Jump to: navigation, search

Eclipse/PMC

Documents

Some documents written and/or used by the PMC:

Meeting Schedule

The Eclipse Project PMC has a weekly phone meeting every Tuesday at 11.00am EST.

Meeting Minutes

May 16, 2017 - Alex, Martin, Dani, Lars

  • Plan Update - updated on Wiki, main plan updated, please provide feedback until tomorrow latest:
    • Has Updates to delivered / not delivered, and Target Operating Environment
  • bug 509922 Performance Tests: Header is wrong, but it really compares against 4.6.1: results are created but report is wrong
    • Code should have been improved through many action enablement changes and startup improvments
    • Sravan will work on making a proper comparison against 4.6
  • Test failures: Some in Platform/UI fail frequently, the most frequent ones should be considered for fixing
    • Browser tests: pick "an available" server, the M7 run was on a newly provisioned server which didn't have all the stuff for Webkit. This should be fixed now.
    • CVS tests fail from time to time since the CVS server isn't stable at the foundation
    • UI tests may fail if Mozilla or similar show a pop-up dialog requesting an update
  • bug 516114 Tabbed Properties View Styling Change
    • Due to the code change, the Tabbed Properties View is now loaded even if it's not used; that may degrade performance
    • The issue is due to how styling is architected; an OSGi flag activates the plugin if any class is "used", the CCS engine just checks for a class without even instantiating, that already triggers the bundle activation
    • That should be discussed for change in 4.7.1 or 4.8 ... maybe use reflection to avoid this, or avoid the need for eager activation of tabbed properties ... Lars did that in the past, but it's a lot of effort
    • Activation by itself would be relatively cheap, but then the code goes through all Singleton instantiations - hard to estimate
      • The impact is likely small compared to PDE for example, which takes 1.5 seconds out of 6 seconds startup time
    • AGREEMENT: In order to be fair to other people who worked on improving startup performance, the change should be reverted. We have a change that benefits few people, but impacts everyone here.
      • Lars: OK to revert, but in the next cycle we should look at improving activations in general. Opening the Resource perspective should not activate PDE + JDT (today, this is likely triggered due to some decorators).
      • McQ would love to see an "optimistic UI" which can render at least the shell even if not all menus/etc are yet filled in
  • Dani is out Thu-Sat, please jump in and help the team if something is needed
  • Alex: Project Structure
    • Platform/Resources doesn't have a lead at the moment. Would like to merge more and more into Platform/UI. It doesn't make sense keeping separate structures just for the sake of creating barriers to contributions.
    • Eventually we'll just have the Platform/UI morph into "The Platform Project" with more co-leads
    • In the end it's not about positions and authorities .. it's all about people who are good
    • Dani: Agrees; in experience, every 2nd contribution is causing some issues, but we do have good people even if they are not official leads
    • McQ: What we really need is high-level overview and authority to recognize changes that might negatively impact others..
    • Could eventually end up with JDT/PDE/SWT/Platform. Lars and McQ are not so sure any more if SWT really needs to be separate, had several overlapping contributions - 9 active SWT committers also work on Platform code. Consider moving forward with this after the Oxygen release...
    • AGREEMENT to move forward on this, with a Platform Leadership group. We need to advance to leadership of Eclipse along with advancing the code. Leadership change should also lead to growing community.

May 9, 2017 - Alex, Martin, Dani, McQ

  • Dani: JSR376 Jigsaw: 10 said yes and 13 said no; a new spec to be provided within 30 days
    • From a technical perspective, should be possible to resolve within 30 days
    • JDT team implementing a compiler; IBM implementing a JVM, that's completely separate concerns!
    • For implementing a compiler, the last spec is quite old, that is not good enough
    • JDT team votes via Mike Milinkovich (Eclipse Foundation), Mike considered JDT team and RH team and probably others
  • Dani: M7/RC0 - test passes upcoming; compressed scheduled due to the moved date for Devoxx
    • When using the RC, watch out for 2 things:
      1. Action contribution enablement fix -- look out for menu and toolbar items not properly enabled
      2. Section forms layout
  • Dani: httpcore 4.4.1 -> 4.4.6 ECF Update and API Compatibility
    • Annotations got removed - no binary breakage, but breaks source compatibility
    • People will prefer newer httpcore; API breakage due to annotations is unlikely
    • Also, source breakage gives developers a chance to just recompile, so accepting the new version is better
    • Alex: Are we forcing our dependencies to not break API? - We don't have that power anyways?
    • Alex: httpclient has security issues every few months ... better consume the latest now, and we'll likely have to update again. So if breaking compatibility, better do it now
    • AGREEMENT to do the update and thus accept the source breakage.
  • Flooding in Ottawa: Restricted to a fairly small area
  • Dani: Release Review, does anybody look at the full document or can it be simplified?
    • All of the information in the RV is public (eg bug statistics, # contributors, ...)
    • McQ: It's useful for consumers to see that new things are happening, that is the N&N; beyond that, the use is limited. Effort around half a day.
    • Many bloggers talk about new things anyways
    • Migration Guide and N&N will get into the RV Material; beyond that, Dani offers adding stuff that people request
    • AGREEMENT to keep it simple, and send add-on requests to Dani if they come up.

May 2, 2017 - Dani, Lars, Martin, Alex, McQ

  • Lars: Regex Evaluator Plugin
    • Basically a view where one can input a regex and text, and see if it matches
    • Would like to add to e4 incubator, goal to eventually add into Platform/Text or similar
    • McQ: Where it the boundary between e4 and Platform? How do people kick off little projects .. e4 incubator?
    • AGREEMENT to move code to Eclipse.org (into the incubator) and proceed from there. Might not end up in Platform, but will be good having the code at Eclipse.org.
  • Dani: Endgame plan sent - Oxygen looking good, not too many bugs open
  • Dani: JSR Jigsaw Ballot
    • Asking for finalizing the Specs on JSR376 before voting Yes, since otherwise life is hard

Apr 25, 2017 - Dani, Alex, Martin

  • Dani: Neon.3a
    • Respin done -- problem: Updating from Neon.3 to Neon.3a , the problematic bundles are not removed
      • AI Alex tell Jeff to ask Pascal or Tom about how to disable the bundles on update
      • Martin: If this problem can't be resolved could live with it ... better not take too much risk trying to resolve this very special problem

Apr 18, 2017 - Dani, Lars, Martin, Alex, McQ

  • Dani: RH and Java9 Jigsaw
    • Very good write-up linked from the AC Mailinglist - see also follow-up
    • Eclipse JDT team concerns that the Spec is not complete brought forward to Wayne
  • Dani: Neon.3 Respin
    • Last action was with Ed Merks to validate that the Fix from Tom Watson actually works - no update so far
  • Lars: Glance Migration to e4 and/or Platform
    • e4 incubator was always intended to have very low entry barrier - can give the Glance team access quickly
  • Lars
    • JUnit 5 licensing - EPL + Apache

Apr 11, 2017 - McQ, Lars, Alex, Dani

  • Dani: Update from last Planning Council call:
  • Dani: Asked Lars to review the Dark theme bug. Lars, said, he's not a user of it, but has one in his team. Dani: Ask him to review the change.

Apr 04, 2017 - Dani, Martin, Alex

  • Martin: bug 514257 and cross-project Neon.3 wiring issue
    • Problem: External libs don't deal with versions the same way we do
    • Dani thinks that the problem is that Release Train projects don't agree on one version to use
    • Martin saw a similar problem related to some bundles doing import-package and others doing require-bundle:
      • Commons Logging is not a Singleton
    • Dani has a call with Tom later today; hoping a respin can be done after fixing in Equinox
  • Dani: Heads up on upcoming Planning Council discussion re: Java 9 and Junit 5
  • Dani: PPC - No response on McQ's message regarding PPC, Dani will update the Plan after another week

