[gentoo-dev] Is there a process for marking ebuilds stable?

[gentoo-dev] Is there a process for marking ebuilds stable?

[Brad Laue]

Given the increasing size of the portage tree I'm becoming concerned
about the rate at which ebuilds move from the unstable ~arch keyword to
the stable one.

Has a formalized process been discussed for this? The first thing that
comes to mind is a set of tinderboxes designed to build packages with
predictable flags sending reports to each ebuild maintainer.

The second is more practical and within reach; advocacy of
stable.gentoo.org, and a policy of accepting a package as stable when
five or more users have vouched for it and two weeks have passed without
a bug report.

These are rough ideas, I'd love to hear some input on them. Any
thoughts?

--
// -- http://www.BRAD-X.com/ -- //


--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Rainer Groesslinger]

On Monday 14 April 2003 00:25, Brad Laue wrote:
> Given the increasing size of the portage tree I'm becoming concerned
> about the rate at which ebuilds move from the unstable ~arch keyword to
> the stable one.

correct, same here.
Additionally there are many ebuilds in the tree that should be removed 
again...for example most of the game mods (not because I don't like gamers 
just because e.g. osp for quake3 isn't maintained although quite some time 
passed already since the last osp release(s) and in a multiplayer game it's 
useless to have an old version of something ;)

> Has a formalized process been discussed for this? The first thing that
> comes to mind is a set of tinderboxes designed to build packages with
> predictable flags sending reports to each ebuild maintainer.

Problem: Gentoo doesn't have maintainers !
It has been discussed to introduce a MAINTAINER="xxxx@gentoo.org" thing in the 
ebuilds but it seems like the idea got dropped by the core developers (or 
didn't even get attention, I don't know).

The only real maintainer is carpaski for portage, most other packages are 
worked on by more or less "Freelancers"...
Sure, many people are related to something, but you can't see who is the 
maintainer of a certain package.
Just imagine...there are some packages where version 0.1 was submitted by dev 
A, 0.2 by dev B and 0.2.1 by dev C and 0.3 again by dev B etc.

Not, that this is bad at all, but it would be much better to have "real" 
maintainers like almost every other distribution has, too.

> The second is more practical and within reach; advocacy of
> stable.gentoo.org, and a policy of accepting a package as stable when
> five or more users have vouched for it and two weeks have passed without
> a bug report.

stable.gentoo.org is _great_ ! Thanks so much to blizzy (unfortunatly he's not 
in the dev team any more). The problem here is that this site must be pushed 
quite hard because there are packages in the tree only a few people use and 
if those people don't use stable.gentoo.org they won't be stable anytime soon 
or might - in a bad case - be pushed into stable because nobody complain 
although it's just because nobody uses stable.gentoo.org


Rainer Groesslinger

--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Rainer Groesslinger]

> The only real maintainer is carpaski for portage, most other packages are
> worked on by more or less "Freelancers"...
> Sure, many people are related to something, but you can't see who is the
> maintainer of a certain package.
> Just imagine...there are some packages where version 0.1 was submitted by
> dev A, 0.2 by dev B and 0.2.1 by dev C and 0.3 again by dev B etc.
>
> Not, that this is bad at all, but it would be much better to have "real"
> maintainers like almost every other distribution has, too.

A little addition about my use of "Freelancers" - I don't know how it's going 
on in the developer team, I just stated what it looks like from a user 
perspective ! Please devs, don't blame me for stating this, if that's not the 
case please correct me.

Rainer Groesslinger



--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Jon Portnoy]

On Mon, Apr 14, 2003 at 01:00:51AM +0200, Rainer Groesslinger wrote:
> On Monday 14 April 2003 00:25, Brad Laue wrote:
> > Given the increasing size of the portage tree I'm becoming concerned
> > about the rate at which ebuilds move from the unstable ~arch keyword to
> > the stable one.
> 
> correct, same here.
> Additionally there are many ebuilds in the tree that should be removed 
> again...for example most of the game mods (not because I don't like gamers 
> just because e.g. osp for quake3 isn't maintained although quite some time 
> passed already since the last osp release(s) and in a multiplayer game it's 
> useless to have an old version of something ;)

Submit an updated ebuild.

