Gentoo Forums :: View topic - On the comparative merits of package managers.

On the comparative merits of package managers.

View unanswered posts
View posts from last 24 hours

Goto page 1, 2, 3 Next

Gentoo Forums Forum Index

Gentoo Chat

View previous topic :: View next topic

Author

Message

hasufell
Retired Dev

Joined: 29 Oct 2011
Posts: 429

Posted: Wed Jul 23, 2014 2:07 am Post subject:

swathe wrote:

I'm coming back (AGAIN) from Arch. Whilst I never have the breakages on it a lot of people get I find the quality control lacking and the amount of abandoned AUR packages is pretty bad. I've come to the conclusion that the only pros for me with arch are install time and the AUR. I don't hate Arch, I just think Gentoo is better.

I recommend to try paludis instead of portage. Portage is not in good shape.

The only problem with paludis is... it's not developed by gentoo devs, so filing bugs or feature requests is a different thing. The paludis developers will not implement things just because gentoo users want it. It has to make sense for them. Additionally, a lot of gentoo devs don't accept ebuild bug reports when they see you were using paludis. To make it even worse, some ebuild writers do things wrong and cause breakage for paludis, but do not notice since it happens to work under portage by accident.

Still, it is a lot less broken than portage and has some interesting features as well. Both can be installed in parallel anyway.

HungGarTiger
Apprentice

Joined: 04 Feb 2014
Posts: 180
Location: /nz/auckland

Posted: Thu Jul 24, 2014 8:17 pm Post subject:

hasufell wrote:

swathe wrote:

I recommend to try paludis instead of portage. Portage is not in good shape.

...

Still, it is a lot less broken than portage and has some interesting features as well. Both can be installed in parallel anyway.

I'm considering moving over to Gentoo on a new build I am planning, so this post interests me.

How is Portage broken? Does this mean it's no longer being maintained? / Should I rethink my ideas around moving off of Arch?

hasufell
Retired Dev

Joined: 29 Oct 2011
Posts: 429

Posted: Thu Jul 24, 2014 8:45 pm Post subject:

HungGarTiger wrote:

hasufell wrote:

swathe wrote:

Portage currently has ~2 developers, most of them are quite new to that task, because zmedico basically left. The code is pure crap since a decade, just look at it. Getting in basic QA and style would already be a major task.

Also, portage has broken dependency calculation in many ways and gives incomplete, wrong or unusable suggestions to the user on how to fix various problems. But people are so used to these problems and workarounds that they often don't realize it anymore. Recently it was suggested that a problem (one of many) in the dependency calculation can be easily fixed by simply disabling dynamic deps, but even then we have developers who oppose this and don't understand the situation or in what shape portage is.

IMO, we should just abandon portage and revive pkgcore or fork paludis.

Anyway, portage is still less broken than pacman, which is broken by design and they call it KISS (as in: let the user figure out conflicts on his own).

krinn
Watchman

Joined: 02 May 2003
Posts: 7471

Posted: Fri Jul 25, 2014 7:48 am Post subject:

hasufell wrote:

IMO, we should just abandon portage and revive pkgcore or fork paludis.

When people says "yeah bad code", and prefer start new code because the old one cannot be understand, then people don't understand why something was made and goes their way.
You just generally endup with brand new code and a brand new variation of........... crap.
And you endup with two different kind of crap, but it's still crap.

So no it's not better to drop portage and redo another crap product, define the structure, what are portage problems... and let devs fix them.

And that is a good example :

hasufell wrote:

So, it mean everyone has left zmedico done crap in portage? zmedico was doing what he wants with no goal or method and out of any control?
Or does it mean zmedico was ask to add new functionalities to portage instead of fixing the decade crappy code of portage zmedico inherit?

From what i know i think one of the new dev is dolsen, having try porthole i have no doubt on his skills at python. But if dolsen is asked to add new EAPI or feature to portage... instead of fixing portage with a cleaner code and then must hack old broken and crappy functionalities in portage to met the need of new functionalities; then i don't think dolsen is making shit, but Gentoo is.

What let you think a new tool instead of fixing the existing one will not endup with crap if the policies that let portage gone crap are kept?

Dr.Willy
Guru
Guru

Joined: 15 Jul 2007
Posts: 547
Location: NRW, Germany

Posted: Fri Jul 25, 2014 10:26 am Post subject:

krinn wrote:

When people says "yeah bad code", and prefer start new code because the old one cannot be understand, then people don't understand why something was made and goes their way.

So, add "bad documentation" to the list of portage's deficiencies.
I know it's a boring task, but "we did foo because of reason01 and reason02; we did not do bar because of reason03 and reason04" is quite important information.

hasufell
Retired Dev

Joined: 29 Oct 2011
Posts: 429

Posted: Fri Jul 25, 2014 12:04 pm Post subject:

krinn wrote:

hasufell wrote:

IMO, we should just abandon portage and revive pkgcore or fork paludis.

Sorry, that is nonsense.

You don't end up with the same crap if you
a) start things over
b) fix the workflow
c) come up with strict policies for coding style, performance regression tests etc

Did you ever really look at portage code? It has indentation level up to ~11.

Pkgcore is NOT another crap product. Paludis neither. Both were coded with strict philosophy in mind. Please read up before making such claims.

krinn wrote:

I think it was already in that shape when zmedico came on board. I don't want to blame him, he just tried to keep things going. But yeah, we didn't have anyone except ferringb who pushed for QA and performance improvements and those didn't come from hacking on portage, but from a rewrite from scratch (pkgcore).

krinn wrote:

From what i know i think one of the new dev is dolsen, having try porthole i have no doubt on his skills at python. But if dolsen is asked to add new EAPI or feature to portage... instead of fixing portage with a cleaner code and then must hack old broken and crappy functionalities in portage to met the need of new functionalities; then i don't think dolsen is making shit, but Gentoo is.

Dol-sen is sick and will take a long time to recover, so we basically have one dev left who came on board a few months ago.

Sure, the shit doesn't only come from the portage code, but also from random features introduced by EAPIs, as well as bad practices in ebuild writing that people think are good.

krinn wrote:

What let you think a new tool instead of fixing the existing one will not endup with crap if the policies that let portage gone crap are kept?

Those tools are not new, they are already there. So I don't know what you are referring to.

krinn
Watchman

Joined: 02 May 2003
Posts: 7471

Posted: Fri Jul 25, 2014 12:46 pm Post subject:

hasufell wrote:

Those tools are not new, they are already there. So I don't know what you are referring to.

I never said pkgcore or paludis are crap.

What i said is that : even if you fork paludis and pkgcore that were done with your a)b)c) good policies and they are indeed good tools ; they won't be handle with a)b)c) policies in gentoo : the result is that your pkgcore/paludis fork will endup.... crap.

portage wasn't made crappy, it has became crappy.
Fix the rules that let portage goes wrong and stop portage get crappier instead of forking a good tool to make it another crap tool the next year.
As it only seems portage have no a)b)c) rules and the result is what we have today.

edit: i'm sorry if it start derive a bit too much from main thread, i'm not against my portage speech taken away to its own thread.

hasufell
Retired Dev

Joined: 29 Oct 2011
Posts: 429

Posted: Fri Jul 25, 2014 1:29 pm Post subject:

krinn wrote:

hasufell wrote:

Those tools are not new, they are already there. So I don't know what you are referring to.

What i said is that : even if you fork paludis and pkgcore that were done with your a)b)c) good policies and they are indeed good tools ; they won't be handle with a)b)c) policies in gentoo : the result is that your pkgcore/paludis fork will endup.... crap.

Erm, I didn't say "let's start over with the same broken workflow".

However, fixing the workflow is _not_ enough. The codebase is just too broken. It would be silly to try to reverse-fix those things when there are already far better working approaches.

