What needs to be improved in Gentoo?

dundas · Guru Joined: 16 Dec 2004 Posts: 317 Location: China, Earth

just a crazy idea, if gentoo could have a version or option that allows newbies to install and play just more or less like a binary distro (like ubuntu, I donno, maybe more binary packages for BIG softwares, or a binary only portage so we won't scare ppl off of big compilation, but sitll utilize the management framework of source-base gentoo?), so it's much more easier for gentoo to attract new users and create largers communities? :evil:

p.s. I know livecd can do the demo part, but if they update or installs packages, do we have the quick binarys to solve the prob in a more lighter way?which could put gentoo at a much lower level to start wiz.
_________________
Appreciate Gentoo: Best Devs, Best Forums. YOU could help too: Help Answer

Cinquero · Apprentice Joined: 24 Jun 2004 Posts: 249

We should record the use of unstable (all?) packages and maybe even detect formal errors (ie. crashes etc.) automatically and report them. The reported data could then be analyzed using statistical methods and the results could give an indication of the stability of the package.

Why I am proposing that? Because I stumble over and over again on packages that are declared as stable but aren't stable at all (produce crashes when playing back sounds or such stupid stuff) while packages declared as unstable work flawlessly.

The point about statistical information analysis is that it _could_ be quite generic in the sense that one does not need to know too much about the system running the package: such "artefacts" are possibly "smeared out".

sageman · Posted: Thu Aug 10, 2006 1:39 pm Post subject:

ColdWind · Posted: Thu Aug 10, 2006 1:41 pm Post subject: Paludis, tagging and stabilizing

@Dralnu: maybe the C-Portage discussion should end with the Paludis existence ;-)

As for tagging (make your own system and world-like groups of packages) it would be interesting and I guess It's easy to implement.

I have a new feature in mind (i'm implementing it in a Python script), first I have to try it myself to see if it's actually usefull in the practice but I'll explain it so people can give some pre-feedback. The feature is "stabilizing", if you have a stable system you'll probably end installing a good number of ~arch packages that you don't want to update in ~arch forever. The "stabilizing" feature would check if there's a stable version of the program equal or higher to the one you put on package.keywords, install it and then remove it from package.keywords.

I'll give an example:
Some months ago I installed X.org 7.0, I had a huge list of packages in p.keywords, but I want to stay with the stable ones as soon as they get stable. So I have to check which is the latest stable version and remove the package from p.keywords when it's finally stable. The stabilizer could do it automatically with a given list of packages.

Root Moose · Tux's lil' helper Joined: 29 Jan 2004 Posts: 112

sageman

Root Moose · Tux's lil' helper Joined: 29 Jan 2004 Posts: 112

SteveBallmersChair · Posted: Thu Aug 10, 2006 2:08 pm Post subject: Gentoo enhancements

Here's a couple of things that I would like to see:

1. Better distcc handling.
I have a fast desktop and a slow laptop running Gentoo. Needlessly to say, I have distcc up and running so that a small update does not take an hour on the laptop. Some packages do not like distcc for some reason and won't build using it. I'd like to see the ability of portage to automatically try to install the package NOT using distcc if the ebuild dies. Something like a "Retrying emerge WITHOUT using distcc in 5 seconds, press ESC to cancel..." would be great. Also, if a package is known to not like distcc, automatically disable the use of distcc in that ebuild without requiring the user to export FEATURES or comment distcc out in the make.conf. That could be overridden with the USE="force-distcc" flag or something similar if the user wants to have a shot at it anyway.

2. Reverse dependency checking upon installation of a package
Yeah, you guys have heard this one a lot and it'd be a pain, but it would surely be a help...

3. Outputting the "need to do" things at the end of the entire emerge.
The little notifications like "run perl-cleaner" or "run revdep-rebuild" that pop up in the middle of emerging/upgrading several packages would be best to gather and display them all at the end so that I do not have to scroll up through 8500 lines of GCC output and such to see them. That way I can better act on it.
_________________
Unix is user friendly- it is just picky who its friends are.

ColdWind

Dralnu · Veteran Joined: 24 May 2006 Posts: 1919

sageman · Posted: Thu Aug 10, 2006 2:33 pm Post subject:

Root Moose · Tux's lil' helper Joined: 29 Jan 2004 Posts: 112

beandog · Posted: Thu Aug 10, 2006 2:47 pm Post subject:

beandog · Posted: Thu Aug 10, 2006 2:49 pm Post subject:

beandog · Posted: Thu Aug 10, 2006 2:51 pm Post subject: Re: Gentoo enhancements

boroshan · l33t Joined: 16 Apr 2003 Posts: 730 Location: upside down

Dralnu · Veteran Joined: 24 May 2006 Posts: 1919

derverstand · Guru Joined: 15 Dec 2005 Posts: 511 Location: /dev/null

Hi,

In my opinion the most important thing is: It shoud be possible to install gentoo withou having internet access. Therefore a gentoo mirror should contain ALL sources. There are many external references in the ebuilds (like xorg etc). For my cluster installation I had to write a script, which parses all ebuilds, extracts the URLs from them and downloads them manually. This was the only possibility to create a standalone mirror...

Best regards!

adsmith · Veteran Joined: 26 Sep 2004 Posts: 1386 Location: NC, USA

ELOG was discussed in the June 19 GWN, and all these questions about CONFIG_PROTECT and the proper use of ACCEPT_KEYWORDS have been well documented.

Gentoo is probably the best documented linux distro in history, so it really astonishes me that people keep bringing up these topics. The docs are there -- read them!

The problems that really bother me are architectual problems with the whole portage system. Discussions keep arising about the best filesystem for portage as well as DB or C/C++ options or reverse-deps for portage. These all come back to a fundamentally broken model for the way portage understand dependencies.

I personally have tried to write a DB backend for portage, but I gave up after two weeks of coding the thing together, trying to figure out how to sort out USE flag logic. Lots of people have tried to do C or C++ versions of portage, the most recent of which is this intruiging "paludis" program.

What bothers me most is that these efforts keep happening from outside, either from non-developers like myself or ex-developers. Portage has evolved more slowly than its user base. I run portage on just 5 personal machines, 3 of which are laptops, and it can really be horrendous. I can't imagine how painful it must be for small companies who try to run Gentoo.

The result is a thousand little tools and web pages (eix, gentoo-portage, esearch, euse, equery, q, ...) which keep duplicating the same efforts but never solve the fundamental problem that searching and sorting dependencies and flags in the tree is just too hard and too slow.

Chaosite · Posted: Thu Aug 10, 2006 3:32 pm Post subject:

Suicidal · l33t Joined: 30 Jul 2003 Posts: 959 Location: /dev/null

boroshan · l33t Joined: 16 Apr 2003 Posts: 730 Location: upside down

beandog · Posted: Thu Aug 10, 2006 3:54 pm Post subject:

sageman · Posted: Thu Aug 10, 2006 3:58 pm Post subject: