Fork provenance register completion + inherited-test floor green
completed
fork-port-connection-watcher
fork-rederive-queue-ops
narrow
low
component
implementation
wave-2
substrate
Description
Close out the fork's bookkeeping and verify the test floor end-to-end
(ADR-018's adoption discipline; ADR-011's tests clause):
PROVENANCE.md delta register: entries for every delta the port
actually made, category-tagged per ADR-018's categories — the three
watcher deltas (W-1, W-2, dead-man's switch), the table-family
rename, the bootstrap re-keying, the drops (recorded as
not-ported), the re-derivation (new code, contract-derived names),
and any mechanical adaptations the port surfaced that the plan
didn't anticipate. Each entry: what, why, the decision citation.
License/provenance cross-check: the notice matches the actual
ported content; the register's revision citation (f4e53c6) matches
what was actually diffed.
Test floor: run the full engine-crate test suite (inherited +
contract-property) as one gate; fix or explicitly record any
inherited test that the deltas legitimately changed (e.g. the
dead-man's-switch panic test becomes a watcher-fatal-death test).
Lineage-diff sanity pass: produce (locally, not committed) the
diff of the substrate subtree vs. the honker-core sources at
f4e53c6 and skim it for accidental divergence — anything not
covered by a register entry is either fixed or registered. This is
the wave-2 review's primary input.
Acceptance Criteria
PROVENANCE.md register complete: every delta categorized and
cited; no unregistered divergence from the lineage
The lineage read of record is quality-read-honker-core.md §6 —
not a diff. Task lines 30–31 cite quality-read §6 for the
lineage vs. ported-tree comparison. §6 is the prose keep / re-derive
/ drop scope register; a mechanical full-coverage line-by-line diff
does not exist as a document. The gate was therefore performed as:
§6's prose scope (plus docs/research/reference-honker-machinery.md
for line-cited verification) read first, the ported tree read
against it in full coverage (not spot sampling), then a mechanical
diff -u of the extracted lineage sources against their closest
substrate counterparts as the corroboration pass. Both methods
agreed on every divergence; the findings below cite the actual
artifacts (subagent report + local diff files, /tmp/opencode/,
not committed).
Lineage-diff findings (all now register-addressable):
scheduler pause/resume/list/update unregistered as a drop → D-21;
two kept-machinery in_savepoint regression tests whose
re-derive-side drivers weren't ported → D-22; two lineage
durability tests (python-child crash recovery, soak) not ported →
D-23; micro-deltas (error-text re-owning, thread name,
clippy lint shape) → D-24/25/26; and a substrate-side orphan
pressure_deadline_is_bounded placeholder test (asserted only that
time moves — a tautology, never part of the lineage or the port
plan) → removed from ops.rs's pressure suite. That removal — the
gate's only code change — is a test-deletion, not a behavior
delta; it is recorded here rather than as a register entry because
the register records divergences from lineage, and this added
nothing.
D-12's enumeration was annotated in place (per ADR-018 §2's
amendment-permitted class — correcting the record, not the story)
to pin the scheduler surface's register/tick/soonest/unregister
shape and cite D-21 for pause/resume/list/update.
Acceptance criterion 3 ("changed inherited tests documented in the
register") — the dead-man's-switch test example in the task text is
D-07's recorded landing (update_watcher_dies_cleanly_on_file_ replacement asserts join() == Ok where the lineage asserted a
panic payload); the dropped non-adapted inherited tests are the
D-22/D-23 rows. All inherited floors that the delta work changed
are therefore register-visible.
Test floor: the single-run gate was exercised repeatedly —
workspace runs (23 core + 3 contract-suite + 83 substrate = 109
tests green) plus a 3× repetition of the substrate crate run and a
5×-grep sweep with zero failures. One transient first-combined-run
substrate failure did not reproduce across 4 subsequent full runs
and 4 additional cargo test -p alkstore-sqlite repetitions; the
flake is unattributed (temp-file-based concurrency tests are the
suspect class). Test floor is green; the watch item is recorded
here for wave-2 review, not as a blocker.
Summary
Gate-time completion of the fork's bookkeeping: the lineage diff
(honker-core @ f4e53c6f4e53c6b83ad2d6979c04098d576db024ce7e98d
vs. the ported substrate) was produced locally and skimmed end-to-end
by a dedicated full-coverage read plus mechanical diff -u
corroboration; every divergence found is register-addressable.
PROVENANCE.md completed: D-21..D-26 appended (scheduler
pause/resume/list/update drop → D-21; dropped in_savepoint
regression-test drivers → D-22; dropped python-child crash-recovery
and soak durability tests → D-23; error-text/thread-name/clippy-shape
micro-deltas → D-24/25/26), D-12's scheduler enumeration annotated
in place, gate-time completion note added to the delta-register
header. License/provenance cross-check: the bundled LICENSE's
Apache-2.0 and MIT texts verified byte-verbatim against upstream
LICENSE-APACHE/LICENSE-MIT via git show f4e53c6:…; the full
fork-point hash, date, and package name node-v0.5.1-10-gf4e53c6
re-verified against the reference checkout; workspace license = MIT
OR Apache-2.0 = upstream — the notice statement matches the tree.
Test floor: workspace cargo test (23 + 3 + 83 = 109 green in one
run) with repetition sweeps clean; one unreproducible first-run
transient noted in the register's watch items, not blocking. One
substrate test removed (pressure_deadline_is_bounded, a tautology
placeholder with no lineage counterpart). The register now carries
D-01..D-26 plus an empty cherry-picks section (append-only), and
every divergence from the lineage is register-entered. Verified:
cargo build, workspace cargo test, cargo clippy --all-targets -- -D warnings, cargo fmt --check all clean.