I really hope people give up on portage. It would be better for gentoo.

steveL
Watchman

Joined: 13 Sep 2006
Posts: 5153
Location: The Peanut Gallery

Posted: Sun Jul 27, 2014 3:52 pm Post subject:

hasufell wrote:

krinn wrote:

Perhaps but it's a necessary precondition, or as krinn said, we'll just end up back in the same place, with a new set of (old) problems. Much of what you're talking about are cultural issues, and that's the kind of thing most geeks don't want to talk about, same as they like to pretend that it's possible to be apolitical, or asocial.

I'd take this more seriously if you were actually going to do something about any of this, but I've been waiting years for the devs to sort themselves out, so with respect I won't be holding my breath. We hear this kind of thing from (various) devs about once a year/18 months. I for one wouldn't want to work in such a culture.

hasufell
Retired Dev

Joined: 29 Oct 2011
Posts: 429

Posted: Sun Jul 27, 2014 4:27 pm Post subject:

steveL wrote:

hasufell wrote:

krinn wrote:

Perhaps but it's a necessary precondition, or as krinn said, we'll just end up back in the same place, with a new set of (old) problems.

That's basically what I said.

steveL wrote:

I'd take this more seriously if you were actually going to do something about any of this

Do what? Write a new PM? We already have them. I'm telling people to stop using and developing portage, that's what I do about it.

desultory
Bodhisattva

Joined: 04 Nov 2005
Posts: 9410

Posted: Thu Jul 31, 2014 3:30 am Post subject:

Split from "Anyone here moved away from Arch Linux?", because krinn was right

swathe
n00b
n00b

Joined: 04 Jul 2011
Posts: 73

Posted: Thu Jul 31, 2014 4:48 am Post subject:

I'm probably coming from a less experienced point of view not being able to code myself and never helped with the development of anything but why not drive to have more active developers working on portage (or any other package) to address these shortcoming/breakages in it? Not being a smart ass, but jumping ship or forking ad infinitum, where does it stop?

steveL
Watchman

Joined: 13 Sep 2006
Posts: 5153
Location: The Peanut Gallery

Posted: Fri Aug 01, 2014 5:35 am Post subject:

steveL wrote:

hasufell wrote:

krinn wrote:

hasufell wrote:

That's basically what I said.

No it's not, because you cut out my main point (fixed that for you), in order to avoid it again, the same way you did with krinn, which is what I explained.

Next time do us all the courtesy of not cutting out someone's main point, in order to make your answer appear clever. It's tedious enough having to deal with that kind of nonsense on the developer list, without having to deal with it in the forums.

Quote:

steveL wrote:

I'd take this more seriously if you were actually going to do something about any of this

Do what? Write a new PM? We already have them. I'm telling people to stop using and developing portage, that's what I do about it.

Do something about your sociopolitical problems, but that's hard I know. Especially when you're now advocating paludis, whose author is let's face it, one of the most egregious examples of poison people.

But you knew that, ofc, which is why you're avoiding it. I'm just filling in the others, should they wish to inform themselves of the history of the Ciaran McCreesh saga, in order to understand why he's considered a poisonous person. You can deal with your cognitive dissonance how you like, just like you do with the behaviour of Gentoo devs as a collective.

Oh and paludis is ofc an awful lot slower than pkgcore, as Patrick delighted in verifying pre-EAPI5. He told us recently it's even slower than portage, and that takes some doing.

hasufell
Retired Dev

Joined: 29 Oct 2011
Posts: 429

Posted: Fri Aug 01, 2014 12:24 pm Post subject:

steveL wrote:

hasufell wrote:

krinn wrote:

hasufell wrote:

That's basically what I said.

You probably missed the spot where I said that gentoo is socially broken.

So no idea why you keep pounding here.

steveL wrote:

hasufell wrote:

steveL wrote:

I'd take this more seriously if you were actually going to do something about any of this

Do what? Write a new PM? We already have them. I'm telling people to stop using and developing portage, that's what I do about it.

