Jump to: navigation, search

Eclipse/PMC/Minutes 2011

< Eclipse‎ | PMC
Revision as of 10:54, 29 February 2012 by Martin.oberhuber.windriver.com (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Meeting Minutes

December 14, 2011: - McQ, Dani, John, Martin, Andrew

  • API Tooling Baseline comparison 3.7 vs 4.2 -- problematic since API was copied
  • bug 343150 git denyNonFastForward ie forbid commits that rewrite history ... looking for general policy ... Eclipse Project to adopt immediately.
    • Andrew: Denis may roll out the change to all repos
    • Resolution: Wait until after Christmas, if not rolled out by then move forward ourselves
  • McQ (Tom Watson): Leadership for p2 team
  • John: 4.2 Builder switching to Hudson for M5 - not all of the UI tests compile
    • Lot of test failures
    • Just flip the switch to Hudson, and live with failures at some point ... (and commit to fixing test failures as fast as we can)
    • Fixing a components' test is the component team's job.
  • Vacations: Skip next 2 weeks

December 7, 2011: - Dani, John, Andrew, Martin

  • Martin: bug 365908 with 4.2m4 not picking up extensions from dropins/
  • John / Dani - XML problems breaking the build due to a "Help Index Generation" ant task
  • John / Anrew - Hudson Failures, Platform fortunately not there
    • Reasons: Hudson can't tag and can't deploy to downloads
    • John: It makes sense that the Hudson user can't write to the Repo
    • Orion: Start from a shellscript (for tagging), then kick off a Hudson build partway through

November 30, 2011: - Dani, John, Andrew, Martin, McQ

  • Andrew - Long Term Support. McQ - Overlap is in the build technology, can't be done for Juno (still busy with git + hudson), but happy to consume what's being developed
    • Foundation contracted with Sonatype (Igor) for moving to Tycho, but just analyzing the current build may take 4 months
    • Linuxtools currently using a small ant wrapper which just calls the Platform build (generate a source tarball such that linux distros can consume Platform from source;
  • Martin - User Assistance Situation
    • Chris G moved to Jazz - Rest of the IBM team completely maxed already, would fix critical security defects only
    • Some work has been done by other IBM committers.. could release patches, but can't be responsive to fixes
    • There's an opportunity for others in the Community to step in and set the direction
    • McQ: The only one most important issue at the moment is compliance with the new US Government accessibility regulations
    • Newer Lucene; Help Search Stemming problems; nobody actively tracking the inbox
    • John: Have been moving "committer-less" components into the general Platform component where 12+ committers have commit rights
    • McQ: Platform project Zombies is OK, but having somebody step up would be better
    • John: 3rd party dependencies... help has a few, eg upgrading to Jetty (managed by Chris before he left)
      • Lucene: Merged with SOLR, then 3 releases came out in the past 3 months, no Foundation CQ's in ... John interested in adopting Orion
      • Been in contact with Gunnar Wagenknecht who is also interested, but newer Lucene pulls in more dependencies than it needs
    • Failing Search : would need an exact bug report where the search is failing, then McQ could get product teams excited
    • John: Orion problem with older lucene is that it's leaking file handles ... less of a problem for Platform where we index only once
  • John - bug 363948 Eclipse 3.8 to 4.2 update
    • Fix would involve changing metadata in p2 such that a 3.8 update wouldn't find 4.2 automatically, unless a specific update is requested
    • That fix leverages what Pascal from p2 considers a bug
    • Products could manage the problem by not adding the Juno repository.. but there's lots of things in Juno that people would want (eg egit, cdt, ...)
      • Fix that John thinks about would be on the product level

November 23, 2011: - Dani, John, Andrew

  • Discussion of whether to turn on auto-tagging in maintenance builds. General agreement this is a good idea. We would not have a separate "integration" branch for maintenance streams.
  • Some discussion of EclipseCon submissions. PC voting is underway and program should be decided in the next two weeks. There is a great program of submissions so far, and it will be tough to select the talks that will make the cut.

