Notice: This Wiki is now read only and edits are no longer possible. Please see: https://gitlab.eclipse.org/eclipsefdn/helpdesk/-/wikis/Wiki-shutdown-plan for the plan.
Eclipse/PMC/Minutes 2012
Meeting Minutes
December 19, 2012 - McQ, Andrew, Martin?, John, Dani
- Dani: Do we want to remove CVS from the SDK? (bug 359466)
- Decided to keep CVS in the SDK. Reasons are:
- Those who want to start without CVS can download the 'Platform Runtime Binary' and then install Git (or whatever team providers) into it.
- Besides a smaller download, removing CVS does not bring any value but can break existing users of the SDK download.
- The Eclipse packages (EPP) are the correct solution to group different Eclipse deliverables into a meaningful download. Unfortunately, some of them are currently missing Git. Dani to file bugs against EPP.
- The Eclipse project does not want to start adding more team providers to the SDK.
- Decided to keep CVS in the SDK. Reasons are:
- No call next week
December 12, 2012 - McQ, Andrew, Martin, John, Dani
- Dani: Project restructuring hickups
- John: EclipseCon - schedule going out - some interesting Platform talks related to e4
- Some off-program sessions planned, eg CBI - it's easy to build Platform on your own PC now ... room set up
December 5, 2012 - McQ, Dani, Andrew, Martin, John
- John: 4.2 Endgame Plan
- Martin: Some news on e4/Scripting.
- Dani: JDT Core leadership potential change by Jan 1
November 28, 2012 - McQ, Dani, Andrew, Martin, John
- Project Reorg - just need to go through committer list and nominate
- McQ: Support for versions < 3.6 -- Will copy a CVS snapshot to inhouse for 3.5.x builds; 1 change every 2-3 months; most fixes go into multiple Streams
- Had been checking into CVS for older streams, now CVS has some newer stuff than git on the old streams
- Source modifications are attached as a patch to the bug to satisfy the EPL
- Patch on the bug is likely the best, since it provides traceability to the problem (target milestone eg "3.5.2+")
- John: CVS Migration - basebuilder is the only active module, rest could be made read-only any time.
November 21, 2012 - McQ, Dani, Martin, John
- Dani: Replace ICU Base from Platform Downloads with Link to Orbit
- Website Migration, Eclipsecon submission reviews
- jsr14 - no longer needed, there's a hope that Java 8 is going to provide smaller profiles
- John: 4.2 Performance Patch going to be available soon
November 14, 2012 - McQ, Dani, John, Martin
- McQ: Anybody still consuming SWT for Windows CE ? - Don't know, eRCP project state also unknown; will likely drop WinCE, but asked on the mailing list.
- Dani: ICU4J v 51 - In the past, we slightly tweaked ICU4J to keep it backward compatible for CDC (only 4 small changes)
- Minimal discussion happened; no big technical issues, but one change is breaking (return type)
- Breaking methods used nowhere in the SDK
- Resolution: Not going to patch ICU4J. Upstream responsible for their stuff. If in doubt, adopters can go back to the older ICU4J.
- John: Website git conversion - Denis test conversion and check for size of repo (should move big stuff to download)
November 7, 2012 - McQ, Dani, Martin
- No topics
October 31, 2012 - McQ, Andrew, John, Martin, Dani
- John: bug 393195 Kepler Plan Updates: Removed XP from Plan, Mac from 10.6 to 10.8 (Means Oracle Java)
- How to position GTK3 ?
- John: Committer Group Granularity - Google doc listing #active committers per group
- Resolved to perform the following merges:
- Platform resources + Platform team
- Platform runtime + UI
- All PDE components into one project
- Resolved to perform the following merges:
- Martin: CVS readme.txt replacement - not yet implemented by Platform, create scripts if it's done
October 24, 2012 - Dani, Andrew, Martin, McQ
- Windows 8 and Eclipse : Seems to run fine so far - no requirements known so far for Metro
- Use of space in Metro "Modern UI" looks similar to what Orion does ... maybe some potential there
- Supporting "Modern UI" with SWT would require complete API rework (not just a skin)
- Classic desktop apps look less shiny than on Win7
October 17, 2012 - Canceled, not enough members available
October 10, 2012 - McQ, John, Martin, Andrew, Dani
- bug 389175 dual 3.x and 4.x stream releases - good discussion, good resolution by the Planning Council
- Can we do more frequent maintenance releases ? - PMC said no
- Orion Graduation Review - great video
- Dani: remove 'org.junit' version 3.8.2 from SDK - PMC says "yes"
October 3, 2012 - McQ, Dani, Martin, John
- Debugging issues with Java Heap
- John: Commit Rights Granularity - avoid components at risk of being orphaned
- JDT still good. John going to look over components.
