Mailing List Archive

Why arch-specific make.conf files?
Hi,

Was just about to finally commit the elog related config stuff into
make.conf just to notice (again) that there are 14 (in words: fourteen)
different make.conf files there, with almost all of them just differing
in CFLAGS and CHOST (only exception is make.conf.mac which isn't used
anymore in any way AFAICT).
From my POV those vars should be set in the profiles instead, and a
quick scan shows that indeed most (maybe all? didn't count them)
profiles set them already, so there isn't really a point in having them
in make.conf too, except to make it easy for users to change them. For
CHOST this seems to be a bad idea, not sure about CFLAGS.
So what's the general opinion about this? Having all these different
files makes it harder to add config changes, not by much but noticably,
so personally I'd like to get rid of them, but if there is a good
reason for them to stay I can live with that.

Marius

--
Public Key at http://www.genone.de/info/gpg-key.pub

In the beginning, there was nothing. And God said, 'Let there be
Light.' And there was still nothing, but you could see a bit better.
Re: Why arch-specific make.conf files? [ In reply to ]
On Tuesday 15 November 2005 20:19, Marius Mauch wrote:
> From my POV those vars should be set in the profiles instead, and a
> quick scan shows that indeed most (maybe all? didn't count them)
> profiles set them already, so there isn't really a point in having them
> in make.conf too, except to make it easy for users to change them
Little note: with Gentoo/FreeBSD I tried avoiding providing CHOST in
make.conf, as to change to non-i686 CHOST you need to rebuild everything, as
the stage is currently i686-centric, I'm sorry of that, I'll try to
automatize a more complete building when I'll have time.

The problem of this is that distcc-config looks inside make.conf for CHOST
instead of using portageq envvar CHOST, so it just breaks :P
I think other things might do the same assumption of finding CHOST in
make.conf, and beside being plainly wrong, I'm not sure if I want to break
everything ;)

--
Diego "Flameeyes" Pettenò - http://dev.gentoo.org/~flameeyes/
Gentoo/ALT lead, Gentoo/FreeBSD, Video, AMD64, Sound, PAM, KDE
Re: Why arch-specific make.conf files? [ In reply to ]
On Tue, 2005-11-15 at 20:19 +0100, Marius Mauch wrote:
> Hi,
>
> Was just about to finally commit the elog related config stuff into
> make.conf just to notice (again) that there are 14 (in words: fourteen)
> different make.conf files there, with almost all of them just differing
> in CFLAGS and CHOST (only exception is make.conf.mac which isn't used
> anymore in any way AFAICT).

Where are these files that you're even talking about?

> From my POV those vars should be set in the profiles instead, and a
> quick scan shows that indeed most (maybe all? didn't count them)
> profiles set them already, so there isn't really a point in having them
> in make.conf too, except to make it easy for users to change them. For
> CHOST this seems to be a bad idea, not sure about CFLAGS.

Well, the stages have a make.conf that is catalyst generated.

> So what's the general opinion about this? Having all these different
> files makes it harder to add config changes, not by much but noticably,
> so personally I'd like to get rid of them, but if there is a good
> reason for them to stay I can live with that.

Without knowing which files these are, I cannot comment further.

--
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux
Re: Why arch-specific make.conf files? [ In reply to ]
On Tue, 2005-11-15 at 20:26 +0100, Diego 'Flameeyes' Pettenò wrote:
> On Tuesday 15 November 2005 20:19, Marius Mauch wrote:
> > From my POV those vars should be set in the profiles instead, and a
> > quick scan shows that indeed most (maybe all? didn't count them)
> > profiles set them already, so there isn't really a point in having them
> > in make.conf too, except to make it easy for users to change them
> Little note: with Gentoo/FreeBSD I tried avoiding providing CHOST in
> make.conf, as to change to non-i686 CHOST you need to rebuild everything, as
> the stage is currently i686-centric, I'm sorry of that, I'll try to
> automatize a more complete building when I'll have time.
>
> The problem of this is that distcc-config looks inside make.conf for CHOST
> instead of using portageq envvar CHOST, so it just breaks :P
> I think other things might do the same assumption of finding CHOST in
> make.conf, and beside being plainly wrong, I'm not sure if I want to break
> everything ;)

CHOST doesn't have to match what is in the profile. In fact, I can
think of a lot of cases where it does not. While I agree that it
shouldn't be required to have CHOST in make.conf, it *is* currently a
requirement, and has been for as long as I can remember.