> 
> > Has a formalized process been discussed for this? The first thing that
> > comes to mind is a set of tinderboxes designed to build packages with
> > predictable flags sending reports to each ebuild maintainer.
> 
> Problem: Gentoo doesn't have maintainers !
> It has been discussed to introduce a MAINTAINER="xxxx@gentoo.org" thing in the 
> ebuilds but it seems like the idea got dropped by the core developers (or 
> didn't even get attention, I don't know).
> 
> The only real maintainer is carpaski for portage, most other packages are 
> worked on by more or less "Freelancers"...
> Sure, many people are related to something, but you can't see who is the 
> maintainer of a certain package.
> Just imagine...there are some packages where version 0.1 was submitted by dev 
> A, 0.2 by dev B and 0.2.1 by dev C and 0.3 again by dev B etc.
> 

Read changelogs.

> Not, that this is bad at all, but it would be much better to have "real" 
> maintainers like almost every other distribution has, too.

We have real maintainers. For example, I maintain a handful of packages.

Could you be more specific about what constitutes a "real" maintainer?

> 
> > The second is more practical and within reach; advocacy of
> > stable.gentoo.org, and a policy of accepting a package as stable when
> > five or more users have vouched for it and two weeks have passed without
> > a bug report.
> 
> stable.gentoo.org is _great_ ! Thanks so much to blizzy (unfortunatly he's not 
> in the dev team any more). The problem here is that this site must be pushed 
> quite hard because there are packages in the tree only a few people use and 
> if those people don't use stable.gentoo.org they won't be stable anytime soon 
> or might - in a bad case - be pushed into stable because nobody complain 
> although it's just because nobody uses stable.gentoo.org
> 

-- 
Jon Portnoy
avenj/irc.freenode.net

--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Rainer Groesslinger]

> Submit an updated ebuild.

why should I update an ebuild I have nothing to do with ? 
Then we would have the result gentoo is currently experiencing with many 
packages IMHO.
I don't use osp, but the ebuild is useless for people who do use osp...If 
there would be a maintainer he could update it or people (like me) report 
that there is a new version, but if the dev who is in mood for osp updates it 
sometime nobody feels responsible (like it seems it's the case with some 
not-so-popular packages).
Also if a developer leaves the team (and that happens from time to time ;) the 
ebuild is just dead if no new maintainer is assigned to the ebuild (also 
seems to happen to a few packages)

> Read changelogs.

You won't believe it but in fact, I really do so.
But what's the point about making a changelog entry which tells me "version 
bump" though a diff shows some more changes and should I assume that the one 
who subbmitted the most recent ebuild is also the maintainer ?

> We have real maintainers. For example, I maintain a handful of packages.
>
> Could you be more specific about what constitutes a "real" maintainer?

that means e.g. having something like MAINTAINER="xxxx@gentoo.org" in the 
ebuild.
I believe that you are maintaining packages, many people do so, too. But the 
maintainers should be "hardcoded" in some way just like debian does (they 
have also about 9.000 packages...they have some non-maintained packages, too 
but then you _know_ that the maintainer isn't working on it anymore or 
something like that in gentoo you don't know that...Additionally I hear 
complains that some bugs get assigned to people on bugs.gentoo.org that have 
nothing to do with the bug/package, just because they submitted a new ebuild 
for a package to fix something and then they are meant to be the maintainer ?

There are packages where you can't see what developer carrys about them...you 
often have the choice between two or more (or sometimes none since all people 
mentioned in the changelog aren't in the dev team anymore ;)

Again, I don't know how it's working internally but from a users perspective 
gentoo's maintaining system is quite poor (if you take the packages in the 
tree...recently we broke the 40k files to consider barrier...if you have that 
many files (sure, changelogs, patches etc. are included but anyway ;) a 
"I-care-about-what-I-like" doesn't work for not-so-popular packages...
I don't say this is the case for most packages...It just happens to quite a 
few and not only two or three.

Rainer

--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Fredrik Jagenheim]

On Mon, Apr 14, 2003 at 01:00:51AM +0200, Rainer Groesslinger wrote:
> stable.gentoo.org is _great_ ! Thanks so much to blizzy
> (unfortunatly he's not in the dev team any more). The problem here
> is that this site must be pushed quite hard because there are
> packages in the tree only a few people use and if those people don't
> use stable.gentoo.org they won't be stable anytime soon or might -
> in a bad case - be pushed into stable because nobody complain
> although it's just because nobody uses stable.gentoo.org

I'm running a mix of 'stable' and 'unstable' packages. I try to report
if they work as much as possible to the stable.gentoo.org, but it's
hard to keep track of what I'm running as stable, and what I'm running
as unstable... The problem is, that when I've just installed an
unstable package, I can't say much about the 'stableness' of it, I can
just say if it emerges or not. And if it didn't emerge properly, I
would use bugs.gentoo.org, not stable.gentoo.org. And after running my
unstable package for a week, I can witness of the stability of the
package, but by then, I have forgotten all about reporting it...

