Re: [gentoo-dev] Unused ebuild built_with_use cleanup

2009-10-29 Thread Doug Goldstein
On Wed, Oct 28, 2009 at 4:51 AM, Petteri Räty betelge...@gentoo.org wrote:
 Mike Frysinger wrote:
 On Tuesday 27 October 2009 14:46:31 Petteri Räty wrote:
 Normally old versions are not kept around as already said if you read
 the thread.

 normal is not the same thing as always.  unless you're the maintainer, 
 you
 have no idea whether old versions are kept there on purpose.  ive had people
 delete older versions of packages on me simply because they made this invalid
 assumption without talking to the maintainer.  the rest of the thread is
 irrelevant as this point was not made.


 Yes I have no idea. That's why I asked on gentoo-dev-announce for
 maintainers to tell me if they are kept on purpose so the point was made
 already at the very start.

 Regards,
 Petteri



I see several of my packages on there as well and there's absolutely
no way you're culling them since they have a definite use since you're
script clearly failed to take into consideration various profiles and
SLOTs. Biggest reason that my packages still use built_with_use is the
lack of the --missing option for EAPI=2.

-- 
Doug Goldstein



Re: [gentoo-dev] Unused ebuild built_with_use cleanup

2009-10-28 Thread Petteri Räty
Mike Frysinger wrote:
 On Tuesday 27 October 2009 14:46:31 Petteri Räty wrote:
 Normally old versions are not kept around as already said if you read
 the thread.
 
 normal is not the same thing as always.  unless you're the maintainer, 
 you 
 have no idea whether old versions are kept there on purpose.  ive had people 
 delete older versions of packages on me simply because they made this invalid 
 assumption without talking to the maintainer.  the rest of the thread is 
 irrelevant as this point was not made.
 

Yes I have no idea. That's why I asked on gentoo-dev-announce for
maintainers to tell me if they are kept on purpose so the point was made
already at the very start.

Regards,
Petteri



signature.asc
Description: OpenPGP digital signature


Re: [gentoo-dev] Unused ebuild built_with_use cleanup

2009-10-28 Thread Alexis Ballier
On Wed, 28 Oct 2009 11:51:34 +0200
Petteri Räty betelge...@gentoo.org wrote:

 Mike Frysinger wrote:
  On Tuesday 27 October 2009 14:46:31 Petteri Räty wrote:
  Normally old versions are not kept around as already said if you
  read the thread.
  
  normal is not the same thing as always.  unless you're the
  maintainer, you have no idea whether old versions are kept there on
  purpose.  ive had people delete older versions of packages on me
  simply because they made this invalid assumption without talking to
  the maintainer.  the rest of the thread is irrelevant as this point
  was not made.
  
 
 Yes I have no idea. That's why I asked on gentoo-dev-announce for
 maintainers to tell me if they are kept on purpose so the point was
 made already at the very start.


Still, it has already been stated but: please file bugs! A big tracker
would probably be messy, one bug / herd or category could be better.
You did file some bugs at some point, where I was in the assigned herd
and did proceed quickly I think. Why did you stop? Some were just
straight removal, some did need some other removal in order to be clean
(this may be due to the fact that packages didnt work with latest
versions but didn't have restricted deps on the package that would have
to be removed).

I am not following -dev very carefully and am too lazy to dig in a list
posted on -dev-announce to find out if I need to do something or not. A
bug makes that easy to track and is the usual way of getting ebuild
changes done in the tree. We rant on users who report bugs on -dev, I'm
afraid the the same applies to developers :)


Also, remark that your script is flawed, I see some freebsd stuff in
your list which are certainly the best visible version on some
profiles. Again, this will be better tracked in a bug.


Alexis.


signature.asc
Description: PGP signature


Re: [gentoo-dev] Unused ebuild built_with_use cleanup

2009-10-27 Thread Mike Frysinger
On Tuesday 27 October 2009 14:46:31 Petteri Räty wrote:
 Normally old versions are not kept around as already said if you read
 the thread.

normal is not the same thing as always.  unless you're the maintainer, you 
have no idea whether old versions are kept there on purpose.  ive had people 
delete older versions of packages on me simply because they made this invalid 
assumption without talking to the maintainer.  the rest of the thread is 
irrelevant as this point was not made.

a quick check of your list shows wine/uclibc shouldnt be blindly culled.

 Live ebuilds shouldn't really have been in the original list
 with my intended logic. For them I will usually just migrate them to
 EAPI 2 like with other packages we have been touching.

