[gentoo-dev] Preparing profiles for EAPI 3 IUSE strictness

[gentoo-dev] Preparing profiles for EAPI 3 IUSE strictness

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 2654 bytes --]

EAPI 3 makes IUSE strict: flags not listed in IUSE can't be used in
dependency strings, use queries and so on. However, the specification
includes ways of implicitly adding things to the effective value of
IUSE via the base profile make.defaults. What the specification doesn't
say, and what hasn't been formally decided, is what exactly will be
implicit.

First up is ARCH. Most people don't seem to want to explicitly list
IUSE="x86" etc to make 'use x86' and 'x86? ( ... )' work. Also, people
seem to want to continue the existing unprefixed behaviour rather than
starting to write 'arch_x86'. So we'll need:

    USE_EXPAND_UNPREFIXED="ARCH"
    USE_EXPAND_IMPLICIT="ARCH"

In addition, all the implicit values need to be listed. According to
arch.list, the values are:

    USE_EXPAND_VALUES_ARCH="alpha amd64 amd64-fbsd arm hppa ia64 m68k \
       mips ppc ppc64 s390 sh sparc sparc-fbsd x86 x86-fbsd ppc-aix \
       x86-freebsd x64-freebsd ia64-hpux x86-interix mips-irix \
       amd64-linux ia64-linux x86-linux ppc-macos x86-macos x64-macos \
       m68k-mint x86-netbsd ppc-openbsd x86-openbsd x64-openbsd \
       sparc-solaris sparc64-solaris x64-solaris x86-solaris x86-winnt"

I've no idea whether that list is accurate. With EAPI 3 it'll matter.

Of the normal USE_EXPAND flags, some appear to be routinely listed in
IUSE anyway, and some pretty much never are. I'd imagine USERLAND,
KERNEL and ELIBC will want to be implicit:

    USE_EXPAND_IMPLICIT="${USE_EXPAND_IMPLICIT} USERLAND KERNEL ELIBC"

And again, the implicit values will have to be listed. According to the
desc files (which could be full of lies):

    USE_EXPAND_VALUES_USERLAND="GNU BSD"
    USE_EXPAND_VALUES_KERNEL="AIX Darwin FreeBSD freemint linux HPUX \
        Interix IRIX NetBSD OpenBSD SunOS"
    USE_EXPAND_VALUES_ELIBC="AIX Darwin DragonFly FreeBSD glibc HPUX \
        Interix IRIX mintlib NetBSD OpenBSD SunOS uclibc"

Are there any other USE_EXPANDs that people want implicit behaviour
for? If so, which ones, and are the desc files accurate?

Finally, there's room to include plain old flags in IUSE automatically.
This was added to the specification as a hypothetical "we might want
this, and it's easy to specify and implement" rather than a "we'll
definitely be using this". Flags that I'm aware of that regularly get
abused are:

    IUSE_IMPLICIT="build debug"

Are people wanting to make those implicit? Are there any other flags
that people really don't want to put in IUSE? Bear in mind that any
flag that's implicit can't ever be used for a use dependency default.

-- 
Ciaran McCreesh

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] Preparing profiles for EAPI 3 IUSE strictness

[Andrew D Kirch]

Ciaran,

I've talked with the pkgcore people and they don't use the EAPI's (or
PMS) in the first place.  This essentially leaves you writing documents
you're requiring for paludis support.  As this seems to be mostly a PM
issue, it should be taken elsewhere.  To that end, here is a
gentoo-portage-dev mailing list that is more appropriate for minor
process issues such as those brought up below.
I think that ending this discussion here and moving it over to a forum
more appropriate to package manager development would reduce the
temperature around your proposals and get them implemented (as Zac seems
willing to do so).  While I realize this is a general purpose mailing
list, it is general purpose, and there is a mailing list specifically
for portage development.

Andrew

Ciaran McCreesh wrote:
> *SNIP*

Re: [gentoo-dev] Preparing profiles for EAPI 3 IUSE strictness

[Brian Harring]

[-- Attachment #1: Type: text/plain, Size: 803 bytes --]

On Tue, Jul 07, 2009 at 02:08:01AM -0400, Andrew D Kirch wrote:
> Ciaran,
> 
> I've talked with the pkgcore people and they don't use the EAPI's (or
> PMS) in the first place.  

No clue who you talked to, but they weren't speaking for pkgcore- I 
speak for pkgcore pretty much solely.  Pkgcore utilizes EAPIs- hell I 
added the original support to portage.  Not sure what info you got 
fed, but it was a bit off.

> I think that ending this discussion here and moving it over to a forum
> more appropriate to package manager development would reduce the
> temperature around your proposals and get them implemented (as Zac seems
> willing to do so).

I don't particularly care one way or another (subscribed to both 
MLs), just mostly correcting one large misstatement...

