On Sun, Nov 21, 2010 at 04:22:52PM +0100, Sebastian Pipping wrote:
> Diego,
>
>
> On 11/21/10 15:29, Diego Elio Pettenò wrote:
> > The reason why many of them are in p.mask is usually because _I_ added
> > them there as they didn't mask with KEYWORDS="", and simply dropping
> > keywords would have users angry.
>
> Why does KEYWORDS="" on live ebuilds make users angry?
>
> Where can I find the rest of this thread?
Ehh, maybe on gentoo archives?
http://archives.gentoo.org/gentoo-dev/msg_4934999b1188cf3ecc53fea784054afb.xml

Is that what you are asking for?
>
> Where can I find documentation of the _current_ policy?
>
Here

http://devmanual.gentoo.org/ebuild-writing/functions/src_unpack/cvs-sources/index.html
http://devmanual.gentoo.org/ebuild-writing/functions/src_unpack/svn-sources/index.html
> Thanks!
>
>
>
> Sebastian
>

--
Markos Chandras (hwoarang)
Gentoo Linux Developer
Web: http://hwoarang.silverarrow.org
Key ID: 441AC410
Key FP: AAD0 8591 E3CD 445D 6411 3477 F7F7 1E8E 441A C410

On Sun, 21 Nov 2010 13:11:53 +0000
Markos Chandras <hwoarang@gentoo.org> wrote:

> Users interpret this as a 'double masking' which in fact it is since
> they need to touch two files before they are able to use the package.

Isn't that the point? People should be discouraged in every way not to use
live ebuilds. I'd add a third if we had one.

But yes, if I had to pick only one I'd go with dropping keywords over
package.mask. In fact it looks like I have some live ebuilds in the tree
that do exactly that.


--
fonts, gcc-porting, it makes no sense how it makes no sense
toolchain, wxwidgets but i'll take it free anytime
@ gentoo.org EFFD 380E 047A 4B51 D2BD C64F 8AA8 8346 F9A4 0662

On Sun, Nov 21, 2010 at 01:00:03PM -0600, Ryan Hill wrote:
> On Sun, 21 Nov 2010 13:11:53 +0000
> Markos Chandras <hwoarang@gentoo.org> wrote:
>
> > Users interpret this as a 'double masking' which in fact it is since
> > they need to touch two files before they are able to use the package.
>
> Isn't that the point? People should be discouraged in every way not to use
> live ebuilds. I'd add a third if we had one.
>
> But yes, if I had to pick only one I'd go with dropping keywords over
> package.mask. In fact it looks like I have some live ebuilds in the tree
> that do exactly that.
>
Actually not. Users are already familiar with the -9999 concept so there
is no point to add extra obstacles in their way. I am trying to find out
corner cases where double masking makes sense. Otherwise it makes no
sense to me. Actually the majority of users get confused when a package
is double masked. Just drop by forums etc and you will see

--
Markos Chandras (hwoarang)
Gentoo Linux Developer
Web: http://hwoarang.silverarrow.org
Key ID: 441AC410
Key FP: AAD0 8591 E3CD 445D 6411 3477 F7F7 1E8E 441A C410

On Sun, 21 Nov 2010 19:05:44 +0000
Markos Chandras <hwoarang@gentoo.org> wrote:

> > Isn't that the point? People should be discouraged in every way not to use
> > live ebuilds. I'd add a third if we had one.
> >
> > But yes, if I had to pick only one I'd go with dropping keywords over
> > package.mask. In fact it looks like I have some live ebuilds in the tree
> > that do exactly that.
> >
> Actually not. Users are already familiar with the -9999 concept so there
> is no point to add extra obstacles in their way. I am trying to find out
> corner cases where double masking makes sense. Otherwise it makes no
> sense to me. Actually the majority of users get confused when a package
> is double masked. Just drop by forums etc and you will see

Again, that's the point. If you can't figure out how to get around a
double mask then you have no business installing live ebuilds.

