[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [cross-project-issues-dev] Neon.3 Update Problems: To Fix and How to Fix?

Hi Ed,

Linux tools spun a 5.3.1 release which now has a 2.3.1 version of docker tooling. The Linux tools download site has update-docker-2.3.1 and
update-docker, both which have 2.3.1 versions of the docker.core plug-in and docker feature. Not sure why you are not seeing this.

-- Jeff J.

On Wed, Apr 19, 2017 at 11:13 AM, Ed Merks <ed.merks@xxxxxxxxx> wrote:

Frederic,

There seem to have been no notes/minutes taken during the meeting:

 https://wiki.eclipse.org/Planning_Council/April_05_2017

I recall agreeing to provide steps for reproducing the problem so that Thomas Watson could test if the wiring resolution fix he made for Oxygen also solves the problem for Neon.3. The fact that he encountered "the mirroring problem" didn't help in that regard:

 https://bugs.eclipse.org/bugs/show_bug.cgi?id=515213

In the end, he sent me a note saying (and I quote):

I see that now there is the same number of httpcomponents bundles as there was in the messed up Oxygen M6 builds. But here my back port of the resolver fix does not seem to have fixed the issue. I'm unsure if that is because it gave up with the sheer number of bundles or if something else is going wrong. But at this point the backport of the resolver fix does not seem to be the solution to the problem.

I assumed (wrongly I guess) that Thomas would investigate a more general fix to address the wiring problem.Â

In the end, I also wasn't sure which version of the docker tools is proposed for contribution to Neon.3a. I tried to search for update sites containing it like this:

Nothing looks like a new version of 2.3. Goodness knows where one should find what's being proposed for contribution...

In any case, the proposed "solution" (A) really just changes the version of httpclient to be one that's not broken (missing packages), but it doesn't change the wiring problem in any fundamental way. There will still be the four versions that can all be installed simultaneously, so we really should expect the same wiring problem(s). In fact, I believe Oxygen M6 has effectively the same four httpcomponents.httpclient bundle as does Neon.3, so I'm a little suspicious whether the wiring problem is in fact really fixed even for Oxygen. We won't know until M7 and that's a month away. It doesn't give me warm fuzzy feelings.

So at this point it remains unclear the nature of the wiring problem(s). Is it a bug? Is it fixable? Does the knowledge, will, and capacity to fix it exist?ÂÂ

Without a fix to the wiring problem I think we can eliminate A as a solution, leaving B, C, and D (i.e., focus on problem avoidance approaches). But I think if the wiring problem is a bug, it will come back, and it will raise its ugly head again when users install various technologies from various sources. To my thinking, fixing the bug seems important.

Regards,
Ed


On 19.04.2017 12:49, Frederic Gurr wrote:
Hi Ed,

In the last planning-council meeting you offered to evaluate if the
fixed Linux Tools package works as expected and if there are still
wiring issues.

Can you give us an update on the current state?

Regards,

Fred

On 31.03.2017 11:14, Ed Merks wrote:
Hi,

The original thread is fractured into many threads so its kind of
impossible to follow each thread with a reply but I'll try at the bottom
of this note, i.e., below the ===========

But before doing that, I'd like to re-focus on the most important
questions: *We currently have a problem with Neon.3, will we fix it, and
if so how will we fix it?*

The discussion has quickly digressed (constructively) into solving the
issue of how Orbit dependencies should be managed by projects and by the
release train.  Unfortunately I see this as a world hunger issue; not
one that is easily addressed and I believe not one we can wait for in
order to solve the Neon.3 problem.  Let's face it, we've not been able
to produce a proper Oxygen milestone in months, we still don't have one
now, and we won't have one until next month, we hope. 

For Neon we've done three maintenance releases.  Neon.1 needed a respin
and Neon.3 looks to be in need of the same thing.  Clearly something is
seriously wrong.  But if we spend our time on solving the Orbit world
hunger issue, will we arrive at a solution in time for Oxygen, let alone
in time to fix Neon.3?  I am very, very doubtful.