~harring

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] Preparing profiles for EAPI 3 IUSE strictness

[Alec Warner]

On Mon, Jul 6, 2009 at 11:08 PM, Andrew D Kirch<trelane@trelane.net> wrote:
> Ciaran,
>
> I've talked with the pkgcore people and they don't use the EAPI's (or
> PMS) in the first place.  This essentially leaves you writing documents
> you're requiring for paludis support.  As this seems to be mostly a PM
> issue, it should be taken elsewhere.  To that end, here is a
> gentoo-portage-dev mailing list that is more appropriate for minor
> process issues such as those brought up below.
> I think that ending this discussion here and moving it over to a forum
> more appropriate to package manager development would reduce the
> temperature around your proposals and get them implemented (as Zac seems
> willing to do so).  While I realize this is a general purpose mailing
> list, it is general purpose, and there is a mailing list specifically
> for portage development.

I think you are slightly misreading his intent (or I am misreading
it).  I believe Ciaran is talking about "what gentoo should put in
make.defaults such that the user experience doesn't totally suck for
this new EAPI3 thing" and provided some ideas.  I'm pretty sure
whether or not PM foo has implemented these things is a moot point as
far as this thread is concerned.

-A

>
> Andrew
>
> Ciaran McCreesh wrote:
>> *SNIP*
>
>

Re: [gentoo-dev] Preparing profiles for EAPI 3 IUSE strictness

[Andrew D Kirch]

Yeah, that was definately a misunderstanding of something bonzaikitten said.

Mea culpa.

Andrew

Brian Harring wrote:
> On Tue, Jul 07, 2009 at 02:08:01AM -0400, Andrew D Kirch wrote:
>   
>> Ciaran,
>>
>> I've talked with the pkgcore people and they don't use the EAPI's (or
>> PMS) in the first place.  
>>     
>
> No clue who you talked to, but they weren't speaking for pkgcore- I 
> speak for pkgcore pretty much solely.  Pkgcore utilizes EAPIs- hell I 
> added the original support to portage.  Not sure what info you got 
> fed, but it was a bit off.
>
>   
>> I think that ending this discussion here and moving it over to a forum
>> more appropriate to package manager development would reduce the
>> temperature around your proposals and get them implemented (as Zac seems
>> willing to do so).
>>     
>
> I don't particularly care one way or another (subscribed to both 
> MLs), just mostly correcting one large misstatement...
>
> ~harring
>   

Re: [gentoo-dev] Preparing profiles for EAPI 3 IUSE strictness

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 1255 bytes --]

On Tue, 07 Jul 2009 02:08:01 -0400
Andrew D Kirch <trelane@trelane.net> wrote:
> I've talked with the pkgcore people and they don't use the EAPI's (or
> PMS) in the first place.

If that were true, which it isn't, then by Council decision pkgcore
should have been package.masked.

> This essentially leaves you writing documents you're requiring for
> paludis support.

No, it leaves me writing documents used by Portage, Pkgcore, Paludis
and at least two more independent under-development third party package
management libraries.

> As this seems to be mostly a PM issue, it should be taken elsewhere.

This thread is necessary for Portage support for EAPI 3, which is being
worked on. It also has considerable developer impact, since depending
upon the decisions made, certain existing conventions may no longer
apply to EAPI 3 things.

Given that your stated intention is for "Paludis to fail", and that
"opposing [me] and everything [I] do was an initiative [you] started
only after careful consideration", I'll kindly ask you to stop randomly
jumping out and flinging turds, since it adds nothing to the discussion
at hand and only serves to make it harder for Gentoo to function as a
community.

-- 
Ciaran McCreesh

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-dev] Preparing profiles for EAPI 3 IUSE strictness

[Nirbheek Chauhan]

I don't see the point for this reply seeing that Andrew admitted he
was in error half-an-hour ago in this very thread.

On Tue, Jul 7, 2009 at 7:22 PM, Ciaran
McCreesh<ciaran.mccreesh@googlemail.com> wrote:
> On Tue, 07 Jul 2009 02:08:01 -0400
> Andrew D Kirch <trelane@trelane.net> wrote:
>> I've talked with the pkgcore people and they don't use the EAPI's (or
>> PMS) in the first place.
>
> If that were true, which it isn't, then by Council decision pkgcore
> should have been package.masked.
>
>> This essentially leaves you writing documents you're requiring for
>> paludis support.
>
> No, it leaves me writing documents used by Portage, Pkgcore, Paludis
> and at least two more independent under-development third party package
> management libraries.
>
>> As this seems to be mostly a PM issue, it should be taken elsewhere.
>
> This thread is necessary for Portage support for EAPI 3, which is being
> worked on. It also has considerable developer impact, since depending
> upon the decisions made, certain existing conventions may no longer
> apply to EAPI 3 things.
>
> Given that your stated intention is for "Paludis to fail", and that
> "opposing [me] and everything [I] do was an initiative [you] started
> only after careful consideration", I'll kindly ask you to stop randomly
> jumping out and flinging turds, since it adds nothing to the discussion
> at hand and only serves to make it harder for Gentoo to function as a
> community.
>
> --
> Ciaran McCreesh
>



-- 
~Nirbheek Chauhan

Re: [gentoo-dev] Preparing profiles for EAPI 3 IUSE strictness

[Arfrever Frehtes Taifersar Arahesis]

[-- Attachment #1: Type: text/plain, Size: 589 bytes --]

2009-07-07 01:01:11 Ciaran McCreesh napisał(a):
> ...
> Finally, there's room to include plain old flags in IUSE automatically.
> This was added to the specification as a hypothetical "we might want
> this, and it's easy to specify and implement" rather than a "we'll
> definitely be using this". Flags that I'm aware of that regularly get
> abused are:
> 
>     IUSE_IMPLICIT="build debug"
> 
> Are people wanting to make those implicit?

