Testing and Quality Gates

SOF treats test strategy as part of the architecture, not just part of CI.

Test Pyramid

  1. unit tests in the owning slice
  2. integration tests at composition boundaries
  3. fuzz targets for parser and state-machine edge cases

Required Local Checks

The main contributor gate is:

cargo make ci

That runs:

  • formatting check
  • architecture boundary check
  • clippy matrix
  • test matrix

For dependency-policy checks as well:

cargo make ci-full

Fast Commands During Development

cargo check -p sof
cargo test -p sof-tx
cargo make arch-check

Fuzzing Expectations

Fuzz targets exist for both sof and sof-gossip-tuning.

Run the bounded workspace smoke suites with:

cargo make fuzz-smoke
cargo make fuzz-smoke-gossip-tuning

For deeper campaigns:

cd crates/sof-observer
cargo +nightly fuzz run <target>

When fuzzing finds a crash:

  • reproduce it deterministically
  • add a regression test in the owning module
  • minimize and save the corpus seed

Policy For Behavior Changes

  • every bug fix gets a regression test
  • invariant-bearing constructors need positive and negative tests
  • flaky tests are defects and should be fixed before merge

SOF 2026