[gentoo-releng] USE flags and stages 1 and 2

[gentoo-releng] USE flags and stages 1 and 2

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 1665 bytes --]

I'm starting to get more and more bugs from users whom have decided,
against the recommendation of those of us in Release Engineering, to do
a stage1 or stage2 installation using their custom USE flags, and
finding that their installation is unable to complete properly.  The
problem stems from more and more packages adding more and more USE
flags.  We already recommend that all users perform a stage3
installation, but this doesn't appear to be enough to resolve the
problem.  What it really boils down to is the inability for us to test
with all of the possible USE flag combinations and ever get a release
out the door.

As it stands, I only see a few options, none of which sound very good:

1.  Only release stage3 tarballs
2.  Inform users that only stage3 will be supported
3.  Change the documentation to recommend users not change USE flags
until after the completion of "emerge -e system"

In pretty much every case, the real answer is "quit using stage1 if you
don't know what you're doing" but unfortunately, we're going to get the
pointless "but Gentoo is about choice" "argument" that really has no
bearing on the truth of what is Gentoo.  Gentoo is *not* about choice.
Gentoo is about empowering the user to make the system as he sees fit.
This means there is a certain expectation that when you start fiddling
with stuff, you're going to pick up the pieces on your own.  At any
rate, the problem is only going to get worse as more and more USE flags
are added.

What can we do to curb this problem?

-- 
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux

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

Re: [gentoo-releng] USE flags and stages 1 and 2

[M. Edward (Ed) Borasky]

Chris Gianelloni wrote:

>I'm starting to get more and more bugs from users whom have decided,
>against the recommendation of those of us in Release Engineering, to do
>a stage1 or stage2 installation using their custom USE flags, and
>finding that their installation is unable to complete properly.  The
>problem stems from more and more packages adding more and more USE
>flags.  We already recommend that all users perform a stage3
>installation, but this doesn't appear to be enough to resolve the
>problem.  What it really boils down to is the inability for us to test
>with all of the possible USE flag combinations and ever get a release
>out the door.
>
>As it stands, I only see a few options, none of which sound very good:
>
>1.  Only release stage3 tarballs
>2.  Inform users that only stage3 will be supported
>3.  Change the documentation to recommend users not change USE flags
>until after the completion of "emerge -e system"
>
>In pretty much every case, the real answer is "quit using stage1 if you
>don't know what you're doing" but unfortunately, we're going to get the
>pointless "but Gentoo is about choice" "argument" that really has no
>bearing on the truth of what is Gentoo.  Gentoo is *not* about choice.
>Gentoo is about empowering the user to make the system as he sees fit.
>This means there is a certain expectation that when you start fiddling
>with stuff, you're going to pick up the pieces on your own.  At any
>rate, the problem is only going to get worse as more and more USE flags
>are added.
>
>What can we do to curb this problem?
>  
>
Well ... at one time, Gentoo *was* about choice, but it's evolved to 
being about quality IMHO. As long as you're willing to make all the 
stage3 tarballs required, I see no reason to continue to support stage1 
installs. The past couple of installs I did were stage1 -- I wanted to 
see how long it took, among other things. And I've never done a stage2 
... I've never seen the point in it.

So my recommendation would be to eliminate stage1 and stage2 and just do 
stage3 installs. They're a lot faster too -- I think getting through 
stage1, stage2 and stage3 took about 6 hours on a 1 GHz P3, hardly a 
productive use of time. I think you might want to put a few more goodies 
in stage3, though -- vim for sure, and niceties like ufed, esearch, 
slocate, genkernel and grub at a minimum. When I do an install, the 
things I *always* bring in, whether desktop or server are:

vim grub metalog vixie-cron slocate coldplug hotplug gentoo-sources 
genkernel ntp
zip unzip esearch gentoolkit rkhunter chkrootkit sysstat procinfo 
mirrorselect ufed



-- 
M. Edward (Ed) Borasky

http://linuxcapacityplanning.com

-- 
gentoo-releng@gentoo.org mailing list

Re: [gentoo-releng] USE flags and stages 1 and 2

[Michael Crute]

[-- Attachment #1: Type: text/plain, Size: 1652 bytes --]

On 10/31/05, Chris Gianelloni <wolf31o2@gentoo.org> wrote:
>
> As it stands, I only see a few options, none of which sound very good:
>
> 1. Only release stage3 tarballs


This is the worst of all these options. I for one have never done a stage3
install and never plan/want to. I am pretty sure there are others out there
like me that would be extremely ticked off to see our stage1 tarballs
disappear.

