From: John Arthorne
[mailto:John_Arthorne@xxxxxxxxxx]
Sent: May-28-10 3:06 PM
To: mike.milinkovich@xxxxxxxxxxx; eclipse-pmc@xxxxxxxxxxx
Subject: Re: Build Eclipse at Eclipse [WAS RE: [eclipse-pmc] Fw:
[platform-releng-dev] [eclipse-build]Build I20100526-1423 (Timestamp:
201005261423): Build failed]
Yes, we're well
along that path already. Under our "build anywhere" plan item in
Helios, Kim made the changes in our build system to enable running our build on
any machine. We now run occasional 3.x stream test builds on build.eclipse.org,
and all 4.x stream builds run on there. So far, every e4/Eclipse 4.0 build has
been run on build.eclipse.org, and hopefully we'll keep it that way. I think
the two main issues right now are:
1) Lack of test
machine infrastructure. I know Denis and Kim are working to set this up so we
should be in good shape. We might still need to run our performance tests on
IBM test machines because virtualized machines on eclipse.org won't be
consistent enough. Test machine coverage for Mac is also missing. Still, this doesn't
prevent us running the rest of the build on eclipse.org, and do performance and
Mac testing as a separate step afterwards.
2) Build
infrastructure stability. The eclipse.org Hudson install in particular has been
a sore spot for many projects. It seems Hudson is upgraded or extended at
arbitrary times which can break project builds. The current install is
completely unstable and requires regular restarts. Once this infrastructure is
on our critical path we will want to see more careful control here. For
example, sandbox any changes to build software on a separate machine (Hudson,
JVMs, etc), and make sure it is stable and working for all projects before
moving it onto the real build environment. I've seen Denis make comments to
this effect in bugs, so I think this is a known issue and hopefully we'll see
improvements in this area after Helios
John
"Mike
Milinkovich" <mike.milinkovich@xxxxxxxxxxx>
Sent by:
eclipse-pmc-bounces@xxxxxxxxxxx
05/28/2010
12:27 PM
Please respond to
mike.milinkovich@xxxxxxxxxxx; Please respond to
eclipse-pmc@xxxxxxxxxxx
|
|
To
|
<eclipse-pmc@xxxxxxxxxxx>
|
cc
|
emo@xxxxxxxxxxx,
webmaster@xxxxxxxxxxx
|
Subject
|
Build
Eclipse at Eclipse [WAS RE: [eclipse-pmc] Fw: [platform-releng-dev]
[eclipse-build]Build I20100526-1423 (Timestamp:
201005261423): Build failed]
|
|
Thanks to the recent and generous donations
of hardware to the Eclipse
Foundation, we feel that we now have sufficient machine capacity to discuss
moving the Eclipse platform builds onto our infrastructure. We would like
to
suggest that we work together on a feasibility analysis and a plan to make
that happen in time for Indigo.
Obviously this would be a big step, but if the PMC is willing to discuss it
the Eclipse Foundation staff will do everything we can on our side to make
this a big win for the community. Imagine how much faster the builds might
run if we're not shuttling those bits back and forth between CVS and IBM
:-)
Mike Milinkovich
Office: +1.613.224.9461 x228
Mobile: +1.613.220.3223
mike.milinkovich@xxxxxxxxxxx
> -----Original Message-----
> From: eclipse-pmc-bounces@xxxxxxxxxxx [mailto:eclipse-pmc-
> bounces@xxxxxxxxxxx] On Behalf Of Jeff McAffer
> Sent: May-27-10 9:41 PM
> To: eclipse-pmc@xxxxxxxxxxx
> Subject: Re: [eclipse-pmc] Fw: [platform-releng-dev]
[eclipse-build]Build
> I20100526-1423 (Timestamp: 201005261423): Build failed
>
> In general our builds need to be either much finer grained or more
> monolithic. Since the latter is not really possible, we need
more
> flexibility to build smaller increments. As was pointed out, as we use
more
> and more things from other projects there is a circularity. Trying to
force
> this in to a sequential stack will be challenging to say the least. We
see
> this all the time in realworld projects that are not as neatly
structured
> as Eclipse.
>
> For Indigo we need to do something about this. Perhaps one of the
many
> parties interested in build systems for OSGi would be willing to solve
the
> problem? For now we just have to be careful.
>
> Jeff
>
> On 2010-05-27, at 8:56 AM, Daniel Megert wrote:
>
> >> So rather than coming up with rules to prevent ppl to
consume other
> > bundles for other components, let's work on making the process of
> consuming
> > external jars smoother.
> > Fine with me if we can do this but otherwise I think it makes
sense to
> > adjust the train to reality i.e. some projects are before us and
we are
> > then +0.5 or +1.0. As you said it not just caused breakage but
also
> stress
> > to actually bring in that change so short before the end.
