The Postgres Operator is an open source project primarily authored and maintained by the team at Crunchy Data. All contributions are welcome: the Postgres Operator uses the Apache 2.0 license and does not require any contributor agreement to submit patches.
Our contributors try to follow good software development practices to help ensure that the code that we ship to our users is stable. If you wish to contribute to the Postgres Operator, be it code or documentation, please follow the guidelines below.
Thanks! We look forward to your contribution.
All ongoing development for an upcoming release gets committed to the
master branch. The master branch technically serves as the "development"
branch as well, but all code that is committed to the master branch should be
considered stable, even if it is part of an ongoing release cycle.
All fixes for a supported release should be committed to the supported release
branch. For example, the 4.3 release is maintained on the REL_4_3 branch.
Please see the section on Supported Releases for more information.
Ensure any changes are clear and well-documented. When we say "well-documented":
-
If the changes include code, ensure all additional code has corresponding documentation in and around it. This includes documenting the definition of functions, statements in code, sections.
-
The most helpful code comments explain why, establish context, or efficiently summarize how. Avoid simply repeating details from declarations,. When in doubt, favor overexplaining to underexplaining.
-
Code comments should be consistent with their language conventions. For example, please use GoDoc conventions for Go source code.
-
Any new features must have corresponding user documentation. Any removed features must have their user documentation removed from the documents.
-
Do not submit commented-out code. If the code does not need to be used anymore, please remove it.
-
While
TODOcomments are frowned upon, every now and then it is ok to put aTODOto note that a particular section of code needs to be worked on in the future. However, it is also known that "TODOs" often do not get worked on, and as such, it is more likely you will be asked to complete the TODO at the time you submit it. -
Write clear, descriptive commit messages. A guide for this is featured later on in the documentation.
Please provide unit tests with your code if possible. If you are unable to provide a unit test, please provide an explanation as to why in your pull request, including a description of the steps used to manually verify the changes.
Ensure your commits are atomic. Each commit tells a story about what changes are being made. This makes it easier to identify when a bug is introduced into the codebase, and as such makes it easier to fix.
All commits must either be rebased in atomic order or squashed (if the squashed commit is considered atomic). Merge commits are not accepted. All conflicts must be resolved prior to pushing changes.
All pull requests should be made from the master branch unless it is a fix
for a specific supported release.
Once a major or minor release is made, no new features are added into the
release branch, only bug fixes. Any new features are added to the master
branch until the time that said new features are released.
Commit messages should be as descriptive and should follow the general format:
A one-sentence summary of what the commit is.
Further details of the commit messages go in here. Try to be as descriptive of
possible as to what the changes are. Good things to include:
- What the changes is.
- Why the change was made.
- What to expect now that the change is in place.
- Any advice that can be helpful if someone needs to review this commit and
understand.
If you wish to tag a Github issue or another project management tracker, please do so at the bottom of the commit message, and make it clearly labeled like so:
Issue: #123
Issue: [ch1234]
All work should be made in your own repository fork. When you believe your work is ready to be committed, please follow the guidance below for creating a pull request.
Ongoing work for new features should occur in branches off of the master
branch. It is suggested, but not required, that the branch name should reflect
that this is for an upcoming release, i.e. upstream/branch-name where the
branch-name is something descriptive for what you're working on.
While not required, it is recommended to make your branch name along the lines
of: REL_X_Y/branch-name where the branch-name is something descriptive
for what you're working on.
Overall, release tags attempt to follow the semantic versioning scheme.
"Supported releases" (described in the next section) occur on "minor" release
branches (e.g. the x.y portion of the x.y.z).
One or more "patch" releases can occur after a minor release. A patch release is used to fix bugs and other issues that may be found after a supported release.
Fixes found on the master branch can be backported to a support release
branch. Any fixes for a supported release must have a pull request off of the
supported release branch, which is detailed below.
When a "minor" release is made, the release is stamped using the vx.y.0 format
as denoted above, and a branch is created with the name REL_X_Y. Once a
minor release occurs, no new features are added to the REL_X_Y branch.
However, bug fixes can (and if found, should) be added to this branch.
To contribute a bug fix to a supported release, please make a pull request off
of the supported release branch. For instance, if you find a bug in the 4.3
release, then you would make a pull request off of the REL_4_3 branch.
When a release is no longer supported, the branch will be renamed following the
pattern REL_X_Y_FINAL with the key suffix being FINAL. For example,
REL_3_2_FINAL indicates that the 3.2 release is no longer supported.
Nothing should ever be pushed to a REL_X_Y_FINAL branch once FINAL is on
the branch name.
At any point in the release cycle for a new release, there could exist one or more alpha, beta, or release candidate (RC) release. Alpha, beta, and release candidates should not be used in production environments.
Alpha is the early stage of a release cycle and is typically made to test the
mechanics of an upcoming release. These should be considered relatively
unstable. The format for an alpha release tag is v4.3.0-alpha.1, which in this
case indicates it is the first alpha release for 4.3.
Beta occurs during the later stage of a release cycle. At this point, the
release should be considered feature complete and the beta is used to
distribute, test, and collect feedback on the upcoming release. The betas should
be considered unstable, but as mentioned feature complete. The format for an
beta release tag is v4.3.0-beta.1, which in this case indicates it is the
first beta release for 4.3.
Release candidates (RCs) occur just before a release. A release candidate should
be considered stable, and is typically used for a final round of bug checking
and testing. Multiple release candidates can occur in the event of serious bugs.
The format for a release candidate tag is v4.3.0-rc.1, which in this
case indicates it is the first release candidate for 4.3.
After a major or minor release, no alpha, beta, or release candidate releases are supported. In fact, any newer release of an alpha, beta, or RC immediately deprecates any older alpha, beta or RC. (Naturally, a beta deprecates an alpha, and a RC deprecates a beta).
If you are testing on an older alpha, beta or RC, bug reports will not be accepted. Please ensure you are testing on the latest version.
We greatly appreciate any and all testing for the project. When testing, please be sure you do the following:
- If testing against a release, ensure your tests are performed against the latest minor version (the last number in the release denotes the minor version, e.g. the "3" in the 4.3.3)
- If testing against a pre-release (alpha, beta, RC), ensure your tests are performed against latest version
- If testing against a development (
master) or release (REL_X_Y) branch, ensure your tests are performed against the latest commit
Please do not test against unsupported versions (e.g. any release that is marked final).
There are several ways to help with the testing effort:
- Manual testing: testing particular features with a series of manual commands or custom scripts
- Writing unit tests: testing specific sections of the code
- Writing integration tests: automatically testing scenarios that require a defined series of steps, such as end-to-end tests
- Environmental & workload testing: testing the code against specific workloads, deployment platforms, deployment models, etc.