Haiku Format Bot

The Haiku project has coding guidelines which includes the coding style. Part of the code review process is to review and enforce that style. In order to make it easier for contributors, the haiku-format tool has been built. It extends clang-format and codifies the Haiku style.

The goal of this tool is to apply the haiku-format tool to changes to the Haiku code base sent in for code review, and if there are style fixes necessary, then post those as suggestions to Gerrit so that users can further decide what to do next.

Currently, the tool is considered to be experimental, both in the sense that it is likely that the haiku-format configuration needs to be further tweaked over time, as well as the newness of this tool itself.

About haiku-format

For formatting, a customized clang-format called haiku-format is used. It is currently maintained on Github.

General Design of the Format-Check-Bot

The tool's functionality is implemented in the following modules:

1. The core.py module implements the main logic for evaluating a specific Gerrit change. That means it fetches a change from Gerrit, runs haiku-format over the changes, and posts them back to Gerrit.
2. The gerrit.py module implements the logic to exchange data with Gerrit. GET requests are converted into the models that are used by other parts of this tool. POST requests are serialized from the internal models into the json format Gerrit expects.
3. The llvm.py module is based on the LLVM code that helps users run clang-format over a diff. It contains the tools to identify which parts of a file have been patched, and it has the logic of running clang-format/haiku-format, and then capturing which segments should be reformatted.
4. The models.py module contains the internal models that codify business logic when processing Gerrit changes, as well as the data structures that are sent to Gerrit when posting the reviews.
5. The runners.py module implements the logic to select changes for Gerrit that need to be formatted, apply the formatting, and submit the reviews. It includes error handling, in order to let the system skip changes that cannot be reformatted due to a bug in this bot, or at any other layer.

Development and Testing

In order to do local development or testing, you require:

• Python 3 (version 3.11 is recommended)
• Requests module
• A build of haiku-format in a location in PATH

In order to use the functionality to reformat a Gerrit change, you run the following command from the root of the project:

# Run the haiku format bot on change 1000. The result is stored in a file called `review.json`
python3 -m formatchecker.core 1000

# Run the haiku format bot for a single and submit it
export GERRIT_USER=user
export GERRIT_PASSWORD=http_password_for_user
python3 -m formatchecker.core --submit 1000

# Run the haiku format bot on all changes in the past 3 days. If they have been checked before, they are excluded.
export GERRIT_USER=user
export GERRIT_PASSWORD=http_password_for_user
python3 -m formatchecker.runner --submit

The test suite can be run using python3 -m unittest in the root of the project.

Docker Build

The repository comes with a Dockerfile packaged. It is compatible with docker and podman. The build runs in two steps. The first step builds a specific version of haiku-format, which is then copied over to an image that contains the format check bot.

Some example commands:

# Build the image
podman build -t format-check-bot:dev .

# Run the format check for change 1000
podman run format-check-bot:dev python3 -m formatchecker.core 1000

# Run the format check for change 1000 and submit
podman run -e GERRIT_USERNAME=user -e GERRIT_PASSWORD=http_password_for-user format-check-bot:dev python3 -m formatchecker.core --submit 1000

# Run the format check in daemon mode
podman run -e GERRIT_USERNAME=user -e GERRIT_PASSWORD=http_password_for-user format-check-bot:dev python3 -m formatchecker.runner --daemon --submit