RFC: reverse the name change for org.oe.stable

RFC: reverse the name change for org.oe.stable

[Rolf Leggewie]

Hi,

I am concerned about the loss of reputation that OE is incurring because
the .stable guys are willfully ignoring the official bug tracker.  I
think it should be renamed back to angstrom$whateveritwasbefore.  We
can't seriously be telling guys all the time "report your problems to
the BTS" and then have the "official" stable branch completely and
willfully ignore it.  I'd go as far as requesting it should be denied
the org.openembedded.* namespace in monotone.

http://thread.gmane.org/gmane.comp.handhelds.openembedded/18400/focus=18462

Regards

Rolf

Re: RFC: reverse the name change for org.oe.stable

[Leon Woestenberg]

Hello Rolf,

On Sat, May 10, 2008 at 3:53 PM, Rolf Leggewie
<no2spam@nospam.arcornews.de> wrote:

> I am concerned about the loss of reputation that OE is incurring because
> the .stable guys are willfully ignoring the official bug tracker.  I
> think it should be renamed back to angstrom$whateveritwasbefore.  We
> can't seriously be telling guys all the time "report your problems to
> the BTS" and then have the "official" stable branch completely and
> willfully ignore it.  I'd go as far as requesting it should be denied
> the org.openembedded.* namespace in monotone.
>



-- 
Leon

Re: RFC: reverse the name change for org.oe.stable

[Leon Woestenberg]

Hello,

On Sat, May 10, 2008 at 3:53 PM, Rolf Leggewie
<no2spam@nospam.arcornews.de> wrote:

> I am concerned about the loss of reputation that OE is incurring because
> the .stable guys are willfully ignoring the official bug tracker.  I

I'm not ignoring the bug tracker, I just don't have time to work on
stable currently.

> think it should be renamed back to angstrom$whateveritwasbefore.  We
> can't seriously be telling guys all the time "report your problems to
> the BTS" and then have the "official" stable branch completely and
> willfully ignore it.  I'd go as far as requesting it should be denied
> the org.openembedded.* namespace in monotone.
>
I understand your problem. Some thoughts:

The only way to have the OE community to maintain .stable, is to have
the OE community use it. Currently, we are in a chicken and egg
situation there.

Personally, I do not have the time to maintain .stable, but I will
probably use the next stable branch and maintain it.

Telling people to report problems to the bugtracker is sane, whether
we have a stable branch or not.

Maintaining recipes that are _cvs or _svn are not feature freeze.
Stable does not match with the SCM versions and should not be chosen
in any stable distro.

Regards,
-- 
Leon

Re: RFC: reverse the name change for org.oe.stable

[Koen Kooi]

Leon Woestenberg wrote:
> Hello,
>
> On Sat, May 10, 2008 at 3:53 PM, Rolf Leggewie
> <no2spam@nospam.arcornews.de>  wrote:
>
>> I am concerned about the loss of reputation that OE is incurring because
>> the .stable guys are willfully ignoring the official bug tracker.  I
>
> I'm not ignoring the bug tracker, I just don't have time to work on
> stable currently.
>
>> think it should be renamed back to angstrom$whateveritwasbefore.  We
>> can't seriously be telling guys all the time "report your problems to
>> the BTS" and then have the "official" stable branch completely and
>> willfully ignore it.  I'd go as far as requesting it should be denied
>> the org.openembedded.* namespace in monotone.
>>
> I understand your problem. Some thoughts:
>
> The only way to have the OE community to maintain .stable, is to have
> the OE community use it. Currently, we are in a chicken and egg
> situation there.

Not quite, we already have a number of vendors releasing products based 
on the stable branch. They just prefer be anonymous and push patches 
through developers instead of mail/bugzilla/whatever. Since "loss of 
reputation" was mentioned: stabbing vendors in the eye is one way of 
accomplishing that.

> Personally, I do not have the time to maintain .stable, but I will
> probably use the next stable branch and maintain it.

I have enough time to review patches people send to the stable mailing 
list. The time has decreased a bit since I started using a faster 
computer for the simulations I need to run during daytime :)

> Telling people to report problems to the bugtracker is sane, whether
> we have a stable branch or not.

There was never any question where to report problems, the bugtracker is 
the canonical place for that. The 'question' was how to get the 
resulting patches applied to the stable branch. The answer is simple and 
documented:

Send them to the stablebranch mailinglist for review.