Mar 28, 2017 - McQ, Alex, Lars, Dani

  • Replacement for Sergey
    • let's wait and see how things go
    • next candidate would be Tom Watson once Equinox is moved back to the Eclipse top-level project
  • Dani: PPC strategy: drop 32-bit, provide 64-bit with Oxygen (4.7) and drop 64-bit BE with Photon (4.8)
    • all agree
    • McQ will send a note to cross-project-issues-dev
  • Lars will nominate Mikaël Barbero for Platform UI
  • Lars plans to bring a IDE search plug-in to e4 (http://ystrot.github.io/glance/)

Mar 21, 2017 - Dani, Martin, Alex

  • Dani: Dirk Fauth Requests on eclipse-pmc
    • Alex agrees with Dani
    • Where does the "0.x version before graduation" requirement come from? - Development Process, but maybe didn't apply to plugin versions
    • Dani thinks that forcing a 0.x version causes forced breakage once something goes API thus doesn't make much sense
  • Dani: Szymon Ptasiewicz

Mar 14, 2017 - Dani, Martin, Alex

  • Dani: Java 9 readiness - see cross-project-issues-dev
    • Step 1: For testing, just download a Java 9 beta JVM, add the module -- all testing is possible from commandline
      • Platform team has an install, but Mike was concerned about legal issues - a bug is open to allow all simrel access
    • Step 2: Running the JDeps tools as per the Java 9 Readiness wiki can show violations
      • Can even run with Java 8, Dani ran it on simrel, it's looking mostly good, only Orbit shows violations
      • Platform team has some scripts; for Hudson there's even a Mojo, but it aborts when it finds a violation.
    • Step 3: Reflection - Will only figure out when knowing the code or testing
    • All these 3 tests can be done without the Java 9 patch in PDE. Only for debugging, one needs to self-host.
      • This can be done either with Ed's Oomph support, or via the Marketplace
    • Platform team is still negotiating options for better support with Oracle
  • Dani: JUnit 5 considers cutting down features for aligning with Oxygen
    • Decision to align, contributing support to Oxygen Update 1 in September
  • Dani: 2 Platform Issues in M6
    • platform.ui reexports everything including jface - lower bundles not updated - JDT may fail if installed (fixed for M7)
    • major version uprev: interface not marked as @noimplement; reverting the version, people who loaded Friday's M6 would never get any updates any more; decision to stick to the version uprev, since the respective bundle is not used a lot
      • In general, accidental major version uprev should be reverted because the error would stay around for long and might break existing binary bundles who can't be fixed any more
      • In this case, it was OK to keep since the bundle isn't used much
  • Alex: Generic Editor and Testing
    • Looking very good already - only 1/3 of the code needed for getting the same feature set as with existing methods for creating editors
    • Looking for features to add missing documentation (workflows, use-cases like converting existing editors): End goal is to provide everything that the JDT Editor is providing
    • Dani: Likes replacement of pages on multipage editors, and some new editors; would not consider replacing the JDT editor
    • In the Platform.ISV docs bundle, there should be a section for "how to write your own editor"

Mar 7, 2017 - Dani, Martin, Alex, Sergey, Lars

  • Dani: 4.6.3 Updates - Looking good, Equinox bugs moved out by Tom
  • Dani: 4.7M7 Update - Infrastructure issues but testing looks OK (except Mac which is not built yet)
  • Dani: Java 9 / Planning Council Update - Postponing Oxygen was discussed but denied; late July update being considered
  • Lars: Platform.Resources Leadership after Sergey Leaving
    • Dani pinged Szymon, will continue discussion when we hear back; merging with platform.ui is one option
    • Sergey: The "Team" and "Resources/Filesystem" parts could be considered separately. Consider merging Team into UI.
    • Alex: One top Platform Project would make sense now...
    • Dani: reconsider when McQ is back. In the past, SWT was seen as separate.
    • Dani: Releng needs to be separate and protected because Releng committers can also change Hudson jobs, shut down servers etc where no history is kept

