Patch #308: Protecting Integrity and Powering Production: AI Training Ban & New Build Pipeline

As the Wave language ecosystem grows, we are committed to two fundamental principles: protecting the creative work of our contributors and providing developers with a professional-grade toolchain. Our latest update introduces a strict AI/ML training prohibition and a complete redesign of the wavec CLI to support complex, real-world compilation workflows.

1. Protecting Our Source: AI/ML Training Prohibition

In an era of ubiquitous data scraping, we believe it is essential to explicitly define how our source code can be used. We have implemented a project-wide policy to disallow the use of Wave’s repository for AI/ML training without prior authorization.

• ai.txt: A new manifest in the repository root clearly prohibits crawling, scraping, pre-training, and fine-tuning. This includes notices in multiple languages to ensure global clarity.

• Universal Headers: We have applied an AI TRAINING NOTICE header to nearly every file in the project—spanning Rust, Wave, Python, Shell scripts, and even our Dockerfiles.

This step ensures that Wave remains a project built by humans, for humans, protecting the intellectual property and intent of our community.

2. A Professional Build Pipeline: BuildPlan Architecture

We have replaced the basic "run and build" logic with a sophisticated BuildRequest and BuildPlan architecture. This transforms wavec into a highly configurable compiler capable of handling everything from quick syntax checks to complex systems engineering.

Granular Control with --emit

You can now stop the compilation process at any stage and inspect the output. The new --emit flag supports:

• ast: View the Abstract Syntax Tree.

• ir / bc: Inspect LLVM Intermediate Representation or Bitcode.

• asm: Generate assembly code for specific architectures.

• obj: Produce object files.

• bin: Create the final executable.

• check: Perform rapid syntax and semantic validation (aliased as the wavec check command) without the overhead of code generation.

Advanced Linking & Systems Support

Wave is now better equipped for low-level tasks like OS development and library creation. New flags include:

• --freestanding: Compile without standard entry points or libraries.

• --static / --shared: Control how your binaries are linked.

• --linker-script & --entry: Define custom memory layouts and entry points.

• --no-start-files: Skip standard OS startup code.

3. Multi-Input & Toolchain Interop

wavec is no longer limited to .wave files. By using the --input-type flag, you can now use the Wave toolchain to link and compile external LLVM IR, Bitcode, or raw Assembly files. This allows Wave to serve as a central hub for multi-language projects that utilize the LLVM/Clang ecosystem.

4. Preview and Inspect

To help developers debug their build processes, we’ve added powerful introspection tools:

• --dry-run: Preview the exact commands and steps the compiler would take. You can output this in a human-readable format or as JSON (--error-format=json) for integration into other tools.

• The print Command: Query the toolchain for detailed information, such as target-list, cpu-list, sysroot, and specific target-features.

5. Backend & CI/CD Enhancements

The LLVM backend has been refined to support specific hardware requirements:

• Code & Relocation Models: Control these via -C code-model and -C relocation-model (e.g., pic, static).

• Exit Codes: wavec now returns specific exit codes based on the error type (e.g., syntax vs. backend failures), making it much easier to integrate into automated CI/CD pipelines.

Conclusion

With these changes, Wave takes a significant step toward becoming a professional, production-ready systems language. We are establishing a space where developers have total control over their build process and where their source code is respected and protected.

Check out the new wavec --help to explore the full range of new build options!

Link

• Pull request

• GitHub

• Community