This is being highly undifferentiated.

First of all, I did not advocate Ciaran in any way (I criticized him before on the ML and in this forum and demanded that ComRel will ban him from the mailing lists... guess if they did). If you look close enough, I said that it's problematic, because it is not developed by gentoo devs and that the solution would be to actually fork it. Do not mix up software with the author (I hope you don't use ReiserFS).
Calling this cognitive dissonance is just silly to put it diplomatic. I'm also confused how you think you are helping gentoo culture with these kind of statements.

So what is the real problem here is that people seem to be emotionally attached to some some piece of software to the point that they refuse to realize that it's broken (this also happened recently with the dynamic deps discussion). In addition we have the NIH syndrome. Both points are highly unprofessional.

And the last point you got wrong is that paludis is slower than portage. First of all it highly depends on what you are actually doing. Most of the people are probably referring to the dependency resolution that is slower. Well, it is a LOT more accurate than portage, so you also have to talk about the effectiveness of dependency resolution, not just blindly compare the time two operations of different package managers with different algorithms take. That's not how you compare software.
Besides that... it does not take more time with '-z', afais (I haven't done professional measurement, but you really have to get familiar with paludis and it's options to even make a comment about it's performance).

krinn
Watchman

Joined: 02 May 2003
Posts: 7471

Posted: Fri Aug 01, 2014 3:43 pm Post subject:

hasufell wrote:

Well, it is a LOT more accurate than portage, so you also have to talk about the effectiveness of dependency resolution, not just blindly compare the time two operations of different package managers with different algorithms take. That's not how you compare software.

For a dev maybe,

But poll users about option they can add enable in portage with :
- Portage resolve deps like the Master of Universe and at only a 1% speed cost than today. An additional feature can show the algo at work if you wish.
- Portage resolve deps fast (and i mean really faster than the shame speed we have today) but is still inaccurate as of today, preserved-libs is disable and portage will not have algo view option.

Despite your thinking, i'm sure everyone will pickup fast option.
So yes, you can blindly compare two implementations without even knowing the algorithm in use, the technical details just sucks, who cares when you wait minutes to get an answer if portage/paludis use the latest conceptual freaking awesome algo?

We can make the poll if you really need. And even in the same "bias" manner with one option adding feature and the other removing some to get more speed.

But my main question remain : instead of trying to get users drop portage, why not pushing to have portage fix? If you get your paludis fork (that have some advantage, for me, at least python free, so if python is broken you can use paludis. But also a huge problem, paludis use its own "language" and it sucks ball you cannot just emerge paludis and run "cave" instead of "emerge" and it works! Options aren't the same, config files...) ; your fork will endup like portage because the policies that allow portage to become what it is today remain.

And if you fix the policies why not apply the new good ones to portage and let portage get fixed?
(by policies i mean rules, paper or QA... anything that define how portage should goes and goals within margin limit in speed/time... Instead of what it seems the chaotic rules that are use today : some kind of no rules, do what you want and what you can and nobody cares, and nobody is watching it anyway).
All projects i see have alpha/beta and release timer or dead line, while portage get out if a bird fly over one dev house, or when a cat cross his car or some other supernatural sign message...
The "when it's done" deadline is not really one anyone use (except for Duke Nukem Forever).
Isn't it really the reason why nobody wants dev portage? Because without any rules, you cannot manage to have many people doing collaborative work to met the same goal as nobody knows really what the other are doing or if anyone can help him ; something that may actually works pretty good if you're one person working on it, but something that gives a huge mess if more are working on it.

hasufell
Retired Dev

Joined: 29 Oct 2011
Posts: 429

Posted: Fri Aug 01, 2014 7:50 pm Post subject:

krinn wrote:

hasufell wrote:

I can't exactly follow you. Portage currently does not even give you the option to do dependency calculation like paludis does. However, paludis has several options that impact the depth and time dependency calculation takes including a fast option.

krinn wrote:

So yes, you can blindly compare two implementations without even knowing the algorithm in use, the technical details just sucks, who cares when you wait minutes to get an answer if portage/paludis use the latest conceptual freaking awesome algo?

