4.4 KiB
Independent package readback
2026-09-25. PASS after one documentation correction; no unresolved package discrepancy. Source/artifact readback completed against 97a8703adc75467a7f8e75f5c6cb00ca542b707c before releasing the worktree for subsequent work. This is local-only evidence, not publication or release acceptance.
Verified evidence.zip: 66 unique members, 606,451 bytes, SHA256 252ea94e7c7f500a2afb3d0fa47c43a051f39d9c844328337289eea0c463becd. CRC passes; member names are relative and contain no parent traversal; every member length/hash and the complete name set match manifest.json (SHA256 f85e85306916440ca17b096404a7ed2b16309cac8f85336fb91171fff6e761b2). All 64 file members match original local records; both patches exactly match immutable Git diffs from 8d52dab97c92cc31baae9f20d8c324b55818aa5d to initial 6b0ef99eca6f82a93e2cfc9eb16185c4f12403a7 and final source. All nine readable mirrors match both archive and original files. Baseline records also match the pinned emission ZIP, SHA256 8d46b31379d75567b195b639cb3bc289ad1c9d26df53d29dbfd99ba22e38c5b9.
Independently checked both candidates' complete 556 declared inputs against scoped Git tree membership and immutable blobs; final inputs also match raw disk bytes. Final HEAD and clean tracked source were confirmed. All gate receipt/log/input/launcher hashes reconcile, and the README's five readable command argument arrays match actual receipts. Both candidates contain six new focused passes, 275 listed names, and full 259 passed / 16 failed. All 269 previous outcomes and ten existing stack controls retain their results. Initial strict Clippy exits 101 with the documented lint; final strict all-target Clippy exits 0. Both format logs remain byte-identical to the predecessor, SHA256 c1c6b12861c8117ce6ffc6d6742ebf92618aafb10160ec6f99d370ca72ebab2d.
Failure-body comparison independently confirms 15 blocks identical after numeric thread-identifier normalization, plus exactly the single named missing backtrace-hint line in both initial and final logs. Its normalized old/new hashes and surrounding bytes match CHECKER-NOTE. The archived checker amendment allows only that named, hashed delta; the original stricter checker is preserved. The successful-caught-panic explanation remains an inference. This is neither 16 identical bodies nor a passing full suite.
The actual final executable was read without execution: 42,614,784 bytes, SHA256 41da1a4d788e2bd7392e69a724b7de259752d9f813ae6b42eadf580ec682a54a, matching focused/full/list receipts. No current-byte verification of the overwritten initial artifact is claimed. Diagnostic helper hashes and source/artifact identities reconcile. Recorded prologues confirm the selected +1,136-byte total, unchanged 2,520/11,720-byte binding/prefix reservations, and the distinct 56-byte test shim. These are partial compiler reservations, not total live stack or headroom. Post-freeze predecessor observations belong to this archive; predecessor extra-process repetitions are not current-source runs.
The archived source review is SHA256 61d93dab9307c105b5eaefcc7d01d71149c5321fc3a9baa10047824e28408dc1; test review 75b02f2392b85943590380bd647b292341e45efe9aa2e16011efc4b1dd14ba66; literal oracle 8387fc13f4cd6b98491b80a5ac1849d9617b915f5557874de80fbfce0ebd326a. The E offset erratum preserves the original error and correct token34..72/diagnostic34..35 without test changes. Readable README hash at inspection: 3b69a888241059ae11ef8086881c6dc143fd4cdcf95b4b526c3427236065da3c.
One finding was corrected by the parent: DECISIONS initially placed independent reconciliation before resolver text transfer. It now correctly describes transferring retained text once into the ledger and independently reconciling the inventory. Correction read back; no archive change was needed. README/STATE/DECISIONS preserve the private prerequisite, local-only/server-disk-full publication status, inherited failures, ordinary-harness scope of new tests, and pending producer/output-acceptance obligations. All 100 inspected relative local Markdown targets resolve. No remote publication was independently checked or claimed.
Audit used read-only Git, file inspection, hashing and independent in-memory Python assertions. No archived helper, compiler, test/product executable, debugger or network operation was run. ZIP, manifest, mirrors, gate records and source were not modified; this report is outside the archive. Parent owns documentation integration and publication.