diff --git a/docs/engineering/VALIDATION.md b/docs/engineering/VALIDATION.md new file mode 100644 index 0000000..28fd733 --- /dev/null +++ b/docs/engineering/VALIDATION.md @@ -0,0 +1,71 @@ +# Validation report + +Status: **no validation run has been recorded for this tree.** Nothing in this +document reports an executed check. Every entry below is a required check with +its exact command and an explicit not-run marker, per +[testing strategy](TEST_STRATEGY.md): unexecuted configurations are not +validation. + +This file is the record required by ROADMAP task 0.8 and by AGENTS.md. Replace +the status markers with real results, exact tool versions and observed +limitations at the end of each foundation change. Do not mark a row passed +unless that command actually ran on the commit being recorded. + +## Toolchain actually used + +| Item | Required | Recorded version | +|---|---|---| +| CMake | 3.28+ | Not recorded | +| Generator | Ninja | Not recorded | +| Linux compiler | GCC 14 (initial CI target) | Not recorded | +| Windows compiler | MSVC 19.44+ | Not recorded | +| clang-format / clang-tidy | major 19 baseline | Not recorded | +| Host OS / CPU | — | Not recorded | +| Commit under test | — | Not recorded | + +## Check status + +| Check | Command | Status | +|---|---|---| +| Debug configure | `cmake --preset debug` | Not run | +| Debug build | `cmake --build --preset debug` | Not run | +| Debug tests | `ctest --preset debug` | Not run | +| Replay verification | `evo-sim --seed 123456 --entities 8 --ticks 1000 --verify-replay` | Not run | +| Save/load continuation | `evo-sim ... --save state.evo` then `--load state.evo` | Not run | +| Format check | `cmake --build --preset debug --target format-check` | Not run | +| clang-tidy analysis | `cmake --preset analysis && cmake --build --preset analysis` | Not run | +| ASan/UBSan | `cmake --preset asan && cmake --build --preset asan && ctest --preset asan` | Not run | +| TSan | `cmake --preset tsan && cmake --build --preset tsan && ctest --preset tsan` | Not run | +| Release build and tests | `cmake --preset release && cmake --build --preset release && ctest --preset release` | Not run | +| Benchmark | `./build/release/evo-benchmark` | Not run | +| Snapshot fuzz (optional) | Clang libFuzzer snapshot restore target | Not run | +| Container build | `deploy/docker/Dockerfile` | Not run | + +## Expected reference fixture + +Documented in [testing strategy](TEST_STRATEGY.md), **not verified here**: seed +123456, 8 particles created at (0,0), 1,000 ticks, rules/format 1 yields +canonical payload FNV-1a `d61ad5bf0c928c76`. When recording a real run, state +whether the observed hash matched this value. + +## Known scope limits + +These are absent from the tree by design, not failures, and no gate below may +be reported as passing: + +- No biological regression evidence. Synthetic particles are not an ecosystem; + this category begins at M2. +- No network, protocol or load evidence. No network service exists. +- No disaster recovery or backup restore evidence. No persistent service or + database exists. +- No Kubernetes deployment is running. `deploy/` is scaffold only. +- Benchmarks carry no portable wall-clock threshold; record seed, count, ticks, + time and compiler instead. + +## Recording procedure + +For each foundation change, record: the commit hash, exact tool versions, each +command as invoked, its observed result, and any environmental limitation that +prevented a check from running (missing compiler, absent container daemon, +unavailable sanitizer). An environmental limit is recorded as a limit, never +silently omitted and never reported as a pass.