The Eclipse Project PMC has a weekly phone meeting every wednesday at 10.30am EST.
Apr 14, 2010: - John, Dani, Martin, Jeff
- Martin - Startup Performance tests: Cold start after reboot is 20% slower in 3.6m6 compared to 3.5.2 (while warm start is about the same)
- Manual test: Reboot a minimal WinXP system, then start into a fresh workspace with 1 JDT project (20 files) 1 open file in the editor.
- 28 sec in 3.5.2 but 36 sec in 3.6m6. Will file a bug tomorrow.
- John: There are 2 startup performance tests in the suite, but they are unreliable. In 3.6, changed the way tests are installed (director rather than dropins), thus baseline is not helpful.
- Jeff: Try have a look at bug 308157 Jarfile cache now limited to 100. Reason might just be cycling through more Jar's.
- John, Dani - Polish List
- John - e4 plan update. Waiting on McQ, wants an accurate list on what's graduating.
- Jeff - once something is in 4.0 you cannot remove in 4.x so better think twice before graduating.
- John - Eclipse SDK 4.0 has a minimal API exposed, most new stuff is under the covers so this is not so much of an issue.
- Jeff - More important to have Eclipse SDK 4.0 rock solid than have it feature complete. Do few things well rather than many things poorly.
- John - Self-hosting a day on Eclipse 4.0 without blocking issues! (But much to be polished, bugs, errors in the log etc).
- Jeff - Runtime SDK's: The label SDK is ambiguous. Want to install tooling + target platform together, but cannot do that today.
- Today, we use "SDK" for (a) tools+source+docs, or (b) runtime+source+docs. None of both is really an SDK.
- Better call the target stuff just "targets".
- Developer docs as part of the tooling is wrong ... should be associated with targets instead.
- John: p2 does have the ability to install into multiple profiles (plan = multiple profiles)... might be (mis)used for this, is it a hack?
- Jeff: Much target provisioning was deferred off 3.6
Apr 7, 2010: - McQ, Dani, Martin, Jeff, John
- Builds - short call
Mar 31, 2010: - McQ, Dani, Martin, John
- Martin - EclipseCon Report
- General industry trend pointing up (as perceived on the exhibition floor); e4 rover great success! other strong topics included build (b3, buckminster, maven / tycho / nexus, athena...)
- e4 message in general very well positioned and received; git / egit was another hot topic
- Modeling and RT projects in an up trend, other projects seem to go slightly down in terms of Community interest as well as commercial involvement
- API Tutorial very well received, Martin going to work on a "Wiki" version of checklists and guidance, will notify AC when done
- John - Eclipse 4.0: Timing for graduating e4 incubation material into the Eclipse proper
- We cannot ship an Eclipse SDK out of the e4 project
- Want a clear message what Eclipse 4.0 is... probably "includes incubating components" like some EPP packages
- AI John talk to Mike and Ian
Mar 17, 2010: - McQ, Dani, Martin
- McQ: git vs CVS: Should there be contributions by Platform on Egit?
- Martin: Don't know how well egit proceeded recently, Boris might know more... important point is that the major workflows are perfect. Looks like the major workflows have been identified already.
- Martin: WebkitGTK / MiniBrowser: In addition to the recent WebkitGTK discussion, perhaps work on a "Minibrowser" API that can live with published frozen Mozilla API only? Many apps may not need the full feature-richness of today's Browser.
- McQ unsure whether this is worthwile, since all industry trends go towards more web integration. AI Martin follow up with Grant
- Dani: Performance and Polish passes
- All teams need to fix the issues that Frederic finds. M7 is the performance and polish pass. Prioritize items.
- Dani: Freeze Plan
- Suggest a 2-day test pass (mon/tue) before the RC's, ie move 1 day from RC2 into M7
Mar 10, 2010: - McQ, Jeff, Martin, John
- John: Provisional API guidelines (removing the requirement to have "internal" in the package name) - important for e4 which will have a lot of provisional API.
- See also bug 261874 and Wiki Provisional API Guidelines Update Proposal
- "Old School" wanted to make provisional API deliberately painful. Migrating "provisional" to real without renaming will make breakage more subtle
- The game has changed: Adding x-internal, friends and API Tooling works much better than before, making it clear where API is provisional
- Martin: Much in favor of this, do we have any markup beyond x-internal for (a) making provisional API more explicitly visible or (b) work on a smaller granularity such as just a class?
- McQ: granularity smaller than package makes it too easy to pollute API
- Jeff: would like x-api-status:=provisional markup instead of x-internal:=true ... better do it right than half-baken. Could probably come to a fairly fast consensus on MANIFEST markup
- Resolution Lets agree now that x-internal is sufficient for provisional API, and discuss further approvements in parallel. AI John to search existing bugs about provisional API markup and initiate a discussion on the eclipse-pmc mailing list.
- Jeff: Target Provisioning and PDE: Target components in Galileo (which cannot be installed into the host) - came up with sort of a hack which still confuses users
- Want just a little bit help in PDE to make target provisioning just a little bit better .. a number of PDE bugs related to this, many been deferred .. there will be new bugs coming to capture what can be done in the short term
- Resolution: will mark up those new bugs where they request PMC involvement
- McQ: Build Quality: There is traditionally a drop in quality around this time of year (API Freeze and Eclipsecon), plus infrastructure problems. It's not really bad but we need to be careful now.