2. Inform users that only stage3 will be supported


This may actually be your best route. Any bugs that come in because users
customized use flags on a non-stage3 install should be closed RTFM.

3. Change the documentation to recommend users not change USE flags
> until after the completion of "emerge -e system"


You could try this first but if people think they are smart enough to change
the use flags (even if they really aren't) then there isn't much of a chance
you're going to talk them out of it.

I am not sure how the other (non-minimal) cd's are setup but you could just
include stage3's on the cds and make people go out to the mirrors to get
their stage1 or 2 tarballs. Basically make it a little bit more difficult to
get them.

Another thought would be an approach similar to the installer irc channel
where you have to read through a really long FAQ before you are told how to
download the stage1 or 2 tarballs.

Just my thoughts as an end user.

-Mike

--
________________________________
Michael E. Crute
Software Developer
SoftGroup Development Corporation

Linux, because reboots are for installing hardware.
"In a world without walls and fences, who needs windows and gates?"

[-- Attachment #2: Type: text/html, Size: 2411 bytes --]

Re: [gentoo-releng] USE flags and stages 1 and 2

[Jory A. Pratt]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Chris Gianelloni wrote:
> I'm starting to get more and more bugs from users whom have decided,
> against the recommendation of those of us in Release Engineering, to do
> a stage1 or stage2 installation using their custom USE flags, and
> finding that their installation is unable to complete properly.  The
> problem stems from more and more packages adding more and more USE
> flags.  We already recommend that all users perform a stage3
> installation, but this doesn't appear to be enough to resolve the
> problem.  What it really boils down to is the inability for us to test
> with all of the possible USE flag combinations and ever get a release
> out the door.
> 
> As it stands, I only see a few options, none of which sound very good:
> 
> 1.  Only release stage3 tarballs
> 2.  Inform users that only stage3 will be supported
> 3.  Change the documentation to recommend users not change USE flags
> until after the completion of "emerge -e system"
> 
> In pretty much every case, the real answer is "quit using stage1 if you
> don't know what you're doing" but unfortunately, we're going to get the
> pointless "but Gentoo is about choice" "argument" that really has no
> bearing on the truth of what is Gentoo.  Gentoo is *not* about choice.
> Gentoo is about empowering the user to make the system as he sees fit.
> This means there is a certain expectation that when you start fiddling
> with stuff, you're going to pick up the pieces on your own.  At any
> rate, the problem is only going to get worse as more and more USE flags
> are added.
> 
> What can we do to curb this problem?
> 
I am in favor of doing nothing but putting a warning on stage1 and stage
2 saying only useflags that will be supported are ntpl ntplonly
everything can and should be default IMHO. Then if they run into
problems it is their problem and not ours. As much as we are about
quality we still have a choice and if I choice a stage1 I do it for my
own personal reasons. I am basically favoring do not edit useflags until
stage3 if you make changes at this point you can use emerge -euD system
- --newuse to rebuild packages to support your USE flags.

That is just a small piece of my opinion.


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFDZpJXGDfjNg8unQIRAj6+AJ9j8xAGVA8XVCtmCec2PuiHjNa+igCeLawn
uBtOw+At0NitqN5SbYleoEc=
=Z4UA
-----END PGP SIGNATURE-----
-- 
gentoo-releng@gentoo.org mailing list

Re: [gentoo-releng] USE flags and stages 1 and 2

[Michael Crute]

[-- Attachment #1: Type: text/plain, Size: 3650 bytes --]