Feb 28, 2017 - Sergey, Alex, Martin, McQ

  • Alex: Short update on Neon.3
    • 2 jdt.core bugs asking for PMC approval - regressions compared to Neon.2, approved
    • PPC rebuild is in, so looking good
  • Sergey will leave Eclipse work after end of March
  • Martin: RCPTT / Reddeer / Testing
    • Reddeer: No tests that check Platform only - everything is on higher level today (WTP, server connectors etc)
    • Runs on Neon.2 right now inside RH - the infrastructure isn't built for consuming daily builds
    • Would take some effort reducing to Platform only - no time for this in the RH team at the moment
    • AI Martin will look at integrating RCPTT as time permits; would like bug 505826 resolved ideally
  • McQ will be out next 2 weeks

Feb 21, 2017 - Dani, Alex, Martin, Lars, Sergey

  • Lars: bug 512273 Allow any committer to retrigger Gerrit validation, and bug 512319 allow rebase
    • Agreement to allow retrigger for anyone, but rebase should be done by committers working on a contribution
    • Dani: Not clear where to retrigger (Sergey: it's discoverable once permission is there)
    • AI Lars follow up on the bug
  • Martin: UI Testing
    • RCPTT is completely Open Source; commercial server could be used for load balancing tests but this is not necessary
    • Alex: Interested in Reddeer for upstreaming tests that already exist at RH; good reports about stability and scalability at RH
    • When we decide which way we go, we need a plan who's doing the work (create the tests). Maintenance would depend on their structure.
    • Some Reddeer tests already upstreamed with WTP and Linuxtools
    • Lars: Where would the tests live? - In a common project accessible to all, or with the component they test ?
      • Lars, Dani: Since these are functional end-to-end, should live in their own repository, accessible by all committers.
    • AI Martin follow-up with Jubula. Try RCPTT on Windows and Linux. Give feedback till next week.
    • AI Alex share list of Reddeer tests existing at RH
    • AI Lars play with the Reddeer API
  • Dani: Update on PPC drop 32-bit, build 64-bit both le and be on RH VMs bug 512224
    • Alex: ppc64 currently built on Fedora 25; could build on REL7 if foundation requests a machine from RH Brno Farm AI Alex and Dani ask Denis
  • Dani: Neon.3 RC3 please keep an eye on builds
  • Dani: Away this Friday afternoon and whole next week

Feb 15, 2017


Feb 8, 2017 - Dani, Alex, Martin, Lars

  • Dani: bug 509412 - Adopt newer JSch for Platform 4.6.3
    • Alex is fully consumed with Webkit work for Oxygen
    • Need to move target platform to the latest Orbit in order to pick up the new JSch; that might cause other changes, since Orbit retention policy keeps only one version
    • Dani: Cherry-pick the new JSch recipe to the Orbit Maintenance Branch ?
    • Martin: Looks like in the past, new Orbit R-Builds were made for Maintenance ... but with the move of Orbit to git, it is unclear how branch builds for Neon.x would be made
  • Dani: Update on RT PMC (Equinox) coming to Platform
    • Tom Watson is in favor, reached out to Wayne for process
    • Potentially move entire IP log, but IP team is currently busy with CQs for Oxygen
  • Dani: Finally Unblocked on Java Language Server
    • Dirk from Microsoft signed the committer agreement
  • Dani: Scenarios for UI Testing
    • Initial Scenario from the PMC Notes can already be used; once that works, could extend to more
  • Dani: Sign Up for Security Rep
    • See Wayne's message to PMCs requesting PMC attendance
    • John Arthorne used to be the Security Rep;
    • Tasks: Follow the security mailing list; currently low volume but might get higher with IoT getting up speed
    • RESOLUTION: Alex agrees to sign up since already following security issues
  • Dani+Alex: libswt-gtk3 natives for PPC
    • Alex: Trying to limit the list of supported GTK versions to maximum 2; now at the latest GTK2 version
    • Currently, only Intel builds do both GTK2 and GTK3; PowerPC only does GTK2; the current hardware is unable to install REL7
    • Would prefer building at the Foundation over using custom company hardware; cost of hardware is around 50K
    • Primary Platforms should be built at the Foundation, secondary builds can be contributed ... should Linux-PPC be considered primary ?
    • Alex could easily provide a Fedora Platform, but it would be much newer than anything else that's around; ARM32 and ARM64 builds are already provided by Fedora
    • If somebody (IBM) needs an older version like CentOS, it would need to be provided by such adopters
    • AI Dani+McQ find out on priority of Linux-PPC (32bit, 64bit)
  • Alex: Update on Reddeer Testing Framework
    • A new version is being written, that no longer uses SWTBot underneath; there's a plan making this an Eclipse project
    • Engineers claim that after the rewrite it's a lot more stable recognizing widgets than on top of SWTBot
    • Martin has been looking at RCPTT which looks promising initially (easy to record tests) but unclear how stable such tests would be.
    • (call had to be closed at this point)