October 26, 2011: - Dani, John, Andrew, McQ

  • Approve Paul Webster as Platform UI project lead
    • Committer vote here.
    • The PMC members unanimously approve this nomination
    • John to send note to EMO
  • Some discussion of the Eclipse Platform UI Git disaster - bug 361707
    • Currently any committer can delete everything from the repository permanently
    • The commit hook prevents committers from restoring commits made by others that they have in their local clone

October 12, 2011: - Dani, John, Martin, McQ

  • PMC upstaffing - looking for more Community input
  • egit bug backlog
  • e4 builds
  • p2 downloadable archived repos
  • Moving to 4.x
    • Community is moving ... any product that's not planned for EOL needs to move

October 5, 2011: - Dani, John, McQ

  • Some discussion of removing CVS from the SDK (bug 359466). There is no strong argument for removing it, so it is hard to argue for disruption to user community. In the end we would like CVS to be removed and would like it to be just as easy to install any SCM into the platform. We would like to see the p2 discovery UI used to make it trivial to add SCM support into SDK (Help > Add Version Control > CVS). Once we had that in place we would be in a better position to remove CVS by default.
  • We also discussed the Git line delimiter problems. Current recommendation:
    • core.autocrlf=false - not honoured by JGit so this makes things worse
    • Use Global workspace setting to use Unix line feeds by default
    • Encourage projects to add a project-specific setting to use Unix line feeds

September 21, 2011: - Dani, John, Martin

  • Dani: git - Mac / Linux using scripts, tooling not yet ready for Windows ... productivity lacks
    • painful when working across multiple repositories ... problematic workflow for reviewing incoming changes
    • John: egit could fetch automatically in the background (without merge)
  • John: used platforms poll

September 14, 2011: - John, McQ, Dani, Martin

  • Martin: PMC membership - still looking for people to ask after Steve and Jeff left
  • John: Jetty 8 - should be using Jetty from Eclipse (Jetty 7 ist stable, Jetty 8 is current) but API changes after Jetty 6
    • eg standalone help - AI John to ping Chris G;
    • No longer re-exporting Jetty API since 3.6 (deprecated api, introduced replacements)
    • releng, would need Jetty at -1 if using current from train .. makes builds a bit more messy again
    • check with Equinox

September 7, 2011: - John, McQ

  • State of building Juno at eclipse.org: we are close but still some issues with the tests. Transition to Git halted our progress on this. Once the Git migration is complete, we should just make the switch over and live with the pain of some failing tests. Component teams will need to help with diagnosing problems with their tests.

August 31, 2011: - Dani, John

  • We need to pull the new ICU4J out of maintenance release because the fix didn't pass review in time. We will defer to Indigo SR2
  • Some discussion of the treatment of tags in Git vs CVS. Committers need to be aware they can't checkout the contents of a particular build into their workspace all at once.

August 24, 2011: - Dani, John

  • EGit is far behind on level of polish, usability, robustness, conforming to UI guidelines, etc. We may need to make contributions to avoid it hurting productivity
  • Proposed workflow changes in bugzilla 4.0: bug 337245. Proposing removing NEW, ASSIGNED and CLOSED states which would lose valuable information for us.

August 10, 2011: - Dani, Martin, McQ, John

  • Dani / Martin - Debug Flexible Hierarchy "provisional.api" namespace
    • Cannot be moved into an "API" namespace without breaking external non-SDK clients since types are leaked from internal packages
    • Q1: Move framework to a place outside Debug?
    • Q2: Declare "internal" packages as API? - API Tools won't be usable
    • McQ: Is it really impossible to add new API in parallel while keeping the old one in place for now?
      • Issues: (1) Performance, (2) bugs due to delegating, (3) is it worth doing that work? CDT already marked x-friends - Debug could offer the same for other clients.
    • McQ: Do we anticipate a huge Community out there just waiting for stuff to become full API? Then it's probably worth it...
      • But at this point any potential consumers already consume, so it's likely not worth the pain of renaming
      • Dani's suggestion: Go into a separate bundle and re-export, such that clients don't need debug
    • John: Unless packages were intended to be API from the beginning, it's likely not API quality (that's likely also why Boris didn't want it in Jface)
    • Martin: Would really want to see namespace refactored... quality standards..
    • John: Had the experience in p2, it was much more than just package renaming
      • Very painful for user community, but in the end produced a much better results (but p2 was very young at that time)
    • McQ: Can we provide backward compatible versions in parallel to the new namespace? We CANNOT break consumers.
      • Renaming only makes sense if mirroring/proxying is possible.
      • Lots of consumers in the old form, how many new consumers would we gain in the new form?
    • Martin: Create a fork / clone on the UI side for new consumers, as real API?
      • McQ: Only makes sense if there's a community just waiting to pick this up.
      • Dani: Debug itself could not move to the new framework without breaking clients ... so we'd not have any clients inside the SDK
    • McQ: It's non-trivial work, and before starting it we'd need to verify there's consumers for it
    • Learnings from the past: Not keep stuff in provisional for so long any more
    • Resolution: Pawel to find consumers for a refactored new bundle, then decide if / how to do the work
  • Dani: Java 7 - Segfault due to loop optimizations (details)
    • Occurs a lot inside Apache Lucene, how to communicate that?
    • McQ: send to eclipse-dev and cross-project
    • Oracle aware of the issue - working on an early refresh, workaround exists (disable loop optimizations via -XX:-UseLoopPredicate)
    • Another JIT bug was fixed by Oracle before GA because we reported it
    • AI Dani Talk to Oracle when an update can be expected, soon send a message to eclipse-dev / cross-project
  • Vacation notices:
    • Martin: away from August 14 to September 1
    • McQ: away from August 22 to September 5