Thus, I propose an extension to emerge that would look through your
system and see which packages are marked as 'unstable' and ask the
user interactively if they think the packages are OK. The responses
could be of the type:
1) Yes, I've used it extensively and it works.
2) Yes, I've used it somewhat and it seems to work.
3) No idea, I don't think I've used it, but nothing is broke.
4) No idea, I haven't used it at all yet.
5) No, it doesn't work and I've used bugzilla to report the bug.

You hopefully get the idea...

Combine it with some kind of cgi-gateway on stable.gentoo.org,
similar to the one at the stats.gentoo.org and it would be quite
simple for users to start tracking their unstable packages so they
could be adopted for stable.

I know, talk isn't worth anything and code is everything. ;)

//H
-- 
To segfault is human; to bluescreen moronic. 

--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Michael Kohl]

[-- Attachment #1: Type: text/plain, Size: 2255 bytes --]

On Mon, 14 Apr 2003 09:32:18 +0200
Fredrik Jagenheim <humming@pobox.com> wrote:

> Thus, I propose an extension to emerge that would look through your
> system and see which packages are marked as 'unstable' and ask the
> user interactively if they think the packages are OK. The responses
> could be of the type:

Maybe a seperate package similar to gentoo-stats really is a good idea.
Actually I was thinking about doing something like this to finally get
myself to learn Perl (I believe to remember that blizzy did the
backend of gentoo stable in Perl too), but then university exams somehow
came between me and this project. After that Ruby distracted me from
learning Perl, but I'm way to OT here...

But: how does the script know _for sure_ that the package was installed
while marked as unstable? Example: you install package foo while it's
still "~x86". Some days go by where you test the newly installed package
(which is maybe a good idea if you want to comment on it being stable or
not).  The day you use gentoo-stable (or whatever the script is
called) you synced before using it and now it's "x86". Would there still
be a way to know if it was installed while being "~x86" masked?

Hm, I had a short look, seems like the ebuild which a package used for
installation is stored in /var/db/pkg/$CATEGORY/$APP/$APP.ebuild (BTW
there's quite a lot of useful information in /var/db/pkg anyway).

Also it maybe would be a nice thing to have a switch for this script to
automatically submit all packages which it finds where installed using
~ARCH as "emerged sucessfully" (which I suppose they were if they land
in /var/db/pkg). This way it would be possible to set it up as a cronjob
so people (who are generally lazy from my experience) wouldn't have to
rate each package individually. Althoug this information is not quite as
good as "the full thing" but somehow I doubt all this people would be
running gentoo-stats if they had to do more then set up the crontab
once...

Sorry for this rather long a quite unfocussed post, mid-term exams are
on the horizon again so I maybe should just shut up and grab my books...

Michael

-- 
www.cargal.org 
GnuPG-key-ID: 0x90CA09E3
Jabber-ID: citizen428 [at] cargal [dot] org
Registered Linux User #278726

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Fredrik Jagenheim]

On Mon, Apr 14, 2003 at 05:00:39PM +0800, Michael Kohl wrote:
> Also it maybe would be a nice thing to have a switch for this script to
> automatically submit all packages which it finds where installed using
> ~ARCH as "emerged sucessfully" (which I suppose they were if they land
> in /var/db/pkg). This way it would be possible to set it up as a cronjob
> so people (who are generally lazy from my experience) wouldn't have to
> rate each package individually. Althoug this information is not quite as
> good as "the full thing" but somehow I doubt all this people would be
> running gentoo-stats if they had to do more then set up the crontab
> once...

You have several valid points. I agree that people are lazy and they
would probably not run gentoo-stats if it wasn't croned. Although I
don't see the need of a script that would check for if a package was
emerged correctly. This could as well be done by emerge when it
detects it is installing an ~ARCH package. This way you don't have to
wonder if it's a 'current' package or not.

Heck, you could use the same mechanism for reporting 'stable' packages
when they fail.

However, a successful emerge is only one of the critieria for an
unstable package to move to stable. It not crashing runtime is
another...