OK

 Using a tracker bug makes sense if you expect some action from individual
 maintainers which is not the case here as they can just leave the job to
 people nuking built_with_use like me.

i dont plan on fixing my ebuilds soonish since this isnt a terribly important 
issue, but they'll get fixed at some point
-mike


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


Re: [gentoo-dev] Unused ebuild built_with_use cleanup

2009-10-09 Thread Petteri Räty
Patrick Lauer wrote:
 
 And that's with all the forced migrations for features like use-deps or the 
 removal of built_with_use. So unless there's some strongly needed features 
 there's no need for it. I can't remember any feature in the EAPI 3 list that 
 really looked useful to me, so not adding it now now now doesn't bother me at 
 all. Just causes more confusion for no real benefit. So who cares if it is 
 delayed by a few timeunits, there's much more important stuff to do.
 

Here's two features that by themselves are important enough to get EAPI
3 implemented. Using pkg_pretend it should be possible to eliminate
expected dies from build time and as such improving user experience. An
example is two use flags that conflict with each other. Use dependency
defaults make the life of ebuild writers easier as you don't need to be
careful with version restrictions any more if you have a case where
something has been on by default and then becomes a use flag for
example. This should eliminate cases like causing glibc downgrade in the
depgraph by being careless.

Regards,
Petteri



signature.asc
Description: OpenPGP digital signature


Re: [gentoo-dev] Unused ebuild built_with_use cleanup

2009-10-09 Thread Petteri Räty
Tomáš Chvátal wrote:
 On čtvrtek 08 Říjen 2009,  23:34:10  Petteri Räty wrote:
 Even this is wrong because:
 Hi
 ...
 betelge...@pena ~ $ portageq metadata / ebuild sys-libs/glibc-2.2.5-r10
 IUSE nls

 For most packages old versions are not kept around so just doing
 =cat/foo-X.Y[use] is fine and EAPI 3 is not needed. I haven't come
 across a case that couldn't be done with EAPI 2 yet. Granted the atoms
 can be a bit cleaner with EAPI 3 but considering how much zmedico slacks
 in implementing it, it's best to do migrating now with EAPI 2 than EAPI
 3 in the far future.
 This is not exactly nice of you. And taking in account that you are actualy 
 the council member it makes me feel not entirely happy.
 If we just simply take look onto this:
 http://cia.vc/stats/author/zmedico/
 we can count that Zac commit something into portage every 3 hours. It does 
 not 
 look entirely like slacking...
 So you are basicaly proposing that maintaining the current codebase and 
 improving what we already have is less important than providing new features, 
 that is also not good.
 

I am not suggesting that the work Zac does is worthless. I am saying
that implementing EAPI 3 is not a colossal amount of work and if it was
a priority to him it would have already been implemented. If he feels
offended by my original comment, I have no problem apologizing to him.
Not having EAPI 3 implemented in general is not his fault as many of us
have the needed skills to start helping on the Portage code base. The
reality just is that he is the most likely person to implement it and as
such a very important factor on when it happens.

Regards,
Petteri



signature.asc
Description: OpenPGP digital signature


Re: [gentoo-dev] Unused ebuild built_with_use cleanup

2009-10-08 Thread Petteri Räty
Marijn Schouten (hkBst) wrote:
 Petteri Räty wrote:
 I wrote a script to check which ebuilds use built_with_use and have
 keywords in never versions making the ebuild unused. This means that
 neither arch or ~arch users are likely to install the ebuild. The script
 and the list of ebuilds is attached. I plan on removing all these
 ebuilds two weeks from now unless a reason is given why not to. If you
 see an ebuild on the list that should be kept, please migrate it to EAPI
 2. If you need assistance in migrating, I can help. With these gone
 built_with_use usage will be down to about 600:
 
 I have some ebuilds on the list: plt-scheme, stklos, lilypond. I usually clean
 up unused ebuilds some time after a new version has gone stable.
 
 Anyway my question is: what is the point of removing unused versions in the
 proposed manner? If the newer version is not ported to EAPI 2 and also uses
 built_with_use what do we gain? Even if it is already ported, do we gain
 anything by the propsed removal? Are all unused ebuilds evil?
 

It saves my time when removing built_with_use from packages that still
have active versions using it. Many packages are also unmaintained so
the versions with built_with_use don't get removed without doing
something like this.

 If built_with_use is to be eliminated from the tree I propose that effort is
 directed towards porting and stabling ebuilds that still use it. After the
 stable version has begun using EAPI 2 use deps, then all uses of 
 built_with_use
 in other versions can be considered obsolete and those ebuilds can be removed 
 in
 one fell sweep if need be.
 

It will be eliminated eventually. I am in the process of doing so but as
you can see from the numbers it takes quite a lot of work to get rid of
them.