Feb 1, 2017 - cancelled


Jan 24, 2017 - Dani, Lars, Martin, Alex

  • Dani: New Jetty Version - Alex will look at it after FOSDEM (Feb 7)
  • Dani: javax.xml - Plan to announce removal from the feature on cross-project for M6
    • Cannot remove from the compile prerequisites, because plugins still on Java 5 need it
  • Alex: Reddeer Testing Framework
    • JBoss Tools uses Reddeer for overall user story validation
    • Lars: Why wasn't the framework added to SWTBot? -
      • Reddeer is not a framework, it's meant as a test harness library - smaller scope than SWTBot or Jubula
      • API Stability across multiple Eclipse versions is out of scope, thus created outside Eclipse for now
    • Martin: Any improvements in object recognition compared to plain SWTBot?
      • Reddeer tries to make use of the best mechanisms from SWTBot for object recognition (Finder, or callbots). That saves users from making newbie mistakes in SWTBot, but apart from that it's no better
      • No solution for external (non-Eclipse) dialogs ... that's not doable in SWT
    • Lars: EclipseSource people have been talking about RCPTT, it might be able to deal with native tooltips etc
    • Dani and Lars think Jubula is too heaviweight; might work for Integration Builds, but not for individual developer's builds. Local setup must be really easy. Also, setup looks non-trivial, needs resources to work on ... and, doesn't seem to add much benefit over SWTBot (assuming that Jubula agent doesn't hook into native libs like win32, Cocoa or GTK).
      • Alex: Especially integration in Maven may be hard.
    • Alex: Main limitation is that we don't have resources to actually create tests. Suggest reaching out for help asking people who would actually create tests ... then use whatever framework that people would like to use.
    • Dani and Lars won't be able to come up with a workflow scenario before end of next week --> Alex will come up with an idea since he has meetings on Monday.
      • Dani: Launch Eclipse, choose a workspace, open the package explorer, create a Java project (helloworld). Expand later, maybe into quick assist.
      • AI Alex to demo Reddeer in 2 weeks
      • AI Martin to try installing Jubula until next week
      • AI Martin ping Sergey re: asking for help on Platform/Resources.