August 3, 2011: - Dani, Martin, McQ

  • Dani - successfully moved Java 7 support to 3.7.1, 3.8, 4.1.1 and 4.2 streams



July 27, 2011: - Dani, John, McQ

  • Discussed how many versions back PDE should support development on. Historically we used N-2 but currently it is N-7. The layers of compatibility are starting to become hard to manage. We agreed dropping support for versions prior to 3.6 is ok if there is a benefit in doing so (cleans up PDE, makes the situation less confusing for end user, etc). However PDE must handle existing plugins with old version numbers in the plugin.xml (possibly with a warning and a quick fix to update/remove the number).
  • We have some rough days ahead: first release train milestone on 4.x, ongoing build pain from Git migration, merge of Java 7 into mainline, and lots of people on holidays. Hold onto your hats!



July 20, 2011: - Dani, Martin, John, Jeff

  • Dani - p2 Installer bug 317785 affects Java 7 users - when they want to update it looks like they are dead
    • Regression test only fails once every 100 times .. might be more often in the wild
    • Might need some note on the download page (clickthrough) - people dont read the readme
  • John - git migration - parts from git / parts from CVS isn't that bad for now



July 13, 2011: - McQ, Dani, Martin

  • Want more Community involvement in the Eclipse PMC - "outside representation" in addition to companies with a commercial interest
  • nice Java 7 webcast last week - Mike Milinkovich mentions our Eclipse Java 7 Support (BETA)

July 6, 2011: - John, Dani, Martin

  • Dani - on track with Java 7 work; Oracle interested in having good Eclipse support
  • git migration - Platform UI likely going to move this week
    • Paul has a tool for gathering build notes from git (currently a commandline script)

June 15, 2011: - McQ, John, Dani, Martin, Jeff

  • John - git migration: should egit be part of the SDK ?
    • Build-time circular dependencies. Keep SDK small, but make it a very simple and obvious step to obtain egit.
    • It's already easy to update - milestone to mileston SDK update is a single step for SDK + egit.
  • John - git migration and the mapfiles
    • In orion, just tag HEAD of the build branch
    • Dani: need to make sure that build notes can still be created

June 1, 2011: - McQ, John, Dani

  • Some discussion of proposed Eclipse Foundation security policy
  • We still want to see the disclosure time-frame being 3 months after the release containing the fix. Not all of our consumer community has the build technology or expertise to apply source patches to their products. Making sure the fix is in a maintenance release is the only practical way for some people to consume fixes
  • It would be nice for critical security fixes to appear in the simultaneous release repository even after the SR2 time period. p2 would need some small amount of work to be able to flag critical/security patches and automatically notify the user about them being available. Nobody has signed up to do this work.
  • Some discussion of Git migration. There is no need to move everything at once since our build can handle homogeneous inputs from both CVS and Git. One possibility is migrating Equinox as a first step. The most complicated piece is migrating the repository containing the builder itself. We will need to decide on where map files and documentation go: likely some kind of "common" repository for Eclipse TLP.

