[gentoo-dev] EAPI 2 is brokened :(

[gentoo-dev] EAPI 2 is brokened :(

[Ciaran McCreesh]

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

Unfortunately Portage and Pkgcore have broken EAPI 2 implementations.
So far as I can see:

Portage:

* doesn't implement the 'default' function correctly for src_prepare.

Pkgcore:

* doesn't implement src_prepare
* doesn't have a working default
* default_src_configure and default_src_prepare are the only
default_s that work

Unfortunately, neither appears to have comprehensive unit tests
covering all new EAPI 2 functionality, so these got through into
package managers that made it into the tree.

This is a bit of a mess. Possible solutions are:

* Ignore it, and tell anyone using broken package managers to upgrade,
and deal with the resulting mess.

* Create a new EAPI 3 which is identical to EAPI 2. Make really really
sure Portage and Pkgcore implement it correctly before they go into the
tree. Deprecate EAPI 2, and convert everything using it to EAPI 3.

* Avoid using any broken feature, and retroactively remove the
definitions from EAPI 2. Unfortunately, this probably isn't viable
because of the src_prepare thing in Pkgcore (were it merely an
unusable 'default' we could get away with this).

None of these are very nice. Better options would be encouraged... We
have to do *something*, though, because this is hitting users already
(see bug 240684 for one).

-- 
Ciaran McCreesh

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

Re: [gentoo-dev] EAPI 2 is brokened :(

[Doug Goldstein]

Ciaran McCreesh wrote:
> Unfortunately Portage and Pkgcore have broken EAPI 2 implementations.
>   
<snip>

Ciaran, I would think at this point you know this since you've seen this
brought up hundreds of times on this list. The mailing list is not an
appropriate place to file bug reports. The proper place would be in
http://bugs.gentoo.org for Portage and http://www.pkgcore.org for pkgcore.

Re: [gentoo-dev] EAPI 2 is brokened :(

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 730 bytes --]

On Thu, 09 Oct 2008 17:47:36 -0400
Doug Goldstein <cardoe@gentoo.org> wrote:
> Ciaran McCreesh wrote:
> > Unfortunately Portage and Pkgcore have broken EAPI 2
> > implementations. 
> <snip>
> 
> Ciaran, I would think at this point you know this since you've seen
> this brought up hundreds of times on this list. The mailing list is
> not an appropriate place to file bug reports. The proper place would
> be in http://bugs.gentoo.org for Portage and http://www.pkgcore.org
> for pkgcore.

Uh... The bugs have already been reported. This is supposed to be a
discussion about how to handle the fallout -- what do you feel you
gain by leaping to erroneous conclusions and making accusations?

-- 
Ciaran McCreesh

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

Re: [gentoo-dev] EAPI 2 is brokened :(

[Zac Medico]

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

Ciaran McCreesh wrote:
> Unfortunately Portage and Pkgcore have broken EAPI 2 implementations.
> So far as I can see:
> 
> Portage:
> 
> * doesn't implement the 'default' function correctly for src_prepare.

This is fixed and released in sys-apps/portage-2.2_rc12.

> None of these are very nice. Better options would be encouraged... We
> have to do *something*, though, because this is hitting users already
> (see bug 240684 for one).

Like I said in the council thread, I think you may be overreacting.
No version of sys-apps/portage that supports EAPI 2 has been marked
stable yet so the fact that some unstable versions were a little
buggy is not a big problem since unstable users tend to upgrade
relatively frequently and sys-apps/portage is promoted to the from
of the merge list anyway.
- --
Thanks,
Zac
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.9 (GNU/Linux)

iEYEARECAAYFAkjug2IACgkQ/ejvha5XGaOhFgCg6OhVpG5ZZjS00Z9GsU79JZUF
aD8Aniy1cTFm3oXrQQ128cQJCfgS2Az/
=eTvZ
-----END PGP SIGNATURE-----

Re: [gentoo-dev] EAPI 2 is brokened :(

[Brian Harring]

[-- Attachment #1: Type: text/plain, Size: 1974 bytes --]