On 10/31/05, M. Edward (Ed) Borasky <znmeb@cesmail.net> wrote:
>
> Chris Gianelloni wrote:
> >I'm starting to get more and more bugs from users whom have decided,
> >against the recommendation of those of us in Release Engineering, to do
> >a stage1 or stage2 installation using their custom USE flags, and
> >finding that their installation is unable to complete properly. The
> >problem stems from more and more packages adding more and more USE
> >flags. We already recommend that all users perform a stage3
> >installation, but this doesn't appear to be enough to resolve the
> >problem. What it really boils down to is the inability for us to test
> >with all of the possible USE flag combinations and ever get a release
> >out the door.
> >
> >As it stands, I only see a few options, none of which sound very good:
> >
> >1. Only release stage3 tarballs
> >2. Inform users that only stage3 will be supported
> >3. Change the documentation to recommend users not change USE flags
> >until after the completion of "emerge -e system"
> >
> >In pretty much every case, the real answer is "quit using stage1 if you
> >don't know what you're doing" but unfortunately, we're going to get the
> >pointless "but Gentoo is about choice" "argument" that really has no
> >bearing on the truth of what is Gentoo. Gentoo is *not* about choice.
> >Gentoo is about empowering the user to make the system as he sees fit.
> >This means there is a certain expectation that when you start fiddling
> >with stuff, you're going to pick up the pieces on your own. At any
> >rate, the problem is only going to get worse as more and more USE flags
> >are added.
> >
> >What can we do to curb this problem?
>
>
> Well ... at one time, Gentoo *was* about choice, but it's evolved to
> being about quality IMHO. As long as you're willing to make all the
> stage3 tarballs required, I see no reason to continue to support stage1
> installs. The past couple of installs I did were stage1 -- I wanted to
> see how long it took, among other things. And I've never done a stage2
> ... I've never seen the point in it.
>
> So my recommendation would be to eliminate stage1 and stage2 and just do
> stage3 installs. They're a lot faster too -- I think getting through
> stage1, stage2 and stage3 took about 6 hours on a 1 GHz P3, hardly a
> productive use of time. I think you might want to put a few more goodies
> in stage3, though -- vim for sure, and niceties like ufed, esearch,
> slocate, genkernel and grub at a minimum. When I do an install, the
> things I *always* bring in, whether desktop or server are: [snip]
>

I REALLY dislike the idea of eliminating stage1 tarballs. I am personally
very picky about my system and what gets installed and from my point of view
installing from a stage3 is as bad as installing Fedora. Plus it would be a
real pain in the butt to have to go backwards, installing a bunch of stuff
up front and then going back by hand and removing it. I do share Ed's
sentiments about stage2 though, you could eliminate those as I don't see
much point to them. You either want it all or you want the bare minimum, I
want the bare minimum.

Another thought is release the tarballs but do not support anyone with
problems doing a stage1 install. If you are brave enough to attempt it, you
go it alone.

I think I've put in about $0.04 now ;)

-Mike

--
________________________________
Michael E. Crute
Software Developer
SoftGroup Development Corporation

Linux, because reboots are for installing hardware.
"In a world without walls and fences, who needs windows and gates?"

[-- Attachment #2: Type: text/html, Size: 4276 bytes --]

Re: [gentoo-releng] USE flags and stages 1 and 2

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 1710 bytes --]

On Mon, 2005-10-31 at 13:43 -0800, M. Edward (Ed) Borasky wrote:
> Well ... at one time, Gentoo *was* about choice, but it's evolved to 
> being about quality IMHO. As long as you're willing to make all the 
> stage3 tarballs required, I see no reason to continue to support stage1 
> installs. The past couple of installs I did were stage1 -- I wanted to 
> see how long it took, among other things. And I've never done a stage2 
> ... I've never seen the point in it.

Actually, it was a few very vocal users who made these claims long ago
and many people simply picked up on it.  It's *never* been about choice
for the sake of choice.  It has always been about empowerment.

> So my recommendation would be to eliminate stage1 and stage2 and just do 
> stage3 installs. They're a lot faster too -- I think getting through 
> stage1, stage2 and stage3 took about 6 hours on a 1 GHz P3, hardly a 
> productive use of time. I think you might want to put a few more goodies 
> in stage3, though -- vim for sure, and niceties like ufed, esearch, 
> slocate, genkernel and grub at a minimum. When I do an install, the 
> things I *always* bring in, whether desktop or server are:

We would absolutely not add any bloat to the stage3 tarballs.  The
stages are supposed to be minimal sets of things that are absolutely
required based on the default USE flags and nothing more.

Yes, I would love to see vim and gentoolkit in all of the stages, along
with distcc and ccache, but it isn't ever going to happen and I happen
to agree with the reasoning for *not* doing it.

-- 
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux

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

Re: [gentoo-releng] USE flags and stages 1 and 2

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 4155 bytes --]

On Mon, 2005-10-31 at 16:51 -0500, Michael Crute wrote:
> On 10/31/05, Chris Gianelloni <wolf31o2@gentoo.org> wrote:
>         As it stands, I only see a few options, none of which sound
>         very good:
>         
>         1.  Only release stage3 tarballs
> 
> This is the worst of all these options. I for one have never done a
> stage3 install and never plan/want to. I am pretty sure there are
> others out there like me that would be extremely ticked off to see our
> stage1 tarballs disappear.

Honestly, no offense, but I don't care.

You are wasting your time and don't even realize how badly you are doing
it.  Ever since we added full vdb support into the stages (2005.0) it is
a complete waste of time to do a stage1 installation.  Exactly what do
you gain from doing a stage3, editing make.conf, then doing an emerge -e
world?