When the algorithms do different things (as in... the output is different), then you definitely cannot blindly compare them.

krinn wrote:

But my main question remain : instead of trying to get users drop portage, why not pushing to have portage fix? If you get your paludis fork (that have some advantage, for me, at least python free, so if python is broken you can use paludis. But also a huge problem, paludis use its own "language" and it sucks ball you cannot just emerge paludis and run "cave" instead of "emerge" and it works! Options aren't the same, config files...) ; your fork will endup like portage because the policies that allow portage to become what it is today remain.

Why would it? It depends on the team that starts the fork. Everything else is guesswork.

krinn wrote:

And if you fix the policies why not apply the new good ones to portage and let portage get fixed?

I could write a plethora of reasons, but in the end this can be explained by looking at the code. Some people try to stop the regressions and fix random bugs, but really fixing the design of the lowlevel codebase is basically a dream. It would technically be a rewrite. And why do that if there are alternatives. I think it's a huge waste of time based on gentoo history and bad blood instead of technical reasons.

krinn wrote:

Isn't it really the reason why nobody wants dev portage?

No. The reason is really the shape of the code. It makes you angry.

steveL
Watchman

Joined: 13 Sep 2006
Posts: 5153
Location: The Peanut Gallery

Posted: Sat Aug 02, 2014 9:45 am Post subject:

LMAO I'll just let the self-evident contradictions stand. You sound just like TomWij, though on a good day. Great way to help the Gentoo culture, that.

hasufell
Retired Dev

Joined: 29 Oct 2011
Posts: 429

Posted: Sat Aug 02, 2014 2:01 pm Post subject:

steveL wrote:

LMAO I'll just let the self-evident contradictions stand. You sound just like TomWij, though on a good day.

Great way to help the Gentoo culture, that.

I'm not sure if you realize the contradiction in your own post.

Answering with "LMAO" to technical arguments after you have shown that you barely know about paludis and then going further with an ad hominem argument.

Seriously?

steveL
Watchman

Joined: 13 Sep 2006
Posts: 5153
Location: The Peanut Gallery

Posted: Sat Aug 02, 2014 5:50 pm Post subject:

Yes seriously. When someone is incapable of seeing the gaping contradictions in the statements they have made, there is very little that can be done to help them, apart from a timeout. If you're too stubborn to admit you're human, there's even less chance of you ever getting any self-insight, so there's even less point bothering with your nonsense. EOD between you and me, afaic. You don't respond to the points in the other thread, and you ignore the points made to you in this one. So by all means go and ahead and get the last word: I'll still be laughing at you. Good day.

hasufell
Retired Dev

Joined: 29 Oct 2011
Posts: 429

Posted: Sat Aug 02, 2014 8:26 pm Post subject:

I don't know what to say. Maybe I have misunderstood you somewhere, but that wasn't on purpose then. English isn't my mother language and I am sometimes drunk. I have no idea where I said that I am not human or that I am doing everything right. I'v written a very long mail a few months back on a closed gentoo ML explaining the vicious circle you easily get into when trying to change and improve gentoo. I'm not excluding myself here in any way. But your repeated efforts to come after me personally is pretty confusing. If you have a problem, contact me privately. In contrary to a lot of gentoo devs, I answer even private mails and I have never had a problem to say "sorry" in case I messed up in one or another way. Other than that, I fail to see technical arguments to your "paludis is slower" claims. This thread is more about the technical side of package managers. I don't mind a heated discussion, as long as it's still technical acute.

yagami
Apprentice

Joined: 12 May 2002
Posts: 269
Location: Leiria, Portugal

Posted: Sun Aug 03, 2014 9:48 am Post subject:

hasufell wrote:

No. The reason is really the shape of the code. It makes you angry.