On Thu, Oct 09, 2008 at 10:53:13PM +0100, Ciaran McCreesh wrote:
> On Thu, 09 Oct 2008 17:47:36 -0400
> Doug Goldstein <cardoe@gentoo.org> wrote:
> > Ciaran McCreesh wrote:
> > > Unfortunately Portage and Pkgcore have broken EAPI 2
> > > implementations. 
> > <snip>
> > 
> > Ciaran, I would think at this point you know this since you've seen
> > this brought up hundreds of times on this list. The mailing list is
> > not an appropriate place to file bug reports. The proper place would
> > be in http://bugs.gentoo.org for Portage and http://www.pkgcore.org
> > for pkgcore.
> 
> Uh... The bugs have already been reported. This is supposed to be a
> discussion about how to handle the fallout -- what do you feel you

Interestingly enough, I don't see a bug filed in either of those 
ticket systems.  I see an irc msg that's vague on details from you 
proceeding this email by an hour or so though.

Pkgcore EAPI2 support was released ~2 days ago; portage EAPI2 support 
was released 09/26, ~12 days ago.  So... not exactly a huge window.  
In addition, all *3* versions supporting EAPI2 are all unstable; 
meaning fairly rapid upgrade cycle by consumers of unstable, and 
not exactly likely to have versions lingering long term in peoples 
vdb.

So where exactly is this "sky is falling" issue you're worried about?  
Bugs happen.  I'd expect zac will fix whatever issue there is and 
knock out a release w/in next day or so.  I know pkgcore will be 
fixed/released w/in next 12 (post work primarily).

Frankly you're overreacting on this- and that is assuming you *are* 
overreacting instead of just going for a bit of a public smear 
via bugs.

Either way, my vote is fix the bugs, leave EAPI2 as is, and in the 
future kindly file bugs properly (preferably w/out the noise, but 
I'll take usable bug reports in almost any form).

I'd assume zac's opinion will be the same, although he obviously 
speaks for himself.
~brian

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

Re: [gentoo-dev] EAPI 2 is brokened :(

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 1309 bytes --]

On Thu, 9 Oct 2008 15:22:19 -0700
Brian Harring <ferringb@gmail.com> wrote:
> So where exactly is this "sky is falling" issue you're worried
> about? Bugs happen.

It means anyone using EAPI 2 now is going to encounter severe
breakages with Pkgcore. Simply put, all your Pkgcore users are going to
get screwed over very messily as soon as they try to use any EAPI 2
things. Is this not something we should be caring about?

> Frankly you're overreacting on this- and that is assuming you *are* 
> overreacting instead of just going for a bit of a public smear 
> via bugs.

Bah. If you want me to lecture you on how you're being blatantly
irresponsible and incompetent then I will do, although by the way you
rush on the defensive and start trying to cover your ass by throwing
accusations at me it looks like you already know it. But what I care
about is getting the mess fixed in the most painless way possible.

This is a real issue and developers need to know the implications.

> Either way, my vote is fix the bugs, leave EAPI2 as is, and in the 
> future kindly file bugs properly (preferably w/out the noise, but 
> I'll take usable bug reports in almost any form).

If you want bug reports via trac instead of IRC, get your trac to
respond faster.

-- 
Ciaran McCreesh

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

Re: [gentoo-dev] EAPI 2 is brokened :(

[Brian Harring]

[-- Attachment #1: Type: text/plain, Size: 3578 bytes --]

On Thu, Oct 09, 2008 at 11:34:59PM +0100, Ciaran McCreesh wrote:
> On Thu, 9 Oct 2008 15:22:19 -0700
> Brian Harring <ferringb@gmail.com> wrote:
> > So where exactly is this "sky is falling" issue you're worried
> > about? Bugs happen.
> 
> It means anyone using EAPI 2 now is going to encounter severe
> breakages with Pkgcore. Simply put, all your Pkgcore users are going to
> get screwed over very messily as soon as they try to use any EAPI 2
> things. Is this not something we should be caring about?

You're massively overreacting, either due to ignorance or due to 
trying to stir things up.

Tree currently has 308 ebuilds out of 25500 that are EAPI=2.  ~1.1% 
of the tree.

Of that 308, number of ebuilds that either inherit java-utils (which 
adds src_prepare), define their own src_prepare, or even *match* 
default via grepping in the ebuild is 20.

I reiterate.  20 out of 308 EAPI2 consumers, meaning that pkgcore 
misbehaved on .078% of the tree (all unstable ebuilds in addition, 
and haven't even counted how many are masked), and it did this for a 
window of a few days.