It's not rocket science.

> Maintaining recipes that are _cvs or _svn are not feature freeze.
> Stable does not match with the SCM versions and should not be chosen
> in any stable distro.

I have trouble parsing that statement. Are you saying "do not use _<scm> 
recipes in a stable distro" or something else?

Re: RFC: reverse the name change for org.oe.stable

[Leon Woestenberg]

Hello Koen,

On Sat, May 10, 2008 at 5:44 PM, Koen Kooi <k.kooi@student.utwente.nl> wrote:
> Leon Woestenberg wrote:
> > Maintaining recipes that are _cvs or _svn are not feature freeze.
> > Stable does not match with the SCM versions and should not be chosen
> > in any stable distro.
>
>  I have trouble parsing that statement. Are you saying "do not use _<scm>
> recipes in a stable distro" or something else?
>

Yes, do not use _<scm> recipes in stable stuff, a known checkout of
the OE stable branch should either build or not for a particular
machine/distro/... combo.

If I checkout today's stable tree in one year time, it should build. Period.

That isn't going to work with _<scm> recipes being active in stable.

(and sorry for the badly formed statement - I don't take enough time
to write down what I would like to say - maybe I should just shut up).

Leon.





-- 
Leon

Re: RFC: reverse the name change for org.oe.stable

[Khem Raj]

But we also have sane* files where these versions are defined. So these
revisions will be frozen as well. I dont see a problem in using a given
revision if it is tested and is stable enough.
-Khem

On Sun, May 11, 2008 at 3:46 PM, Leon Woestenberg <
leon.woestenberg@gmail.com> wrote:

> Hello Koen,
>
> On Sat, May 10, 2008 at 5:44 PM, Koen Kooi <k.kooi@student.utwente.nl>
> wrote:
> > Leon Woestenberg wrote:
> > > Maintaining recipes that are _cvs or _svn are not feature freeze.
> > > Stable does not match with the SCM versions and should not be chosen
> > > in any stable distro.
> >
> >  I have trouble parsing that statement. Are you saying "do not use _<scm>
> > recipes in a stable distro" or something else?
> >
>
> Yes, do not use _<scm> recipes in stable stuff, a known checkout of
> the OE stable branch should either build or not for a particular
> machine/distro/... combo.
>
> If I checkout today's stable tree in one year time, it should build.
> Period.
>
> That isn't going to work with _<scm> recipes being active in stable.
>
> (and sorry for the badly formed statement - I don't take enough time
> to write down what I would like to say - maybe I should just shut up).
>
> Leon.
>
>
>
>
>
> --
> Leon
>
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
>

Re: RFC: reverse the name change for org.oe.stable

[Richard Purdie]

On Mon, 2008-05-12 at 00:46 +0200, Leon Woestenberg wrote:
> Yes, do not use _<scm> recipes in stable stuff, a known checkout of
> the OE stable branch should either build or not for a particular
> machine/distro/... combo.
> 
> If I checkout today's stable tree in one year time, it should build. Period.
> 
> That isn't going to work with _<scm> recipes being active in stable.

Floating SCM versions won't be reproducible, fixed SCM versions will. If
SCM versions are used they need to be locked to a version.

Cheers,

Richard

Re: RFC: reverse the name change for org.oe.stable

[Leon Woestenberg]

Hello,

On Mon, May 12, 2008 at 9:19 AM, Richard Purdie <rpurdie@rpsys.net> wrote:
> On Mon, 2008-05-12 at 00:46 +0200, Leon Woestenberg wrote:
>  > Yes, do not use _<scm> recipes in stable stuff, a known checkout of
>  > the OE stable branch should either build or not for a particular
>  > machine/distro/... combo.
>  >
>  > If I checkout today's stable tree in one year time, it should build. Period.
>  >
>  > That isn't going to work with _<scm> recipes being active in stable.
>
>  Floating SCM versions won't be reproducible, fixed SCM versions will. If
>  SCM versions are used they need to be locked to a version.
>

Agreed, so I would like to propose the stable branch should not need
to support floating SCM versions.

Only full lock-down (and tarball mirrors) will make our stable trees
useful in the long term.

It's an important property of a stable tree, that third parties using
them can rely on them to be able to perform a bug fix in say, one year
time.