> >
> > Dani
> >
> > |------------>
> > | From: |
> > |------------>
> >>
------------------------------------------------------------------------
>
--------------------------------------------------------------------------|
> > |Pascal Rapicault <pascal@xxxxxxxxxxxx>
> |
> >>
------------------------------------------------------------------------
>
--------------------------------------------------------------------------|
> > |------------>
> > | To: |
> > |------------>
> >>
------------------------------------------------------------------------
>
--------------------------------------------------------------------------|
> > |eclipse-pmc@xxxxxxxxxxx
> |
> >>
------------------------------------------------------------------------
>
--------------------------------------------------------------------------|
> > |------------>
> > | Date: |
> > |------------>
> >>
------------------------------------------------------------------------
>
--------------------------------------------------------------------------|
> > |27.05.2010 14:35
> |
> >>
------------------------------------------------------------------------
>
--------------------------------------------------------------------------|
> > |------------>
> > | Subject: |
> > |------------>
> >>
------------------------------------------------------------------------
>
--------------------------------------------------------------------------|
> > |Re: [eclipse-pmc] Fw: [platform-releng-dev]
[eclipse-build]Build
> I20100526-1423 (Timestamp: 201005261423): Build failed
> |
> >>
------------------------------------------------------------------------
>
--------------------------------------------------------------------------|
> >
> >
> >
> >
> >
> > It is definitely not an issue with ECF. ECF fixed a bug (
> > https://bugs.eclipse.org/bugs/show_bug.cgi?id=313519),
recommended p2 to
> > consume a new version, I did it because it was useful to the our
end
> users.
> >
> > In this case the main issue is the build system which makes it
complex
to
> > consume other jars, which is made worse when JavaDoc generation
is
> involved
> > since most of the classpath information of the plug-ins as to be
> replicated
> > in some obscure file contained in platform.doc.isv. Now, for the
record,
> > the last failure after I reverted the ECF changes were caused by
a
> problem
> > at the foundation).
> >
> >> Eclipse PMC must forbid this for the future: either they can
> >> provide their latest stuff earlier (lets say one week before
we have a
> >> deadline) or we go with what we have.
> > ECF has always (at least through 3.6) been providing
their stuffs
> > early. However when it comes to the end-game the iterations are
so short
> > that we don't really have much time. Remember that in Helios the
complete
> > train is shipped in ONE week.
> > The reality here is that Eclipse SDK is not really +0 but +0.5,
and
> > overtime, I can see it become a +1 (for example how would the SDK
consume
> a
> > last minute change in EMF for Eclipse 4.0, or how in the future
will the
> > SDK consume the equinox built that will be run separately).
> > So rather than coming up with rules to prevent ppl to
consume other
> > bundles for other components, let's work on making the process of
> consuming
> > external jars smoother.
> >
> >
> > On 2010-05-27, at 2:40 AM, Daniel Megert wrote:
> >
> >>
> >> Oh boy! We are in RC3 and all three recent builds failed and
I expect
> >> I20100527-0800 (our projected RC3 build !!!) to fail as well.
This is
> > just
> >> not acceptable and it is not the first time that switching to
the
newest
> >> ECF code at a very late time burns us. I'm not saying it's a
failure in
> > ECF
> >> alone: it also seems that just the process of switching is
error prone.
> I
> >> think we the Eclipse PMC must forbid this for the future:
either they
> can
> >> provide their latest stuff earlier (lets say one week before
we have a
> >> deadline) or we go with what we have.
> >>
> >> BTW: I could not find a bug with the appropriate approvals in
Eclipse
or
> > RT
> >> for this change - Jeff, did I miss something?
> >>
> >> Dani
> >> ----- Forwarded by Daniel Megert/Zurich/IBM on 27.05.2010
08:18 -----
> >> |------------>
> >> | From: |
> >> |------------>
> >>>
> >
-------------------------------------------------------------------------
>
-------------------------------------------------------------------------|
> >
> >> |Kim Moir <Kim_Moir@xxxxxxxxxx>
> > |
> >>>
> >
-------------------------------------------------------------------------
> -------------------------------------------------------------------------|
> >
> >> |------------>
> >> | To: |
> >> |------------>
> >>>
> >
-------------------------------------------------------------------------
> -------------------------------------------------------------------------|
> >
> >> |"Eclipse platform release engineering list."
> > <platform-releng-dev@xxxxxxxxxxx>
> > |
> >>>
> >
-------------------------------------------------------------------------
> -------------------------------------------------------------------------|
> >
> >> |------------>
> >> | Date: |
> >> |------------>
> >>>
> >
-------------------------------------------------------------------------
>
-------------------------------------------------------------------------|
> >
> >> |26.05.2010 22:21
> > |
> >>>
> >
-------------------------------------------------------------------------
>
-------------------------------------------------------------------------|
> >
> >> |------------>
> >> | Subject: |
> >> |------------>
> >>>
> >
-------------------------------------------------------------------------
>
-------------------------------------------------------------------------|
> >
> >> |Re: [platform-releng-dev] [eclipse-build]Build
I20100526-1423
> > (Timestamp: 201005261423): Build failed
> > |
> >>>
> >
-------------------------------------------------------------------------
>
-------------------------------------------------------------------------|
> >
> >>
> >>
> >>
> >>
> >>
> >>
> >> I investigated this for quite a while and don't understand
why the
build
> > is
> >> failing. I've asked Pascal to revert the ecf change to
allow the build
> > to
> >> proceed for other teams while I continue to debug.
