[gentoo-project] rfc: live ebuilds, masking in p.mask or with empty keywords

[gentoo-project] rfc: live ebuilds, masking in p.mask or with empty keywords

[William Hubbs]

[-- Attachment #1: Type: text/plain, Size: 1498 bytes --]

All,

I am moving this to a new thread so the thread about agenda items does
not get too long.

On Sat, Nov 02, 2013 at 03:25:24PM -0400, Rick "Zero_Chaos" Farina wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On 11/02/2013 02:52 PM, William Hubbs wrote:

* snip *

> > Can we discuss and maybe vote on how we want live ebuilds in the tree? I
> > see three possibilities:
> > 
> > 1) empty keywords (this appears to be what most people are doing)
> > 2) package.mask (not required, the way I see it, because of 1 and
> > because package.mask shouldn't be permanent)
> > 3) both package.mask and empty keywords (this would be double masking,
> > and again shouldn't be necessary)
> > 
> > Thoughts?
> 
> Personally, I prefer option 1.  That said, there is a reason for Options
> 2 and 3.
> 
> When using a minor arch, a lot of packages are not keyworded for that
> arch, which then requires me to install them with KEYWORDS="**" and that
> pulls in live ebuilds all the time.  Personally, I'm fine dealing with
> things like that, but that would be a valid reason for requiring
> package.mask.  That said, if we want to persue that, I would say that we
> should start adding keywords to live ebuilds (~arch obviously) and
> p.mask them so we know what arches it is expected to work on.

Have you tried specifying the version of the package you want, e.g.

=app-misc/foo-x.y.z **

should pull in the exact version you want to test.

William


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

Re: [gentoo-project] rfc: live ebuilds, masking in p.mask or with empty keywords

[Ben de Groot]

On 3 November 2013 03:35, William Hubbs <williamh@gentoo.org> wrote:
> All,
>
> I am moving this to a new thread so the thread about agenda items does
> not get too long.
>
> On Sat, Nov 02, 2013 at 03:25:24PM -0400, Rick "Zero_Chaos" Farina wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> On 11/02/2013 02:52 PM, William Hubbs wrote:
>
> * snip *
>
>> > Can we discuss and maybe vote on how we want live ebuilds in the tree? I
>> > see three possibilities:
>> >
>> > 1) empty keywords (this appears to be what most people are doing)
>> > 2) package.mask (not required, the way I see it, because of 1 and
>> > because package.mask shouldn't be permanent)
>> > 3) both package.mask and empty keywords (this would be double masking,
>> > and again shouldn't be necessary)
>> >
>> > Thoughts?
>>
>> Personally, I prefer option 1.  That said, there is a reason for Options
>> 2 and 3.
>>
>> When using a minor arch, a lot of packages are not keyworded for that
>> arch, which then requires me to install them with KEYWORDS="**" and that
>> pulls in live ebuilds all the time.  Personally, I'm fine dealing with
>> things like that, but that would be a valid reason for requiring
>> package.mask.  That said, if we want to persue that, I would say that we
>> should start adding keywords to live ebuilds (~arch obviously) and
>> p.mask them so we know what arches it is expected to work on.
>
> Have you tried specifying the version of the package you want, e.g.
>
> =app-misc/foo-x.y.z **
>
> should pull in the exact version you want to test.
>
> William
>

We already have an existing policy, which is option 1.

-- 
Cheers,

Ben | yngwin
Gentoo developer

Re: [gentoo-project] rfc: live ebuilds, masking in p.mask or with empty keywords

[Rich Freeman]

On Sun, Nov 3, 2013 at 12:03 AM, Ben de Groot <yngwin@gentoo.org> wrote:
>>
>
> We already have an existing policy, which is option 1.

Citation?

I've been masking live ebuilds with option 2 ever since QA filed a bug
telling me to - I don't think that policy was written down either.

