On Wed, 2007-05-30 at 17:27 -0400, Mike VanRoy wrote:
bisco wrote:
Johannes Winkelmann ha scritto:
Hi Bisco,
On Tue, May 29, 2007 at 15:55:14 +0200, bisco wrote:
Jesse Kokkarinen ha scritto:
[...]
A few of the crux "ports" you refer to are already linked [1]. They are unofficial and do not fall under the responsibility of crux itself.
[...]
the _real_ question is: why are these unofficial projects yet?
Probably because these projects are developed completely independent from the CRUX project. Different people, different release schedules, and in some cases different goals.
if you remember, we tried some times to talk about merging these projects. We had no results because no one wanted to change its own position. So, other teams developed indipendently, but this is the wrong way, imho.
What do you think would be the benefits of merging the CRUX PPC projects into the CRUX (crux.nu) project, especially for the CRUX x86 side? I ask this especially since I can't remember seeing patches/generic fixes from the PPC developers sent to crux-devel.
I think merging related projects to the main one can give more interest into the distro. Most of the popular distros have official porting projects: why CRUX cannot have them?
Also, would you be willing to give up the non-platform dependent differences you made to CRUX PPC (like the different version scheme etc.), or would you want to keep that independence?
The initial idea was to give the same packages' release (when possible) than the x86 one. Then, due to our (both x86 and ppc) hard positions, we've decide to take our choices indipendently. But we still wish and support a common development for Crux for every architecture!
we can join the teams and choose a common way to develop distros' ports, a multiarch port system, and so on...
We've been there a couple of times before, yet so far there's no model for this multiarch port system which was generally accepted, and it may be hard to find it, considering that the different flavours of CRUX have slightly different goals.
If my memories aren't wrong, the problem was because the proposed solutions haven't a KISS style.
One of the things I'd like to stress here is that previously, the main problem was communication. If you want to be considered as future team members, you'll somehow need to get into a position to make the crux x86 devs want to work together with you, and trust you. That won't happen overnight, and not as long as you think (and publically express) the problems are just with "the others".
Yes, the problem was communication, but I think that it's interest of both on having a unified (or, if you think it's better, synchronized) team. But as far as I know, the same problem was shared with other teams working with other architecture than x86.
Case in point: Crux64 aims at multiarch/multilib. I, however prefer a "Crux-pure64" version, and have been using such for about a year. For me it is much simpler (KISS) and the majority of x86 ports work "as-is". And the rest only need minor patching, (the worst exceptions being the mozilla plug-ins). I had kept quiet mainly to avoid such discussions, as they seem somewhat counter-productive.
By having a "unified (x86) team", the system can grow and be maintained more consistently, while leaving open the ease of customizations. I guess my point is that the beauty of crux is its simplicity and flexibility, while facilitating the individuality of users within the community.
It's not a matter of aiming for multiarch/multilib, more about providing it. I've found many useful features of the multiarch implementation I am playing with. For example, I wanted to add mysql support to lighttpd. To do this I added a mysql directory in the lighttpd port, Copied the build function from the official port to the mysql directory added mysql functionality and rebuilt it with 'pkgmk -a mysql'. This same method could be used to modify existing ports for a pure64 environment. It's at about this time I can hear people howling USE USE USE. But for me it just makes sense to keep all your ports together, and it makes maintaining compat32 ports a hell of a lot easier. A bit of history on the multiarch system. Originally I was just editing the original Pkgfiles but it made merging with git a complete nightmare. So I put arch specific files in their own directory and then it didn't really make sense to keep a complete port for the multiarch port so I decided to just include the information that differs from the original. Some are simply a matter of adding 'export CFLAGS="$CFLAGS -fPIC"' to the x86_64/Pkgfile. Now I'm not saying the way is The Best Way (TM), but it works for me...well mostly, there are some compatability issues with prt-get. But IMHO unifying the efforts of the i686 and other projects is not necessarily a bad thing. If anyone is interested in working on a miltiarch repository please contact me and we can experiment some more. -- Lucas Hazel