- The terminology we use is based on the Integrator Workflow
- Use a branching workflow similar to the one described at http://progit.org/book/ch3-4.html.
- Keep your own "main" branch in sync with the mainline repository's "main" branch. Specifically, do not push your own commits directly to your "main" branch.
- Any commit should pass all tests (see Running Tests).
- See the An Example section below for a full walk through
- In addition to a review of the algorithmic and logical changes in your contribution, it will be reviewed on a variety of levels including such things as style, documentation, tests, etc. While such review can appear tedious, these aspects are important for the long term success of the project.
- Please make sure you describe the changes you made to the code in the CHANGELOG.
- When you are ready to move changes from one of your topic branches into the "main" branch, it must be reviewed and accepted by another developer.
- You may want to review this tutorial before you make a pull request to the main branch.
- Look over the code.
- Check that it meets our style guidelines.
- Make inline review comments concerning improvements.
- Wait for the Continuous Integration service to show full test passage
- Click the green "Merge Pull Request" button
- Note: if the button is not available, the requester needs to merge or rebase from the current HEAD of the mainline "main" branch
- You can run the tests yourself using:
for Cyclus:
$ cyclus_unit_tests
for Cycamore:
$ cycamore_unit_tests
- DO NOT rebase any commits that have been pulled/pushed anywhere else other than your own fork (especially if those commits have been integrated into the blessed repository). You should NEVER rebase commits that are a part of the 'main' branch. If you do, we will never, ever accept your pull request.
As this type of workflow can be complicated to converts from SVN and very complicated for brand new programmers, an example is provided.
For the sake of simplicity, let us assume that we want a single "sandbox" branch in which we would like to work, i.e. where we can store all of our work that may not yet pass tests or even compile, but where we also want to save our progress. Let us call this branch "Work". So, when all is said and done, in our fork there will be three branches: "Master", "Develop", and "Work".
We begin with a fork of the mainline Cyclus repository. After initially forking the repo, we will have the main branch in your fork.
A fork is your copy of Cyclus. Github offers an excellent
tutorial on how to set one up. The rest of this
example assumes you have set up the "upstream" repository as cyclus/core. Note that git
refers to your fork as "origin".
First, let's make our "work" branch:
.../cyclus_dir/$ git branch work .../cyclus_dir/$ git push origin work
We now have the following situation: there exists the mainline copy of the main branch, there exists your fork's copy of the main and working branches, AND there exists your local copy of the main and working branches. It is important now to note that you may wish to work from home or the office. If you keep your fork's branches up to date (i.e., "push" your changes before you leave), only your local copies of your branches may be different when you next sit down at the other location.
Now, for the workflow! This is by no means the only way to perform this type of workflow, but I assume that you wish to handle conflicts as often as possible (so as to keep their total number small). Let us imagine that you have been at work, finished, and successfully pushed your changes to your Origin repository. You are now at home and want to continue working a bit. To begin, let's update our home's local branches:
.../cyclus_dir/$ git checkout main .../cyclus_dir/$ git pull upstream main .../cyclus_dir/$ git push origin main .../cyclus_dir/$ git checkout work .../cyclus_dir/$ git pull origin work .../cyclus_dir/$ git rebase main .../cyclus_dir/$ git push origin work
Perhaps a little explanation is required. We first want to make sure that this new local copy of the main branch is up-to-date with respect to the remote origin's branch and remote upstream's branch. If there was a change from the remote upstream's branch, we want to push that to origin. We then follow the same process to update the work branch, except:
- we don't need to worry about the upstream repo because it doesn't have a work branch, and
- we want to incorporate any changes which may have been introduced in the main branch update.
As time passes, you make some changes to files, and you commit those changes (to your local work branch). Eventually (hopefully) you come to a stopping point where you have finished your project on your work branch AND it compiles AND it runs input files correctly AND it passes all tests! Perhaps you have found Nirvana. In any case, you've performed the final commit to your work branch, so it's time to make a pull request online and wait for our main friends to review and accept it.
Sometimes, your pull request will be held by the reviewer until further changes are made to appease the reviewer's concerns. This may be frustrating, but please act rationally, discuss the issues on the GitHub space made for your pull request, consult the style guide, email the developer listhost for further advice, and make changes to your topic branch accordingly. The pull request will be updated with those changes when you push them to your fork. When you think your request is ready for another review, you can reopen the review yourself with the button made available to you.
A good description of a git workflow with good graphics is available at http://nvie.com/posts/a-successful-git-branching-model/
If you are going through a release of Cyclus and Cycamore, check out the release procedure notes here and on the website.