That reminds me, we should lock-down (and include) the corresponding
bitbake version in the stable tree as well.

Regards,
-- 
Leon

Re: RFC: reverse the name change for org.oe.stable

[Koen Kooi]

Leon Woestenberg wrote:

> That reminds me, we should lock-down (and include) the corresponding
> bitbake version in the stable tree as well.

I proposed that a while ago, but it got NAKed, time to propose it again? 
I see no reason why we shouldn't have a known working bitbake copy 
inside the stable tree

Re: RFC: reverse the name change for org.oe.stable

[Rod Whitby]

Koen Kooi wrote:
> Leon Woestenberg wrote:
> 
>> That reminds me, we should lock-down (and include) the corresponding
>> bitbake version in the stable tree as well.
> 
> I proposed that a while ago, but it got NAKed, time to propose it again? 
> I see no reason why we shouldn't have a known working bitbake copy 
> inside the stable tree

+1 from me - when nslu2-linux creates a binary release source tarball we 
always include bitbake in the tarball.

-- Rod

Re: RFC: reverse the name change for org.oe.stable

[Marcin Juszkiewicz]

Dnia Monday, 12 of May 2008, Leon Woestenberg napisał:

> Agreed, so I would like to propose the stable branch should not need
> to support floating SCM versions.
>
> Only full lock-down (and tarball mirrors) will make our stable trees
> useful in the long term.

What define which recipe in stable is supported? For example - 
does "asleap 2.1" supported? It is not listed in conf/checksums.ini so I 
suppose that it is not maintained by anyone and that no one tried to 
build it.

Keeping sources mirrored is one thing - but what use if of mirroring if 
there is no way to check does files are proper ones? Who will tell does 
asleap-2.1.tgz has md5 starting with "1234" or with "f38b" if it is not 
fetchable from other place then 3 random mirrors?

I think that for stable branch someone whould at least fetch all sources 
to generate checksums so it will be known that asleap-2.1.tgz md5sum 
really starts with "1234".

-- 
JID: hrw-jabber.org
OpenEmbedded developer/consultant

   Programming consists of 50% planning, 50% coding and 30% mathematics

Re: RFC: reverse the name change for org.oe.stable

[Holger Freyther]

On Monday 12 May 2008 09:19:21 Richard Purdie wrote:
> On Mon, 2008-05-12 at 00:46 +0200, Leon Woestenberg wrote:
> > Yes, do not use _<scm> recipes in stable stuff, a known checkout of
> > the OE stable branch should either build or not for a particular
> > machine/distro/... combo.
> >
> > If I checkout today's stable tree in one year time, it should build.
> > Period.
> >
> > That isn't going to work with _<scm> recipes being active in stable.
>
> Floating SCM versions won't be reproducible, fixed SCM versions will. If
> SCM versions are used they need to be locked to a version.

The reality is that SRCDATE is completely meaningless for at least CVS and 
SVN. Even worse with CVS and SVN the history can be altered and even if you 
checkout e.g. the glibc repo by date you get a different snapshot that you 
got last time you tried.

The sad reality is that only SCMs with a hash as revision number can be 
trusted, for the others we would have to make sure that the mirrors carry the 
checkout as we used it to build the package and to disable checking stuff out 
by date/revnumber in bitbake for a real stable branch.


just my two cents, I might be off topic.
	z.

Re: RFC: reverse the name change for org.oe.stable

[Sergey 'Jin' Bostandzhyan]

Hi,

On Sat, May 10, 2008 at 03:53:52PM +0200, Rolf Leggewie wrote:
> I am concerned about the loss of reputation that OE is incurring because
> the .stable guys are willfully ignoring the official bug tracker.  I
> think it should be renamed back to angstrom$whateveritwasbefore.  We
> can't seriously be telling guys all the time "report your problems to
> the BTS" and then have the "official" stable branch completely and
> willfully ignore it.  I'd go as far as requesting it should be denied
> the org.openembedded.* namespace in monotone.
> 
> http://thread.gmane.org/gmane.comp.handhelds.openembedded/18400/focus=18462
>

I've been following the threads on oe.stable and this leads to some questions:

what is the "official" definition of the oe.stable branch?

at first I had the impression that it is intended for product development
(hence - stable), is this the case or was this a misconception?

is there any point in submitting bugs to BTS vs. oe.stable (i.e. can at least
some help regarding those bugs be expected)?

is the stable branch actively maintained (well, the answer to this probably
results from the anwer to the previous question)?


