Skip to content

Latest commit

 

History

History
42 lines (31 loc) · 2.67 KB

upstreaming.md

File metadata and controls

42 lines (31 loc) · 2.67 KB
Raw

Upstreaming packages

Getting your packages to upstream

The easiest way to do this is to create your own fork of Spack. Create a new branch starting from the latest develop branch and copy your changes there. Your package recipe and any patches should go under var/spack/repos/builtin/packages/.

You'll need to:

  • make sure a recent (preferably the latest) version is present in the recipe
  • if specifying a version linked to a git branch, ensure the version name and branch name are the same
  • make sure dependencies are correct and complete
    • give them the correct type. For Python dependencies type=("build", "run") in most cases
    • for Python packages: no need to specify py-wheel or py-pip, those come with the parent class
    • when adding multiple versions, keep in mind that different versions can have different dependencies, versions of dependencies, ...
  • ensure your code is properly formatted (black) and the header is correct (including the copyright year)
  • an upstreamed package should not depend on anything internal - it should install from your spack fork without any bluebrain-specific modifications

Create your pull request from your fork to the Spack develop branch and go through the review process.

Backporting upstreamed changes

Once your pull request has been merged, you'll need to backport your changes to our own spack repository. In order to do this, there's a script available: bluebrain/bin/upstream.sh. If you call it without arguments, it will explain how to use it.

Prepare to run by creating the branch you'd like to commit your changes to.

WARNING: if you have a local branch called upstream-develop it will be reset to match upstream/develop

The first argument is the name of the package you've upstreamed The optional second argument can be 'update' to update your upstream-develop branch from upstream/develop. Typically you'll update upstream only on the first package in your batch.

The script will:

  • if necessary, create the upstream remote pointing to github.com:spack/spack
  • if asked (through the 'update' argument), checkout upstream-develop tracking upstream/develop, git pull and go back to the branch you were on when you started
  • git checkout the upstream/develop version of your upstreamed package (the first argument) in var/spack/repos/builtin/packages
  • for whatever exists of your package under bluebrain/repo-{patches,bluebrain}/packages/: git rm and rm -rf

Now you can commit, push and file a pull request to our own Spack repository.