Regards,
Petteri



signature.asc
Description: OpenPGP digital signature


Re: [gentoo-dev] Unused ebuild built_with_use cleanup

2009-10-08 Thread Petteri Räty
Stelian Ionescu wrote:
 On Tue, 2009-09-29 at 16:32 +0300, Petteri Räty wrote:
 I wrote a script to check which ebuilds use built_with_use and have
 keywords in never versions making the ebuild unused. This means that
 neither arch or ~arch users are likely to install the ebuild. The script
 and the list of ebuilds is attached. I plan on removing all these
 ebuilds two weeks from now unless a reason is given why not to. If you
 see an ebuild on the list that should be kept, please migrate it to EAPI
 2. If you need assistance in migrating, I can help. With these gone
 built_with_use usage will be down to about 600:
 
 built_with_use shouldn't be removed until EAPI=3 gets approved because
 currently there's no good way to emulate --missing true|false|die
 yes, I can use something like this in sbcl:
 || ( sys-libs/glibc-2.6[nptl] =sys-libs/glibc-2.6 )
 but not all its use cases may be this simple
 

Even this is wrong because:

betelge...@pena ~ $ portageq metadata / ebuild sys-libs/glibc-2.2.5-r10 IUSE
nls

For most packages old versions are not kept around so just doing
=cat/foo-X.Y[use] is fine and EAPI 3 is not needed. I haven't come
across a case that couldn't be done with EAPI 2 yet. Granted the atoms
can be a bit cleaner with EAPI 3 but considering how much zmedico slacks
in implementing it, it's best to do migrating now with EAPI 2 than EAPI
3 in the far future.

Regards,
Petteri



signature.asc
Description: OpenPGP digital signature


Re: [gentoo-dev] Unused ebuild built_with_use cleanup

2009-10-08 Thread Jeremy Olexa
On Thu, Oct 8, 2009 at 4:34 PM, Petteri Räty betelge...@gentoo.org wrote:
 Stelian Ionescu wrote:
 On Tue, 2009-09-29 at 16:32 +0300, Petteri Räty wrote:
 I wrote a script to check which ebuilds use built_with_use and have
 keywords in never versions making the ebuild unused. This means that
 neither arch or ~arch users are likely to install the ebuild. The script
 and the list of ebuilds is attached. I plan on removing all these
 ebuilds two weeks from now unless a reason is given why not to. If you
 see an ebuild on the list that should be kept, please migrate it to EAPI
 2. If you need assistance in migrating, I can help. With these gone
 built_with_use usage will be down to about 600:

 built_with_use shouldn't be removed until EAPI=3 gets approved because
 currently there's no good way to emulate --missing true|false|die
 yes, I can use something like this in sbcl:
 || ( sys-libs/glibc-2.6[nptl] =sys-libs/glibc-2.6 )
 but not all its use cases may be this simple


 Even this is wrong because:

 betelge...@pena ~ $ portageq metadata / ebuild sys-libs/glibc-2.2.5-r10 IUSE
 nls

 For most packages old versions are not kept around so just doing
=cat/foo-X.Y[use] is fine and EAPI 3 is not needed. I haven't come
 across a case that couldn't be done with EAPI 2 yet. Granted the atoms
 can be a bit cleaner with EAPI 3 but considering how much zmedico slacks
 in implementing it, it's best to do migrating now with EAPI 2 than EAPI

Comments like these are not acceptable. Zac works his tail off on
portage. Please refrain from such comments in the future.
-Jeremy

 3 in the far future.

 Regards,
 Petteri





Re: [gentoo-dev] Unused ebuild built_with_use cleanup

2009-10-08 Thread Petteri Räty
Jeremy Olexa wrote:
 On Thu, Oct 8, 2009 at 4:34 PM, Petteri Räty betelge...@gentoo.org wrote:
 Stelian Ionescu wrote:
 On Tue, 2009-09-29 at 16:32 +0300, Petteri Räty wrote:
 I wrote a script to check which ebuilds use built_with_use and have
 keywords in never versions making the ebuild unused. This means that
 neither arch or ~arch users are likely to install the ebuild. The script
 and the list of ebuilds is attached. I plan on removing all these
 ebuilds two weeks from now unless a reason is given why not to. If you
 see an ebuild on the list that should be kept, please migrate it to EAPI
 2. If you need assistance in migrating, I can help. With these gone
 built_with_use usage will be down to about 600:
 built_with_use shouldn't be removed until EAPI=3 gets approved because
 currently there's no good way to emulate --missing true|false|die
 yes, I can use something like this in sbcl:
 || ( sys-libs/glibc-2.6[nptl] =sys-libs/glibc-2.6 )
 but not all its use cases may be this simple

 Even this is wrong because:

 betelge...@pena ~ $ portageq metadata / ebuild sys-libs/glibc-2.2.5-r10 IUSE
 nls

 For most packages old versions are not kept around so just doing
 =cat/foo-X.Y[use] is fine and EAPI 3 is not needed. I haven't come
 across a case that couldn't be done with EAPI 2 yet. Granted the atoms
 can be a bit cleaner with EAPI 3 but considering how much zmedico slacks
 in implementing it, it's best to do migrating now with EAPI 2 than EAPI
 
 Comments like these are not acceptable. Zac works his tail off on
 portage. Please refrain from such comments in the future.
 -Jeremy
 

He has said himself that he is not especially interested in implementing
EAPI 3 so slack at least to me seems like a good term.