Actually, the more I think of it, the whole point is moot...
stable.gentoo.org might be a good idea, but why not replace it with
stats.gentoo.org? If I submit a stat report which tells me that I have
the latest version of a package installed, you could assume that it
works for me. If not, I'd post a bugreport. And if I couldn't be arsed
to set up gentoo-stats and post bugs about things that doesn't work,
why would I use stable.gentoo.org anyway?

Please, don't see this as an attack on stable.gentoo.org, just trying
to see it in another perspective... And I probably missed something in
that generalization...

> Sorry for this rather long a quite unfocussed post, mid-term exams are
> on the horizon again so I maybe should just shut up and grab my books...

Exams are important, but so is gentoo. ;)

//H
-- 
To segfault is human; to bluescreen moronic. 

--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Mikael Andersson]

On Monday 14 April 2003 11.04, Fredrik Jagenheim wrote:
>
> However, a successful emerge is only one of the critieria for an
> unstable package to move to stable. It not crashing runtime is
> another...
>
> Actually, the more I think of it, the whole point is moot...
> stable.gentoo.org might be a good idea, but why not replace it with
> stats.gentoo.org? If I submit a stat report which tells me that I have
> the latest version of a package installed, you could assume that it
> works for me. If not, I'd post a bugreport. And if I couldn't be arsed
> to set up gentoo-stats and post bugs about things that doesn't work,
> why would I use stable.gentoo.org anyway?
>
I don't think stable.gentoo.org is not a good solution since it's too much
manual work included from a user for (apparently nothing) in return.

I think the most efficient way to mark packages stable is statistics based.

If you compare number of installations of a  package against number of bugs
filed and their severity i think you should get pretty decent stability 
figures for most packages. The exception to this is packages with few users
but the users of such packages is probably more interested in 'voting' for 
their packages.

This is only an initial suggestion, please comment and improve :)

1) Successful Emerges/Bugs
  a) Count package downloads and bugs filed. If no blocker/critical bugs 
exists after a week or two mark as stable. For important packages this rule
could be made more stringent.
  b) Count real merges/unmerges of packages and not only package downloads, 
this should to opt-in since it would in some way need to post information 
back to gentoo.org

2) Voting console
A console application which finds all stable packages installed more than
a week or two ago and asks you to determine how much you've used them.
As default it only needs to show packages with few users since simple 
statistics should be enough to validate packages with many users. At the
end of the emerge remind a user of 'unpopular' packages to run the voting
console to mark as stable.

> Please, don't see this as an attack on stable.gentoo.org, just trying
> to see it in another perspective... And I probably missed something in
> that generalization...
>
> > Sorry for this rather long a quite unfocussed post, mid-term exams are
> > on the horizon again so I maybe should just shut up and grab my books...
>
> Exams are important, but so is gentoo. ;)
>
+1

> //H


/Mikael Andersson

--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Alec Berryman]

[-- Attachment #1: Type: text/plain, Size: 1236 bytes --]

On Mon, 2003-04-14 at 11:29, Mikael Andersson wrote:
> This is only an initial suggestion, please comment and improve :)
> 
> 1) Successful Emerges/Bugs
>   a) Count package downloads and bugs filed. If no blocker/critical bugs 
> exists after a week or two mark as stable. For important packages this rule
> could be made more stringent.
>   b) Count real merges/unmerges of packages and not only package downloads, 
> this should to opt-in since it would in some way need to post information 
> back to gentoo.org
> 
> 2) Voting console
> A console application which finds all stable packages installed more than
> a week or two ago and asks you to determine how much you've used them.
> As default it only needs to show packages with few users since simple 
> statistics should be enough to validate packages with many users. At the
> end of the emerge remind a user of 'unpopular' packages to run the voting
> console to mark as stable.

http://bugs.gentoo.org/show_bug.cgi?id=12785 was opened at the end of
December.  Blizzy left the dev team and currently there's no one in
charge of gentoo-stable and gentoo-stats as far as I know - the position
is still open.

-- 

Alec Berryman <alec@lorax.wox.org>

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Brad Laue]

On Mon, 2003-04-14 at 16:31, Alec Berryman wrote:
> http://bugs.gentoo.org/show_bug.cgi?id=12785 was opened at the end of
> December.  Blizzy left the dev team and currently there's no one in
> charge of gentoo-stable and gentoo-stats as far as I know - the position
> is still open.

