STEM/Git Migration

From Eclipsepedia

Jump to: navigation, search

Contents

Background

In April 2013, the STEM Project initiated its move from Subversion (SVN) to Git. This process is expected to take about a month and impacts all users of the STEM source code.

The Eclipse Foundation has encouraged projects to migrate from CVS and SVN, including shutting down the CVS servers late last year.


About Git

There are countless resources for learning about Git. Some useful ones:

Details

Our migration is being tracked in Bugzilla under Bug 404584

Migration Timeline

  • March 21: Initial discussion on committer call re/ moving to Git (complete)
  • March 28: Further discussion on committer call re/ moving to Git. Committer motion to develop migration plan and initiate process passes. Most committers have experience with Git.
  • March 29: Request Git migration from Eclipse Webmaster
  • April 1: Create test repositories and publish to Eclipse for testing, begin migrating infrastructure
  • April 2: Public announcement of migration with STEM 2.0 milestone 1
  • April: test runs of repository split and import, prepare infrastructure (build scripts, CI server, project sets, etc) and update documentation
  • (Tentative) April 30: Freeze SVN, execute import, roll out new Git repositories
  • (Tenative) May 1: Declare migration complete

Repository Layout

Due to the nature of distributed versioning system like Git, an entire repository must be "cloned" onto a developer's machine before it can be worked with. While Git's compression is impressive, it still requires significant network transfer time and local storage for large repositories - even if you only need to work with a small subset. In STEM's case, we store a large amount of denominator in our SVN repository, on the order of 2 GB.

STEM has will split our code base into multiple repositories that can be fetched independently as needed. The proposed repositories are:

  • org.eclipse.stem.git - Core repository, including majority of source code and epidemiological modeling components
  • org.eclipse.stem.data.git - Common denominator data, including administrative borders and human population data
  • org.eclipse.stem.data.earthscience.git - Less common Earth Science denominator data

The idea is you can work independently with STEM without needing the additional repositories and they can be added as needed (although most developers will benefit from having the org.eclipse.stem.data repository cloned).

In July 2011, STEM did a logical refactor that changed the layout of our SVN repository from a flat to hierarchical structure reflecting components. We're happy with this layout and will not be changing the overall structure of our repository.