September 26, 2012 - McQ, Martin, John, Andrew
- Members Meeting / 4.x Performance
- Are there any things that prevent commercial adoption ? - 4.x is the path forward and here's where we invest. Beyond that, 3.x is still around.
- Andrew: Papercuts Feedback - small problems that people run into ... will think about useful ways in which Community can contribute small things
September 19, 2012 - McQ, Martin, John, Andrew, Dani
- Martin: Juno patch feature for bug 381407 and bug 383890 ?
- This got fixed by a re-spin of RC4
- McQ: 4.x Performance, and Platform committers
- Team is working on issues... like Jochen Krause's blog
- LTS going to improve things
September 12, 2012 - Andrew, Dani, John, Martin
- John: CBI Build - good progress with Thanh from EMO visiting Platform Team's offices this week
- Focus on Juno Stream because LTS builds need to get done
- Can produce an SDK but cannot produce many of the other artifacts (That's good enough for LTS which only needs a p2 repo)
- Can claim readiness only when binary bits are the same (p2 comparator tool from Olivier - which does byte code level compare of class files, and binary comparison of other files)
- John: 4.x Performance - Several people on the team actively looking at issues, some problems identified, some not yet reproduced
- Some work started on re-enabling Performance tests
- start running on virtual machines to see whether dedicated machines are needed or not
- How to improve the process integrating people / resources / donations from companies (EMO pushes towards more contributions)
- Need more committers for longer term, not only short-term patch contributors
- Andrew: FOSDEM Europe (Brussels) coming up - Ian Skerret going to reserve a room, asking for people to join, good conference, lots of cross-pollination
September 5, 2012 - Andrew, Dani, Martin
- Martin: Windows 8 testing ? - Filed bug 388560 to get "Windows 8" added to bugzilla
- Dani: Initial testing by SWT team didn't reveal anything; Dani in favor of getting W8 deployed on Hudson once it is out officially
- Should consider adding to the Kepler plan; consider backports to Juno SR for issues found
- Martin: 3.8 vs 4.2 - posting on cross-project list
- Need to look at stack traces in detail eg bug 385272. Watch how that E-Mail thread evolves.
- Dani: Go/No-Go decision process for RC's - to be discussed in the Arch call
August 29, 2012: - John, Dani, Martin
- John: Bugzilla's new auto-duplicate-find ... Dani likes it but moved suggestions somewhere else with Greasemonkey to avoid interfering with typing
- Tom Watson: Big changes coming in Equinox for Java modularity - thought about Kepler but will wait for another year in a branch
- Java 8 also won't be in Kepler ... so will likely be kind of a Maintenance Release
- Dani: Version Numbering document is a bit unclear about upreving from a n.n.2
- RESOLUTION: desire is clarifying the next is n.n.100
- Dani: Delete files from CVS projects and add a README ?
- Yes if it can be done in an automated way .. to avoid confusion about outdated sources
- eg README points to a Wiki page
- John: Keep the .project in place such that jumping from version to version doesn't break a workspace
- Martin: Acceptance criterion that importing from scmurl in MANIFEST.MF still works (when choosing the tag ... choosing HEAD should bring up the README)
- Yes if it can be done in an automated way .. to avoid confusion about outdated sources
- John: Plan Update upcoming
August 15, 2012: - John, Dani, Andrew
- Do we support removal of bugzilla CLOSED state? Resolved: yes we do
- Discussion of long term release plans beyond Kepler
August 8, 2012: - McQ, Dani, Martin, Andrew
- Dani: News from Jeff
- McQ vacation from August 13 to August 31
July 25, 2012:
- CBI Build
- For SR1, the build can produce what the Release Train needs ... but not all the extra validation
- Need to run both builds for now ... until the CBI team's bytecode verifier produces the same output
- CBI team continues to look at the list of diffs and improve their builder
- Platform team would be OK dropping the legacy build once the bytecode validator has no diffs
- Platform team can't commit more resources to CBI now (Paul Webster already heavily involved)
- Losing Commits on the Orion git repo
- Dozen commits or so
- Might be a defect in the Orion git tools, but the server side should prevent that !!!
- DenyNonFastForward should have protected ... need an extra layer of protection ?
June 27, 2012: - McQ, Dani, Martin, John
- Eclipse Demo Camps - none in Ottawa
- Going to skip the arch calls for all of July - leave the PMC slot in place.