What is required to run gentoo-stable?

Brad


--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Alec Berryman]

[-- Attachment #1: Type: text/plain, Size: 532 bytes --]

On Mon, 2003-04-14 at 16:48, Brad Laue wrote:
> On Mon, 2003-04-14 at 16:31, Alec Berryman wrote:
> > http://bugs.gentoo.org/show_bug.cgi?id=12785 was opened at the end of
> > December.  Blizzy left the dev team and currently there's no one in
> > charge of gentoo-stable and gentoo-stats as far as I know - the position
> > is still open.
> 
> What is required to run gentoo-stable?

Gentoo-stable is in PHP (probably with MySQL).  I understand that
gentoo-stats is in perl.

-- 

Alec Berryman <alec@lorax.wox.org>

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Mikael Andersson]

On Monday 14 April 2003 21.58, Alec Berryman wrote:
> On Mon, 2003-04-14 at 16:48, Brad Laue wrote:
> > On Mon, 2003-04-14 at 16:31, Alec Berryman wrote:
> > > http://bugs.gentoo.org/show_bug.cgi?id=12785 was opened at the end of
> > > December.  Blizzy left the dev team and currently there's no one in
> > > charge of gentoo-stable and gentoo-stats as far as I know - the
> > > position is still open.

I might be interested in helping with this but i'll have too look at it a bit 
more in detail before i volunteer some part of my time to this.

> >
> > What is required to run gentoo-stable?
>
> Gentoo-stable is in PHP (probably with MySQL).  
As far as i can see in cvs[1] it's in Java unless this isn't the code
that is actually running Gentoo-stable.

>I understand that gentoo-stats is in perl.
Where is the source to the web-frontend/server to be found ? I know 
(now) that the client is in portage. But i guess it's somewhere in cvs but i 
can't find any information on where the cvs repository is to do a co and try 
to find it myself. The web interface to cvs is not the best tool to find a 
package you don't know where it resides .)

[1] http://cvs.gentoo.org/cgi-bin/viewcvs.cgi/gentoo-src/blizzy/gentoo-stable/

/Mikael Andersson

--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Brad Laue]

On Mon, 2003-04-14 at 12:29, Mikael Andersson wrote:
> I don't think stable.gentoo.org is not a good solution since it's too much
> manual work included from a user for (apparently nothing) in return.

Agreed.

> I think the most efficient way to mark packages stable is statistics based.
> 
> If you compare number of installations of a  package against number of bugs
> filed and their severity i think you should get pretty decent stability 
> figures for most packages. The exception to this is packages with few users
> but the users of such packages is probably more interested in 'voting' for 
> their packages.

A tinderbox would be good to work around problems with 'unpopular'
packages. Over the course of this thread I've seen several problems
which an automated build and report system would solve.

> This is only an initial suggestion, please comment and improve :)
> 
> 1) Successful Emerges/Bugs
>   a) Count package downloads and bugs filed. If no blocker/critical bugs 
> exists after a week or two mark as stable. For important packages this rule
> could be made more stringent.

Good idea.

>   b) Count real merges/unmerges of packages and not only package downloads, 
> this should to opt-in since it would in some way need to post information 
> back to gentoo.org

The package system should probably be self-sufficient; the userbase is
too ephemeral to rely on for something like this; stable.gentoo.org is
evidence of that; a small fraction of the userbase is a) aware of it, b)
interested in using it, and c) interested in using it often enough.

Probably the only input a package should receive from a person is from
the maintainer itself.

Which leads me into another problem; currently there are no official
maintainers for a large number of the ebuilds in the tree. This prevents
the above from being doable; no one is around to represent and vouch for
the functionality of those ebuilds, just the bug reporting system.

So, is a tinderbox doable? One would have to be volunteered for each
supported architecture, or at least x86, sparc and powerpc to begin
with. One foreseeable complication would be the nearly infinite number
of combinations of USE flags, but I'm sure with discussion a way around
this could be found.

Thoughts?

Brad


--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Brad Laue]

On Mon, 2003-04-14 at 03:32, Fredrik Jagenheim wrote:
> Thus, I propose an extension to emerge that would look through your
> system and see which packages are marked as 'unstable' and ask the
> user interactively if they think the packages are OK. The responses
> could be of the type:
> 1) Yes, I've used it extensively and it works.
> 2) Yes, I've used it somewhat and it seems to work.
> 3) No idea, I don't think I've used it, but nothing is broke.
> 4) No idea, I haven't used it at all yet.
> 5) No, it doesn't work and I've used bugzilla to report the bug.