Now, think about this.  When doing a stage1 installation, you have to
bootstrap, which compiles a minimalized version of the toolchain.  After
this, you run an emerge -e world, which recompiles *all* of the
toolchain, along with a ton more packages.  So you are compiling the
entire toolchain twice, the first time giving you no benefit,
whatsoever, other than allowing you to enable a few options which can
honestly be turned on at any time, such as nptl.

What takes less time, compiling the entire toolchain from scratch, or
unpacking a tarball?  In either case, you're running an "emerge -e
system" to get your customized system.

>         2.  Inform users that only stage3 will be supported
> 
> This may actually be your best route. Any bugs that come in because
> users customized use flags on a non-stage3 install should be closed
> RTFM.

This is honestly probably going to be the route we're forced into, even
though it is not the optimal route for either the developers or the
users, simply because of the stubbornness of many of our developers and
vocal users that will refuse the dropping of stages 1 and 2 from our
mirrors.  I used to be one of these developers until I came in and
modified the way catalyst worked when producing stages.

Now both stages 1 and 2 are completely redundant.

>         3.  Change the documentation to recommend users not change USE
>         flags
>         until after the completion of "emerge -e system"
> 
> You could try this first but if people think they are smart enough to
> change the use flags (even if they really aren't) then there isn't
> much of a chance you're going to talk them out of it.

Except we would probably mark the bugs as INVALID.

> I am not sure how the other (non-minimal) cd's are setup but you could
> just include stage3's on the cds and make people go out to the mirrors
> to get their stage1 or 2 tarballs. Basically make it a little bit more
> difficult to get them. 

We currently only ship the stage3 tarballs on the Universal InstallCD.

> Another thought would be an approach similar to the installer irc
> channel where you have to read through a really long FAQ before you
> are told how to download the stage1 or 2 tarballs.

My personal impression is that this will piss off the end user.  This
works for the installer *channel* but would never work for a CD
installation.  It's simply asinine to ask the user to jump through hoops
to do what they want.

Even if we stopped shipping stage1 tarballs, there's nothing stopping
some inventive user from grabbing a stage3 tarball and making a stage1
tarball from it using catalyst.  It only takes a couple hours on a
decent machine.

My point is that once again, we're not *really* removing choice even if
we were to drop the earlier stages altogether, we're just making the
user do the work themselves and removing one more abysmal headache and
QA nightmare from our already enormous and growing list of stuff that
has been delaying the past few releases well beyond our scheduled
release dates.

> Just my thoughts as an end user.

They're appreciated.

Thanks.

-- 
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux

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

Re: [gentoo-releng] USE flags and stages 1 and 2

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 2850 bytes --]

On Mon, 2005-10-31 at 15:53 -0600, Jory A. Pratt wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Chris Gianelloni wrote:
> > I'm starting to get more and more bugs from users whom have decided,
> > against the recommendation of those of us in Release Engineering, to do
> > a stage1 or stage2 installation using their custom USE flags, and
> > finding that their installation is unable to complete properly.  The
> > problem stems from more and more packages adding more and more USE
> > flags.  We already recommend that all users perform a stage3
> > installation, but this doesn't appear to be enough to resolve the
> > problem.  What it really boils down to is the inability for us to test
> > with all of the possible USE flag combinations and ever get a release
> > out the door.
> > 
> > As it stands, I only see a few options, none of which sound very good:
> > 
> > 1.  Only release stage3 tarballs
> > 2.  Inform users that only stage3 will be supported
> > 3.  Change the documentation to recommend users not change USE flags
> > until after the completion of "emerge -e system"
> > 
> > In pretty much every case, the real answer is "quit using stage1 if you
> > don't know what you're doing" but unfortunately, we're going to get the
> > pointless "but Gentoo is about choice" "argument" that really has no
> > bearing on the truth of what is Gentoo.  Gentoo is *not* about choice.
> > Gentoo is about empowering the user to make the system as he sees fit.
> > This means there is a certain expectation that when you start fiddling
> > with stuff, you're going to pick up the pieces on your own.  At any
> > rate, the problem is only going to get worse as more and more USE flags
> > are added.
> > 
> > What can we do to curb this problem?
> > 
> I am in favor of doing nothing but putting a warning on stage1 and stage
> 2 saying only useflags that will be supported are ntpl ntplonly
> everything can and should be default IMHO. Then if they run into
> problems it is their problem and not ours. As much as we are about
> quality we still have a choice and if I choice a stage1 I do it for my
> own personal reasons. I am basically favoring do not edit useflags until
> stage3 if you make changes at this point you can use emerge -euD system
> - --newuse to rebuild packages to support your USE flags.