- Should keep an eye on the Community feedback for Eclipse 4
June 13, 2012: - McQ, Dani, Martin, John
- Discussed the ECF rebuild request (bug 382206). Arguments in favor: it is a safe change, it reverts the situation to where it was in RC3 and earlier. Arguments against summarized in https://bugs.eclipse.org/bugs/show_bug.cgi?id=382206#c20.
- We will need to tag the release but we can wait until we are officially done (week of June 25th).
June 6, 2012: - McQ, Dani, Martin, John, Andrew
- Discussion of how to handle test failures for final build. Most of the remaining failures are infrastructure issues related to our move to Hudson for tests. We will document the failures through bugs, and include a note on the build results page for the final build explaining the failures
- Discussion of whether we need a 3.9 release. Unanimous conclusion was that we do not see a need for it. Reducing to one stream will greatly reduce our development overhead and encourage consolidation around the single stream.
- Discussion of how to handle the disappearance of org.eclipse.equinox.concurrent from our RC3 builds. This was never deliberately included in the platform, but was present due to a greedy optional dependency from ECF. By very close 3/2 decision PMC decided to leave it alone and document the disappearance in porting guide, readme file, release review material, and cross-project list.
May 23, 2012: - McQ, Dani, Martin
- McQ: CBI / Tycho builder: Paul Webster driving Platform efforts (AI Martin notify about bug 359719)
- Martin: bug 380984 ant / Orbit respin? - Unclear for now
- McQ: Finally figured out some long-standing bug eg "Variables view sometimes empty"
- Dani: It's 2 issues: First-time switch to Perspective deferred to 4.2.1, but bug 379818 is finally fixed
- The actual underlying defect had been in there for years
May 16, 2012: - McQ, John, Dani, Martin
- Dani: Version Number in branding features gets auto-generated and 3.8 version is now higher than 4.2 (since 3.8 build runs after 3.8)
- John thinks it's not a big deal
- John: CBI Patches - CBI has lots of patches downstream in their POM's ... RC2 is the last chance getting that in
- Auto-generating the POM.XML would be nice but that's not how things work today
May 9, 2012: - McQ, John, Dani
- John: Classic update can still be enabled via capabilities - should we remove it?
- Dani: in general "yes", but too late now
- McQ: too late, but make sure we do it for Kepler
May 2, 2012: - McQ, Dani, Martin, Andrew, John
- John: Forking of SWT 3.8 and 4.2 are now separate Streams with CTabFolder implementation being the only difference
- Dani: Icons (not discussed)
April 18, 2012: - McQ, Dani, Martin, Andrew, John
- Martin: E4/Scripting initiative
- McQ thinks the E4 Application Model should be exposed as a DOM to scripting .. anything other than utilizing the e4 DOM for hiding a view feels wrong
- Martin: Approach is very pragmatic - don't see anybody investing into new Frameworks at this point but converging on what already exists
- If this goes towards macro recording, the recorder should utilize the e4 app model
- Signing up on the eclipse-scripting-dev mailing list will keep people in the loop (expect low volume)
- McQ thinks the E4 Application Model should be exposed as a DOM to scripting .. anything other than utilizing the e4 DOM for hiding a view feels wrong
- Dani: API Approval Process Approval on Bug plus sent to eclipse-pmc list (like last year)
- John: PDE Committers - perception of not being open
- McQ would like to see a plan item for next year "actively promote more committers from outside"
- Martin: Asterisk - will need a dialin for Dani in Switzerland; AI Dani forward his dialin to Martin for the time being
April 11, 2012: - McQ, Dani, Martin, Andrew, John
- Martin: LTS vs Quality/Fixes in Current Stream - Like John's reply making the link between the two
- McQ/John: Builds - Moving forward
- McQ: CBI Builds - Platform not going to take any ownership but will be a consumer like others - still shooting for Juno SR1
- Andrew: 4.2 Builds in Fedora 17 - Testing now
- McQ: IBM Ottawa Team moving to new building - expect delays during next couple days
April 4, 2012: - McQ, Dani, Martin, Andrew
- Dani: VM Requirements - Help no longer works on Java 5, this surprised some people... but Project Plan talks about a Java 6 req now
- Martin suggestion: on the "Required JVM" table in the Project Plan, add a 2nd column with closure of BREE including all pre-reqs (continue having 1st column the BREE of the bundle itself auto-generated)
- Dani: Jeff still mentioned as PMC member
- AI Dani send E-Mail
- Martin: Concern about Platform Quality - eg SWT Redraws on 3.7.x - Old-time committers moving away
- McQ: Platform team's focus will continue being maintenance, and new Platform support ... team's capacity is going down, but SWT is not affected mosted
- Martin: Code contributions are easy, hi-quality triage and review are harder
- Martin: Re-think the idea of having committers paid by the Foundation and Member Companies paying their share
- Example: Why would Pascal review p2 in his sparetime ... but if he gets paid per bug his experience can be extremely valuable
- Be more creative on resourcing side
- McQ: In favor of putting more structure in place similar to LTS ("hire people for fixes") ... just continue with LTS ?