There is a difference between not caring about breakage, and gauging 
the affect of the breakage and dealing with it appropriately 
(triaging).

Regardless, it's worth noting that 0.4.7.11 was cut, and commited to 
the tree (thank you zac) already fixing the issues you so kindly 
pointed out.  So those 25 ebuilds pkgcore broke, are now addressed.  
Frankly I'd be amazed if anyone even spotted it yet (both portage and 
pkgcore) due to the minimal slice.


> > Frankly you're overreacting on this- and that is assuming you *are* 
> > overreacting instead of just going for a bit of a public smear 
> > via bugs.
> 
> Bah. If you want me to lecture you on how you're being blatantly
> irresponsible and incompetent then I will do, although by the way you
> rush on the defensive and start trying to cover your ass by throwing
> accusations at me it looks like you already know it. But what I care
> about is getting the mess fixed in the most painless way possible.

s/painless/painful to targets/.

And to be clear, just like the last time you reported bugs in pkgcore 
via ml, yes, there were bugs in EAPI2.

No ass covering.

Also, if you want to bitch at me and call me incompetent do it via 
your blog.  No one this ml cares to watch us trade blows, let 
alone deal with personal BS that bleeds into your posts.


> This is a real issue and developers need to know the implications.
> 
> > Either way, my vote is fix the bugs, leave EAPI2 as is, and in the 
> > future kindly file bugs properly (preferably w/out the noise, but 
> > I'll take usable bug reports in almost any form).
> 
> If you want bug reports via trac instead of IRC, get your trac to
> respond faster.

Actually trac is responsive.  Now if you work your way into the 
trac-bzr bits (source browsing), yeah, it's slower then I'd like 
(patches welcome of course).

Regardless, bugs.gentoo.org exists; if it's EAPI related, I'd expect 
folks wouldn't mind if a pkgcore bug or two wound up there (it's not 
like it's a feature request for pkgcore after all).

Frankly, the sky is not falling.  A max of 20 freaking ebuilds 
misbehaving for pkgcore (<20 for portage also) doesn't warrant 
mangling EAPI2 let alone going to the ml to stir up shit.

In the future, kindly just file bugs.  If upon investigation it is an 
issue, sure escalate it.  Essentially focus on getting it straightened 
out instead of going for drama.

~brian

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

Re: [gentoo-dev] EAPI 2 is brokened :(

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 724 bytes --]

On Thu, 9 Oct 2008 16:38:56 -0700
Brian Harring <ferringb@gmail.com> wrote:
> Of that 308, number of ebuilds that either inherit java-utils (which 
> adds src_prepare), define their own src_prepare, or even *match* 
> default via grepping in the ebuild is 20.

Of those, and those in overlays, and those that are going to be
committed over the next few weeks, how many use src_prepare to apply
security related patches?

> Also, if you want to bitch at me and call me incompetent do it via 
> your blog.  No one this ml cares to watch us trade blows, let 
> alone deal with personal BS that bleeds into your posts.

I'm not the one going around levelling personal attacks at everyone.

-- 
Ciaran McCreesh

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

Re: [gentoo-dev] EAPI 2 is brokened :(

[Mart Raudsepp]

[-- Attachment #1: Type: text/plain, Size: 1265 bytes --]

On Fri, 2008-10-10 at 00:55 +0100, Ciaran McCreesh wrote:
> On Thu, 9 Oct 2008 16:38:56 -0700
> Brian Harring <ferringb@gmail.com> wrote:
> > Of that 308, number of ebuilds that either inherit java-utils (which 
> > adds src_prepare), define their own src_prepare, or even *match* 
> > default via grepping in the ebuild is 20.
> 
> Of those, and those in overlays, and those that are going to be
> committed over the next few weeks, how many use src_prepare to apply
> security related patches?

A round zero. Security patches are going stable soon after entering
portage tree, and EAPI=2 ebuilds can not go stable yet, as there is no
package manager supporting EAPI=2 that is going to be stable in the next
week or two (so maintainers make sure they don't use EAPI=2 for security
fix revisions).
And if the bug is there and properly filed to the appropriate bugzilla,
they wouldn't go stable before that bug is fixed (which I read are
already fixed).
I can not understand why this is dragged on. It was a bug, it is fixed.
The sky is not falling and EAPI-2 is not broken - there was a bug in the
implementation that is fixed.


