Re: [gentoo-dev] iuse defaults example

2007-07-15 Thread Mike Frysinger
On Sunday 15 July 2007, Arfrever Frehtes Taifersar Arahesis wrote:
> 2007-07-15 21:22:07 Mike Frysinger napisał(a):
> > On Sunday 15 July 2007, Thomas de Grenier de Latour wrote:
> > > the day you switch from IUSE="nocxx" to IUSE="+cxx", will you
> > > remember that, as a consequence, you have to fix hardened/2.6/minimal
> > > profile?
> >
> > there is no "nocxx" reference anywhere in the profiles/ tree
>
> grep -r "USE.*-\*" /usr/portage/profiles

profile deserves what it gets then
-mike


signature.asc
Description: This is a digitally signed message part.


Re: [gentoo-dev] iuse defaults example

2007-07-15 Thread Arfrever Frehtes Taifersar Arahesis
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1

2007-07-15 21:22:07 Mike Frysinger napisał(a):
> On Sunday 15 July 2007, Thomas de Grenier de Latour wrote:
> > the day you switch from IUSE="nocxx" to IUSE="+cxx", will you
> > remember that, as a consequence, you have to fix hardened/2.6/minimal
> > profile?
>
> there is no "nocxx" reference anywhere in the profiles/ tree
> -mike

grep -r "USE.*-\*" /usr/portage/profiles

- -- 
Arfrever Frehtes Taifersar Arahesis
-BEGIN PGP SIGNATURE-
Version: GnuPG v2.0.5 (GNU/Linux)

iD8DBQFGmoKZ/axNJ4Xo/ZERAv5dAJ9f+8hMkuRtrJtlYfy+KgdxYc3jFACeM3fn
UZ2vEVoPm150Pe0vD+pKtnI=
=GhHz
-END PGP SIGNATURE-
-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-dev] iuse defaults example

2007-07-15 Thread Mike Frysinger
On Sunday 15 July 2007, Thomas de Grenier de Latour wrote:
> My point is just that it doesn't work that well with the USE_ORDER that
> have been chosen. Even keeping the "-* in make.conf" case appart
> (obviously my opinion on how it should behave was not widely shared, i
> can live with that), there is still a problem with -* in make.defaults
> files:

there are ways to make the USE=nocxx -> USE=cxx transition nice and i plan on 
going that route

> the day you switch from IUSE="nocxx" to IUSE="+cxx", will you 
> remember that, as a consequence, you have to fix hardened/2.6/minimal
> profile?

there is no "nocxx" reference anywhere in the profiles/ tree
-mike


signature.asc
Description: This is a digitally signed message part.


Re: [gentoo-dev] iuse defaults example

2007-07-15 Thread Thomas de Grenier de Latour
On 2007/07/15, Marius Mauch <[EMAIL PROTECTED]> wrote:

> On Sun, 15 Jul 2007 11:53:08 +0200
> Thomas de Grenier de Latour <[EMAIL PROTECTED]> wrote:
> 
> > My point is just that it doesn't work that well with the USE_ORDER
> > that have been chosen. Even keeping the "-* in make.conf" case
> > appart (obviously my opinion on how it should behave was not widely
> > shared, i can live with that), there is still a problem with -* in
> > make.defaults files: the day you switch from IUSE="nocxx" to
> > IUSE="+cxx", will you remember that, as a consequence, you have to
> > fix hardened/2.6/minimal profile? 
> 
> Well, it's just like any other renaming of USE flags in that regard.

But it shows that the "we shouldn't care about per-ebuild defaults in
profiles" argument doesn't really stand, which is unfortunate because
Mike is probaly right that it would have been a good thing.

> And while I can see why people would want IUSE defaults to have a
> higher priority than USE in make.defaults and/or make.conf, I suspect
> the vast majority of users would get completely lost in finding out
> where a flag was enabled/disabled (the current system is already
> confusing to a lot of people until they get a detailed explanation).

I don't think it's something which would be that hard to explain to
users.  All it takes is having "emerge -pv" to clearly shows that
something unusual is happening when a flag value is overidden by an
IUSE-default, for instance with an exclamation mark suffix, and to
document that in the man page, with the rest of the --verbose output:
 ! suffix = profile's global default value for this flag is
  overidden by an ebuild-specific setting.  You can still enable / 
  disable it in your own configuration (make.conf or package.use)
  if you really want to.
Maybe i am over-estimating the average user, but to me it doesn't sound
that complicated or obscure.