--
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux
Re: Why arch-specific make.conf files? [ In reply to ]
On Tue, 2005-11-15 at 20:19 +0100, Marius Mauch wrote:
> From my POV those vars should be set in the profiles instead, and a
> quick scan shows that indeed most (maybe all? didn't count them)
> profiles set them already, so there isn't really a point in having them
> in make.conf too, except to make it easy for users to change them. For
> CHOST this seems to be a bad idea, not sure about CFLAGS.

Oh yeah, if you happen to change any of this stuff, please be sure to
inform the catalyst team, along with Release Engineering. We don't want
something like this surprising us when we're in the middle of release
building.

--
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux
Re: Why arch-specific make.conf files? [ In reply to ]
On Tue, Nov 15, 2005 at 02:52:28PM -0500, Chris Gianelloni wrote:
> On Tue, 2005-11-15 at 20:19 +0100, Marius Mauch wrote:
> > Was just about to finally commit the elog related config stuff into
> > make.conf just to notice (again) that there are 14 (in words: fourteen)
> > different make.conf files there, with almost all of them just differing
> > in CFLAGS and CHOST (only exception is make.conf.mac which isn't used
> > anymore in any way AFAICT).
>
> Where are these files that you're even talking about?

before catalyst started nuking make.conf, they were the standard
/etc/make.conf files ... now though, you can find them at
/etc/make.conf.example
-mike
--
gentoo-dev@gentoo.org mailing list
Re: Why arch-specific make.conf files? [ In reply to ]
On Tue, 2005-11-15 at 20:01 +0000, Mike Frysinger wrote:
> On Tue, Nov 15, 2005 at 02:52:28PM -0500, Chris Gianelloni wrote:
> > On Tue, 2005-11-15 at 20:19 +0100, Marius Mauch wrote:
> > > Was just about to finally commit the elog related config stuff into
> > > make.conf just to notice (again) that there are 14 (in words: fourteen)
> > > different make.conf files there, with almost all of them just differing
> > > in CFLAGS and CHOST (only exception is make.conf.mac which isn't used
> > > anymore in any way AFAICT).
> >
> > Where are these files that you're even talking about?
>
> before catalyst started nuking make.conf, they were the standard
> /etc/make.conf files ... now though, you can find them at
> /etc/make.conf.example

Ahh... and there's different ones installed based on some criteria,
rather than a single example? Now it makes sense.

--
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux
Re: Why arch-specific make.conf files? [ In reply to ]
On Tue, Nov 15, 2005 at 04:01:07PM -0500, Chris Gianelloni wrote:
> On Tue, 2005-11-15 at 20:01 +0000, Mike Frysinger wrote:
> > On Tue, Nov 15, 2005 at 02:52:28PM -0500, Chris Gianelloni wrote:
> > > On Tue, 2005-11-15 at 20:19 +0100, Marius Mauch wrote:
> > > > Was just about to finally commit the elog related config stuff into
> > > > make.conf just to notice (again) that there are 14 (in words: fourteen)
> > > > different make.conf files there, with almost all of them just differing
> > > > in CFLAGS and CHOST (only exception is make.conf.mac which isn't used
> > > > anymore in any way AFAICT).
> > >
> > > Where are these files that you're even talking about?
> >
> > before catalyst started nuking make.conf, they were the standard
> > /etc/make.conf files ... now though, you can find them at
> > /etc/make.conf.example
>
> Ahh... and there's different ones installed based on some criteria,
> rather than a single example? Now it makes sense.

congrats, the last horse just came in
-mike
--
gentoo-dev@gentoo.org mailing list
Re: Why arch-specific make.conf files? [ In reply to ]
On Tue, 15 Nov 2005 14:54:01 -0500
Chris Gianelloni <wolf31o2@gentoo.org> wrote:

> On Tue, 2005-11-15 at 20:26 +0100, Diego 'Flameeyes' Pettenò wrote:
> > On Tuesday 15 November 2005 20:19, Marius Mauch wrote:
> > > From my POV those vars should be set in the profiles instead, and
> > > a quick scan shows that indeed most (maybe all? didn't count them)
> > > profiles set them already, so there isn't really a point in
> > > having them in make.conf too, except to make it easy for users to
> > > change them
> > Little note: with Gentoo/FreeBSD I tried avoiding providing CHOST
> > in make.conf, as to change to non-i686 CHOST you need to rebuild
> > everything, as the stage is currently i686-centric, I'm sorry of
> > that, I'll try to automatize a more complete building when I'll
> > have time.
> >
> > The problem of this is that distcc-config looks inside make.conf
> > for CHOST instead of using portageq envvar CHOST, so it just
> > breaks :P I think other things might do the same assumption of
> > finding CHOST in make.conf, and beside being plainly wrong, I'm not
> > sure if I want to break everything ;)
>
> CHOST doesn't have to match what is in the profile. In fact, I can
> think of a lot of cases where it does not. While I agree that it
> shouldn't be required to have CHOST in make.conf, it *is* currently a
> requirement, and has been for as long as I can remember.

The portageq way would scan all make.* files, so you *could* still set
CHOST in make.conf if you want to.

Marius

--
Public Key at http://www.genone.de/info/gpg-key.pub

In the beginning, there was nothing. And God said, 'Let there be
Light.' And there was still nothing, but you could see a bit better.
Re: Why arch-specific make.conf files? [ In reply to ]
On Tue, 15 Nov 2005 14:52:28 -0500
Chris Gianelloni <wolf31o2@gentoo.org> wrote:

> On Tue, 2005-11-15 at 20:19 +0100, Marius Mauch wrote:
> > Hi,
> >
> > Was just about to finally commit the elog related config stuff into
> > make.conf just to notice (again) that there are 14 (in words:
> > fourteen) different make.conf files there, with almost all of them
> > just differing in CFLAGS and CHOST (only exception is make.conf.mac
> > which isn't used anymore in any way AFAICT).
>
> Where are these files that you're even talking about?

http://viewcvstest.gentoo.org/viewcvs.py/portage/main/trunk/cnf/
Installed as make.conf.example by portage.

Marius

--
Public Key at http://www.genone.de/info/gpg-key.pub

In the beginning, there was nothing. And God said, 'Let there be
Light.' And there was still nothing, but you could see a bit better.
Re: Why arch-specific make.conf files? [ In reply to ]
On Wed, 2005-11-16 at 03:30 +0100, Marius Mauch wrote:
> On Tue, 15 Nov 2005 14:54:01 -0500
> Chris Gianelloni <wolf31o2@gentoo.org> wrote:
>
> > On Tue, 2005-11-15 at 20:26 +0100, Diego 'Flameeyes' Pettenò wrote:
> > > On Tuesday 15 November 2005 20:19, Marius Mauch wrote:
> > > > From my POV those vars should be set in the profiles instead, and
> > > > a quick scan shows that indeed most (maybe all? didn't count them)
> > > > profiles set them already, so there isn't really a point in
> > > > having them in make.conf too, except to make it easy for users to
> > > > change them
> > > Little note: with Gentoo/FreeBSD I tried avoiding providing CHOST
> > > in make.conf, as to change to non-i686 CHOST you need to rebuild
> > > everything, as the stage is currently i686-centric, I'm sorry of
> > > that, I'll try to automatize a more complete building when I'll
> > > have time.
> > >
> > > The problem of this is that distcc-config looks inside make.conf
> > > for CHOST instead of using portageq envvar CHOST, so it just
> > > breaks :P I think other things might do the same assumption of
> > > finding CHOST in make.conf, and beside being plainly wrong, I'm not
> > > sure if I want to break everything ;)
> >
> > CHOST doesn't have to match what is in the profile. In fact, I can
> > think of a lot of cases where it does not. While I agree that it
> > shouldn't be required to have CHOST in make.conf, it *is* currently a
> > requirement, and has been for as long as I can remember.
>
> The portageq way would scan all make.* files, so you *could* still set
> CHOST in make.conf if you want to.

I wasn't disputing that. I was only stating that currently, there are
things that *require* CHOST in make.conf that would need to be adjusted.

