Current-origin amendment execution

Radio bc8b6331d3c76cbc011210e13f81e6ea351ef06f, Hemera d5f0a09539360934fb32db9b2e15a91995607db6; previous Radio344 candidate and audit remain untouched. At this initial research point production had no changes in the owned worktree.

Fresh raw origins, archive/lock preparation, actual toolchain identities, default and all-feature dependency metadata and test-only current witness are under /tmp/radio-empty-status.20261011-60036-oi3wr7. Every runner records full argv, environment, cwd, time, exit and exact stdout/stderr hashes.

witness-empty-root-debug: actual1.95, exit101,17 tests:15 pass/2 Fs failures. Both retain the ordinary public Complete{0} assertion and return NotFound. Published empty/plain domains, native current Hemera behavior and nonempty controls pass. Only cfg(test) module wiring and three test files were overlaid in the private witness; current upstream constant/docs/equality are unchanged.

witness-enumeration-probe: exit0, one test, current fresh/after-empty-add lists both empty and computed key canonical. Its temporary test append was removed before further gates; prior source bytes and overlay identities are retained.

witness-empty-root-release: actual1.95, exit101,17 tests:15 pass/2 Fs failures, the same NotFound versus Complete{0} assertions. This separate execution preserves all optimized-output diagnostics; it is not inferred from debug.

At the end of research, new PLAN and subsequent CODE approval were pending. No guard, commit, push, version, release or current broad-gate success is claimed.

Fresh PLAN review — before implementation

Grok4.7 APPROVE, blockers none, raw review SHA256 fdc5e14c5d7c4a424004001cb05b95f1dea6637e53635bc36b7a30d395c960e6; exact review retained in current scaffold review/plan-review.txt. Frozen58-section PLAN pack stays immutable. Accepted: unconditional canonical EMPTY guard before the database, one synthetic availability comment, current pinned vector revision in the test comment, actual no-default measurement and current local docs/reverse-consumer gates. No optional legacy-row test: existing public fresh/reopen/force-delete controls exercise this contract; adding private seeded data would expand the approved17-test scope.

One reviewer observation is factually wrong: committed .config/nextest.toml SHA256 ef588b7e34ddd981c24a523ba359ec76a616a45b40f3f3f525692ae809aea590 includes [profile.default] slow-timeout = { period = "10s", terminate-after = 3 }. Default runtime uses that explicit30s termination. CI profile inheritance will be verified against pinned nextest0.9.80 before claiming its effective bound; any additional local profile will be retained and labeled as local policy.

Local --locked --offline strengthens literal CI argv; raw receipts distinguish that from hosted CI. The MSRV command is workspace/all-targets with default features despite its step title. At most2 Cargo jobs and20GB staging/30GB active budget; no old344 gate result is current evidence. Root owns CODE review, commit and merge. Old source/audit/CODE verdict remains historical, immutable and superseded by origin PR35. Current CODE review is still pending.

Pinned nextest0.9.80 source at19c03ce27d861e12b73df4590881d1e97e5ea07f confirms custom ci falls back to the default slow timeout (config_impl.rs676–679). Termination begins after3×10s, then default10s graceful shutdown may extend wall time. No additional timeout profile is needed or used. Exact archive/config/source hashes and raw download/tag receipts are retained in nextest-bound.json.

Implementation applied: four-line unconditional status guard plus synthetic-availability comment, cfg(test) wiring and three independent tests. Upstream constant/docs/equality unchanged. Vector provenance comment now pins current d5f0a095; assertions unchanged. First current all-feature run exits0:17 pass; previous current witness was15 pass/2 fail. Candidate source freeze SHA256 751d7471250a7899ed4a7a6af35df9200c404448dc08633208d74fc96852213d.

Scoped new-test rustfmt check initially used edition2024 in error and returned import-order differences. Owning iroh-blobs manifest actually declares edition2021; corrected exact same check with edition2021 passes. Both raw commands are retained; no test source changed. Workspace formatter result remains independently red.

Current workspace default cargo test --workspace --no-run --locked --offline completed exit0. Orchestrator then hit its20GB staging threshold (run plus shared tools) before launching the consumers gate; no consumers command or result exists for this stop. Completed release artifacts had already been removed; after the stopped orchestrator all Cargo jobs were gone. Cleaned only completed debug/incremental cache (raw before/removal/after receipts), retaining binaries, source and logs. With root instruction, subsequent gates and their counterparts use CARGO_INCREMENTAL=0 as an explicit local storage limit. Completed gates are not repeated: resume validates argv/toolchain and exact raw stream hashes.