--
TGL.
-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-dev] iuse defaults example

2007-07-15 Thread Marius Mauch
On Sun, 15 Jul 2007 11:53:08 +0200
Thomas de Grenier de Latour <[EMAIL PROTECTED]> wrote:

> On 2007/07/10, Mike Frysinger <[EMAIL PROTECTED]> wrote:
> 
> > for some flags yes ... for others, i dislike that idea for the exact
> > same reason for the other profile-based suggestions: these defaults
> > should live in the ebuild, not the profile
> 
> I agree that putting per-package defaults in ebuilds is far more
> elegant than putting them in profiles.  
> 
> My point is just that it doesn't work that well with the USE_ORDER
> that have been chosen. Even keeping the "-* in make.conf" case appart
> (obviously my opinion on how it should behave was not widely shared, i
> can live with that), there is still a problem with -* in make.defaults
> files: the day you switch from IUSE="nocxx" to IUSE="+cxx", will you
> remember that, as a consequence, you have to fix hardened/2.6/minimal
> profile? 

Well, it's just like any other renaming of USE flags in that regard.

> And also, in bug #61732 there is this comment from Zac about "-foo"
> not being supported because pkginternal is at the bottom of the stack.
> Imho, that's missing a great opportunity to make users' life a bit
> easier... Take the "gtk" flag, which is on by default in usual desktop
> profiles, but as the drawback to trigger GTK+-1.2 installation just
> for a few CLI programs which comes with an optional obsolete GUI:
> wouldn't it be nice if said packages could state IUSE="-gtk", so that
> the default behavior would be to install only GTK+-2.x GUIs?  I'm
> pretty sure it would save a tenth of /etc/portage/package.use entries
> for many users.

IIRC that has been added a little while ago, but with the current
default USE_ORDER it's more or less useless. And while I can see why
people would want IUSE defaults to have a higher priority than
USE in make.defaults and/or make.conf, I suspect the vast majority of
users would get completely lost in finding out where a flag was
enabled/disabled (the current system is already confusing to a lot of
people until they get a detailed explanation).

Marius
-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-dev] iuse defaults example

2007-07-15 Thread Thomas de Grenier de Latour
On 2007/07/10, Mike Frysinger <[EMAIL PROTECTED]> wrote:

> for some flags yes ... for others, i dislike that idea for the exact
> same reason for the other profile-based suggestions: these defaults
> should live in the ebuild, not the profile

I agree that putting per-package defaults in ebuilds is far more
elegant than putting them in profiles.  

My point is just that it doesn't work that well with the USE_ORDER that
have been chosen. Even keeping the "-* in make.conf" case appart
(obviously my opinion on how it should behave was not widely shared, i
can live with that), there is still a problem with -* in make.defaults
files: the day you switch from IUSE="nocxx" to IUSE="+cxx", will you
remember that, as a consequence, you have to fix hardened/2.6/minimal
profile? 

And also, in bug #61732 there is this comment from Zac about "-foo" not
being supported because pkginternal is at the bottom of the stack.
Imho, that's missing a great opportunity to make users' life a bit
easier... Take the "gtk" flag, which is on by default in usual desktop
profiles, but as the drawback to trigger GTK+-1.2 installation just for
a few CLI programs which comes with an optional obsolete GUI: wouldn't
it be nice if said packages could state IUSE="-gtk", so that the
default behavior would be to install only GTK+-2.x GUIs?  I'm pretty
sure it would save a tenth of /etc/portage/package.use entries for many
users.

--
TGL.
-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-dev] iuse defaults example

2007-07-10 Thread Mike Frysinger
On Tuesday 10 July 2007, Thomas de Grenier de Latour wrote:
> On 2007/07/10, Mike Frysinger <[EMAIL PROTECTED]> wrote:
> > the no* flags were introduced more to address default behavior than
> > the -* case, so yes we can kick many of the no* USE flags
>
> To address only the default behavior, adding "foo" to the profile USE
> instead of using a "nofoo" flag would have been enough.  This could
> have been done long ago, but my understanding has always been that it
> was not considered a good enough solution because it was not -*-proof.

for some flags yes ... for others, i dislike that idea for the exact same 
reason for the other profile-based suggestions: these defaults should live in 
the ebuild, not the profile
-mike


signature.asc
Description: This is a digitally signed message part.


Re: [gentoo-dev] iuse defaults example

2007-07-10 Thread Thomas de Grenier de Latour
On 2007/07/10, Mike Frysinger <[EMAIL PROTECTED]> wrote:

> 
> the no* flags were introduced more to address default behavior than
> the -* case, so yes we can kick many of the no* USE flags
> 

To address only the default behavior, adding "foo" to the profile USE
instead of using a "nofoo" flag would have been enough.  This could
have been done long ago, but my understanding has always been that it
was not considered a good enough solution because it was not -*-proof.

