5.0 KiB
Private mixed-syntax Windows evidence
Final source a578ac98a06772a60bec00ae287e65650680d119. The packager checks every closed receipt, actual test name/outcome, artifact hash and immutable Git input before emitting this archive. Its results concern this private syntax checkpoint; they do not establish public source8 or application completion.
The complete core unit suite remains FAILED: 178 passed / 16 known public-header failures, exit 101. All 152 historical statuses are unchanged; all 42 later tests pass. The corrected listing contains 194 tests, retaining the previous 173; 85 typed tests pass within the full run, including the earlier stack controls, and the focused mixed run passes its exact 20 listed members. Strict core all-target Clippy passes. These counts are package acceptance constraints, verified from raw output before this report is emitted.
The optional formatting gate FAILED, exit 1 at both intermediate and final pins. Its nine diff blocks affect model.rs, references.rs, semantic/catalog.rs, session.rs, transitions.rs and validation.rs. Read-only Git comparison proves all six complete blobs identical at a578ac9, eeb1069, prior10d7f33 and historical f1d79df. The original eeb formatting note is preserved unchanged; the verifier independently checks final equality and both raw formatting outputs. This is inherited formatting debt, not a formatting pass or proof that fmt was run historically. No unrelated file was reformatted, and there is no blanket all-gates-pass claim.
The original 754a9b4 listing compile failure is preserved, including E0063 at semantic_source.rs:191. Successor eeb1069 adds only the missing test-only mixed: None initializer. All five eeb gates remain: listing194 PASS, focused20 PASS, full diagnostic178/16 FAILED101, fmt FAILED1, strict Clippy FAILED101 for unnecessary_unwrap/collapsible_if. Final a578 changes only the two equivalent test-only guards plus the exact shared AGENTS update. Of541 scoped inputs,540 are unchanged; AGENTS is outside that scope. Historical source-only handoffs/reviews retain their temporal claims.
Final strict Clippy passed before the original Cargo listing. Following the host clock rollback, Cargo rebuilt for the focused run and produced a different binary hash: original listing ae4974da…, focused 089d7b03…, both 39,706,624 bytes. Their equivalence is not inferred. A fresh actual direct listing of 089d7b03… returned the same ordered194 names; focused20 and the subsequent unfiltered full direct run bind that exact executable. The original Cargo listing remains separate evidence. The direct recorder uses the package-root working directory, captures real executable output and checks all541 committed inputs plus executable/bound-evidence hashes before and after. It invokes no Cargo and fabricates no Cargo log line. Full178/16 is an actual libtest execution, not a new cargo test invocation. The archived review explains this protocol; no cache or source timestamps were changed.
The host UTC clock moved backward during the original Cargo listing. Its receipt retains reversed start/end timestamps and 768.422 monotonic elapsed seconds. The verifier permits that timestamp ordering only for this exact hash-bound receipt; source, artifact, output and all other closure checks remain required. See the archived clock disposition. No wall-clock elapsed duration is inferred.
Run from a checkout containing the named source and documentation Git objects:
python -B verify.py --repo /path/to/ARCADIA-Studio
Only read-only Git commands run. No archived Python, Rust, launcher, binary or test is executed. Source input bytes are checked against complete immutable Git input membership (541 at each of754/eeb/a578,538 at prior listing,534 at historical full run). Baseline raw receipts are independently bound to their published Git objects. Every present launcher is preserved once by its content hash; different runs may use different immutable launchers. No binaries, build cache or profile files are included.
The verifier establishes evidence consistency, not independent runtime re-execution or cryptographic authenticity of receipt authorship. Preserve the package's Git commit or external checksums when transporting it. Extraction is unnecessary; ZIP members are checked without execution. The separately supplied verifier is reviewed code, not an archived helper selected by the manifest.
No public source8 activation, complete semantic witness, mixed occurrence index/compositor, full model acceptance, checked-save or application/runtime completion is claimed. Private syntax still defers nested model-record/ordinary-step admission while retaining the original policy for that later boundary.
The parent reran this packaged verifier successfully. Independent real-archive controls accepted the unchanged archive and refused all 15 corrupted copies. Their raw results preserve an initial host-timeout attempt separately; no interrupted attempt is counted as a successful control.