-- 
Mart Raudsepp
Gentoo Developer
Mail: leio@gentoo.org
Weblog: http://planet.gentoo.org/developers/leio

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [gentoo-dev] EAPI 2 is brokened :(

[Alec Warner]

On Thu, Oct 9, 2008 at 3:34 PM, Ciaran McCreesh
<ciaran.mccreesh@googlemail.com> wrote:
> On Thu, 9 Oct 2008 15:22:19 -0700
> Brian Harring <ferringb@gmail.com> wrote:
>> So where exactly is this "sky is falling" issue you're worried
>> about? Bugs happen.
>
> It means anyone using EAPI 2 now is going to encounter severe
> breakages with Pkgcore. Simply put, all your Pkgcore users are going to
> get screwed over very messily as soon as they try to use any EAPI 2
> things. Is this not something we should be caring about?

I think everyone appreciates the forewarning (even if not everyone
appreciates the manner in which it was delivered).  I think we do care
and we are fixing it.  I believe the developers of said packages have
a different idea of the risks involved than you and I don't expect
everyone to agree on specific software development or release
processes.

>
>> Frankly you're overreacting on this- and that is assuming you *are*
>> overreacting instead of just going for a bit of a public smear
>> via bugs.
>
> Bah. If you want me to lecture you on how you're being blatantly
> irresponsible and incompetent then I will do, although by the way you
> rush on the defensive and start trying to cover your ass by throwing
> accusations at me it looks like you already know it. But what I care
> about is getting the mess fixed in the most painless way possible.
>
> This is a real issue and developers need to know the implications.
>

If you want to call people names do it on your own lists.

>> Either way, my vote is fix the bugs, leave EAPI2 as is, and in the
>> future kindly file bugs properly (preferably w/out the noise, but
>> I'll take usable bug reports in almost any form).
>
> If you want bug reports via trac instead of IRC, get your trac to
> respond faster.
>
> --
> Ciaran McCreesh
>

Re: [gentoo-dev] EAPI 2 is brokened :(

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 1168 bytes --]

On Fri, 10 Oct 2008 09:32:44 +0300
Mart Raudsepp <leio@gentoo.org> wrote:
> > Of those, and those in overlays, and those that are going to be
> > committed over the next few weeks, how many use src_prepare to apply
> > security related patches?
> 
> A round zero. Security patches are going stable soon after entering
> portage tree, and EAPI=2 ebuilds can not go stable yet, as there is no
> package manager supporting EAPI=2 that is going to be stable in the
> next week or two (so maintainers make sure they don't use EAPI=2 for
> security fix revisions).

Oh really? So you're absolutely certain there aren't and won't soon be
any EAPI 2 bumps of non-EAPI 2 versions that include security patches?
And you're absolutely certain that there aren't, say, any packages that
sed a broken chmod in a makefile in src_prepare?

> I can not understand why this is dragged on. It was a bug, it is
> fixed. The sky is not falling and EAPI-2 is not broken - there was a
> bug in the implementation that is fixed.

The point of EAPI is to avoid these kinds of problems. The process is
failing and the fallout needs to be handled.

-- 
Ciaran McCreesh

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

Re: [gentoo-dev] EAPI 2 is brokened :(

[Robert Buchholz]

[-- Attachment #1: Type: text/plain, Size: 1029 bytes --]

On Friday 10 October 2008, Ciaran McCreesh wrote:
> On Fri, 10 Oct 2008 09:32:44 +0300
>
> Mart Raudsepp <leio@gentoo.org> wrote:
> > > Of those, and those in overlays, and those that are going to be
> > > committed over the next few weeks, how many use src_prepare to
> > > apply security related patches?
> >
> > A round zero. Security patches are going stable soon after entering
> > portage tree, and EAPI=2 ebuilds can not go stable yet, as there is
> > no package manager supporting EAPI=2 that is going to be stable in
> > the next week or two (so maintainers make sure they don't use
> > EAPI=2 for security fix revisions).
>
> Oh really? So you're absolutely certain there aren't and won't soon
> be any EAPI 2 bumps of non-EAPI 2 versions that include security
> patches?

This would not affect stable users, as Mart pointed out. ~arch users 
should keep their systems up-to-date anyway (including their PM), and 
should be aware their security support is limited to providing updates.


Robert

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 835 bytes --]