4.4 KiB
Contribution guidelines
Firstly, thank you for your interest in contributing to this project! This project is powered by the community and relies on hackers across the globe to report bugs, send suggestions and contribute code to move this project forward.
Ways to contribute
Coding guidelines
Special comments
FIXME(@<username>)
: Use this when you have made an implementation that can be improved in the future, such as improved efficiencyHACK(@<username>)
: Use this when the code you are using a temporary workaroundTODO(@<username>)
: Use this when you have kept something incompleteUNSAFE(@<username>)
: Use this to explain why theunsafe
block used is safe
Comment style
We prefer to use a 'mixed' commenting style: a cross between the C and C++ styles. Generally, if a comment can be explained in one-line, then follow the C++ commenting style. In other cases, use the C style.
Formatting
-
All Rust code should be formatted using
rustfmt
-
Imports in Rust source files are to be merged into a single
use
block. For example, instead of:use devtimer::SimpleTimer; use devtimer::ComplexTimer; use serde::{Serialize, Deserialize}; use crate::something;
this mode of styling is preferred:
use { devtimer::{SimpleTimer, ComplexTimer}, serde::{Serialize, Deserialize}, crate::something, };
Parts of the project
The main parts (ignorning CI scripts, stress test suite, test harness and custom compiler macros) are:
cli
: interactive shellskysh
server
: server daemonskyd
sky-bench
: benchmark tool
The pkg
folder contains scripts used to build packages for different targets. The examples
folder
contains example configuration files.
Branches
The next
branch is the kind of stable branch which contains the latest changes. However, for most purposes, you should always download sources from the tags.
Pushes are made directly to next if they don't change things significantly; for example, changes in documentation comments and general optimizations. If however the changes are huge, then they must be created on a separate branch, a pull request opened, the CI suite run and finally merged into next.
Pull request guidelines
Typo correction or doc updates
The creation of superflous merge requests is generally discouraged. Such examples include the creation of multiple PRs to correct single typos, update comments or update docs.
It would be far better if you could fix a considerable subset (if not all) of these issues in one pull request (it's fine to create multiple commits in the same PR).
This is because we don't want to utilize compute capacity or multiply our git history for changes
that do not have any behavioral impact. It is recommended that you open up a single PR and correct a substantial subset of these inconsistencies (if not all), while also adding [skip ci]
or [ci skip]
in your
commit messages to avoid triggering the workflows.
Steps
- Fork the repository
- Make your changes and start a pull request
- Sign the CLA (if you haven't signed it already)
- One of the maintainers will review your patch and suggest changes if required
- Once your patch is approved, it will be merged into the respective branch
- Done, you're now a contributor 🎉!
Development environment setup
Skytable uses a Makefile for builds and running the test suite along with a number of tools. To set these up:
- Install the latest Rust toolchain (stable)
- Install a C Compiler, Make, Perl and the libssl-dev package on platforms where they are required
- On Windows:
- You might need to perform additional steps to add
perl
to ourPATH
variable - You may also need to set up powershell if it isn't set up already as it is used by the test script
- You might need to perform additional steps to add
Building
- For debug (unoptimized) builds:
make build
- For release (optimized) builds:
make release
TIP: You can explicitly specify a target triple by setting a
TARGET
environment variable
Testing
Testing is simple: just run this:
make test
NOTE: Make sure port 2003 and 2004 are not used by any applications. Also, make sure your own instance isn't running on any of these ports; if that is the case, you might end up losing data due to conflicting entity names! The test suite creates multiple spaces and some models within it to run all the tests.