- IBM will have a need for fixes for at least another 8 years .. but that's a very self-serving focus
-
AI Martin send E-Mail to Mike to keep the ball rolling
- Dani: bug 375292 Reference to Marketplace in the SDK
- Martin: "Classic" is a Building Block but "Packages" are for end users
- McQ: The purity of the SDK in itself isn't relevant
- Martin: Why not just create a new package "Classic" which is at +3 and pulls in egit + marketplace ... the process for building is all in place, may just need a package reviewer
- AI Martin mail Markus
March 21, 2012: - McQ, Dani, Andrew, Martin
- McQ Oracle Java on Mac
- Dani denyNonFastForward
- McQ JDT Team might see some losses
- Andrew Running OpenJDK tests now on ARM :)
March 14, 2012: - McQ, Andrew, Dani, John
- Losing Eclipse project release engineer next week
- David Williams is stepping in to help with our builds until Juno
- Most immediate issue is that our build needs elevated privileges to tag, sign, and promote builds
- kmoir was the only committer with sufficient privileges
- We either need to give another committer sufficient rights, or give build user access it needs
- Could use hooks to ensure build user isn't doing anything other than tagging
- All agreed the build user approach was the best option. John to open bug against foundation to carry it out
March 7, 2012: - McQ, Andrew, Martin, John
- John: API Deletions bug 370248 - PMC agrees.
- McQ: Common Build Infrastructure at the Foundation
- Shipping Platform Juno bits built by PDE ... but Foundation wants EPP packages built by Tycho; Platform can't support this
- Reason might be that 4.2 is planned to be the start of LTS and VLTS .. ok for now if LTS bits are different since not publicly available
- Plus, LTS really starts after Juno SR2 ... Platform might be in a better position for Tycho by then
- In Linux, Fedora also builds by themselves ... but assumption is that people come to Fedora support first
- Martin, Andrew really concerned about fragmentation and duplication of effort ... eg what if the qualifiers don't match
- Should be treated as proof of concept but not publicly available until Juno SR2, in order to avoid fragmentation !
- Shipping Platform Juno bits built by PDE ... but Foundation wants EPP packages built by Tycho; Platform can't support this
February 29, 2012: - Dani, McQ, John, Martin, Andrew
- John: Orion team received a box full of Black Ducks to honour their Black Duck Rookie of the Year award
- Dani: Gerrit Move
- John: Important information needs to stay in Bugzilla ... don't pull from outside git exclusively
- Andrew: Platform Defect Contest
- McQ concern: Lots of duplicate bugs would take focus away from the team
- Martin: Could encourage triage as part of the contest, and only count good/new bugs
- John: Some low-hanging fruit bugs in Platform UI but not necessarily in 4.x ("helpwanted")
- John: Re-introduce the "greatbug" keyword ?
- A greatbug is relevant, new, good description that helps understanding/finding the issue; ideally with a patch.
- Resolution: AI John send note to Mike
- John: PDE API Moved
- Some internal, provisional API was migrated to full API in this release
- Someone from SAP (Dimitar Giormov) Asked adding back the provisional API such that they could support 2 versions of Eclipse with one codebase
- McQ: Would have helped them 4 years ago .. but at the moment don't have enough people
- McQ: Question how pervasive their use of the old API is ... how much Reflection would they need to put int
- One resolution could be OK but only if they provide all the patches
- This kind of duplication is error prone and adds a maintenance burden on the project
February 22, 2012: - Dani, McQ, John, Andrew
- Repeat of PDE build discussion from last week. We decided to nominate John Arthorne as a committer so we at least have someone to accept community patches. We will revisit after Juno and perhaps do a larger restructuring
- Communication with eGit: We have written some patches for eGit that have been growing stale. Dani will review the patches, and then send an email to eGit lead asking if there is a particular reason for not accepting them. If there is no problem with the patches we will track down an eGit committer to release them.