- John: Resist the urge to put in extra fixes. We are past the test pass. Quality over function, especially now.
Mar 3, 2010: - Dani, Martin, McQ
- Martin: Remove org.eclipse.update.configurator and related bundles from SDK? As per bug 304505 it makes Eclipse slow even when off.
- Dani: Might be more than a packaging issue, somebody would have to invest
- McQ: Should fix the Performance issue at any rate, regardless of other issues.
- John sick, Dani vacation next week.
Feb 24, 2010: - Dani, Martin, McQ
- Dani: Remove Java 7 support as a plan item due to (a) legal reasons and (b) Java 7 not being finished when Eclipse 3.6 ships
Feb 17, 2010: - Martin, Dani, John
- Martin: bug 196337 Pushing CDT Spawner into the Platform?
- John, Dani: Platform could only accept it when there is use for it in the SDK. Otherwise it would just bloat the Platform
- Recommended best practice: Keep Spawner living in CDT, but put it into a separate bundle such that it can be used by others out of Helios or other p2 Repos
- The Nexus Project, which was once meant to collect such micro functionality to be shared between projects was never successful. Similar requests (e.g. faceted projects) are consumed as individual bundles through p2 today, no matter in which project they have their home where they are developed.
- Martin: bug 301563 Fast project import from snapshot data - UI or not?
- John: Want some UI in the Platform in order to test it more easily, e.g. an export wizard
- John: Webkit
- Foundation is considering allowing LGPL for exempt prereqs, but not for works-with .. missing a policy for dealing with LGPL works-with
- John: Our original reason for marking works-with is that the SWT browser can use either Mozilla or WebKit. However our long term direction is WebKit-only due to brittleness of the Mozilla API which keeps breaking us. There is an increasing number of distros bundling these WebKit libraries so there is a reasonable chance going forward that the library will already be present on the user's machine.
- Martin: Exempt works-with (optional) prereq is perfectly fine for Webkit, since there is a chance it's already there on a Platform (similar to Mozilla)
- Classifying it as such makes most sense for Product builders, who look at the prereqs to understand what they need to bundle with their Eclipse based product.
- PMC agreed to reclassify these libraries as exempt pre-req.
Feb 10, 2010: - John, Dani, McQ
- We agreed to list WebKitGTK and libsoup 2.4 as works-with prerequisites
- We need to find consensus on bug 243582 (embedding source info in binaries)
- Discussed moving Ubuntu version on the plan from 9.04 to 10.04. It is too early to make this decision because release candidates of 10.04 are not yet available, but we will continue to monitor it and make the decision to move up (or not) later in the 3.6 cycle
Feb 03, 2010: - Dani, Martin
- Dani: bug 301563 - Fast project import from snapshot data
- Has the feature been verified to really return the expected performance gain? - Martin: Yes, Cisco reports 10 minute -> 5 seconds improvement by using the feature on project import on their view (65000 files)
- Is the feature valuable without Index contributions from JDT / CDT? - Martin: Yes, even "plain" projects benefit when there are linked resources pointing to web folders through RSE/EFS since they can be browsed immediately and refresh can be reduced to what's really needed. But most benefit is gained when there is also a shared index to be imported for immediate use.
- Dani proposed checkin into a branch for easier merge / review - Martin: Will start working with patches
- AI Martin: Contact Sharon regarding IP review (reserve a slot)
Jan 27, 2010: - John, Dani, McQ, Martin
- Dani: Markus Keller taking over JDT UI
- John: M6 Splashscreen for Eclipsecon: bug 297355
- McQ: Removing Builds - SWT needs Linux-Motif, so only WPF about to be removed
- In discussions with Microsoft, it turned out that WPF is not required to get full Windows 7 experience under Win32
- XAML for styling was meant to be a cool idea but never got flying
- Socialize people with this -- find whether people are inerested in contributing on this, if yes then we should support them
- Still working the IBM approval process for travelling to Eclipsecon
- Avoid merging major feature work after a milestone's Tuesday test pass
Jan 20, 2010: - John, Dani, McQ, Martin
- McQ: Contacted Steve N, still interested but unlikely to get more energy for investing into Eclipse
- John: 3.5.2 test pass tomorrow, but yesterday's I-build been a mess
- McQ: Message about supporting Open JDK in a blog ... status should be "nice that it works but it's not a reference platform"
Jan 13, 2010: - John, Dani, McQ, Martin
- McQ: U Manitoba students to help with technical communication (documentation, website, ...) for e4
- Dani: New way of contributing Capabilities for Helios... are we OK? - John: yes, Platform Capabilities are in the SDK feature
- FYI: Incubating projects are
- Martin: Documenting the Platforms we routinely test on
- Unittest / Perftest machines are know. When John updated the Reference Platform doc, he made sure that he knows at least one committer on each platform
- A poll to know what Platform(s) are actively used (by committers) on milestone granularity would be very helpful - John going to set that up
Jan 6, 2010: - John, Dani, McQ, Jeff
- Agreed on 3.5.2 freeze plan
- Note RC2 is a week earlier to avoid colliding with Helios M5 week
- Discussed Helios plan updates 2 bug 298200
- Update Java 7 plan item to indicate only working on publicly available bits. Some progress made on getting access to specs but going slowly.
- Update reference JRE's to latest version of each JRE
- Jeff will be away for next six weeks (vacation)
- McQ to contact Steve to see if he still wishes to remain on PMC