Advanced Development with FastLED

LLDB Debugging on Unit Tests

Yes, we have step through debugging with FastLED using LLDB.

Quick Start:

• VSCode with LLDB extension (recommended: CodeLLDB)
• Navigate to one of the tests in tests/ and open it
• Hit F5 to start debugging

Command Line:

• Interactive: uv run clang-tool-chain-lldb <executable>
• For unit tests: uv run clang-tool-chain-lldb .build/meson-debug/tests/runner.exe -- <test_name>.dll

Note: FastLED includes built-in crash handlers that automatically provide excellent stack traces with function names, line numbers, and full call stacks. For most debugging needs, the automatic crash output is sufficient. Use LLDB when you need breakpoints, variable inspection, or step-through debugging.

See docs/agents/lldb-debugging.md for detailed LLDB usage guide.

Enabling 3-second compile times using our web-compiler

• You must have docker installed for fastest compile times. It's free.
• cd <FASTLED FOLDER>
• pip install fastled
• fastled examples/Blink/Blink.ino
  • Cpp changes to the fastled source can be compiled by the live fastled compiler.

Testing your changes

Most of this is in the basic CONTRIBUTING.md guide. But as a reminder

• Unit Testing: ./test
• Linting: ./lint
• Compiling on platforms ./compile uno,teensy41,esp32s3 --examples Blink,Apa102HD

Enabling AI coding

aider.chat is available for advanced and high velocity coding with FastLED. To use it, have your open-ai or Anthropic api key ready. It's recommended to use Anthropic as it's performance is much better than Open-AI for coding.

At the root of the project type:

./ai and follow the prompts. Once the key is installed you will get a prompt that looks like this:

architect>  

There are two modes to use this AI, a watch mode which watches your files for changes and launches automatically, and a slow way which you add target files then instruct it to make changes:

• watch mode (best for implimenting a function or two)
  • Edit any file in the repo. Add a comment with AI! at the end. The ai will see this and start implementing what you just typed.
  • Example: Edit src/fl/vector.h and put in a comment // Add more comments AI!, then say yes to the changes in the prompt.
• Slow mode (much better for bigger changes across the file or several)
  • While you are in the architect> prompt you will add a file to the chat
    • /add src/fl/vector.h
    • Now tell the AI what you want it to do, and it will do it.
• Making the AI fix it's own problems it introduced.
  • At the AI command prompt, have it run the following
    • Linux/Mac: /run ./test
    • On Windows: /run uv run test.py
  • After the test concludes, the AI will ask you if you want to add the output back into the chat, agree to it then let it try to correct it's mistakes.

Every single time you do a change, make sure and thoroughly check it. I recommend VSCodes built in git diff tool.

Although the AI is pretty amazing, it will inject entropy into your code and this is the source of a lot of problems. So watch all changes it makes very thoroughly. Under almost all circumstances you will have to revert unnecessary changes (like comments) line by line.