Named CI runtime uses profileci isolation and inherited timeout policy with a local --test-threads1 resource limit; this is disclosed as a local variation from the literal workflow argv. Native and actual local-consumer comparisons also use their planned serial limit. Named CI adds locked/offline and records2 Cargo build jobs plus CARGO_INCREMENTAL=0 resource policy.

Current ordinary measurements: focused all/default/release17 passes each, no-default15; upstream EMPTY1, root_policy5, export_pairs14, parent_pairs11; cyber-bao71+8+9. Local consumers nextest108 pass/3 skipped (85 docs and23 integration; selected radio-cli compiled with no unit cases). Owning native nextest141 pass/2 skipped,158.864s. Workspace default no-run exit0. All-feature named nextest no-run exits101 at unchanged iroh-gossip api.rs468 (address_lookup returns Result; .add absent). Current owning Clippy/private docs/workspace format are also red; exact current-parent counterparts still pending at this point. No historical344 cause is used as their proof.

After the completed all-feature no-run command, stagingguard prevented the next command. Full completed candidate target was cleaned before resume. Resume verified completed command identities/raw hashes; no successful or red completed matrix was repeated. Cleanup and both stage-stop observations are orchestration outcomes, not product gate passes.

Root narrowed only failing CI runtime verification: compare exact failing test cases under identical features/profile/environment and timeout policy, preserving per-test diagnostics; do not repeat unrelated passing runtime cases. The approved full owning native baseline and exact compiler/lint/docs/fmt/policy counterpart commands remain. This is a verification scope reduction, with no source/behavior change. Each narrowed runtime comparison gets its own mapping, argv and distinct failed-cases receipt; no full parent CI result is implied.

Final current ordinary/MSRV/nightly matrix completed. CI default runtime547 pass/9 skipped; CI no-default546 pass/1 Willow timeout/9 skipped. All seven actual1.89 gates pass, including literal default-feature workspace/all-targets and real local-consumer compilation. Full parent native124 run:122 pass/2 connection-pool timeouts/2 skipped, versus candidate141 pass/2 skipped. Per-name transitions122 PASS→PASS,2 TIMEOUT→PASS,17 new passes; timing improvements have no established cause. Exact failed-case-only parent Willow comparison reports net failed: failed while active (FAIL6.987s), while candidate TIMEOUT30.009s. Its generated proptest regression seed, absent from committed sources, is retained separately and is not a source change or same-cause proof.

Current parent compiler/Clippy feature-loop/private-doc/fmt comparisons match full normalized diagnostics or formatter hunks. Both pinned-nightly commands fail at rustdoc search-index invalid-null generation for cyber-radio/base. Deny36 headings match; the sole full-block difference is synthesized Hemera license column126/129 and matching caret padding from path length. First owning parent Clippy rustc derive_builder was SIGKILLed; one exact follow-up completed with candidate-identical normalized lint blocks. Both raw attempts remain; the signal's cause is unknown. Its wrapper initially propagated the product red exit before recording the follow-up mapping; only mapping/cleanup resumed from the existing verified receipt, without another Clippy execution.

Last-tag semver202 checks pass/58 skipped against f2b1298 and explicitly current Hemera. Initial current-parent semver helper darling rustc received SIGKILL before API comparison; preserve its raw receipt and actual helper lock as inconclusive. One exact serial follow-up is in progress. No unrelated process was stopped and no memory-cause claim is made.

Current-parent semver exact follow-up completed exit0:202 checks pass/58 skipped; its helper sources/lock were retained before cleanup. All approved current gate executions and required narrow counterparts are complete. No further matrix is scheduled. Current five-source freeze remains unchanged; source comparison separately identifies the single generated parent proptest seed. Fresh CODE and immutable audit packaging follow; root owns review/commit/merge.

Fresh CODE completed2026-10-11T17:56:11Z, exit0, APPROVE with no blockers. Root read the full verdict and source delta, independently reconstructed64 CODE sections, replayed the portable979-file capsule and all seven exported source pairs,196 command receipts/392 streams and21 counterpart identities. Exact commands, timestamps and output are outer audit sidecars. Source freeze751d7471 is unchanged. Required broad reds retain wip/decision accounting; this completes the ordinary source repair, not foundation row4 or a release. See review-disposition.md.