[gentoo-dev] autoconf now supports multislots

[gentoo-dev] autoconf now supports multislots

[Mike Frysinger]

[-- Attachment #1: Type: text/plain, Size: 668 bytes --]

for people who work with projects that require specific versions of autoconf 
and nothing else (like ==2.64), i've long left older versions in the tree so 
people could downgrade on the fly.  this isn't so bad as you can do:
	emerge -bk ~autoconf-2.64
	<use it>
	emerge -bk autoconf

but i'm super lazy, so even this manual step is annoying.  as such, i've added 
USE=multislot support to autoconf (just like it is with binutils & gcc).

now you can install multiple versions in parallel and run:
	WANT_AUTOCONF=2.64 autoconf
or more specifically:
	autoconf-2.64
without affecting the default `autoconf`.

the default is, of course, unchanged.
-mike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

Re: [gentoo-dev] autoconf now supports multislots

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 333 bytes --]

On Wed, 17 Apr 2013 14:33:29 -0400
Mike Frysinger <vapier@gentoo.org> wrote:
> but i'm super lazy, so even this manual step is annoying.  as such,
> i've added USE=multislot support to autoconf (just like it is with
> binutils & gcc).

But it's massively illegal and doesn't work correctly in Portage.

-- 
Ciaran McCreesh

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

Re: [gentoo-dev] autoconf now supports multislots

[Michael Mol]

On 4/17/2013 2:48 PM, Ciaran McCreesh wrote:
> On Wed, 17 Apr 2013 14:33:29 -0400
> Mike Frysinger <vapier@gentoo.org> wrote:
>> but i'm super lazy, so even this manual step is annoying.  as such,
>> i've added USE=multislot support to autoconf (just like it is with
>> binutils & gcc).
> 
> But it's massively illegal and doesn't work correctly in Portage.
> 

For the benefit of those of us (well, me) not sufficiently versed in PMS
et al to be able to immediately deduce why it's illegal and why it
wouldn't work correctly in Portage, could you please elucidate?

Re: [gentoo-dev] autoconf now supports multislots

[Mike Frysinger]

[-- Attachment #1: Type: Text/Plain, Size: 673 bytes --]

On Wednesday 17 April 2013 14:48:13 Ciaran McCreesh wrote:
> On Wed, 17 Apr 2013 14:33:29 -0400 Mike Frysinger wrote:
> > but i'm super lazy, so even this manual step is annoying.  as such,
> > i've added USE=multislot support to autoconf (just like it is with
> > binutils & gcc).
> 
> But it's massively illegal and doesn't work correctly in Portage.

if you don't like it, then don't use it.  as i stated already, default 
behavior (USE=-multislot) is unchanged.

i'm not really interested in rehashing the topic with you.  there's a long 
standing bug open on the topic you can easily refer to.  nothing you say is 
going to change behavior today.
-mike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

Re: [gentoo-dev] autoconf now supports multislots

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 1158 bytes --]

On Wed, 17 Apr 2013 14:58:20 -0400
Michael Mol <mikemol@gmail.com> wrote:
> On 4/17/2013 2:48 PM, Ciaran McCreesh wrote:
> > On Wed, 17 Apr 2013 14:33:29 -0400
> > Mike Frysinger <vapier@gentoo.org> wrote:
> >> but i'm super lazy, so even this manual step is annoying.  as such,
> >> i've added USE=multislot support to autoconf (just like it is with
> >> binutils & gcc).
> > 
> > But it's massively illegal and doesn't work correctly in Portage.
> > 
> 
> For the benefit of those of us (well, me) not sufficiently versed in
> PMS et al to be able to immediately deduce why it's illegal and why it
> wouldn't work correctly in Portage, could you please elucidate?

Metadata variables, such as SLOT, are cached and are required to be
invariant. When an ebuild violates that requirement, the package
mangler usually sees the wrong value for the variable when doing the
resolution. This means the package mangler could calculate and display
an invalid resolution for what it ends up doing, or it could ignore the
ebuild's attempts at changing SLOT from what the cache holds, or it
could do something even worse.

-- 
Ciaran McCreesh

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

Re: [gentoo-dev] autoconf now supports multislots

[Chí-Thanh Christopher Nguyễn]

Michael Mol schrieb:
>> But it's massively illegal and doesn't work correctly in Portage.
>>
> 
> For the benefit of those of us (well, me) not sufficiently versed in PMS
> et al to be able to immediately deduce why it's illegal and why it
> wouldn't work correctly in Portage, could you please elucidate?

