| Thanks for the
        attention to Kepler. We
        (Platform Releng) do plan to move to CBI based builds for
        Kepler, as soon
        as possible. 
      
 One question, if I recall, from
        the
        past comments, Tycho itself must be updated to use a particular
        version
        of JDT compiler. In our PDE based builds, we've already move to
        use Kepler
        M4 JDT compiler.
 Anyone know if that's been done
        for
        Tycho yet? And ... more important for my education :) ... how
        can I tell?
 
 Thanks,
 
 
 
 
 
 From:      
         Andrew Ross
        <andrew.ross@xxxxxxxxxxx>
 To:      
         cbi-dev@xxxxxxxxxxx,
 Date:      
         01/09/2013 08:59 AM
 Subject:    
           Re: [cbi-dev]
        Minutes: platform CBI build
 Sent by:    
           cbi-dev-bounces@xxxxxxxxxxx
 
 
 
 
 
 John Arthorne made this point in the meeting... said pushing
        more aggressively
        in Kepler can & should be done. Thanh provided an update
        that Kepler
        was building based on the patches provided (see Bug
            39608). Looks like 3 more to
        merge in.
 
 Also, we talked about whether there is any advantage to focusing
        on Kepler
        first, then Juno. Consensus was unanimous that continuing to
        work in parallel
        (both Juno & Kepler) made sense and wouldn't slow us down.
 
 On 01/09/2013 02:34 AM, Krzysztof Daniel wrote:
 I think the Kepler CBI was discussed, too, but
          I don't
          remember the
 output.
 Andrew,
 what was the final conclusions?
 
 On Wed, 2013-01-09 at 02:19 -0500, Aleksandar Kurtakov wrote:
 
 ----- Original Message -----
 
 From: "Andrew Ross" <andrew.ross@xxxxxxxxxxx>
 To: cbi-dev@xxxxxxxxxxx
 Sent: Tuesday, January 8, 2013 9:44:30 PM
 Subject: [cbi-dev] Minutes: platform CBI build
 
 
 
 Hi All,
 
 The minutes from today's meeting are posted here:
 http://wiki.eclipse.org/CBI/January8_2013
 
 
 As someone heavily interested in CBI why don't we switch
          Kepler builds
          now?
 Let's be honest, backporting build system change to stable
          branch has never
          worked (at least not well). Even if there are known problems
          there will
          be 6 months to fix them for Kepler. If we wait for Juno branch
          to build
          with CBI even the Kepler train will be missed.
 P.S. Example of expected speedup - SWT CBI patches were
          staying in bugzilla
          for looong time because they were supposed to go to stable
          branch and I
          personally would not push possible breakage(considering the
          complexity
          of the build system) to stable branch but Kepler patch is
          something entirely
          different which I would have tested and pushed myself
          probably. I guess
          that many other committers feel the same way.
 
 Alexander Kurtakov
 Red Hat Eclipse team
 
 
 
 As usual, please correct if I have mis-stated anything or
          omitted
 something important.
 
 Thank you to those that participated. For those working on
          bugs,
 we'll be meeting again next week for an update.
 
 Andrew
 
 On 01/08/2013 09:26 AM, Andrew Ross wrote:
 
 
 Hi Everyone,
 
 At 10am EST (about 50 minutes from now), I have booked a
          meeting to
 discuss the state of the Eclipse platform build based on CBI.
          The
 conference bridge coordinates are posted here:
 http://wiki.eclipse.org/CBI/Conference
          - all are welcome of course.
 
 In short:
 1) The issues that popped up over the holidays are fairly
          straight
 forward and should be fixed soon (24-48 hours most likely). So
          good
 news there and no major concerns. We'll update if any of the
          issues
 turn out to be nastier than expected.
 
 However...
 
 2) The list of bugs we need in order to release Juno SR2 built
          with
 CBI is a concern. There is some serious concern if we can
          solve them
 in time. Below is information from Paul Webster regarding the
          list.
 We'll be discussing this and updating our plan. I'd like us to
          be
 fairly ruthless to limit the list to what we absolutely need
          for
 releasing Juno SR2 only.
 