> >>
> >> Kim
> >>
> >>
> >>
> >>
> >> Pascal Rapicault
> >
> >> <pascal@xxxxxxxxxxxx>
> >
> >> Sent by:
> To
> >
> >> platform-releng-dev-bounces@ec
"Eclipse platform release
> >
> >> lipse.org
engineering list."
> >
> >>
<platform-releng-dev@xxxxxxxxxxx>
> >
> >>
cc
> >
> >> 05/26/2010 03:21 PM
> >
> >>
Subject
> >
> >>
Re: [platform-releng-dev]
> >
> >> Please respond to
[eclipse-build]Build
> >
> >> "Eclipse platform release
I20100526-1423 (Timestamp:
> >
> >> engineering list."
201005261423): Build failed
> >
> >> <platform-releng-dev@eclipse.
> >
> >> org>
> >
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >> This looks related to the change I did this morning to
consume the new
> >> version of ECF.
> >> I updated the map files and the java doc generation....
> >> I'll help Kim
> >>
> >> On 2010-05-26, at 3:13 PM, Kim_Moir@xxxxxxxxxx wrote:
> >>
> >>> Build I20100526-1423 (Timestamp: 201005261423): The
following error
> >> occurred while executing this line:
> >>>
>
/builds/I201005261423/org.eclipse.releng.eclipsebuilder/buildAll.xml:185:
> >
> >> The following error occurred while executing this line:
> >>>
>
/builds/I201005261423/org.eclipse.releng.basebuilder/plugins/org.eclipse.pd
> e.build_3.6.0.v20100512/scripts/build.xml:91:
> >
> >> The following error occurred while executing this line:
> >>>
>
/builds/I201005261423/org.eclipse.releng.eclipsebuilder/eclipse/buildConfig
> s/master/customTargets.xml:18:
> >
> >> The following error occurred while executing this line:
> >>>
>
/builds/I201005261423/org.eclipse.releng.eclipsebuilder/eclipse/buildConfig
> s/master/allElements.xml:16:
> >
> >> The following error occurred while executing this line:
> >>>
>
/builds/I201005261423/org.eclipse.releng.basebuilder/plugins/org.eclipse.pd
> e.build_3.6.0.v20100512/scripts/genericTargets.xml:107:
> >
> >> Processing inclusion from feature master-ecf: Bundle
> >> org.eclipse.ecf_3.1.0.v20100526-0200 failed to resolve.:
> >>>
Missing required plug-in
> >> org.eclipse.equinox.common_[3.5.0,3.6.0).
> >>>
Missing required plug-in
> >> org.eclipse.equinox.registry_[3.4.0,3.5.0).
> >>>
Missing required plug-in
> >> org.eclipse.ecf.identity_[3.1.0,3.2.0).
> >>>
> >>>
> >>>
http://download.eclipse.org/eclipse/downloads/drops/I20100526-1423
> >>> _______________________________________________
> >>> platform-releng-dev mailing list
> >>> platform-releng-dev@xxxxxxxxxxx
> >>>
https://dev.eclipse.org/mailman/listinfo/platform-releng-dev
> >>
> >> _______________________________________________
> >> platform-releng-dev mailing list
> >> platform-releng-dev@xxxxxxxxxxx
> >> https://dev.eclipse.org/mailman/listinfo/platform-releng-dev
> >> _______________________________________________
> >> platform-releng-dev mailing list
> >> platform-releng-dev@xxxxxxxxxxx
> >> https://dev.eclipse.org/mailman/listinfo/platform-releng-dev
> >>
> >>
> >> _______________________________________________
> >> eclipse-pmc mailing list
> >> eclipse-pmc@xxxxxxxxxxx
> >> https://dev.eclipse.org/mailman/listinfo/eclipse-pmc
> >
> > _______________________________________________
> > eclipse-pmc mailing list
> > eclipse-pmc@xxxxxxxxxxx
> > https://dev.eclipse.org/mailman/listinfo/eclipse-pmc
> >
> >
> >
> > _______________________________________________
> > eclipse-pmc mailing list
> > eclipse-pmc@xxxxxxxxxxx
> > https://dev.eclipse.org/mailman/listinfo/eclipse-pmc
>
> _______________________________________________
> eclipse-pmc mailing list
> eclipse-pmc@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/eclipse-pmc
_______________________________________________
eclipse-pmc mailing list
eclipse-pmc@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipse-pmc