Skip to content

Latest commit

 

History

History
50 lines (38 loc) · 2.28 KB

README.md

File metadata and controls

50 lines (38 loc) · 2.28 KB

Integration test directory structure

Integration tests for the Noir compiler are broken down into the following directories:

  • compile_failure: programs which are not valid or unsatisfiable Noir code and so the compiler should reject.
  • compile_success_empty: programs which are valid satisfiable Noir code but have no opcodes.
  • compile_success_contract: contracts which are valid Noir code.
  • execution_success: programs which are valid Noir satisfiable code and have opcodes.
  • execution_panic: a temporary location for tests that panic, but may or may not be valid or satisfiable Noir code. These tests should:
    • Each have a corresponding open issue
    • Be moved into another test folder once the panic is addressed
  • execution_stack_overflow: a temporary location for tests that stack overflow, but may or may not be valid or satisfiable Noir code. These tests should:
    • Each have a corresponding open issue
    • Be moved into another test folder once the stack overflow is addressed
    • NOTE: these tests are not currently run, either locally or CI

The current testing flow can be thought of as shown:

flowchart TD

    subgraph compile_failure
        A1[Attempt to compile] --> A2[Assert compilation fails]
    end

    subgraph compile_success_empty
        B1[Attempt to compile] --> B2[Assert compilation succeeds]
        B2 --> B3[Assert empty circuit]
    end

    subgraph execution_success
        C1[Attempt to compile] --> C2[Assert compilation succeeds]
        C2 --> C3[Write circuit to file]
        C3 --> C4[Assert execution succeeds]
        C4 --> C5[Write witness to file]
        
        C6[Publish witness + circuit as artifact]
        C3 --> C6
        C5 --> C6
    end

    subgraph compile_success_contract
        D1[Attempt to compile] --> D2[Assert compilation succeeds]
    end
Loading

execution_success vs compile_success_empty

Note that execution_success and compile_success_empty are distinct as compile_success_empty is expected to compile down to an empty circuit. This may not be possible for some argument-less circuits in the situation where instructions have side-effects or certain compiler optimizations are missing, but once moved to compile_success_empty a program compiling down to a non-empty circuit is a compiler regression.