Skip to content

AGENTS.md

Latest commit

 

History

History
170 lines (117 loc) · 4.99 KB

AGENTS.md

File metadata and controls

170 lines (117 loc) · 4.99 KB

Orchestra

Orchestra is a mature, composable WPF shell and framework built on top of Catel. It provides a robust yet flexible Line of Business (LoB) shell, designed with best practices in mind, to jump-start the development of desktop applications.

Orchestra consists of the following projects:

  • Orchestra.Core — Core shell library with services, commands, and shared WPF functionality.
  • Orchestra.Shell.Ribbon.Fluent — Fluent Ribbon shell implementation.
  • Orchestra.Shell.TaskRunner — Task Runner shell implementation.

Critical Rules (Read First)

These rules are non-negotiable. Violating them causes broken builds, crashes, or downstream breakage.

1. Never Edit Generated Files

Files matching *.generated.cs, *.generated.xaml are auto-generated.

  • NEVER manually edit these files

2. ABI / API Stability

This project maintains stable ABI / API. Breaking changes break downstream apps.

Allowed Never
Add new overloads Modify existing signatures
Add new methods Remove public APIs
Add new classes Change return types

3. Tests Are Mandatory

Building alone is NOT sufficient. Run tests before claiming completion (see Commands).

4. Branch Protection (COMPLIANCE REQUIRED)

Direct commits to protected branches are a policy violation.

Repository Protected Branches
Orchestra master
Orchestra develop

Required workflow:

  1. Create a feature branch FIRST — Use naming convention: feature/issue-NNNN-description
  2. Make all commits on the feature branch — Never commit directly to protected branches
  3. Submit a Pull Request — Changes must be reviewed by a human before merging
# CORRECT — Always create a feature branch first
git checkout -b feature/issue-1234-fix-description

# NEVER DO THIS — Policy violation
git checkout develop && git commit  # FORBIDDEN

# NEVER DO THIS — Policy violation
git checkout master && git commit  # FORBIDDEN

The repository has protected branches that must be respected.

Commands

Single source of truth for all commands:

Task Command
Build dotnet cake --target=build
Test dotnet cake --target=test
Build and test dotnet cake --target=buildandtest

Architecture & Directories

Project Overview

Orchestra.Core                  => Core shell services and WPF utilities (cross-shell)
Orchestra.Shell.Ribbon.Fluent   => Fluent Ribbon shell
Orchestra.Shell.TaskRunner      => Task Runner shell
Orchestra.Shell.Shared          => Shared shell code
Orchestra.Tests                 => Automated test suite

Directory Guide

Directory / Pattern Editable? Notes
*.generated.cs No Leave as-is — auto-generated
*.generated.xaml No Leave as-is — auto-generated
deployment/ No Deployment / build scripts
doc/ Yes Documentation and images
src/Orchestra.Core/ Yes Core shell source
src/Orchestra.Shell.Ribbon.Fluent/ Yes Fluent Ribbon shell source
src/Orchestra.Shell.TaskRunner/ Yes Task Runner shell source
src/Orchestra.Tests/ Yes Automated tests

Writing Code

Anti-Patterns (Never Do This)

Anti-Pattern Why
Modifying method signatures ABI breaking
Manual edits to *.generated.cs, *.generated.xaml Overwritten on regenerate
Using default parameters in public APIs ABI breaking
Reformatting unrelated code Makes diffs harder to review
Skipping failing tests Unacceptable — tests must pass

Testing & Debugging

Running Tests

dotnet cake --target=test

Tests MUST Pass

NON-NEGOTIABLE: Tests must PASS before claiming completion.

  • Do NOT skip failing tests
  • Do NOT claim completion if tests fail
  • Do NOT use SkipException to work around failures

Writing Tests

  1. Use NUnit to write tests
  2. Create a Facts class for a feature
  3. Combine Pascal / Snake case for test methods (e.g. Feature_Does_Work)
[Test]
public void Feature_Does_Work()
{
    var result = 47 - 5;

    Assert.That(result, Is.EqualTo(42));
}

Philosophy: Tests FAIL when wrong, never skip (except missing hardware).

Debugging Methodology

  1. Establish baseline — What's the known-good state?
  2. One change at a time — Verify each change before proceeding
  3. Track changes in a table — Log what you changed and the result
  4. Platform differences are signals — If X works and Y fails, the difference IS the answer
  5. Revert if worse — Don't pile fixes on top of failures

Further Reading

Topic Document
Contributing guidelines CONTRIBUTING.md
Documentation portal https://opensource.wildgums.com
Catel (underlying framework) https://github.com/Catel/Catel
⚡