v5 Capstone

Full pipeline

Lex -> Parse -> Analyze -> CompileModules -> Link -> Encode/Decode -> VM

Stage ownership

• parser builds AST, no linking assumptions
• analyzer validates modules/types/errors only
• compiler emits module artifacts
• linker rewrites/merges
• decode validates artifact compatibility
• VM executes merged bytecode

Error policy

• stage-order precedence: parse > semantic > compile/link/encode/decode > runtime
• analyzer returns many internally, capstone returns first deterministic one

Why encode/decode in path

• catches boundary drift before runtime
• ensures artifact readability
• keeps external tooling compatibility

Determinism

• preserve module and export order
• stable linker rebasing
• same source -> same encode/decode/execute behavior

Practical checks

• module import/export validation happens before runtime
• main remains entry module path in linked program
• no state leakage across VM runs
• only execution failures remain once earlier stages pass