 
 As background, Juno SR2 based on CBI is a key enabler to LTS
          since
 Juno is the first release entering into the LTS program. It is
 highly desirable to release Juno SR2 built with CBI to provide
 utmost confidence that updates delivered with CBI later are
 sane/valid.
 
 The LTS program has subsidized the work on CBI and other work
          which
 does directly benefit the community. The potential here is
          just
 starting to be realized. If you look at the list of projects
          using
 CBI and the feedback from people who can build the platform on
          their
 own systems for the first time this is clear.
 
 Ideally we persevere and have Juno SR2 released built with
          CBI. We
 need your help to make this happen. Thank you kindly, and
          special
 thanks to the team who's tireless efforts have gotten us to
          where we
 are - within striking distance of achieving this.
 
 Andrew
 
 The following is from Paul Webster. Thanks Paul!
 "Here's the CBI bug for the platform build:
 
 Bug 372792 - Status of CBI build for Eclipse Platform
          (dependency
 tree)
 
 Here are the bugs I think are still outstanding and functional
          in
 nature:
 
 
 
 * Bug 394216 - o.e.equinox.executables IUs must be in build
          repo
 (CBI) This is needed to be able to build from PDE using our
          p2
 build repo. It also needs to be there so we can slice up
          our
 build repo into the smaller zipped repos we distribute.
 * Bug 377190 - CBI Platform SDK product compared with
          Upstream
 SDK product ( Eclipse ) This comparison (of the SDK
          product)
 needs to be run again, to make sure our zipped SDKs are
 comparable.
 * Bug 378234 - CBI build should produce a comparable build
          output
 repo ( Eclipse ) We need to run a comparison of the p2
          build
 repos. First at the IU level, and then deeper, as perhaps
 missing dependencies can be causing Bug 395506
 * Bug 381057 - CBI should be able to generate the API tools
 description (CBI) This is still not provided, so our
          bundles
          are
 missing the .api_description file consumed by PDE Tooling
 * Bug 384873 - org/eclipse/core/runtime/IPluginDescriptor
          is
 missing in runtime_registry_compatibility.jar (Eclipse) A
          fix
 for this can be applied now.
 * Bug 385154 - JDT core needs a way to produce ecj jar
          during
 tycho builds (CBI) If we still want to create the ecj jar,
          we
 should use kdaniel's patch or call antrunner
 * Bug 385959 - org.eclipse.jdt.launching uses
 customBuildCallbacks (Eclipse) I think we have a fix for
          this,
 JDT/Debug just has to apply it
 
 Here's our build bug:
 
 Bug 393922 - [CBI] setup a CBI build on build.eclipse.org
 
 I got far enough to be able to produce a build page, like
 http://download.eclipse.org/eclipse/staging/cbi/drops4/M20121231-0716/
 The 3 large outstanding issues on our side are:
 
 
 
 1. We need to launch the tests from one of our build
          scripts
 2. We need to capture the comparator output and have a look
          at
 it. We also need to turn on "replace" to make it
          similar to how
 we're using it in PDE.
 3. Our sliced repos are incorrect like org.eclipse.rcp,
          it's
          .24
 MB(CBI) vs 48 MB(PDE). We need to get those correct, where
          the
 org.eclipse.rcp and delta pack repos are probably the most
 important.
 
 --
 Paul Webster
 Hi floor. Make me a sammich! - GIR"
 
 
 
 _______________________________________________
 cbi-dev mailing list
 cbi-dev@xxxxxxxxxxx
 http://dev.eclipse.org/mailman/listinfo/cbi-dev
 
 
 
 
 _______________________________________________
cbi-dev mailing list
cbi-dev@xxxxxxxxxxx
http://dev.eclipse.org/mailman/listinfo/cbi-dev
 
 |