Have you read the bootstrap script?  Only nptl, nptlonly, nls, and
unicode are currently supported.  Everything else is filtered.

The problem is the "system" target having more and more packages that
have optional components with dependencies on things like a configured
kernel, which isn't done until well after the stage3 process is
completed.

-- 
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux

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

[gentoo-releng] Re: USE flags and stages 1 and 2

[Anthony Gorecki]

[-- Attachment #1: Type: text/plain, Size: 1573 bytes --]

On Monday, October 31, 2005 13:09, Chris Gianelloni wrote:
> 2.  Inform users that only stage3 will be supported
> 3.  Change the documentation to recommend users not change USE flags
> until after the completion of "emerge -e system"

These two options seem to be the most reasonable. If a user doesn't know how 
to deploy a system using stage one or two, they should either be instructed 
to use stage three, or be given alternate (more cautious) instructions for 
the earlier stages. If, on the other hand, the person is familiar with the 
inner workings of Gentoo and its packages, they would be likely to ignore the 
instructions anyway.


> In pretty much every case, the real answer is "quit using stage1 if you
> don't know what you're doing" but unfortunately, we're going to get the
> pointless "but Gentoo is about choice" "argument" that really has no
> bearing on the truth of what is Gentoo.

I agree. A section of documentation could be created to explain that concept, 
and bugs that were filed as a result of inexperience could then be closed and 
redirected to the documentation.

I use USE="-*" in make.conf for all of my deployments because I don't want the 
system's USE flags to be modified by Gentoo developers; I also don't have any 
difficulty deploying such systems. Unfortunately, I see many individuals that 
will leap to configure their systems in ways that cannot be maintained. In 
those cases, the Gentoo developers should not be the ones to correct the 
errors.


-- 
Anthony Gorecki
Ectro-Linux Foundation

[-- Attachment #2: Type: application/pgp-signature, Size: 828 bytes --]

Re: [gentoo-releng] USE flags and stages 1 and 2

[Seemant Kulleen]

Here's my take. The purpose of stage1 is to get you to stage2.  Stage
1's psychological purpose is "look ma, I can boostrap".  If we go to two
stages, take out 1, and let the emerge -e system step be the one that
the customisation starts at (including things like changing
/etc/portage/profile/virtuals to change virtual/editor to vim or emacs
or blah).  You don't do any of that in stage1, it's stage2 that all the
action actually happens.

Having said that, you will see a lot of noise about wanting stage1's.
Stage 3 has the advantage of speed, but the disadvantage of defaults
like nano (which then need to be expressly replaced).

Anyway, I don't think I've added any help in one direction or another,
just putting my thoughts out there.

Thanks,

Seemant
-- 
gentoo-releng@gentoo.org mailing list

Re: [gentoo-releng] USE flags and stages 1 and 2

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 1954 bytes --]

On Mon, 2005-10-31 at 16:59 -0500, Michael Crute wrote:
> I REALLY dislike the idea of eliminating stage1 tarballs. I am
> personally very picky about my system and what gets installed and from
> my point of view installing from a stage3 is as bad as installing
> Fedora. Plus it would be a real pain in the butt to have to go
> backwards, installing a bunch of stuff up front and then going back by
> hand and removing it. I do share Ed's sentiments about stage2 though,
> you could eliminate those as I don't see much point to them. You
> either want it all or you want the bare minimum, I want the bare
> minimum. 

Read my response to him to see why you're not saving yourself any time
or work.

> Another thought is release the tarballs but do not support anyone with
> problems doing a stage1 install. If you are brave enough to attempt
> it, you go it alone.

This was one of my original options, stated in my first email.  The only
issue with this is it still causes a great burden on the developers that
are responsible for release bugs.  We are severely overworked when
release time comes around, and having things such as a stage1 which
essentially allows a user to completely bypass all of our QA is a
logistical nightmare.

Choice be damned, if you can't ever get Gentoo on your system, who cares
if you got to make a few *choices* or not?

The removal of a stage1 tarball does not remove *any* choice from the
user doing the installation.  Anyone who tells you otherwise does not
know what they are talking about.

It *does* however guarantee that you have a minimal working system of
known quality with which to do your installation from, rather than a
completely hacked and limited functionality minimal environment designed
only to support the building of a minimal working system.

-- 
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux

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

[gentoo-releng] Re: USE flags and stages 1 and 2

[Anthony Gorecki]

[-- Attachment #1: Type: text/plain, Size: 497 bytes --]

On Monday, October 31, 2005 14:26, Chris Gianelloni wrote:
> It *does* however guarantee that you have a minimal working system of
> known quality with which to do your installation from, rather than a
> completely hacked and limited functionality minimal environment designed
> only to support the building of a minimal working system.