What is written down is:
"Live" cvs.eclass ebuilds are generally only intended for the
convenience of developers and should always be masked with a ~[arch]
keyword.   [1]

That would be none of the above - half the tree is "masked" with ~arch
keywords.  I'm not suggesting that this is the right way to do it -
only that there certainly is nothing approaching a clear policy on the
matter.

Rich

[1] - http://www.gentoo.org/proj/en/devrel/handbook/handbook.xml?part=2&chap=1&style=printable

Re: [gentoo-project] rfc: live ebuilds, masking in p.mask or with empty keywords

[Tom Wijsman]

[-- Attachment #1: Type: text/plain, Size: 2039 bytes --]

On Sun, 3 Nov 2013 06:05:09 -0500
Rich Freeman <rich0@gentoo.org> wrote:

> On Sun, Nov 3, 2013 at 12:03 AM, Ben de Groot <yngwin@gentoo.org>
> wrote:
> >>
> >
> > We already have an existing policy, which is option 1.
> 
> Citation?

Nobody has found one since the, except for the bug planning to patch it.

> I've been masking live ebuilds with option 2 ever since QA filed a bug
> telling me to - I don't think that policy was written down either.

"Note that CVS ebuilds should not generally be added to the tree
(except under package.mask) for the following reasons: ..."

http://devmanual.gentoo.org/ebuild-writing/functions/src_unpack/cvs-sources/index.html#disadvantages-of-cvs-sources

Although the word "generally" makes it a bit vague, what are exceptions?

There also hasn't been found a link with consensus as far as I followed
the discussion; so, no idea how and why that landed up there.

> What is written down is:
> "Live" cvs.eclass ebuilds are generally only intended for the
> convenience of developers and should always be masked with a ~[arch]
> keyword.   [1]

The last bit "masked with a ~[arch] keyword" can be misinterpreted to
mean two different things; one way to interpret it is 'package.mask
with KEYWORDS="~[arch]"', another way is '~[arch] keyword masked'.

> That would be none of the above - half the tree is "masked" with ~arch
> keywords.  I'm not suggesting that this is the right way to do it -
> only that there certainly is nothing approaching a clear policy on the
> matter.

Exactly, it's also quite a minor issue since we're talking about just a
few entries that are done in the opposite way; but I can imagine it
becoming more major as more people follow the suggestion, with
eventually someone quoting a package.mask commit and asking "why?!".

-- 
With kind regards,

Tom Wijsman (TomWij)
Gentoo Developer

E-mail address  : TomWij@gentoo.org
GPG Public Key  : 6D34E57D
GPG Fingerprint : C165 AF18 AB4C 400B C3D2  ABF0 95B2 1FCD 6D34 E57D

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

Re: [gentoo-project] rfc: live ebuilds, masking in p.mask or with empty keywords

[Matthew Thode]

[-- Attachment #1: Type: text/plain, Size: 1286 bytes --]

On 11/03/2013 05:05 AM, Rich Freeman wrote:
> On Sun, Nov 3, 2013 at 12:03 AM, Ben de Groot <yngwin@gentoo.org> wrote:
>>>
>>
>> We already have an existing policy, which is option 1.
> 
> Citation?
> 
> I've been masking live ebuilds with option 2 ever since QA filed a bug
> telling me to - I don't think that policy was written down either.
> 
> What is written down is:
> "Live" cvs.eclass ebuilds are generally only intended for the
> convenience of developers and should always be masked with a ~[arch]
> keyword.   [1]
> 
> That would be none of the above - half the tree is "masked" with ~arch
> keywords.  I'm not suggesting that this is the right way to do it -
> only that there certainly is nothing approaching a clear policy on the
> matter.
> 
> Rich
> 
> [1] - http://www.gentoo.org/proj/en/devrel/handbook/handbook.xml?part=2&chap=1&style=printable
> 

I feel that we need some way for live ebuilds to be unmasked.  For
instance, there are openstack ebuilds are available that track stable
branches in git.  These are very useful as they tend to get bugs before
a release is made.  Backporting patches is no fun.  I could see 9999
packages getting p.masked but not something like 2013.1.2.9999.