- Andrew asked about process for approving CQs:
- Any PMC member can feel free to approve CQs
- New versions of existing third party libraries are a no-brainer, typically just +1 right away
- For new third party libraries we may need a PMC discussion about why the library is needed, what feature it will live in, etc
- For code contributions just make sure the legal process has been followed and then +1. We generally trust the judgement of the committer who will process the contribution.
February 15, 2012: - Dani, McQ, John
- Some discussion of what to do with PDE Build, which lacks committers. PDE as a whole has barely enough committers to be a functioning project, so maybe we should be looking at a larger restructuring after
Juno rather than one piece at a time. We lacked quorum to make a decision so we will defer discussion to another week.
February 8, 2012: - Dani, Martin, Andrew, McQ
- Andrew: Eclipse stand at FOSDEM in Brussles
- Lots of people just dropped by to say hi and thanks
- Mike Milinkovich was showing off Orion; Gunnar and Andrew did others (Web, EclipseRT, C/C++)
- McQ: 4.2 Quality ... current list of defects looks manageable to tick all off .. any messaging we want to do ?
February 1, 2012: - Dani, Martin, John
- Dani: No UA Committers any more - move UA under Platform UI such that at least somebody can apply patches
- AI John check with Wayne ... UA is a subcomponent now, may not even need a move review.
- Dani: bug 343150 git DenyNonFastForward
- Orion could eventually add tooling to support easy workflows in spite of deny on the repo ... but not in this release
- Ping on the bug again and proceed
- John: Decide when to switch 4.2 builds from re-using 3.x binary bundles vs full 4.x builds
- Today, the 4.x testsuite just tests 4.x specific stuff; things like JDT tests just run against 3.8 today
- Today, cannot generate Javadocs in 4.x today sine it needs the full source
- Today, nightlies are running 3.x only (could change to running 4.x nightlies only)
- Today, 3.x builds run at IBM and 4.x builds run at Eclipse.org
- Only few test failures left in the 4.x builds --> ready to switch
- AI: Kim to send 4.x test results to the mailing list, consider switching right after M5
- Performance Tests: Not yet run on 4.x yet - should be able to run on Denis' vserver
- IBM is going to lose the physical machines for performance tests in 2 weeks ! (Going to migrate to vserver too!)
- Need a new baseline
- Andrew: at FOSSLC in Brussels this week, going to represent Eclipse with Mike Milinkovich, Gunnar Wagenknecht and others
- Gerrit: Platform won't jump in as an early adopter this time, wait for other projects first
- Not all Platform Unittests can run on Hudson today (need at least some fingerprint tests to get value from Gerrit)
January 25, 2012: - McQ, John, Martin, Andrew, Dani
- McQ: Note about 4.2 on Release Train
- Andrew: Eclipse 4.2 for Fedora - Feature note written for Eclipse Juno led to some discussion at Eng.Steering Committee, but finally approved
- Dani: Team down by 1 but no impact (Dani is going to take over responsibilities)
- John: Build Failures in UA, Update, PDE, p2 yesterday after the move to Java 7... nobody watching these, need to find caretakers
- UA tests often fail due to a link validity checker, these can be caused by anyone
- Certificate Validity issue (Jar signing)
January 18, 2012: - McQ, John, Martin, Andrew, Dani
- McQ: 4.2 quality and next steps
- More resources found to enable 4.2 as the LTS base - encouraging, although some items still missing
- Fedora 17 think about shipping 4.2m6 (work ongoing) - that would be a large stack on top of 4.2
- John: Oracle Java 7 as Reference Platform now that the most severe issues are fixed (Java 6 EOL is June 2012 !) - that is Oracle, not OpenJDK
- Not remove Java 6 for now
- Fedora also does all tests on 7
January 11, 2012: - McQ, John, Dani, Martin, Andrew
- McQ: State of Presentation API's
- Different API's in 4.2 and 3.x -- will need to deprecate API in 3.x
- Christian suggested deleting API that's not implemented in 4.x to get compile errors ... John thinks that binary compatibility is more important
- Plugins will be "mostly OK", presentation API is cosmetic anyways
- McQ: get in touch with clients who use presentation API, telling them what's best to do ... often newer 4.x infrastructure is much better than trying to keep 3.x around forever
- Andrew: Cairo and GTK3
- John: 4.2 compiling now
- Martin: Updated JSch-0.1.44 version - how to announce properly
- AI Martin Announce on eclipse-dev and cross-project
- Should also put into migrtion notes and README
January 4, 2012: - McQ, Dani, Martin, Andrew
- McQ: Pascal's note about p2 leadership .. not sure who could step in and make a commitment (may see the same fate as UA)
- Tom Watson looking for candidates, but looks like none in McQ's team