Files
ARCADIA-Studio/harness/experiments/analysis-source-s0-s1-windows-20260913/README.md
T

3.1 KiB

Isolated analysis source S0/S1 Windows gates

Exact tested source: 2c515cf3bdb3f707beee320be8fed90cffb40757, isolated analysis branch. These are historical gate results, not an integrated release or public source8 acceptance. Source/design and the original source reviews remain authoritative in the existing source handoff. The changed local S1 review, including its 2c supplement, and the earlier independent repair review are retained here without changing their historical conclusions.

manifest.json binds 24 exact-byte evidence files (586,875 bytes), preserving one original input manifest per actual receipt. The five completed Cargo gates each report 518 unchanged inputs; archive verification additionally matched every recorded input size and SHA-256 to the immutable tested Git blob. Original absolute paths describe provenance and need adaptation for another machine. Executables and toolchains are omitted; the test artifact identity remains bound in the original receipts. The commands require the tested source and corresponding Windows toolchain/build environment.

Actual gate Result
Initial listing launcher Failed before Cargo spawn or input capture; launcher exit 1, no Cargo result
Corrected listing Exit 0; 114 unique tests
Staged syntax Exit 0; all 11 selected tests passed
Save containment guard Exit 0; all 4 selected tests passed
Full library suite Exit 101; 98 passed, 16 failed
Strict Clippy, all targets Exit 101; clippy::enum_variant_names on private Kind::AsKind

The full suite's exact membership equals the listing's 114 names. All fifteen new focused tests also passed in that full execution. Sixteen individual failure sections identify fourteen unsupported-source8 errors and two legacy definitions-envelope errors. Those are failures, with no waiver or general suite acceptance. The completed strict log contains one reported lint diagnostic and the resulting compile failure; it does not establish absence of further diagnostics after repair.

The first launcher failure and its original helper are preserved alongside the final recorder and admission helper. Admission receipt hashes were checked against the exact result/log/input files. The strict log was copied only after its final receipt existed and the parent confirmed process closure. No compiler, application, source edit or analysis-worktree mutation was performed during archiving.

S0 guards generated source before filesystem staging; its four new tests exercise the shared private production guard with injected formatter output, not the outer public entrypoint with the real formatter. S1 remains test-staged syntax. Full source8 reading/writing, legacy migration, whole-bundle limits, public save/update/reopen and GUI behavior are not established by these passes. Any subsequent lint correction or later gate belongs to another source pin and must retain these failed historical results.