IMHO they shouldn't be implicit.
(And maybe IUSE_IMPLICIT shouldn't be supported at all.)

-- 
Arfrever Frehtes Taifersar Arahesis

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] Preparing profiles for EAPI 3 IUSE strictness

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 1340 bytes --]

On Wed, 8 Jul 2009 05:02:34 +0200
Arfrever Frehtes Taifersar Arahesis <Arfrever@gentoo.org> wrote:
> >     IUSE_IMPLICIT="build debug"
> > 
> > Are people wanting to make those implicit?
> 
> IMHO they shouldn't be implicit.

I should probably explain the rationale behind those two...

Back in the day, various eclasses would do 'if use build' and 'if use
debug' and the like, and at one point eutils had a DEPEND="!build?
( patch )" in there.

I *think* all the major offenders there are gone now. On the other
hand, if they're not, and IUSE_IMPLICIT doesn't include those, it means
EAPI 3 won't be usable with certain fairly common eclasses.

Historically, IUSE was purely a visual thing, and didn't affect package
manager behaviour. With the introduction of the newuse stuff, and
later, use dependencies, that slowly stopped being true, and IUSE
started to matter a lot more.

> (And maybe IUSE_IMPLICIT shouldn't be supported at all.)

Personally I hate the whole implicit thing, and would rather everyone
stuck absolutely everything in IUSE. But a majority of developers
thought otherwise.

There were also calls for some fancy prefix use flags to go in
IUSE_IMPLICIT at some point. Alas, it doesn't look like something we
could have excluded from the specification entirely...

-- 
Ciaran McCreesh

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] Preparing profiles for EAPI 3 IUSE strictness

[Harald van Dijk]

On Wed, Jul 08, 2009 at 05:02:10AM +0200, Arfrever Frehtes Taifersar Arahesis wrote:
> 2009-07-07 01:01:11 Ciaran McCreesh napisał(a):
> > ...
> >     IUSE_IMPLICIT="build debug"
> > 
> > Are people wanting to make those implicit?
> 
> IMHO they shouldn't be implicit.

Agreed. They shouldn't be implicit, if only because you really want
emerge --newuse (or equivalent) to pick up on changes in USE=build or
USE=debug. Also, are USE dependencies possible with implicit IUSE? If
not, things like DEPEND="dev-lang/python[-build]" would have to be
changed even though they're probably a good idea.

Re: [gentoo-dev] Preparing profiles for EAPI 3 IUSE strictness

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 790 bytes --]

On Fri, 10 Jul 2009 19:32:02 +0200
Harald van Dijk <truedfx@gentoo.org> wrote:
> Also, are USE dependencies possible with implicit IUSE? If
> not, things like DEPEND="dev-lang/python[-build]" would have to be
> changed even though they're probably a good idea.

Use dependencies are possible. Use dependency defaults are not. So:

DEPEND="dev-lang/python[-build]" would be ok so long as all versions of
python either have build in IUSE or are EAPI 3 and have build in
IUSE_IMPLICIT.

DEPEND="dev-lang/python[-build(-)]" would be ok so long as build is not
in IUSE_IMPLICIT, regardless of build's status in IUSE.

As you can see, implicit flags mess things up lots... No big deal for
USERLAND et al, but they're a pain in the ass for 'normal' flags.

-- 
Ciaran McCreesh

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-dev] Preparing profiles for EAPI 3 IUSE strictness

[Andrew D Kirch]

Ciaran McCreesh wrote:
> On Tue, 07 Jul 2009 02:08:01 -0400
> Andrew D Kirch <trelane@trelane.net> wrote:
>   
> Given that your stated intention is for "Paludis to fail", and that
> "opposing [me] and everything [I] do was an initiative [you] started
> only after careful consideration", I'll kindly ask you to stop randomly
> jumping out and flinging turds, since it adds nothing to the discussion
> at hand and only serves to make it harder for Gentoo to function as a
> community.
>
>   
[response censored by fmmcor and devrel]

Andrew D Kirch
Funtoo.org

(though I'd note the timing of Ciaran's response is 5.5 hours after a
clear Mea Culpa that I had misspoken)
http://archives.gentoo.org/gentoo-dev/msg_ff98cc6ecb9054a2938d1379d2c78d1f.xml
@ 09:24
http://archives.gentoo.org/gentoo-dev/msg_a933ccf75960bba6d0a133d64454ebff.xml
@ 14:52