--
TGL.
-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-dev] iuse defaults example

2007-07-10 Thread Mike Frysinger
On Tuesday 10 July 2007, Thomas de Grenier de Latour wrote:
> On 2007/07/10, Thilo Bangert <[EMAIL PROTECTED]> wrote:
> > - we could finally kick all the no* USE flags. USE flags are use
> > flags - they determine what should be used. not what should not be
> > used...
>
> Because of the way USE flags stack in Portage (the USE_ORDER variable),
> IUSE defaults are not a solution for dropping no* flags:
> http://thread.gmane.org/gmane.linux.gentoo.devel/43137/focus=43175
> As Zac pointed out in his reply to this post, dropping nocxx and
> friends is more a job for use.force / package.use.force.

the no* flags were introduced more to address default behavior than the -* 
case, so yes we can kick many of the no* USE flags

also, use.force isnt exactly a nice solution ... more like brute force, i'm 
not sure any no* flag would be appropriate
-mike


signature.asc
Description: This is a digitally signed message part.


Re: [gentoo-dev] iuse defaults example

2007-07-10 Thread Thomas de Grenier de Latour
On 2007/07/10, Thilo Bangert <[EMAIL PROTECTED]> wrote:

> - we could finally kick all the no* USE flags. USE flags are use
> flags - they determine what should be used. not what should not be
> used...

Because of the way USE flags stack in Portage (the USE_ORDER variable),
IUSE defaults are not a solution for dropping no* flags:
http://thread.gmane.org/gmane.linux.gentoo.devel/43137/focus=43175
As Zac pointed out in his reply to this post, dropping nocxx and
friends is more a job for use.force / package.use.force.

--
TGL.
-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-dev] iuse defaults example

2007-07-10 Thread Thilo Bangert
Mike Frysinger <[EMAIL PROTECTED]> said:
> On Tuesday 10 July 2007, William Hubbs wrote:
> > On Mon, Jul 09, 2007 at 11:26:19PM +0100, Ciaran McCreesh wrote:
> > > As for IUSE defaults... There were objections against that feature
> > > on the grounds that it's unnecessary and increased maintenance. Do
> > > they really offer any benefit over package.use?
> >
> > Would iuse defaults not be appropriate when a certain use flag is
> > recommended as the default for most users for a package??
>
> other examples that make sense and are a pain with package.use:
>  - local USE flags (suddenly not so local huh)
>  - local USE flags and changing names
>  - defaults based on version (feature sucked <= 1.x and then rocked >=
> 2.x) - developing new ebuilds for personal use
>  - developing new ebuilds for merging into tree (btw: need to update

- we could finally kick all the no* USE flags. USE flags are use flags - 
they determine what should be used. not what should not be used...

/usr/portage/profiles $ grep :no use.local.desc | wc -l
87

Thilo


pgpCq4ecWgN3q.pgp
Description: PGP signature


Re: [gentoo-dev] iuse defaults example

2007-07-10 Thread Mike Frysinger
On Tuesday 10 July 2007, Petteri Räty wrote:
> Mike Frysinger kirjoitti:
> > On Tuesday 10 July 2007, William Hubbs wrote:
> >> On Mon, Jul 09, 2007 at 11:26:19PM +0100, Ciaran McCreesh wrote:
> >>> As for IUSE defaults... There were objections against that feature on
> >>> the grounds that it's unnecessary and increased maintenance. Do they
> >>> really offer any benefit over package.use?
> >>
> >> Would iuse defaults not be appropriate when a certain use flag is
> >> recommended as the default for most users for a package??
> >
> > other examples that make sense and are a pain with package.use:
> >  - local USE flags (suddenly not so local huh)
>
> [EMAIL PROTECTED] /usr/portage/profiles $ cat base/package.use
> # This file requires >=portage-2.1.2 (see bug #61732)
>
> # Strongly recommended, otherwise all logos, icons, etc. appear in b/w.
> app-editors/emacs xpm
> app-editors/emacs-cvs xpm
>
> Seems local to me...

you missed the point ... ideally local USE flags should not appear outside of 
an ebuild.  if i had a solution for it, i'd propose getting rid of 
use.local.desc ...

> >  - local USE flags and changing names
>
> Normally you would only have to change base/package.use

"normally" doesnt cut it.  package.use is stackable and can appear in any 
profile directory which means these flags can be listed anywhere.