As another data point, if I install the egg-laying-wool-milk-pig for
Neon.3.  The following happens.  I'm prompted to accept this license:

    Red Hat, Inc. licenses these features and plugins to you under
    certain open source licenses (or aggregations of such licenses),
    which in a particular case may include the Eclipse Public License,
    the GNU Lesser General Public License, and/or certain other open
    source licenses. For precise licensing details, consult the
    corresponding source code, or contact Red Hat, Attn: General
    Counsel, 100 East Davie St., Raleigh NC 27601 USA.

I'm not sure how this license slipped into the release train.   Aren't
there checks for this?  (Sorry to digress, but this is also unacceptable.)

Launching the final installation comes up like this:

Clearly a disgusting mess, but I've mentioned that before and the same
projects are still doing the same bad things, so we clearly all accept
this situation as normal.

The most important point here is the error log (first attachment) is
full of exactly the problem indications (bundle wiring problems) we
should have expected from the Neon.3 repository's contents, if someone
were to install an arbitrary combination of the repository's contents. 
It's really not so hard to test this!

If I create the same installation with my local build of the Oomph 1.8
installer---which installs my locally built version of Oomph 1.8 so the
Oomph setup plugins are no longer disabled because I made the
userstorage dependency optional and eliminated the strict <=4.4 upper
bound constraints on httpclient, which was such a bad idea I can almost
have a canary to think this done to solve a problem with no anticipation
of the problems it would cause---then I can visit all the preference
pages producing the second attached much larger log.  It seems clear
that proper testing really doesn't happen for far too many projects on
the train.  With distributed responsibility, no one is really responsible...

==================================

Orbit Issues

1) Respinning Linux Tools against Oxygen Mx seems to miss the point that
we should only distribute released versions of bundles,  so no Neon
build should redistribute any unreleased version of anything.  If a new
version of something is needed for security reasons or other reasons, it
should be released first.  And doing that in a maintenance train without
testing the overall impact is clearly something we should never do again
(without waving a bunch of red flags of warning).  And as Martin
Oberhuber asks, is nothing in place to check for this?  So suppose we do
respin with a fixed released version, like what we have for Oxygen M6,
then most likely we'd still have the problems we have in Oxygen M6 so
we'd need a fix to the resolver in Neon.  Better would seem to respin
with the old version(s) of the Orbit bundles, but somehow we can never
delete the broken version from Neon and because it has a higher version
number is likely to slip back in unexpected (though hopefully not, given
that features have pinned their bundle versions).

2) Don't include Orbit bundles in your project's features.  Sounds like
a great idea, but begs endless questions, and while solving a problem
might well introduce more new problems than it solves.  The first
question (as Carsten points out) is how do these things end up in a
repository, and if they are in a repository somehow, how are they
categorized?  It's hard to get them in and once you do, they're
categorized poorly.  The next question is, how do they end up in the
release train, if the projects that need them don't contribute them?  
Directly from Orbit you say?  But which ones should be pulled in from
Orbit and how is that discovered?   Are those the ones the projects have
tested against? Then there is the question of whether an installation is
deterministic if the bundle version isn't pinned?  It's not; it will
depend on what's in the repos that are available at resolve time.  But
Gunnar argues that even packages are not deterministic, which I think is
false: if the feature pins the bundle version and the package requires
the feature, then the pinned bundle is definitely in that package.  But
regardless, Gunnar's important point is that the runtime wiring seems
kind of non-determinstic, and while uses constraints might help, who the
heck understands those well, what tooling produces it correctly for us,
is that nicely integrated in PDE, and will it be properly maintained (in
contrast to lower bound constraints which you can pretty expect will
remain on whatever stale version they were initially set to).  This may
well be the right direction in which to go, but getting there isn't
going to be even half the fun...

Regards,
Ed





_______________________________________________
cross-project-issues-dev mailing list
cross-project-issues-dev@eclipse.org
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/cross-project-issues-dev

_______________________________________________
cross-project-issues-dev mailing list
cross-project-issues-dev@eclipse.org
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/cross-project-issues-dev


_______________________________________________
cross-project-issues-dev mailing list
cross-project-issues-dev@eclipse.org
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/cross-project-issues-dev

PNG image