Skip to content

Latest commit

 

History

History
103 lines (74 loc) · 6.24 KB

CONTRIBUTING.md

File metadata and controls

103 lines (74 loc) · 6.24 KB

Contributing

Welcome! We're excited you want to take part in the CrowdStrike community!

Please review this document for details regarding getting started with your first contribution, tools you'll need to install as a developer, and our development and Pull Request process. If you have any questions, please let us know by posting your question in the discussion board.

Table of Contents

Code of Conduct

Please refer to CrowdStrike's general Code of Conduct and contribution guidelines.

How you can contribute

Working with the Code

To simplify and standardize things, we use GNU Make to execute local development tasks like building and linting the code, running tests and benchmarks, etc.

The table below shows the most common targets:

Target Description
test Runs all tests for the root module and the nested examples module
bench Runs the benchmarks in the nested examples module
generate Generates the Go code in the nested examples module by invoking the appropriate protoc plug-ins
regenerate Regenerates the Go code in the nested examples module by removing the existing generated code and build artifacts then invoking the appropriate protoc plug-ins

Prerequisites

We have tried to minimize the "extra" work you'll have to do, but there are a few prerequisites:

  • GNU Make
  • Go 1.17 or higher
  • The protoc Protobuf compiler

As the exact details of how to install these tools varies from system to system, we have chosen to leave that part to you.

Other Build-Time Tools

All other build-time tooling is installed into a project-level bin/ folder because we don't want to impact any other projects by overwriting binaries in $GOPATH/bin. To accomodate that, Makefile prepends the local bin/ folder to $PATH so that any Make targets will find the build-time tools there. If you choose to not use our Make targets, please ensure that you adjust your environment accordingly.

Another twist: we need to generate code using both the "old" (V1 Protobuf API) and "new" (V2 Protobuf API) versions of protoc-gen-go, so we explicitly rename the resulting binaries in the project-local bin/ folder to protoc-gen-go-v1 and protoc-gen-go-v2, respectively, and use the --plugin option to protoc to point the --go_out=... parameter to the "correct" binary when generating code.

Commit Message Formatting and Hygiene

We use Conventional Commits formatting for commit messages, which we feel leads to a much more informative change history. Please familiarize yourself with that specification and format your commit messages accordingly.

Another aspect of achieving a clean, informative commit history is to avoid "noise" in commits. Ideally, condense your changes to a single commit with a well-written Conventional Commits message before submitting a PR. In the rare case that a single PR is introducing more than one change, each change should be a single commit with its own well-written message.

Benchmarks

The primary purpose of this library is to provide a runtime-independent Protobuf API, but allowing for optimal performance is a close second. In fact, we were able to beat both Google's and Gogo's runtime performance by as much as 40% by using our protoc-gen-fastmarshal code generator.

With that in mind, any changes that negatively impact these benchmarks will likely be rejected. At a minimum, changes that decrease performance will receive pushback and will need to be justified explicitly.

Pull Requests

All code changes should be submitted via a Pull Request targeting the main branch. We are not assuming that every merged PR creates a release, so we will not be automatically creating new SemVer tags as a side effect of merging your Pull Request. Instead, we will manually tag new releases when required.

Code Coverage

While we feel like achieving and maintaining 100% code coverage is often an untenable goal with diminishing returns, any changes that reduce code coverage will receive pushback. We don't want people to spend days trying to bump coverage from 97% to 98%, often at the expense of code clarity, but that doesn't mean that we're okay with making things worse.