> >  - defaults based on version (feature sucked <= 1.x and then rocked >=
> > 2.x)
>
> package.use should accept version atoms
>
> >  - developing new ebuilds for personal use
>
> /etc/portage/package.use
>
> >  - developing new ebuilds for merging into tree (btw: need to update all
> > these other files in profiles/ instead of just committing the one ebuild)
>
> base/package.use

your replies have just backed up my point: it's a [pain]ita when it should be 
a [pleasent]ita.
-mike


signature.asc
Description: This is a digitally signed message part.


Re: [gentoo-dev] iuse defaults example

2007-07-10 Thread Petteri Räty
Mike Frysinger kirjoitti:
> On Tuesday 10 July 2007, William Hubbs wrote:
>> On Mon, Jul 09, 2007 at 11:26:19PM +0100, Ciaran McCreesh wrote:
>>> As for IUSE defaults... There were objections against that feature on
>>> the grounds that it's unnecessary and increased maintenance. Do they
>>> really offer any benefit over package.use?
>> Would iuse defaults not be appropriate when a certain use flag is
>> recommended as the default for most users for a package??
> 
> other examples that make sense and are a pain with package.use:
>  - local USE flags (suddenly not so local huh)

[EMAIL PROTECTED] /usr/portage/profiles $ cat base/package.use
# This file requires >=portage-2.1.2 (see bug #61732)

# Strongly recommended, otherwise all logos, icons, etc. appear in b/w.
app-editors/emacs xpm
app-editors/emacs-cvs xpm

Seems local to me...


>  - local USE flags and changing names

Normally you would only have to change base/package.use

>  - defaults based on version (feature sucked <= 1.x and then rocked >= 2.x)

package.use should accept version atoms

>  - developing new ebuilds for personal use

/etc/portage/package.use

>  - developing new ebuilds for merging into tree (btw: need to update all 
> these 
> other files in profiles/ instead of just committing the one ebuild)
> -mike

base/package.use

Regards,
Petteri



signature.asc
Description: OpenPGP digital signature


Re: [gentoo-dev] iuse defaults example

2007-07-09 Thread Mike Frysinger
On Tuesday 10 July 2007, William Hubbs wrote:
> On Mon, Jul 09, 2007 at 11:26:19PM +0100, Ciaran McCreesh wrote:
> > As for IUSE defaults... There were objections against that feature on
> > the grounds that it's unnecessary and increased maintenance. Do they
> > really offer any benefit over package.use?
>
> Would iuse defaults not be appropriate when a certain use flag is
> recommended as the default for most users for a package??

other examples that make sense and are a pain with package.use:
 - local USE flags (suddenly not so local huh)
 - local USE flags and changing names
 - defaults based on version (feature sucked <= 1.x and then rocked >= 2.x)
 - developing new ebuilds for personal use
 - developing new ebuilds for merging into tree (btw: need to update all these 
other files in profiles/ instead of just committing the one ebuild)
-mike


signature.asc
Description: This is a digitally signed message part.


[gentoo-dev] iuse defaults example

2007-07-09 Thread William Hubbs
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1

On Mon, Jul 09, 2007 at 11:26:19PM +0100, Ciaran McCreesh wrote:
> As for IUSE defaults... There were objections against that feature on
> the grounds that it's unnecessary and increased maintenance. Do they
> really offer any benefit over package.use?


Would iuse defaults not be appropriate when a certain use flag is
recommended as the default for most users for a package??

Here is an example, I think, where an iuse default would make sense.
Espeak uses the portaudio library as a way to actually play sounds.
However, it can also be compiled without portaudio in order to just
generate wav files.

The way espeak is currently in the tree, it requires portaudio, and that
is the way I figure most people would want it to compile.

The issue though is that the way we have it right now it is not possible
to build espeak only to create wav files.  There would be 2 ways I can
see to do that.

1)  Use the portaudio use flag, and put it in IUSE as +portaudio so that
a user can turn it off if they desire, or
2) use a use flag something like "noportaudio", which would turn it off
if the user enables it, but I don't like that because of the reverse
logic -- If a flag is in use, it should enable support for something,
not disable it.

Is there another option that I'm missing, or is something like this a
good case for iuse defaults?

- -- 
William Hubbs
gentoo accessibility team lead
[EMAIL PROTECTED]
-BEGIN PGP SIGNATURE-
Version: GnuPG v1.4.7 (GNU/Linux)

iD8DBQFGkxHwblQW9DDEZTgRAirDAJ4y50ul/NbtMM+nyTHH7y7Y7N3NdQCeMVEW
8rY3UNew30RgEFqW42sMwzE=
=Um4Q
-END PGP SIGNATURE-
-- 
[EMAIL PROTECTED] mailing list