Please don't get me wrong, it is not my intention to get in the middle of the
discussion about reversing names or whatever. I use OE for product development
and I ultimately need to know what branch I should base my work on.

It would also be nice if the status of the branches gets clearly communicated
once this discussion is over. I think it's important that people who do 
product dev know where to start and what to expect.

Again, just to make it clear: my argument is not about usage of 
BTS/maintenance/naming/etc., it's about a clear communication on the status
of things, so that people know what branch to choose.

Kind regards,
Jin

Re: RFC: reverse the name change for org.oe.stable

[Koen Kooi]

Sergey 'Jin' Bostandzhyan wrote:
> Hi,
>
> On Sat, May 10, 2008 at 03:53:52PM +0200, Rolf Leggewie wrote:
>> I am concerned about the loss of reputation that OE is incurring because
>> the .stable guys are willfully ignoring the official bug tracker.  I
>> think it should be renamed back to angstrom$whateveritwasbefore.  We
>> can't seriously be telling guys all the time "report your problems to
>> the BTS" and then have the "official" stable branch completely and
>> willfully ignore it.  I'd go as far as requesting it should be denied
>> the org.openembedded.* namespace in monotone.
>>
>> http://thread.gmane.org/gmane.comp.handhelds.openembedded/18400/focus=18462
>>
>
> I've been following the threads on oe.stable and this leads to some questions:
>
> what is the "official" definition of the oe.stable branch?
>
> at first I had the impression that it is intended for product development
> (hence - stable), is this the case or was this a misconception?
>
> is there any point in submitting bugs to BTS vs. oe.stable (i.e. can at least
> some help regarding those bugs be expected)?

Yes, as has been said, the canonical place to submit bugs is the BTS. 
Just don't expect that patches will get applied without review on the 
mailinglist. We take "stable" seriously. Applying random shit just to 
keep bugs from being closed by an overzealous bugmaster is not taking 
"stable" seriously[2].

> is the stable branch actively maintained (well, the answer to this probably
> results from the anwer to the previous question)?

I do a build from scratch every week from the stable branch to see if it 
still builds and functions on the targets.  For the distro (angstrom 
2007) and targets (various arm and ppc boards) I'm pretty satisfied with 
the current status. Combine that with limited time to work on OE stuff, 
I have no real incentive to actively go looking for bugs (or fixes) 
outside my "area of testing". I suspect that's true for the other 
stablebranch maintainers as well.

> Please don't get me wrong, it is not my intention to get in the middle of the
> discussion about reversing names or whatever. I use OE for product development
> and I ultimately need to know what branch I should base my work on.

"stable" is a branch that has know issues and features an coupled with 
the review policy we try to avoid introducing regressions. So if your 
product is (nearly) ready for the market, it makes sense to use 
"stable". If you want things as packaged-staging or RPs toolchain 
improvements, you want .dev, but that may break anytime (it is broken 
for me atm. with the recent .lai uniqueness check. I like the check, 
just need to fix breakages associated with it (stupid EFL stuff)), and 
moves very quickly ("API and ABI unstable" in code speak).

> It would also be nice if the status of the branches gets clearly communicated
> once this discussion is over. I think it's important that people who do
>product dev know where to start and what to expect.

The status of the stable branch is still the same, so the top news item 
on http://www.openembedded.org is still valid and accurate. The 
mailinglist archive is also open to the public and have a real good 
signal/noise ratio.

>  Again, just to make it clear: my argument is not about usage of
> BTS/maintenance/naming/etc., it's about a clear communication on the status
> of things, so that people know what branch to choose.

I think more people should read the news on http://www.openembedded.org 
and spread less FUD, but that's just the cynic in me acting up. Also 
typing "oe stable branch" or "oe stablebranch" into google returns good 
info. People, it's not rocket science[1].


[1] Even if it is, it's just blowing stuff up and throwing it out the back.
[2] Nor making bugzilla more usefull. Closing bugs that haven't been 
fixed just because your lollypop got stolen actually decreases the value 
of bugzilla.