It's illegal because it calls "use" in global scope and modifies SLOT based
on the result.
It is not correctly handled by portage, as you can test by
# USE="multislot" emerge -pv gcc:4.6.3


Best regards,
Chí-Thanh Christopher Nguyễn

[gentoo-dev] Re: autoconf now supports multislots

[Ryan Hill]

[-- Attachment #1: Type: text/plain, Size: 792 bytes --]

On Wed, 17 Apr 2013 15:03:55 -0400
Mike Frysinger <vapier@gentoo.org> wrote:
> On Wednesday 17 April 2013 14:48:13 Ciaran McCreesh wrote:
> > On Wed, 17 Apr 2013 14:33:29 -0400 Mike Frysinger wrote:

> > > but i'm super lazy, so even this manual step is annoying.  as such,
> > > i've added USE=multislot support to autoconf (just like it is with
> > > binutils & gcc).
> > 
> > But it's massively illegal and doesn't work correctly in Portage.
> 
> if you don't like it, then don't use it.  as i stated already, default 
> behavior (USE=-multislot) is unchanged.

Maybe just add something like "do not enable this unless you know what you're
doing" to the flag description.  Thanks for doing this, it annoyed me too.


-- 
gcc-porting
toolchain, wxwidgets
@ gentoo.org

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 490 bytes --]

Re: [gentoo-dev] Re: autoconf now supports multislots

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 899 bytes --]

On Sat, 20 Apr 2013 00:22:56 -0600
Ryan Hill <dirtyepic@gentoo.org> wrote:

> On Wed, 17 Apr 2013 15:03:55 -0400
> Mike Frysinger <vapier@gentoo.org> wrote:
> > On Wednesday 17 April 2013 14:48:13 Ciaran McCreesh wrote:
> > > On Wed, 17 Apr 2013 14:33:29 -0400 Mike Frysinger wrote:
> 
> > > > but i'm super lazy, so even this manual step is annoying.  as such,
> > > > i've added USE=multislot support to autoconf (just like it is with
> > > > binutils & gcc).
> > > 
> > > But it's massively illegal and doesn't work correctly in Portage.
> > 
> > if you don't like it, then don't use it.  as i stated already, default 
> > behavior (USE=-multislot) is unchanged.
> 
> Maybe just add something like "do not enable this unless you know what you're
> doing" to the flag description.  Thanks for doing this, it annoyed me too.

And mask it.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 966 bytes --]

Re: [gentoo-dev] autoconf now supports multislots

[hasufell]

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

On 04/17/2013 09:05 PM, Ciaran McCreesh wrote:
> On Wed, 17 Apr 2013 14:58:20 -0400 Michael Mol <mikemol@gmail.com>
> wrote:
>> On 4/17/2013 2:48 PM, Ciaran McCreesh wrote:
>>> On Wed, 17 Apr 2013 14:33:29 -0400 Mike Frysinger
>>> <vapier@gentoo.org> wrote:
>>>> but i'm super lazy, so even this manual step is annoying.  as
>>>> such, i've added USE=multislot support to autoconf (just like
>>>> it is with binutils & gcc).
>>> 
>>> But it's massively illegal and doesn't work correctly in
>>> Portage.
>>> 
>> 
>> For the benefit of those of us (well, me) not sufficiently versed
>> in PMS et al to be able to immediately deduce why it's illegal
>> and why it wouldn't work correctly in Portage, could you please
>> elucidate?
> 
> Metadata variables, such as SLOT, are cached and are required to
> be invariant. When an ebuild violates that requirement, the
> package mangler usually sees the wrong value for the variable when
> doing the resolution. This means the package mangler could
> calculate and display an invalid resolution for what it ends up
> doing, or it could ignore the ebuild's attempts at changing SLOT
> from what the cache holds, or it could do something even worse.
> 

So why is this allowed in this ebuild again?

> SLOT=$(usex multislot "${PV}" "2.5")
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.20 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBAgAGBQJSSWUtAAoJEFpvPKfnPDWzFC4H/2Qif+q8g64TJn4oIoKoTsy4
HBelOIQy6673EbWn5Gq1cdUT7+BCbw5Dj5L1CGcdOLf7bdvklIAPa0w7foSj3rDv
KUzyTb7Vog+KWKgXrmZat/r4u9tN2V/2zdJkKcxQtJnT0Drb3ISo8J4SSK2DBhDH
f1o/FarT5W33bTjwp2BxIksB0sQsp8XsR+hBViMCc7EUJvROEgJO4Cg5lcArALg0
UUoH7T9mmLtNPomB8T4mAKt4aknW+ECeK4TNjsZMr/3JAawgus0YzDx5cz5weJ+J
zQ9YQlMu4USuCYJylf+pWp7OMHtSmKvtrHdtUDv1GNIumXgU3EK0z/2Mj57xJrA=
=XtgG
-----END PGP SIGNATURE-----