Steering Committee Meeting: October 17, 2013

Date: 
Thursday, October 17, 2013 - 10:00
  • Eclipse Platform migration from PDE build to CBI - Retrospective
  • EclipseCon Europe planning
Minutes: 

Platform Migration to CBI - Retrospective  
The Eclipse Platform project migrated to CBI during the Kepler release.  This retrospective describes their experience http://wiki.eclipse.org/Platform-releng/Kepler_2013_PDE_to_CBI_retrospective   David Williams met with the the LTS SC today to discuss. 

 
Intent is for the LTS SC to understand their experience, especially problems they ran into, to see if we can smooth the way for other teams in the future.  

Primary points:

  • Biggest surprise was that it required a lot more effort than they expected  - still not completely finished.  Tycho/Maven was a lot more complicated than expected, partly because of paradigm differences.  They are easy to use, quick to get running, but if you deviate at all from the straight path, it gets complicated quickly.   
  • Easy to get a build, but difficult to get a correct build.  The build will look like it succeeded, but didn't build what you thought, which might not be discovered until much later.
  • Recommendation - for LTS, it's not as simple as getting everyone to use the same build infrastructure.   Could have used a lot more education about Tycho/Maven.  It depends on each committer knowing a lot more about release engineering, which was frustrating for developers.   
  • The initial prototype would be provided by the Foundation, but the team didn't realize that it was just a prototype.  It didn't get them to a production level build infrastructure, so they had to invest more effort on their own.   
  • Tycho/Maven not well documented.  Expert users end up looking at the source code to figure out issues, but can't expect the more general user to do that. Not clear on what is or isn't supported.  When there is a problem with the build, it's very difficult to tell whether it's a problem with the build setup, or actually a Tycho bug.  Even for experienced users, this seems to be true.   
  • Tycho bugs - got some good help on some critical bugs, but a lot of bugs were deferred or considered enhancements.  Tycho is still an incubator.   
  • Started this effort with the idea that there were few people who understand the complicated PDE build, so moving to Tycho/Maven would be an improvement because a lot more people understand that - but it turned out not to be true.  Found few people who have good understanding of Tycho/Maven either.  If you run into a problem, you're sort of on your own.  

Discussion:

  • Is the platform build getting better?  Yes, but it's slow, don't have a lot of people devoted to it.  Each committer needs to know a lot more about the build process.   
  • Are the problems more with Tycho or Maven?   Tycho - it's less mature, less well documented.  But sometimes it's hard to distinguish between them.  Maven is not really a substitute for ANT, which developers are already familiar with.   But the Platform builds are working now, and not causing any downstream problems for consumers.
  • Can you provide a list of problems that are still open?  Some are identified in the retrospective document (run the queries).  To help other teams going forward, the biggest thing is to get better documentation, including a migration guide, and better documentation on what is and isn't supported.     
  • It would be good to write up a more formal article on lessons learned, to help other projects that are looking at this migration, to publish somewhere like Eclipse Magazine.  Last month Thanh Ha wrote a good article on how to get started with CBI - maybe this achieves what we want.  http://www.eclipse.org/community/eclipse_newsletter/2013/august/article3...  
  • We should point to this article as well as the Platform's retrospective on our LTS page.  They will help to set expectations, which is key.  The biggest issue for the Platform team was that their expectations were wrong on how much work would be involved with the migration.   If they had known the complexity, they could have planned it better in terms of resources and timeline.  

Planning for EclipseCon Europe

  • Posters
    • LTS Overview - done
    • LTS Infrastructure - finalizing the graphics
    • LTS Business Value - done
    • LTS Marketplace - done
    • LTS Members and maintenance providers poster - in progress
  • Charter update for Guest membership - unanimous approval by LTS SC.  Now also approved by Mike Milinkovich from Eclipse Foundation.
  • Enlisting maintenance providers - note was sent to the mailing list yesterday
  • LTS Website .
    • Updates to the About Us web page - if no further input, Jutta will publish her updates   
    • FAQs - we should add more content  - at a minimum, we can extract key points from our LTS documentation
    • Update the membership section of our charter
  • LTS Support Levels - Andrew is updating the paper to focus on self-service non-CBI projects vs. CBI projects, rather than release train vs. project-level builds.  We want the paper to be reworked before EclipseCon, for discussion at the face to face LTS SC meeting.
Dial In Numbers: 

This meeting is for LTS Steering Committee Members Only.

Location: 

This meeting is for LTS Steering Committee Members Only.

Attendees: 
  • Jutta Bindewald - SAP
  • Pat Huff - IBM
  • Lisa Lasher - IBM
  • Steve Francisco - IBM
  • Mike Haering - CA Technologies
  • Pierre Gaufillet - Airbus, Polarsys IWG
  • Andrew Ross- Eclipse Foundation
  • Ralph Mueller - Eclipse Foundation
  • David Williams - IBM, Eclipse Tools project
  • Denis Roy - Eclipse Foundation
  • Thanh Ha - Eclipse Foundation