It is for that reason that I support keeping stage one tarballs available. 
Cheers for the stubborn.


-- 
Anthony Gorecki
Ectro-Linux Foundation

[-- Attachment #2: Type: application/pgp-signature, Size: 828 bytes --]

Re: [gentoo-releng] USE flags and stages 1 and 2

[Michael Crute]

[-- Attachment #1: Type: text/plain, Size: 2159 bytes --]

On 10/31/05, Chris Gianelloni <wolf31o2@gentoo.org> wrote:
>
> > Another thought would be an approach similar to the installer irc
> > channel where you have to read through a really long FAQ before you
> > are told how to download the stage1 or 2 tarballs.
>
> My personal impression is that this will piss off the end user. This
> works for the installer *channel* but would never work for a CD
> installation. It's simply asinine to ask the user to jump through hoops
> to do what they want.


Well I wasn't really referring to a CD (but in this case read the MS license
agreement when installing Windoze I think that would be the equivalent of
this :). What I was really referring to was putting up some sort of scenario
like that to download the stage1/2 tarballs.

Even if we stopped shipping stage1 tarballs, there's nothing stopping
> some inventive user from grabbing a stage3 tarball and making a stage1
> tarball from it using catalyst. It only takes a couple hours on a
> decent machine.


True but if they fire up catalyst to make a stage1 chances are they aren't
going to be so dumb as to request support when their use flag combinations
don't work. The problem is ignorant users and most of them can't us
catalyst.

My point is that once again, we're not *really* removing choice even if
> we were to drop the earlier stages altogether, we're just making the
> user do the work themselves and removing one more abysmal headache and
> QA nightmare from our already enormous and growing list of stuff that
> has been delaying the past few releases well beyond our scheduled
> release dates.
>

Regardless of how foolish you may think it is to do a stage1 install (and it
may well be) I would still greatly appreciate the stage1 tarballs even if
you refuse to support them. Personally I have never needed support on a
stage1 that I couldn't get by simply RTFM.

-Mike

--
________________________________
Michael E. Crute
Software Developer
SoftGroup Development Corporation

Linux, because reboots are for installing hardware.
"In a world without walls and fences, who needs windows and gates?"

[-- Attachment #2: Type: text/html, Size: 2823 bytes --]

Re: [gentoo-releng] USE flags and stages 1 and 2

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 1575 bytes --]

On Mon, 2005-10-31 at 17:25 -0500, Seemant Kulleen wrote:
> Here's my take. The purpose of stage1 is to get you to stage2.  Stage
> 1's psychological purpose is "look ma, I can boostrap".  If we go to two
> stages, take out 1, and let the emerge -e system step be the one that
> the customisation starts at (including things like changing
> /etc/portage/profile/virtuals to change virtual/editor to vim or emacs
> or blah).  You don't do any of that in stage1, it's stage2 that all the
> action actually happens.

The bugs I am receiving are all in the "emerge -e system" step due to
requirements on configured kernels and improper dependency tracking.
The dependency issue isn't going to resolve itself, but there is nothing
to resolve the "kernel-dependent packages in system" problem other than
not having such a broken system to begin with for the user.

> Having said that, you will see a lot of noise about wanting stage1's.
> Stage 3 has the advantage of speed, but the disadvantage of defaults
> like nano (which then need to be expressly replaced).

Agreed.  There are some drawbacks, but I've found that these are still
minimal compared to compiling an entire USE="bootstrap" toolchain right
before compiling the entire "system" target, including toolchain, all
over again.

> Anyway, I don't think I've added any help in one direction or another,
> just putting my thoughts out there.

They're very much appreciated.

-- 
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux

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

Re: [gentoo-releng] Re: USE flags and stages 1 and 2

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 2867 bytes --]

On Mon, 2005-10-31 at 14:22 -0800, Anthony Gorecki wrote:
> On Monday, October 31, 2005 13:09, Chris Gianelloni wrote:
> > 2.  Inform users that only stage3 will be supported
> > 3.  Change the documentation to recommend users not change USE flags
> > until after the completion of "emerge -e system"
> 
> These two options seem to be the most reasonable. If a user doesn't know how 
> to deploy a system using stage one or two, they should either be instructed 
> to use stage three, or be given alternate (more cautious) instructions for 
> the earlier stages. If, on the other hand, the person is familiar with the 
> inner workings of Gentoo and its packages, they would be likely to ignore the 
> instructions anyway.