Heh, sorry to be offtopic, but this line made me think about system*

Should Gentoo be anti-portage also ?

( its to be read as kind of light humor ... dont be too angry ! I too, as a programmer, think that blinky windows are nice for users, but the longevity of the project is largelly based on code quality )

hasufell
Retired Dev

Joined: 29 Oct 2011
Posts: 429

Posted: Sun Aug 03, 2014 12:03 pm Post subject:

yagami wrote:

Should Gentoo be anti-portage also ?

Yes, it's better if that codebase dies, but as you can see we have too many people who are too stubborn to accept that, so it will not happen.

It's a mix of these
* emotionally attached to portage
* "works for me" argument, although the bug reports keep piling up and ugly user workarounds have become the default
* most people who like portage have never looked at the code
* not-invented-here syndrome when it comes to paludis
* "paludis was written by Ciaran, so we cannot use it" (err)
* "portage is the fastest PM" (just wrong)
* no one seems to have interest or time to put up a monster project like forking/continuing one of the lesser broken PMs and I can't blame anyone for these reasons (including myself)
* people rather considering to write a PM from scratch which usually results in vaporware

After all, a PM is a huge thing and the collaboration model in gentoo is broken anyway. Unless someone wants to use this for his thesis or unless money is involved in one way or another, I doubt we will see action.

Ant P.
Watchman

Joined: 18 Apr 2009
Posts: 6920

Posted: Sun Aug 03, 2014 11:42 pm Post subject:

hasufell wrote:

* "paludis was written by Ciaran, so we cannot use it" (err)

Wow, there are people so stuck in the past they still use the "ciaranm is a big mean doo-doo head" excuse? In 2014!?

Have they noticed how toxic some of Gentoo's current developers are? Some of them seem to spend more time hurling abuse at users and other devs than actually doing anything worthwhile. No doubt a few of them are using ciaranm as a scapegoat to avoid having to behave themselves.

I want to make a comparison to the state of PHP's internal developer circles here, but I've already self-censored most of this post out of respect for the people already present in the thread. Besides, I'm already way over my monthly assholery quota.

Navar
Guru
Guru

Joined: 20 Aug 2012
Posts: 355
Location: usa

Posted: Mon Aug 04, 2014 12:08 am Post subject:

hasufell wrote:

Did you ever really look at portage code? It has indentation level up to ~11.

Wow. I'll admit I have only looked at bits and pieces briefly. Although a part of me just screams inside to resolve dependency resolution performance which seems to have choked in ability in 2014.

hasufell wrote:

Sure, the shit doesn't only come from the portage code, but also from random features introduced by EAPIs, as well as bad practices in ebuild writing that people think are good.

Let's start there. From an outsider looking in, is there some concise and clear references/thread anything somewhere that refutes any issues showing in the existing published developer docs for considered best practices? I'd prefer to avoid a 300 page set of side rants or anything that simply strays away from the technical facts towards personal attacks.

krinn
Watchman

Joined: 02 May 2003
Posts: 7471

Posted: Mon Aug 04, 2014 9:56 am Post subject:

Navar wrote:

hasufell wrote:

Sure, the shit doesn't only come from the portage code, but also from random features introduced by EAPIs, as well as bad practices in ebuild writing that people think are good.

...is there some concise and clear references/thread anything somewhere that refutes any issues showing in the existing published developer docs for considered best practices?...

And no dev ever think creating an "ebuild-validator" that check ebuild and issue warning/hint/errors that might resolve most of them before they are even push in the tree?

Display posts from previous:

	Gentoo Forums Forum Index Gentoo Chat	All times are GMT Goto page 1, 2, 3 Next
Page 1 of 3

Jump to:

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum

Copyright 2001-2025 Gentoo Foundation, Inc. Designed by Kyle Manna © 2003; Style derived from original subSilver theme. | Hosting by Gossamer Threads Inc. © | Powered by phpBB 2.0.23-gentoo-p11 © 2001, 2002 phpBB Group
Privacy Policy