Jan 17, 2017 - Dani, Sergey, Alex, Lars, Martin, McQ

  • Sergey: Merging platform.resources with platform.ui
    • Seems that more contributions to platform.resources are made by platform.ui committers, than others. Only 2 platform.resources committers seem active (and those two are actually platform.ui committers as well)
      • Opening up to Platform UI would encourage more people to actually contribute to resources. Strive for shared access on components that are interrelated
    • Dani: Merging into platform.ui would mean making Dani and Lars the leaders for signoff - that causes too much burden
    • Sergey: Platform/Team has actually more synergy with Platform/UI than resources
    • Alex: For gcc for example, there is a notion of "maintainers" per git repo for signoffs - that's independent of leadership or commit rights
      • McQ: Not sure how adding another role would simplify things ?
    • Sergey: The problem for signoffs is not is much the volume of commits, but the volume of failing tests
    • Alex: The manpower of Platform Resources is so low by now that even bug 509412 was not being filed ... existing committers do not really seem to "own" the project any more, this is an indication for merging
    • Martin: Can't see how merging into a bigger entity would resolve the "ownership" problem. Perhaps make interested contributors Platform/Resources committers instead, and see how to resolve the ownership issue on the smaller scope ?
    • McQ: Propose applying the recent simplification of gaining commit rights (for "known" committers)
  • Lars: Removing javax.xml
    • Was introduced only for very old JRE's , it is part of the JRE now
    • Martin: Might cause a uses constraint issue since not exporting the package with version any more, see http://njbartlett.name/2011/09/02/uses-constraints.html
    • Lars: With javax.annotation, that was solved with smart reexporting
    • McQ: OK with removal when we can confirm we're not breaking things
    • Dani: Whoever requested removal should double-check whether the API is the same.
    • AI Lars send a note to Tom Watson as the submitter.
  • Dani: Project Updates
    • Deadlines upcoming for CQs and APIs - see Endgame Plan, please provide feedback
  • McQ: UI Testing and Jubula

Jan 10, 2017 - Dani, Lars, Martin, Alex, Sergey, McQ - Special Guest: Alex Schladebeck

  • Dani: Platform Bits on Maven Central - Now available thanks to Stephan Herrmann
  • Dani: Board Committer Rep Elections opening
  • Dani: After David Williams' move, Wayne is looking for a Planning Council Chair; Fred Gurr has taken over Simrel Releng
  • Dani: objenesis (Mockito / Easymock dependency) and Java9 - will wait for how upstream is planning to address issues
  • Lars: Equinox move to Platform - no updates - AI Dani talk to Tom again
  • McQ / AlexS: Jubula
    • In the past, with lots of full-time Platform committers, sniff tests just happened automatically giving great confidence. Now, with more part-time work happening, there is more need for end-to-end workflow test automation to give quality confidence
    • Martin: On top of low-level JUnit tests, want at least very few highlevel end-to-end test to avoid "Eclipse looking silly". Need UI tests that are really stable (avoid random failures)
    • Dani: On EGit SWTBot tests, too many libraries to install, how to run on Gerrit HIPP slaves
    • AlexS: Original Jubula goal was to allow non-programmers write tests. Beginning 2015, the client API (for controlling the SUT) was separated from the front-end, was separated from the ITE. This allows writing UI tests programmatically, and avoids the need for a database.
      • For identifying components/actions to write tests, officially an object mapping needs to be done in the front-end (ITE, integrated test environment). But there may be a shortcut available for getting object IDs
      • Compared to SWTBot, Jubula is 90% blackbox (code highlevel actions rather than things like calling setText())
      • UI Tests typically take longer than unittests by definition (due to setup, teardown, ...), so might not be applicable for Gerrit triggers
      • Dani: Would like to run a first test on CentOS first, as part of the integration tests
      • Lars: Robustness of the tests? - SWT on Mac is problematic, but Linux and Windows should be OK.
      • Martin: Dependencies? - Around 5MB Libs, plus the agent needs to be installed (around 100MB). Once that is there, tests are stand-alone on a laptop. There might also be an "embedded agent" but not sure (AI AlexS to check)
      • Getting a first test up and running ? - Maybe around 1 week, depends on what to do ?
        • Lars: Would like something for quick access. Martin would something for "fresh download" (fresh config area + workspace) maybe combined with quick access.
      • AI AlexS send slides
      • AI Dani + Lars propose scenarios. Check possible owners of the initiative on Platform/UI side
      • AI AlexS check possible PlatformUI Contributors from Jubula team to get started

Archive