-- 
-- Matthew Thode (prometheanfire)


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

Re: [gentoo-project] rfc: live ebuilds, masking in p.mask or with empty keywords

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 1506 bytes --]

Dnia 2013-11-03, o godz. 14:13:16
Matthew Thode <prometheanfire@gentoo.org> napisał(a):

> On 11/03/2013 05:05 AM, Rich Freeman wrote:
> > On Sun, Nov 3, 2013 at 12:03 AM, Ben de Groot <yngwin@gentoo.org> wrote:
> >>>
> >>
> >> We already have an existing policy, which is option 1.
> > 
> > Citation?
> > 
> > I've been masking live ebuilds with option 2 ever since QA filed a bug
> > telling me to - I don't think that policy was written down either.
> > 
> > What is written down is:
> > "Live" cvs.eclass ebuilds are generally only intended for the
> > convenience of developers and should always be masked with a ~[arch]
> > keyword.   [1]
> > 
> > That would be none of the above - half the tree is "masked" with ~arch
> > keywords.  I'm not suggesting that this is the right way to do it -
> > only that there certainly is nothing approaching a clear policy on the
> > matter.
> > 
> > Rich
> > 
> > [1] - http://www.gentoo.org/proj/en/devrel/handbook/handbook.xml?part=2&chap=1&style=printable
> 
> I feel that we need some way for live ebuilds to be unmasked.  For
> instance, there are openstack ebuilds are available that track stable
> branches in git.  These are very useful as they tend to get bugs before
> a release is made.  Backporting patches is no fun.  I could see 9999
> packages getting p.masked but not something like 2013.1.2.9999.

You can do:

  <dev-python/foo-9999 **

in package.accept_keywords.


-- 
Best regards,
Michał Górny

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

Re: [gentoo-project] rfc: live ebuilds, masking in p.mask or with empty keywords

[Matthew Thode]

[-- Attachment #1: Type: text/plain, Size: 1699 bytes --]

On 11/03/2013 03:16 PM, Michał Górny wrote:
> Dnia 2013-11-03, o godz. 14:13:16
> Matthew Thode <prometheanfire@gentoo.org> napisał(a):
> 
>> On 11/03/2013 05:05 AM, Rich Freeman wrote:
>>> On Sun, Nov 3, 2013 at 12:03 AM, Ben de Groot <yngwin@gentoo.org> wrote:
>>>>>
>>>>
>>>> We already have an existing policy, which is option 1.
>>>
>>> Citation?
>>>
>>> I've been masking live ebuilds with option 2 ever since QA filed a bug
>>> telling me to - I don't think that policy was written down either.
>>>
>>> What is written down is:
>>> "Live" cvs.eclass ebuilds are generally only intended for the
>>> convenience of developers and should always be masked with a ~[arch]
>>> keyword.   [1]
>>>
>>> That would be none of the above - half the tree is "masked" with ~arch
>>> keywords.  I'm not suggesting that this is the right way to do it -
>>> only that there certainly is nothing approaching a clear policy on the
>>> matter.
>>>
>>> Rich
>>>
>>> [1] - http://www.gentoo.org/proj/en/devrel/handbook/handbook.xml?part=2&chap=1&style=printable
>>
>> I feel that we need some way for live ebuilds to be unmasked.  For
>> instance, there are openstack ebuilds are available that track stable
>> branches in git.  These are very useful as they tend to get bugs before
>> a release is made.  Backporting patches is no fun.  I could see 9999
>> packages getting p.masked but not something like 2013.1.2.9999.
> 
> You can do:
> 
>   <dev-python/foo-9999 **
> 
> in package.accept_keywords.
> 
> 
I generally tell people to do that (except in openstack, where
2013.1.9999 and 2013.2.9999 are very diferent :P).

-- 
-- Matthew Thode (prometheanfire)


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]