Regards,
Petteri



signature.asc
Description: OpenPGP digital signature


Re: [gentoo-dev] Unused ebuild built_with_use cleanup

2009-10-08 Thread Tomáš Chvátal
On čtvrtek 08 Říjen 2009,  23:34:10  Petteri Räty wrote:
 Even this is wrong because:
Hi
...

 betelge...@pena ~ $ portageq metadata / ebuild sys-libs/glibc-2.2.5-r10
 IUSE nls

 For most packages old versions are not kept around so just doing
 =cat/foo-X.Y[use] is fine and EAPI 3 is not needed. I haven't come
 across a case that couldn't be done with EAPI 2 yet. Granted the atoms
 can be a bit cleaner with EAPI 3 but considering how much zmedico slacks
 in implementing it, it's best to do migrating now with EAPI 2 than EAPI
 3 in the far future.
This is not exactly nice of you. And taking in account that you are actualy 
the council member it makes me feel not entirely happy.
If we just simply take look onto this:
http://cia.vc/stats/author/zmedico/
we can count that Zac commit something into portage every 3 hours. It does not 
look entirely like slacking...
So you are basicaly proposing that maintaining the current codebase and 
improving what we already have is less important than providing new features, 
that is also not good.

Just my 2 cents

Tomas



Re: [gentoo-dev] Unused ebuild built_with_use cleanup

2009-10-08 Thread Patrick Lauer
On Friday 09 October 2009 00:22:26 Petteri Räty wrote:

  across a case that couldn't be done with EAPI 2 yet. Granted the atoms
  can be a bit cleaner with EAPI 3 but considering how much zmedico slacks
  in implementing it, it's best to do migrating now with EAPI 2 than EAPI
 
  Comments like these are not acceptable. Zac works his tail off on
  portage. Please refrain from such comments in the future.
  -Jeremy
 
 He has said himself that he is not especially interested in implementing
 EAPI 3 so slack at least to me seems like a good term.

I'm not sold on it either. Most devs barely know the difference between 
different EAPIs (just extrapolating from the many questions I see e.g. on IRC)
(and I think they shouldn't have to know because we should be using one EAPI 
only, but that's just my random opinion)

Most ebuilds are still EAPI0 - rough count gives me:

EAPI 0 - 19654
EAPI 1 -  1651
EAPI 2 -  5497

And that's with all the forced migrations for features like use-deps or the 
removal of built_with_use. So unless there's some strongly needed features 
there's no need for it. I can't remember any feature in the EAPI 3 list that 
really looked useful to me, so not adding it now now now doesn't bother me at 
all. Just causes more confusion for no real benefit. So who cares if it is 
delayed by a few timeunits, there's much more important stuff to do.



Re: [gentoo-dev] Unused ebuild built_with_use cleanup

2009-10-07 Thread Marijn Schouten (hkBst)
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1

Petteri Räty wrote:
 I wrote a script to check which ebuilds use built_with_use and have
 keywords in never versions making the ebuild unused. This means that
 neither arch or ~arch users are likely to install the ebuild. The script
 and the list of ebuilds is attached. I plan on removing all these
 ebuilds two weeks from now unless a reason is given why not to. If you
 see an ebuild on the list that should be kept, please migrate it to EAPI
 2. If you need assistance in migrating, I can help. With these gone
 built_with_use usage will be down to about 600:

I have some ebuilds on the list: plt-scheme, stklos, lilypond. I usually clean
up unused ebuilds some time after a new version has gone stable.

Anyway my question is: what is the point of removing unused versions in the
proposed manner? If the newer version is not ported to EAPI 2 and also uses
built_with_use what do we gain? Even if it is already ported, do we gain
anything by the propsed removal? Are all unused ebuilds evil?

If built_with_use is to be eliminated from the tree I propose that effort is
directed towards porting and stabling ebuilds that still use it. After the
stable version has begun using EAPI 2 use deps, then all uses of built_with_use
in other versions can be considered obsolete and those ebuilds can be removed in
one fell sweep if need be.

Marijn

- --
If you cannot read my mind, then listen to what I say.

Marijn Schouten (hkBst), Gentoo Lisp project, Gentoo ML
http://www.gentoo.org/proj/en/lisp/, #gentoo-{lisp,ml} on FreeNode
-BEGIN PGP SIGNATURE-
Version: GnuPG v2.0.11 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkrMedEACgkQp/VmCx0OL2zCEACeK0xQpwjf1UPGVWz4izqD/km6
6ZoAn2BQAcS8Ir0NKAkaH5ui1U/cN1V3
=rmtP
-END PGP SIGNATURE-