We also get bug reports from users using genkernel to build their own
CDs using --unionfs-dev, which we don't even list anywhere in the
documentation because we know it is broken and it says as much when you
use it.

The plain and simple truth is that if we allow it *at all* then we will
be held responsible for what happens when a user tries to use it,
whether they have fore-knowledge of the brokenness or not.

> > In pretty much every case, the real answer is "quit using stage1 if you
> > don't know what you're doing" but unfortunately, we're going to get the
> > pointless "but Gentoo is about choice" "argument" that really has no
> > bearing on the truth of what is Gentoo.
> 
> I agree. A section of documentation could be created to explain that concept, 
> and bugs that were filed as a result of inexperience could then be closed and 
> redirected to the documentation.

I am currently working to have a <warn> section added to the Handbook
stating that we highly recommend that users not touch stages 1 or 2.

> I use USE="-*" in make.conf for all of my deployments because I don't want the 
> system's USE flags to be modified by Gentoo developers; I also don't have any 
> difficulty deploying such systems. Unfortunately, I see many individuals that 
> will leap to configure their systems in ways that cannot be maintained. In 
> those cases, the Gentoo developers should not be the ones to correct the 
> errors.

I agree.

On the other side of the coin, it is exactly *because* of -* that we
have such hacks like nocxx in USE.  USE flags are designed for enabling
optional support.  We should not have no* USE flags.  While I personally
don't really care either way, it is something that we are currently
*required* to do because we have no way to ensure that it will be
enabled by default, thanks to -*.  Funny enough, I use -* for the same
reason as you, but then came to find out about auto-use, and am now glad
that I've always used -* myself.  ;]

-- 
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux

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

Re: [gentoo-releng] Re: USE flags and stages 1 and 2

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 817 bytes --]

On Mon, 2005-10-31 at 14:27 -0800, Anthony Gorecki wrote:
> On Monday, October 31, 2005 14:26, Chris Gianelloni wrote:
> > It *does* however guarantee that you have a minimal working system of
> > known quality with which to do your installation from, rather than a
> > completely hacked and limited functionality minimal environment designed
> > only to support the building of a minimal working system.
> 
> It is for that reason that I support keeping stage one tarballs available. 
> Cheers for the stubborn.

Huh?

You support keeping a stage1 tarball around because it is a completely
hacked and broken minimal environment capable of *only* supporting the
bootstrap operation?

-- 
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux

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

Re: [gentoo-releng] USE flags and stages 1 and 2

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 3552 bytes --]

On Mon, 2005-10-31 at 17:32 -0500, Michael Crute wrote:
> On 10/31/05, Chris Gianelloni <wolf31o2@gentoo.org> wrote:
>         > Another thought would be an approach similar to the
>         installer irc
>         > channel where you have to read through a really long FAQ
>         before you
>         > are told how to download the stage1 or 2 tarballs.
>         
>         My personal impression is that this will piss off the end
>         user.  This 
>         works for the installer *channel* but would never work for a
>         CD
>         installation.  It's simply asinine to ask the user to jump
>         through hoops
>         to do what they want.
> 
> Well I wasn't really referring to a CD (but in this case read the MS
> license agreement when installing Windoze I think that would be the
> equivalent of this :). What I was really referring to was putting up
> some sort of scenario like that to download the stage1/2 tarballs.

Like I said, alienating users isn't the best idea for future growth.

>         Even if we stopped shipping stage1 tarballs, there's nothing
>         stopping
>         some inventive user from grabbing a stage3 tarball and making
>         a stage1 
>         tarball from it using catalyst.  It only takes a couple hours
>         on a
>         decent machine.
> 
> True but if they fire up catalyst to make a stage1 chances are they
> aren't going to be so dumb as to request support when their use flag
> combinations don't work. The problem is ignorant users and most of
> them can't us catalyst.

Exactly.  That was my entire point.  If they're taking the time and
energy to build the stage themselves, perhaps they would understand some
of the issues and would be better educated.

Consider it a very high barrier of entry on the ability to be able to do
something extremely dumb with your system.

We don't put reiser4 tools on our CD, either.  Why should we put stage1
on there?  Personally, I believe a stage1 to be much more dangerous and
time-wasting for us developers and our users than reiser4.

>         My point is that once again, we're not *really* removing
>         choice even if
>         we were to drop the earlier stages altogether, we're just
>         making the 
>         user do the work themselves and removing one more abysmal
>         headache and
>         QA nightmare from our already enormous and growing list of
>         stuff that
>         has been delaying the past few releases well beyond our
>         scheduled
>         release dates. 
> 
> Regardless of how foolish you may think it is to do a stage1 install
> (and it may well be) I would still greatly appreciate the stage1
> tarballs even if you refuse to support them. Personally I have never
> needed support on a stage1 that I couldn't get by simply RTFM.

