Target parent: c4182664222b89f464cc2453b88898bdd42df09a.
This records a test-only composition of complete retained openings and existing historical public response mathematics, outside both fixed review subjects. It adds no application source, witness store, lookup service, signer, recovery or protected-use adapter. See construction.
| Check | Result |
|---|---|
| Framing regressions with explicit callback controls | 18 passed in 1.939 s; recorded runner 2.038 s; zero skips or failures after correcting the new static source pin. |
| New actual public-worker qualifier | 12 passed in 2.522 s; recorded runner 2.628 s; zero skips or failures. |
| Complete offline Python suite | 1316 passed in 934.505 s; recorded runner 934.799 s. Every discovered method id passed; zero skips, failures and ResourceWarning lines. |
| Complete fixed review inventories | Complete observation subject: 189 files; frozen baseline: 119 files. |
| Artifact and relative-link checks | 690 index/worktree versions across 345 tracked files; 1142 relative file links resolve. |
| Safe local SQLite runtime | SQLite 3.54.0, DELETE journal, synchronous EXTRA (3); authentication and physical entry false. |
The first targeted run executed all 18 framing tests and failed one new static pin assertion: the cached compiled executable digest had been incorrectly used as the Rust source-file digest. The source bytes were verified against the target parent and the new assertion was corrected to the actual source digest. No existing source, fixture, worker or signature changed. This failed run is not counted as a passing qualification result. The separate actual-worker run passed on its first invocation.
Two initial ancillary commands stopped before validation: the inventory command lacked its required immutable-commit argument, and an incorrect SQLite probe command selected no script. The recorded commands were corrected to the retained workflow invocations. Neither correction changed repository source or tests. One private publication-helper preparation also stopped on a syntax error before execution; the corrected private helpers were compiled before use.
Both claims are independently derived from complete openings before either public check. A bad later packet and invalid retained-prefix relationship refuse before even the earlier callback runs. Directed regressions cover exact types, canonical/bounded bytes, complete selections, callback exceptions, altered result digests/flags and defensive unsigned descriptions. Callback controls intentionally forge flags; only the separate actual-worker run supplies public mathematics.
The new actual qualifier checks nine live identical histories, the signed
absent/pending/completed chain, an actual completed-versus-revoked source fork,
three validly signed false states, two signed complete-original collisions,
freshly challenged old absence, unavailable/null facts, stale delivery, coherent
source and consumer restore, zero-signature callback forgery, repeated policy
heads with changed record heads, cached results and changed executable measurement.
Two signed futures can both extend one prior witness without canonical selection.
Restoring both source and consumer knowledge repeats synthetic effects at [2, 2].
These controls grant no current authority, use permission or recovery algorithm.
Rust, Go, public example builds and the seventeen unchanged actual-worker script groups were not rerun locally because their source, fixtures, dependencies and harnesses are byte-preserved. Hosted CI retains their separate execution: 123 Rust tests, 55 top-level Go tests and the previous 152 actual cases; it adds the new 12-case group for 164 cases across eighteen groups. These inherited totals are not claimed as new local execution. The eleven unchanged standalone model CLI comparisons were not rerun locally; their regression tests stay in the full suite and their separate CLI steps stay in hosted CI.
The workflow gains only one post-build qualifier step in the existing adaptor job. Every earlier step, dependency/toolchain/action pin, seven-job matrix and timeout remains unchanged. No warning filter changes. Hosted execution remains separate from independent security/privacy review and release/runtime provenance.
| Object | SHA256 |
|---|---|
| Complete-opening helper | e1e8cdd3f868bce2117c56f1eb2eab90bef8626b7687793b62214f0884805ff6 |
| Retained-prefix helper | 20b7e00c5aef3ec28184d385d7e10098f289a6e476373ab93527deaa7664a622 |
| Actual-snapshot unsigned inputs | 0f0bdf4d916c5371f35eb7c2afee03dbcdef4a319999c01955f4b052d9d8fd01 |
| Public historical responses over those inputs | 2d461a54413ef156df62db26c88ac6f789ed8ea27322f3b8b12800b74f8de1bb |
| Unchanged public worker source | e96a8219e500cd0f79716796c146f64ce22e5f95bd3af26ccd2ad8a208c4834e |
The locally selected cached executable digest is
e3255b96df664a5b5b7a5d4813a3b665d3c9e3cba7916b93091cae09cb2d9073.
It identifies that local compiled selection, not a reproducible release or an
expected hosted platform binary. Hosted builds independently execute the same
unchanged source. Repeated digest measurement is not atomic launch or a sandbox.
No signature, key material, fixture, worker, wire schema or commitment domain is generated. The ten snapshot scenarios and six signed counterclaims, both fixed inventories and their unfilled reports remain unchanged. Complete retained-row disclosure uses public synthetic values and selects no private lookup protocol.
All required local checks passed.
Historical mathematics and consistency against a retained witness are separate from authenticated current heads and nonrollback source/consumer provenance. Source integration, private/application signing, recovery integration, core port, node activation, deployment, wallet access, transaction broadcast, physical entry and funded use remain NO-GO. Both independent reports stay unfilled.