Re: RFC: reverse the name change for org.oe.stable

[Philip Balister]

[-- Attachment #1: Type: text/plain, Size: 922 bytes --]

Rolf Leggewie wrote:
> Hi,
> 
> I am concerned about the loss of reputation that OE is incurring because
> the .stable guys are willfully ignoring the official bug tracker.  I
> think it should be renamed back to angstrom$whateveritwasbefore.  We
> can't seriously be telling guys all the time "report your problems to
> the BTS" and then have the "official" stable branch completely and
> willfully ignore it.  I'd go as far as requesting it should be denied
> the org.openembedded.* namespace in monotone.


Rolf.

Please stop suggesting backward looking actions. It is important that OE 
provide a stable branch for people to use, regardless of what 
distribution people use. We know that we are not a perfect organization 
and the we do need to improve processes. You are making contributions in 
this area, however give people time to work through your ideas and see 
how they work out.

Philip

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/x-pkcs7-signature, Size: 3303 bytes --]

Re: RFC: reverse the name change for org.oe.stable

[Richard Purdie]

On Sun, 2008-05-11 at 03:43 -0400, Philip Balister wrote:
> Rolf Leggewie wrote:
> > I am concerned about the loss of reputation that OE is incurring because
> > the .stable guys are willfully ignoring the official bug tracker.  I
> > think it should be renamed back to angstrom$whateveritwasbefore.  We
> > can't seriously be telling guys all the time "report your problems to
> > the BTS" and then have the "official" stable branch completely and
> > willfully ignore it.  I'd go as far as requesting it should be denied
> > the org.openembedded.* namespace in monotone.
>
> Please stop suggesting backward looking actions. It is important that OE 
> provide a stable branch for people to use, regardless of what 
> distribution people use. We know that we are not a perfect organization 
> and the we do need to improve processes. You are making contributions in 
> this area, however give people time to work through your ideas and see 
> how they work out.

Well said.

We do need a stable branch. The rename made sense and improved several
issues. We're not naive and are going to pretend that everything is
perfect, its not. It is better than what we had previously which was no
stable branch. Some people are trying to make stable work and I'd like
us to find ways to help them rather than hinder them.

The stable branch people have commented on the use of the bugzilla and I
can't really see a problem there. If there are issues, lets keep them
technical rather than emotive and discuss the problems before proposing
inflammatory solutions.

Cheers,

Richard

Re: RFC: reverse the name change for org.oe.stable

[Rolf Leggewie]

Richard Purdie wrote:
> We do need a stable branch.

I guess so.  And I don't question that.  I just raise the flag that if
nobody is indeed interested in actually fixing bugs, we should
reconsider having it.  Right now, we have a snapshot with virtually no
activity, not a stable, actively maintained branch.  IMNSHO, OE should
be honest or it will loose credibility.  If we can't support a stable
branch (because supposedly people are too busy), we should not be
suggesting we do.  The openembedded-stable ml and the bug tracker,
together with the many people here who say "We should have a stable
branch, *but* right now I can't work on it" tell a clear story.

> The stable branch people have commented on the use of the bugzilla and I
> can't really see a problem there. 

They say they want mail, fine.

The problem is that they both block mail from bugzilla to the oe-stbale
list (and I am not going to play human gateway) and ignoring it in the
case of koen (who I added specifically as cc for bp flags raised).  I
don't think I should even be doing this.  If people are supporting
stable as the claim stands, there should be at least some activity in
the bug tracker wrt fixing bugs.

* several weeks if not months of no bug triaging from .stable
  causes concern
* no activity after some bug triaging is done for them
  causes real concern
* still no activity after blocking their meta bugs about a month ago
  causes serious concern
* still no activity after sending several reminder mails to the ml
  causes grave concern
* a "do as you please"-attitude after issuing a last warning and still
  no activity after providing them the patches the stable guys should
  have derived themselves if claims of support are to be taken serious

  this falls back on OE

BTW, still no bug triaging and not even the simple activity of looking
through the bugs that have been meticulously prepared.

> technical rather than emotive and discuss the problems before proposing
> inflammatory solutions.

I am absolutely technical.  The choice is simple

a) support it, including at least some work in the bug tracker
b) rename it
c) risk loss of reputation for OE as not living up to claims made

It is one of the three and really that easy.  Unfortunately, it looks
like everybody choosing c) because "we want stable".