Working during bootstrap
Record the complete command, host, target, executable digest, provider, minimal source and diagnostic for each reproduction. When comparing generations, retain their providers too: a runtime difference can resemble a compiler regression.
Positive fixtures establish accepted behavior and results. Negative fixtures establish rejection and diagnostics. A surface or semantic change must keep parser, AST, semantic analysis, IR, fixtures and documentation aligned.
The V81 dossier records byte-identical B2/C2 compilers within each Windows and Linux/WSL x86-64 host, alongside executed programs, runtime contracts and adversarial checks. Those results apply to the documented profiles; they do not close execution support for every declared architecture.
A useful contribution identifies one behavior: an incorrect example, a missing diagnostic, an unclear contract or a reproducible regression. Include the expected result and the smallest source that demonstrates the issue. Syntax proposals must explain how their meaning survives subsequent compiler phases.
Source references
AGENTS.md: continuity rules and completion criteria.docs/syntax.md: compact syntax and manual memory.crates/ncc-stage0-bridge/build.rs: seed requirements.examples/00_hello/hello.nandexamples/11_machine.docs/NCC_MODULES_V81_DOSSIER.mdanddocs/NCC_MODULES_V81_REPRO.md.
The architecture reference describes the pipeline and interpretation of bootstrap evidence. Release notes and installation documentation carry the operational instructions for each published kit.