But this is getting off topic. If you want to change the policy to recommend
dropping keywords rather than using package.mask then I support it.
package.mask has the disadvantage that it's too easy to accidentally unmask
live versions with >=. And nothing stops someone from doing both if they
want.


--
fonts, gcc-porting, it makes no sense how it makes no sense
toolchain, wxwidgets but i'll take it free anytime
@ gentoo.org EFFD 380E 047A 4B51 D2BD C64F 8AA8 8346 F9A4 0662

On 11/21/10 20:30, Ryan Hill wrote:
>> Actually not. Users are already familiar with the -9999 concept so there
>> is no point to add extra obstacles in their way. I am trying to find out
>> corner cases where double masking makes sense. Otherwise it makes no
>> sense to me. Actually the majority of users get confused when a package
>> is double masked. Just drop by forums etc and you will see
>
> Again, that's the point. If you can't figure out how to get around a
> double mask then you have no business installing live ebuilds.

I know how to do it and still am grateful if I don't have to do both of
them. It's not about users only.


> If you want to change the policy to recommend
> dropping keywords rather than using package.mask then I support it.

+1 for KEYWORDS="". Less effort, less likely to break the tree.



Sebastian

On 11/21/10 17:27, Markos Chandras wrote:
>> Where can I find the rest of this thread?
> Ehh, maybe on gentoo archives?
> http://archives.gentoo.org/gentoo-dev/msg_4934999b1188cf3ecc53fea784054afb.xml
>
> Is that what you are asking for?

In a way, yes, thanks. Should have thought of looking there. I
supposed that you guys were talking somewhere else before: Your mail
starting the thread never got here. Your later mail that Ryan replied
to never got here either. So I guess maybe my spam detection has a problem.

Best



Sebastian

On Sun, Nov 21, 2010 at 01:30:15PM -0600, Ryan Hill wrote:
> On Sun, 21 Nov 2010 19:05:44 +0000
> Markos Chandras <hwoarang@gentoo.org> wrote:
>
> > > Isn't that the point? People should be discouraged in every way not to use
> > > live ebuilds. I'd add a third if we had one.
> > >
> > > But yes, if I had to pick only one I'd go with dropping keywords over
> > > package.mask. In fact it looks like I have some live ebuilds in the tree
> > > that do exactly that.
> > >
> > Actually not. Users are already familiar with the -9999 concept so there
> > is no point to add extra obstacles in their way. I am trying to find out
> > corner cases where double masking makes sense. Otherwise it makes no
> > sense to me. Actually the majority of users get confused when a package
> > is double masked. Just drop by forums etc and you will see
>
> Again, that's the point. If you can't figure out how to get around a
> double mask then you have no business installing live ebuilds.
>
> But this is getting off topic. If you want to change the policy to recommend
> dropping keywords rather than using package.mask then I support it.
> package.mask has the disadvantage that it's too easy to accidentally unmask
> live versions with >=. And nothing stops someone from doing both if they
> want.
>
>
> --
> fonts, gcc-porting, it makes no sense how it makes no sense
> toolchain, wxwidgets but i'll take it free anytime
> @ gentoo.org EFFD 380E 047A 4B51 D2BD C64F 8AA8 8346 F9A4 0662

If the majority of the devs ( at least of those who participate to this
thread ) is positive, then I will commit a patch to devmanual and
possibly migrate the cvs&svn sources pages into a single one.

--
Markos Chandras (hwoarang)
Gentoo Linux Developer
Web: http://hwoarang.silverarrow.org
Key ID: 441AC410
Key FP: AAD0 8591 E3CD 445D 6411 3477 F7F7 1E8E 441A C410

Ryan Hill posted on Sun, 21 Nov 2010 13:30:15 -0600 as excerpted:

> On Sun, 21 Nov 2010 19:05:44 +0000
> Markos Chandras <hwoarang@gentoo.org> wrote:
>
>> > Isn't that the point? People should be discouraged in every way not
>> > to use live ebuilds. I'd add a third if we had one.
>> >
>> Actually not. Users are already familiar with the -9999 concept so
>> there is no point to add extra obstacles in their way. I am trying to
>> find out corner cases where double masking makes sense. Otherwise it
>> makes no sense to me. Actually the majority of users get confused when
>> a package is double masked. Just drop by forums etc and you will see
>
> Again, that's the point. If you can't figure out how to get around a
> double mask then you have no business installing live ebuilds.

