
> pkg-config magic to make it work propely? Does rpath take care of the
> other libraries' symbols automatically?

When cuneiform will be available as a standalone library,
I'll put it in /usr/lib. For now it is a heap of highly tailored libs,
I think. And keeping them in the such public place as /usr/lib
is not good idea. Just my opinion. Correct me, please, if I'm wrong.
The reason of such situation is simple---cuneiform comes from
the world of Windows (:

-- 
 .  Hoc est simplicissimum!
..: maxim.vuets.name



References