Skip to content

Latest commit

 

History

History
140 lines (117 loc) · 7.22 KB

CONTRIBUTING.md

File metadata and controls

140 lines (117 loc) · 7.22 KB
Raw

Contributing

Please help us by finding problems, discussing on the mailing lists, contributing documentation, bug fixes or even features. Below are some brief guidelines.

Reporting a problem

Please use our issue-tracker: https://github.com/SyneRBI/SIRF/issues

Submitting a patch

For contributing any code or documentation that is non-trivial, we require a signed Contributor License Agreement, stating clearly that your conributions are licensed appropriately. This will normally need to be signed by your employer/university, unless you own your own copyright. You will have to do this only once. Please contact us for more information.

Please keep a patch focused on a single issue/feature. This is important to keep our history clean, but will also help reviewing things and therefore speed-up acceptance.

Process

This is our recommended process. If it sounds too daunting, ask for help.

  1. Create a new issue (see above). State that you will contribute a fix if you intend to do so.
  2. Create a fork on github and work from there.
  3. Create a branch in your fork with a descriptive name and put your fixes there. If your fix is simple you could do it on github by editing a file, otherwise clone your project (or add a remote to your current git clone) and work as usual.
  4. If your change is important, add it to CHANGES.md and even UserGuide.md or other documentation files.
  5. Use well-formed commit messages for each change (in particular with a single "subject" line followed by an empty line and then more details).
  6. Push the commits to your fork and submit a pull request (PR) (enable changes by project admins.) Be prepared to add further commits to your branch after discussion. In the description of the PR, add a statement about which Issue this applies to using a phrase such that github auto-closes the issue when merged to master.
  7. Be prepared to add further commits to your branch after discussion. Please by mindful about the resources used by our Continuous Integration (CI) workflows:
  • Group your commits and only push once your code compiles and tests succeed on your machine
  • Use specific keywords in the body of the last commit that you push to prevent CI being run:
    • [ci skip] skips all CI runs (e.g. when you only change documentation, or when your update isn't ready yet)
    • [actions skip] does not run GitHub Actions, see here.
    • [travis skip] does not run Travis-CI, see here.
  1. After acceptance of your PR, go home with a nice warm feeling.

Suggested reading: https://help.github.com/articles/fork-a-repo/, https://git-scm.com/book/en/v2/GitHub-Contributing-to-a-Project or https://guides.github.com/activities/forking/.

A note on copyright dates and notices (and licenses)

(Almost) all SIRF files start with a copyright and license header. Please do this for your files as well. If you modify an existing file, you need to make sure the copyright header is up-to-date for your changes (unless it's a trivial change).

If you copied code from somewhere, you need to preserve its copyright date/notice. If you copied non-SIRF code, you need to make sure its license is compatible with the SIRF license, and indicate clearly what the license of the copied code is (and follow its terms of course).

In addition, you might need to add yourself to NOTICE.txt.

Project rules

  • Only one official, stable, up-to-date branch: master
    • Essentially "latest stable beta version with no known bugs since the last official release version"
    • Never knowingly add a bug to master
  • Any work-in-progress commits should be in their own branches.
  • GitHub assigns a unique number to each issue, c.f. the issue-tracker.
  • A pull request (PR) is an issue with an associated branch, c.f. pull-requests. Even for "internal" development, we prefer a PR for a branch to allow review and discussion.
  • Branches and PRs are kept small (ideally one 'feature' only) and branch from master, not from another branch, unless required. This allows commenting/improving/merging this branch/PR independent of other developments.
  • Discussions on issues and PRs are forwarded to the SyneRBI-DEVEL@jiscmail.ac.uk mailing list daily.
    • Forwarded from github via the googlegroup, which is also a backup in case github dies.
  • Contributions of new features should also update documentation and release notes. After version 1.0, this documentation needs to state something like "introduced after version 1.xxx".
  • We prefer issues to be opened via github due to the following reasons:
    • Ensures issues will never get lost in emails
      • Facilitates issue status tracking
    • Allows focused comments/discussion
      • Easy cross-referencing of related issues, PRs, and commits
    • The mailing list gets notified within 24 hours.

Submodules

[...] A submodule is a repository embedded inside another repository. The submodule has its own history [...]

This project uses submodules to point to infrequently changed/large datasets.

  1. How to pull updates to SIRF
# first time
SIRF$ git clone https://github.com/SyneRBI/SIRF --recursive
# subsequently
SIRF$ git pull
SIRF$ git submodule update --init --recursive
# or
SIRF$ git pull --recurse-submodules
  1. How to push updates to the data submodule in SIRF
SIRF$ cd data
SIRF/data$ # create/change some files
SIRF/data$ git commit -m "add some data" && git push && cd ..
SIRF$ git add --force data  # force required for safety (see 3 below)
SIRF$ git commit -m "update submodule" && git push
  1. What happens when switching branches
    • When switching branches in SIRF, the data submodule will not be touched. git submodule update needs to be manually run in order to ensure the correct submodule commit is also checked out
    • WARNING: This is why git add --all is bad practice (if you forget to update submodules when needed, then committing all modifications will include the current wrong submodule commit)
  2. What happens if upstream changes conflict
    • The data submodule may be force-pushed to by maintainers to reduce its size (removing unneeded data from git history). Consequentially SIRF$ git status may indicate changes in the data submodule despite a user not having touched it. In order to overwrite the local outdated submodule:
SIRF$ cd data
SIRF/data$ git fetch origin && git reset --hard origin/master && cd ..