May 25, 2011: - McQ, John, Dani, Martin

  • Missing CQ's for old legacy stuff in old update sites / repos
    • No problem for ZIP's (we can move to archive)
    • OK with removing some legacy old stuff... How many years back to keep old stuff?
  • Java 7 and 3.7.1
    • Mike M OK with including in 3.7.1 (when it's stable enough)

May 4, 2011: - Jeff, Dani, Martin, McQ, John

  • RC1 test pass went well overall
  • Investigating an SWT blocker preventing checkbox tables from working
  • There is a problem with the Ant mail task with Ant 1.8. Ant mail requires an optional extra jar to be added. In Ant 1.7 this didn't cause build to fail, but in Ant 1.8 it does. Our conclusion was to document in the migration guide but take no other action.
  • Git migration:
    • Kim and DJ are investigating migration of the build to Git. Making progress but don't have a full build working yet
    • We should consider summer 2011 for migration. For those teams with forked code between 3.x/4.x, they are finding it difficult to keep fixes in sync and potentially git can help with that.

April 27, 2011: - Jeff, Dani, Martin, McQ, John

  • Dani still 100 bugs tagged 3.7 - need to remind in the archcall to properly tag target milestones
  • John EDL for Examples
    • Value of examples is they're there to be used - there's no business value in holding on to IP via Copyright so EDL seems right
    • Re-licensing existing examples is difficult (eg SWT examples); but new examples should go under EDL
    • Including examples in distributed features must take care of fixing abouts
  • John 3.7 plan final update: all plan items committed or deferred; update to RH6

April 20, 2011: - Jeff, Dani, Martin, McQ, John

  • Jeff,Martin Builds at Eclipse - Work ongoing. IBM Mac has been donated. Mac testing has problems with some weird vnc bug
    • Martin - could kick a 3.6.2+ build and artifacts OK, but confused by lots of warning messages. Kim responded immediately, but can this be made "more official" ?
    • Public Builds - get the engineering underpinnings in place, then polish afterwards
    • Won't get a lot better in 3.7 but could make it a plan item for 3.8
    • John agrees that improving the warning messages to be more informative is a good thing - Kim's very helpful
  • Dani Polish Bug to be prioritized: bug 342648 In SWT on Linux, RTL mode, Tree.getItem(Point) gives invalid result.
    • Regression ?
    • McQ: Every GTK version needs work - maybe it's specific to a new GTK; might be more than a P2

April 13, 2011: - McQ, John, Martin, Dani, Jeff

  • John: reminder to review proposed project charter update
  • Freeze plan for Indigo is ready
  • Reviewed new polish items
  • Investigating some 4.1 compatibility problems, possibly related to EMF changes

March 30, 2011: - McQ, John, Martin, Dani

  • Discussed and clarified the API approval guidelines. Will be announced in the arch call.
  • Dani - Java 7 deferred from 3.7. GA planned together with 3.7.1.
    • John: is feature work which might need review
  • John - we need to prepare the Git migration
    • John to ask DJ to do investigations

March 23, 2011: - McQ, Dani

  • Prioritized the Polish3.7 list. This is a recurring task for the next few weeks.
  • Discussed and agreed on API change approval process, see PMC mailing list note.

March 9, 2011: - John, McQ

  • Ongoing discussion of direction for 2012 release train. Will continue discussion in email to ensure all PMC members are involved.

March 2, 2011:

  • McQ - Migrating to 4.x. Don't be disruptive when there is no need to. Want to avoid migration pain.
    • New feature work will be in 4.x, but as long as there are committers working on 3.x it will stay alive
    • Make sure that quality keeps up: Releng / Build, make long-term committers continue reviewing contributions
    • Problem with 4.x is just the bugs (impeding productivity), it's not anything architectural
    • Convene again next week and come up with a consistent message, AI McQ get hold of Jeff for a statement
  • John - 3.6.2+ builds set up on Hudson now
    • Kim's bug 338557 Job's mostly ready, Martin is the first tester. Should inform the Community once things go smoothly
    • Updating version numbers like going to a virtual never-existing "3.6.3" release in order to give a clear and consistent message
    • Dani has an (internal) document - Martin: interested for TM and CDT projects - AI Dani send Doc to Martin for for meeting notes
    • Always updating the micro by one only ... so if bundle was not updated in Helios SR1 or SR2, it should go .1 in 3.6.2+

February 23, 2011:

  • Martin - 3.6.2+ M-Builds post 3.6.2 - needed for adopters to get 3.6.2+ bundles signed; on demand only? If scheduled, how often?