With some criteria this would be a good idea.

A big question is whether or not Gentoo should concern itself with the
operational functionality of the package in question. A 'does it build'
criteria would very efficiently mark packages stable.

But what if a new point release of foo introduces a bug that causes a
crash? Should Gentoo be responsible for marking the package stable or
unstable on that basis?

I don't think this is practical, although this is what seems to occur
fairly regularly based on my perusal of bugs.gentoo.org. This tends to
cause ebuilds which otherwise work fine to remain marked unstable even
if the user reporting the bug has made a mistake.

Perhaps a combination. If it builds succesfully with ALSA, KDE, GNOME in
the USE flags and sufficient numbers of people report this, mark it
stable. Also leave three prior release versions available for
installation marked stable in case the latest introduces a flaw caused
by programmer error. Introduce patches and fixes to the ebuilds as bugs
are reported.

--
// -- http://www.BRAD-X.com/ -- //


--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[C. Brewer]

[-- Attachment #1: Type: text/plain, Size: 1520 bytes --]

On 14 Apr 2003 11:07:01 -0400
Brad Laue <brad@brad-x.com> wrote:

> Perhaps a combination. If it builds succesfully with ALSA, KDE, GNOME in
> the USE flags and sufficient numbers of people report this, mark it
> stable. Also leave three prior release versions available for
> installation marked stable in case the latest introduces a flaw caused
> by programmer error. Introduce patches and fixes to the ebuilds as bugs
> are reported.

While I don't oppose your or anyones use of KDE, GNOME, or alsa, some things must be taken into consideration. I, for one, don't use a desktop environment, preferring a window-manager for my tasks, and I'm sure that I am not alone. Therefore you have negated that many testers with this sort of proposal (I for one run completely unstable). Secondly, while alsa may be the sound system of the future, it does not support all cards and some people are restricted to the kernel mods. Shorten your list of testers again. And since KDE and such require X, you leave out all of your CLI users, which will probably be your most critical old-schoolers and ex-UNIX peeps (gross stereotyping here). I think that it should stay that if it builds and runs with a reasonable amount of positive reports vs. bugs then it should go stable. I just wanted to point out that certain matters of selectiveness will negate your amount of testers, and probably work against your ends.


-- 
Chuck Brewer
Registered Linux User #284015
Get my gpg public key at pgp.mit.edu!! Encrypted e-mail preferred.



[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Abhishek Amit]

On 12:39 Mon 14 Apr     , C. Brewer wrote:
> On 14 Apr 2003 11:07:01 -0400
> Brad Laue <brad@brad-x.com> wrote:
> 
> > Perhaps a combination. If it builds succesfully with ALSA, KDE, GNOME in
> > the USE flags and sufficient numbers of people report this, mark it
> > stable. Also leave three prior release versions available for
> > installation marked stable in case the latest introduces a flaw caused
> > by programmer error. Introduce patches and fixes to the ebuilds as bugs
> > are reported.
> 
> While I don't oppose your or anyones use of KDE, GNOME, or alsa, some things must be taken into consideration. I, for one, don't use a desktop environment, preferring a window-manager for my tasks, and I'm sure that I am not alone. Therefore you have negated that many testers with this sort of proposal (I for one run completely unstable). Secondly, while alsa may be the sound system of the future, it does not support all cards and some people are restricted to the kernel mods. Shorten your list of testers again. And since KDE and such require X, you leave out all of your CLI users, which will probably be your most critical old-schoolers and ex-UNIX peeps (gross stereotyping here). I think that it should stay that if it builds and runs with a reasonable amount of positive reports vs. bugs then it should go stable. I just wanted to point out that certain matters of selectiveness will negate your amount of testers, and probably work against your ends.
> 
> 
> -- 
> Chuck Brewer
> Registered Linux User #284015
> Get my gpg public key at pgp.mit.edu!! Encrypted e-mail preferred.
> 
> 


While we may not(we being users of wms only) use desktop enviorments, KDE and GNOME are still in the default USE flags. ALSA is also in there(/etc/make.profile/use.defaults). I think that the defualt flags should be the basis of testing, and note I am writing this from mutt in flxbox.