As a user who regularly uses certain live ebuilds (and contrasting SP),
strongly agreed. If the double-masking is confusing them, they're better
off sticking with standard versioned ebuilds as they're demonstrably not
up to dealing with other difficulties which might arise with a live
package and Gentoo doesn't need the extra bug noise. Double-masking for
live ebuilds in the main tree thus seems to me to be the best policy.

For the main tree, anyway. In overlays, I'd say it's up to the overlay
maintainers, as in many cases, the overlays are overtly experimental
already, and just the fact that it's in the overlay not the main tree has
added a barrier of its own.

It could also be argued whether the general main tree policy should be
"maintainer's discretion" or not. Obviously, it's that way already in
practice. Should we tighten up QA or make the policy overtly
"maintainer's discretion"?

--
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

On Sun, Nov 21, 2010 at 09:22:24PM +0000, Markos Chandras wrote:
> On Sun, Nov 21, 2010 at 01:30:15PM -0600, Ryan Hill wrote:
> > On Sun, 21 Nov 2010 19:05:44 +0000
> > Markos Chandras <hwoarang@gentoo.org> wrote:
> >
> > > > Isn't that the point? People should be discouraged in every way not to use
> > > > live ebuilds. I'd add a third if we had one.
> > > >
> > > > But yes, if I had to pick only one I'd go with dropping keywords over
> > > > package.mask. In fact it looks like I have some live ebuilds in the tree
> > > > that do exactly that.
> > > >
> > > Actually not. Users are already familiar with the -9999 concept so there
> > > is no point to add extra obstacles in their way. I am trying to find out
> > > corner cases where double masking makes sense. Otherwise it makes no
> > > sense to me. Actually the majority of users get confused when a package
> > > is double masked. Just drop by forums etc and you will see
> >
> > Again, that's the point. If you can't figure out how to get around a
> > double mask then you have no business installing live ebuilds.
> >
> > But this is getting off topic. If you want to change the policy to recommend
> > dropping keywords rather than using package.mask then I support it.
> > package.mask has the disadvantage that it's too easy to accidentally unmask
> > live versions with >=. And nothing stops someone from doing both if they
> > want.
> >
> >
> > --
> > fonts, gcc-porting, it makes no sense how it makes no sense
> > toolchain, wxwidgets but i'll take it free anytime
> > @ gentoo.org EFFD 380E 047A 4B51 D2BD C64F 8AA8 8346 F9A4 0662
>
> If the majority of the devs ( at least of those who participate to this
> thread ) is positive, then I will commit a patch to devmanual and
> possibly migrate the cvs&svn sources pages into a single one.

Count me in on this policy change. I think that double masking is a
pain as well.

I would agree that we might want to document somewhere that if
users are using "**" in package.keywords they really get to keep the
pieces, since they will be installing live ebuilds.

William

>>>>> On Sun, 21 Nov 2010, Markos Chandras wrote:

> If the majority of the devs ( at least of those who participate to
> this thread ) is positive, then I will commit a patch to devmanual
> and possibly migrate the cvs&svn sources pages into a single one.

Just for the record, I'm not in favour of such a change.

So far, the policy has been that KEYWORDS would reflect the stability
of the ebuild, whereas package.mask would indicate that the upstream
package is deemed unstable. Why should live ebuilds be handled in a
different way?

Also, for an ebuild with empty KEYWORDS, repoman will not indicate any
problems with dependencies.

Ulrich