I see no reason whatsoever to release anything that we refuse to
support.  It doesn't speak much on our professionalism or our quality if
we're willing to put out release media that we won't support.

I would much rather not release them and deal with the potential flames
and blowback from a user community that has been wrongly trained that
the stage1 tarball is a good thing than to release something completely
unsupported.  All that does is raise more user->developer relations
problems and widens the gap between users and developers.

-- 
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux

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

Re: [gentoo-releng] Re: USE flags and stages 1 and 2

[Sven Vermeulen]

[-- Attachment #1: Type: text/plain, Size: 840 bytes --]

On Mon, Oct 31, 2005 at 05:44:25PM -0500, Chris Gianelloni wrote:
> I am currently working to have a <warn> section added to the Handbook
> stating that we highly recommend that users not touch stages 1 or 2.

Honestly, if we need to update the installation instructions to highly
discourage the use of stage1 or stage2, I'd rather remove stage1/2
alltogether from the guide and inform the interested users elsewhere on
bootstrapping and such.

I mean, we could very well post this in the Gentoo FAQ, with a few
paragraphs about how and why we don't recommend it.

Wkr,
      Sven Vermeulen

-- 
  Gentoo Foundation Trustee          |  http://foundation.gentoo.org
  Gentoo Documentation Project Lead  |  http://www.gentoo.org/proj/en/gdp
  Gentoo Council Member  

  The Gentoo Project   <<< http://www.gentoo.org >>>

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

Re: [gentoo-releng] Re: USE flags and stages 1 and 2

[Chris Gianelloni]

[-- Attachment #1: Type: text/plain, Size: 1708 bytes --]

On Tue, 2005-11-01 at 20:18 +0100, Sven Vermeulen wrote:
> On Mon, Oct 31, 2005 at 05:44:25PM -0500, Chris Gianelloni wrote:
> > I am currently working to have a <warn> section added to the Handbook
> > stating that we highly recommend that users not touch stages 1 or 2.
> 
> Honestly, if we need to update the installation instructions to highly
> discourage the use of stage1 or stage2, I'd rather remove stage1/2
> alltogether from the guide and inform the interested users elsewhere on
> bootstrapping and such.

I would have no issues with that.  The simple truth is that no user
should be using a stage1 tarball for doing an installation.  There's
simply no need for it.  It introduces countless possibilities for
errors, and if you're just unpacking a tarball, running bootstrap.sh,
and then emerge -e system, you're not gaining anything.  It would be a
different story if the user was hand-modifying the bootstrap script to
produce a highly modified base system for building, such as if they were
adding distcc/ccache/cross-compiling support.  As it stands now, most
users are doing it for the rice factor.  Compiling from stage1 increases
the size of their ePenis and gives Release Engineering countless more
bugs to contend with after every release.

> I mean, we could very well post this in the Gentoo FAQ, with a few
> paragraphs about how and why we don't recommend it.

This sounds like a good plan.

I would say that at this point, we're merely discussing it, since I feel
like we're going to have a very hard time convincing some developers.

-- 
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux

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

Re: [gentoo-releng] Re: USE flags and stages 1 and 2

[Gustavo Zacarias]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Chris Gianelloni wrote:

> I would have no issues with that.  The simple truth is that no user
> should be using a stage1 tarball for doing an installation.  There's
> simply no need for it.  It introduces countless possibilities for
> errors, and if you're just unpacking a tarball, running bootstrap.sh,
> and then emerge -e system, you're not gaining anything.  It would be a
> different story if the user was hand-modifying the bootstrap script to
> produce a highly modified base system for building, such as if they were
> adding distcc/ccache/cross-compiling support.  As it stands now, most
> users are doing it for the rice factor.  Compiling from stage1 increases
> the size of their ePenis and gives Release Engineering countless more
> bugs to contend with after every release.

The only other scenario i can figure out that would benefit from a
stage1/2 would be if the user is using his own profile.
But then he could roll his own stages for that too...

- --
Gustavo Zacarias
Gentoo/SPARC monkey
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFDaM2FV3G/IBCn/JARAp+nAKCDCAMPXHQozyiQPOelSXPU1aMGcQCfcSuT
YeUzhvO1H5IjT6NeAEXSc5A=
=aSfP
-----END PGP SIGNATURE-----
-- 
gentoo-releng@gentoo.org mailing list