--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Fredrik Jagenheim]

On Mon, Apr 14, 2003 at 11:21:46PM -0700, Abhishek Amit wrote:
> While we may not(we being users of wms only) use desktop
> enviorments, KDE and GNOME are still in the default USE flags. ALSA
> is also in there(/etc/make.profile/use.defaults). I think that the
> defualt flags should be the basis of testing, and note I am writing
> this from mutt in flxbox.

But, remember that the default flags are what maintainers usually are
running. So a tinderbox system would only test exactly what the
maintainers are testing, and thus wouldn't give any idea of how good
an ebuild was.

I have no statistics or not even investigated it, but I have a hunch
that most ebuilds fail because of;
*) Forgotten dependencies
*) virtual dependencies satisfies another package than the maintainer
are running
*) Strange USE flags

I don't think it would be feasable for a tinderbox to test all the
possible combinations for every new ebuild.

Although, a tinderbox would be nice for maintainers that doesn't have
access to sparc/powerpc architectures to test that their packages
atleast build on them. But gentoo perhaps already has a couple of
servers that can be used to manually test different archs?

//H

Btw, if you're using mutt... Can't you investigate if there are some
way to make your editor do linebreaks? ;)

-- 
To segfault is human; to bluescreen moronic. 

--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[C. Brewer]

[-- Attachment #1: Type: text/plain, Size: 2376 bytes --]

On Mon, 14 Apr 2003 23:21:46 -0700
Abhishek Amit <abhishekamit2000@yahoo.com> wrote:

> On 12:39 Mon 14 Apr     , C. Brewer wrote:
> > On 14 Apr 2003 11:07:01 -0400
> > Brad Laue <brad@brad-x.com> wrote:
> > 
> > > Perhaps a combination. If it builds succesfully with ALSA, KDE, GNOME in
> > > the USE flags and sufficient numbers of people report this, mark it
> > > stable. Also leave three prior release versions available for
> > > installation marked stable in case the latest introduces a flaw caused
> > > by programmer error. Introduce patches and fixes to the ebuilds as bugs
> > > are reported.
> > 
> > While I don't oppose your or anyones use of KDE, GNOME, or alsa, some things must be taken into consideration. I, for one, don't use a desktop environment, preferring a window-manager for my tasks, and I'm sure that I am not alone. Therefore you have negated that many testers with this sort of proposal (I for one run completely unstable). Secondly, while alsa may be the sound system of the future, it does not support all cards and some people are restricted to the kernel mods. Shorten your list of testers again. And since KDE and such require X, you leave out all of your CLI users, which will probably be your most critical old-schoolers and ex-UNIX peeps (gross stereotyping here). I think that it should stay that if it builds and runs with a reasonable amount of positive reports vs. bugs then it should go stable. I just wanted to point out that certain matters of selectiveness will negate your amount of testers, and probably work against your ends.
> > 
> > 
> > -- 
> > Chuck Brewer
> > Registered Linux User #284015
> > Get my gpg public key at pgp.mit.edu!! Encrypted e-mail preferred.
> > 
> > 
> 
> 
> While we may not(we being users of wms only) use desktop enviorments, KDE and GNOME are still in the default USE flags. ALSA is also in there(/etc/make.profile/use.defaults). I think that the defualt flags should be the basis of testing, and note I am writing this from mutt in flxbox.
> 
I realize that those are the default flags, and I can't speak for anyone else, but the first thing I did was "-" all that DE noise, so I don't have those, but I'm still unstable, ans testing, when other won't..so do I not count now?:)

-- 
Chuck Brewer
Registered Linux User #284015
Get my gpg public key at pgp.mit.edu!! Encrypted e-mail preferred.



[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Brad Laue]

On Mon, 2003-04-14 at 15:39, C. Brewer wrote:

> While I don't oppose your or anyones use of KDE, GNOME, or alsa, some things must be taken into consideration. I, for one, don't use a desktop environment, preferring a window-manager for my tasks, and I'm sure that I am not alone. Therefore you have negated that many testers with this sort of proposal (I for one run completely unstable). Secondly, while alsa may be the sound system of the future, it does not support all cards and some people are restricted to the kernel mods. Shorten your list of testers again. And since KDE and such require X, you leave out all of your CLI users, which will probably be your most critical old-schoolers and ex-UNIX peeps (gross stereotyping here). I think that it should stay that if it builds and runs with a reasonable amount of positive reports vs. bugs then it should go stable. I just wanted to point out that certain matters of selectiveness will negate your amount of testers, and probably work against your ends.
> 

Absolutely. One set of CFLAGS would have to logically exclude GNOME and
KDE extensions, to see whether the ebuild builds and merges without
these.