February 16, 2011:


February 9, 2011: - McQ, John, Dani, Martin

  • Dani - Builder bug: Theoretically, the builder bug can cause bad binaries shipped from Eclipse.org
    • But switching the builder now could have other consequences. Most people likely use the 3.6.0 or 3.6.1 builder
    • Platform uses the 3.6.1 builder. Bug appeared in 3.6.0
    • McQ thinks we should recommend using the new builder - John: can never know whether a problematic code construct is there or not
    • Provide a patched version of the 3.6.1 basebuilder as an option for people to use - AI John ask Kim to re-produce a new basebuilder, then respin RC4
  • McQ - Builds at Eclipse.org
    • IBM Ottawa consolidating 7 offices into 2, plan to not have a build lab
    • Kim still frustrated with Foundation facilities; McQ talked to people both inside IBM and the Foundation



February 2, 2011: - Canceled (no topics)


January 26, 2011: - Dani, John, Martin,

  • Dani: bug 330534 ObjectTeams Contribution - JDT might allow using their namespace as an exception, couple p2 bugs pending
    • If the p2 solution cannot be done, OT/J will need to remain off the train
  • Martin: Concerned about Eclipse 3.x quality, many regressions and not being addressed
    • bug 320329 ThreadJob#waitForRun can spin when blockingJob is not RUNNING
      • Platform Runtime: Regression introduced in 3.6, can have severe Performance impact, reported 10-Jul, commenters unsure about approach to fix
      • A buggy contribution was accepted, and then the contributor left
      • John spent a lot of time.. there was a fix that didn't work
      • Dani: Could the contribution be pulled out? - No because it's new API (yield())
      • Contributor was quite good originally but then just switched teams and was no longer available
      • The new testcase from Markus Schorn might help narrowing down the case quickly
    • bug 335153 Regression: p2 downloads are much slower due to picking remote artifacts even when a local file: URL is available
      • p2: Regression introduced in 3.6, reported 24-Jan, testcase + patch attached, response pending
    • bug 332840 Wrong error about API change (Regression; works with 3.7M3)
      • API Tooling: Regression introduced in 3.7M4, reported 17-Dec, Olivier: "I'll take a look" but no response since
    • bug 332838 Bogus potential null pointer access warning (regression; works with 3.6)
      • JDT: Regression introduced in 3.7M4, reported 17-Dec, Ayushman: "I'll investigate" but no response since
    • bug 332507 [regression] incomplete build path reported on o.e.rse.subsystem.shells.local
      • PDE UI: Regression introduced in 3.7M4, reported 14-Dec, initial traffic but no followup since 16-Dec
    • bug 335466 Regression: category.xml processing fails when "name" contains special chars
      • Just detected recently
    • bug 313899 [Progress] Every user-job causes a busy cursor to be shown for a minimum of 250ms
      • Platform UI: Patch provided on 21-May, no response whatsoever from Platform UI
    • John: There had been a fair bit of turnover, new committers getting up to speed ... having less experienced people is a fact
      • Identifying key bugs in the big bug database takes a lot of experience
    • Dani: M5 is the Major feature freeze, so new feature / API work has highest priority ... so quality drop is somewhat expected in M5
    • Even for regressions... if the regression is less critical than new bugs, it goes behind
    • RESOLUTION: Make the important bugs visible to the component leads or PMC. It's natural that we sometimes need to ping on a bug. M5 is a problematic time.
  • John: bug 335374 Redhat 6 as a Reference Platform
    • SWT team doing some final testing, then will put only RH6 on the Ref Platform list
  • John: OpenJDK as a Reference Platform - Oracle is interested and might contribute testing

January 19, 2011: - McQ, John, Jeff, Dani, Martin

  • Jeff will be unavailable for meeting next 6-7 weeks
  • John: How do we interpret "under the direction of the PMC" in Eclipse legal process?
    • McQ: Not about approval, but about PMC being aware
    • Jeff: Goal is to indicate "Working with an Eclipse mindset", ie in close collaboration with a committer ... contributor must be aware of IP due diligence guidelines from the beginning of evolution of the code. We want to avoid introducing IP leaks by accepting legacy code that was written before assuming an Eclipse mindset.
    • McQ: This is about understanding the direction of Eclipse as a whole. Request a PMC +1 on the bug as work is being started. PMC should work with committer to ensure that contributor has been instructed about IP rules.

January 12, 2011: -


January 5, 2011: - McQ, John, Dani, Jeff, Martin

  • Dani: Component Lead progress? - Portal still doesn't have the correct list of components

Archive