--
Chris Gianelloni
Release Engineering - Strategic Lead
x86 Architecture Team
Games - Developer
Gentoo Linux
Re: Why arch-specific make.conf files? [ In reply to ]
On Wed, 16 Nov 2005 07:50:47 -0500
Chris Gianelloni <wolf31o2@gentoo.org> wrote:

> On Wed, 2005-11-16 at 03:30 +0100, Marius Mauch wrote:
> > On Tue, 15 Nov 2005 14:54:01 -0500
> > Chris Gianelloni <wolf31o2@gentoo.org> wrote:
> >
> > > On Tue, 2005-11-15 at 20:26 +0100, Diego 'Flameeyes' Pettenò
> > > wrote:
> > > > On Tuesday 15 November 2005 20:19, Marius Mauch wrote:
> > > > > From my POV those vars should be set in the profiles instead,
> > > > > and a quick scan shows that indeed most (maybe all? didn't
> > > > > count them) profiles set them already, so there isn't really
> > > > > a point in having them in make.conf too, except to make it
> > > > > easy for users to change them
> > > > Little note: with Gentoo/FreeBSD I tried avoiding providing
> > > > CHOST in make.conf, as to change to non-i686 CHOST you need to
> > > > rebuild everything, as the stage is currently i686-centric, I'm
> > > > sorry of that, I'll try to automatize a more complete building
> > > > when I'll have time.
> > > >
> > > > The problem of this is that distcc-config looks inside make.conf
> > > > for CHOST instead of using portageq envvar CHOST, so it just
> > > > breaks :P I think other things might do the same assumption of
> > > > finding CHOST in make.conf, and beside being plainly wrong, I'm
> > > > not sure if I want to break everything ;)
> > >
> > > CHOST doesn't have to match what is in the profile. In fact, I
> > > can think of a lot of cases where it does not. While I agree
> > > that it shouldn't be required to have CHOST in make.conf, it *is*
> > > currently a requirement, and has been for as long as I can
> > > remember.
> >
> > The portageq way would scan all make.* files, so you *could* still
> > set CHOST in make.conf if you want to.
>
> I wasn't disputing that. I was only stating that currently, there are
> things that *require* CHOST in make.conf that would need to be
> adjusted.

You have a list of (some of) these things? From a portage POV those
things are broken, the only thing you can safely assume about
make.conf is that it exists (and even that is debatable).

Marius

--
Public Key at http://www.genone.de/info/gpg-key.pub

In the beginning, there was nothing. And God said, 'Let there be
Light.' And there was still nothing, but you could see a bit better.
Re: Why arch-specific make.conf files? [ In reply to ]
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Marius Mauch wrote:

> Was just about to finally commit the elog related config stuff into
> make.conf just to notice (again) that there are 14 (in words: fourteen)
> different make.conf files there, with almost all of them just differing
> in CFLAGS and CHOST (only exception is make.conf.mac which isn't used
> anymore in any way AFAICT).
> From my POV those vars should be set in the profiles instead, and a
> quick scan shows that indeed most (maybe all? didn't count them)
> profiles set them already, so there isn't really a point in having them
> in make.conf too, except to make it easy for users to change them. For
> CHOST this seems to be a bad idea, not sure about CFLAGS.
> So what's the general opinion about this? Having all these different
> files makes it harder to add config changes, not by much but noticably,
> so personally I'd like to get rid of them, but if there is a good
> reason for them to stay I can live with that.

Speaking for the sparc swamp(tm) i wouldn't mind if CHOST went away from
make.conf* since people thinking "ohhh i have a sparc64" and modifying
them only causes trouble in the mixed user/kernel land we have.

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

iD8DBQFDe2n1V3G/IBCn/JARAjqEAJ99gBPAeVatWj0lgDclU3xYW6f3DwCePIwT
QswmG4b9eHFV48/APSdiJXw=
=SnPs
-----END PGP SIGNATURE-----
--
gentoo-dev@gentoo.org mailing list
Re: Why arch-specific make.conf files? [ In reply to ]
Marius Mauch posted <20051116160434.629519df@sven.genone.homeip.net>,
excerpted below, on Wed, 16 Nov 2005 16:04:34 +0100:

> You have a list of (some of) these things? From a portage POV those
> things are broken, the only thing you can safely assume about
> make.conf is that it exists (and even that is debatable).

Plus, I'm enthusiastically using the still fairly new "source xyzfile"
functionality in portage to break out my portage config into various
modules (use, fs, net, etc). My make.conf itself is simply a collection
of several "source" lines. If whatever is brokenly assuming it can
collect such info from make.conf is sourcing the file using bash, or if it
too understands the source directive, no problem, but if not, it could be
a serious problem!

FWIW, I've experienced no issues here. (I don't use distcc, so that
already mentioned issue doesn't affect me.)

--
Duncan - List replies preferred. No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master." Richard Stallman in
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html


--
gentoo-dev@gentoo.org mailing list