Brad


--
gentoo-dev@gentoo.org mailing list

RE: [gentoo-dev] Is there a process for marking ebuilds stable?

[Todd Wright]

> Which leads me into another problem; currently there are no official
> maintainers for a large number of the ebuilds in the tree. 
> This prevents
> the above from being doable; no one is around to represent 
> and vouch for
> the functionality of those ebuilds, just the bug reporting system.

To add my 2 cents worth on this point...  even when there is a maintainer, the dev team are reluctant to believe them when they vouch for the ebuild or package and would prefer to modify the ebuilds without any reasonable knowledge of the package and the impacts that those modifications may have. I had a lot of trouble getting the dev team to place my unmodified hercules-2.17.1 ebuild in the portage tree, despite the fact that I am involved with the hercules development team. The gentoo developers seem to have a 'we know better than you' attitude.

-- _--_|\ --------- Todd Wright -- wylie@geekasylum.org --------
  /      \            ICQ: 9589981   YIM: mvs38j
  \_.--._*  <---   http://www.geekasylum.org/~wylie/
        v       Mobile: +61-403-796-001    Ph: +61-2-9699-1746
----------------------------------------------------------------




--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Jon Portnoy]

It's not that, although there may be some developers with a bad 
attitude. It's that we can't automatically trust everyone who comes up 
with an ebuild and says "look, I know all about this, just put it in."

Would you prefer we _weren't_ concerned about quality assurance?

On Tue, Apr 15, 2003 at 12:08:22PM +1000, Todd Wright wrote:
> 
> > Which leads me into another problem; currently there are no official
> > maintainers for a large number of the ebuilds in the tree. 
> > This prevents
> > the above from being doable; no one is around to represent 
> > and vouch for
> > the functionality of those ebuilds, just the bug reporting system.
> 
> To add my 2 cents worth on this point...  even when there is a maintainer, the dev team are reluctant to believe them when they vouch for the ebuild or package and would prefer to modify the ebuilds without any reasonable knowledge of the package and the impacts that those modifications may have. I had a lot of trouble getting the dev team to place my unmodified hercules-2.17.1 ebuild in the portage tree, despite the fact that I am involved with the hercules development team. The gentoo developers seem to have a 'we know better than you' attitude.
> 
> -- _--_|\ --------- Todd Wright -- wylie@geekasylum.org --------
>   /      \            ICQ: 9589981   YIM: mvs38j
>   \_.--._*  <---   http://www.geekasylum.org/~wylie/
>         v       Mobile: +61-403-796-001    Ph: +61-2-9699-1746
> ----------------------------------------------------------------
> 
> 
> 
> 
> --
> gentoo-dev@gentoo.org mailing list

-- 
Jon Portnoy
avenj/irc.freenode.net

--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Is there a process for marking ebuilds stable?

[Dylan Carlson]

On Monday 14 April 2003 10:08 pm, Todd Wright wrote:
>  I had a lot of trouble getting the dev team to place my unmodified
> hercules-2.17.1 ebuild in the portage tree, despite the fact that I am
> involved with the hercules development team. The gentoo developers seem
> to have a 'we know better than you' attitude.

That's a bit presumptuous, and offhand rude...

Ultimately the the Gentoo developer who commits the ebuild is responsible 
for the ebuild, not you.  Towards this, we appreciate ebuild submissions 
from users and outside developers, however: every ebuild has to go through 
a minimal QA process.  

We're not going to immediately commit an ebuild because someone tells us 
that they're associated with the dev team of that particular package.  It 
has to be tested just like everything else.  

If you're not happy with that, you have options:  

(x) you can host the ebuild on your website, and point interested parties 
to it until we test it and put it in the tree...

(y) you can switch to another distribution which I can almost guarantee 
will move slower than we do...

(z) you can wait patiently while we do the best we can to satisfy all of 
these requests and maintaining quality control.  

Remember, this is open source development.   You need to check your 
expectations a little.  The world doesn't move just because you snap your 
fingers and demand that it does.  

Cheers,
Dylan Carlson

Public Key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x708E165F
Key fingerprint = 3AEA DE38 FE42 15A6 C0E2 730E 3D04 BCC1 708E 165F

--
gentoo-dev@gentoo.org mailing list