pg queue/tx/notify dedupe + doc truth-telling: job_from_row has one owner (queue.rs, tx.rs imports it) and get_job_tx reuses live_columns()/dead_columns() instead of inlining the 18-column lists — the contract-pinned Job shape now has exactly one decode owner (ADR-012 §2); sweep_expired's doc states the moved + retention-deleted sum in the in_tx multi-statement frame (the sum the SQLite twin returns — the doc was the only liar); notify.rs's closed-store listen error routes through the shared database_error helper; bridge_capacity() is a const with its rationale doc carried. No behavior change — identical SQL strings, error shapes, and capacity. (task pg-fix-dedupe-cleanup, review 002 minor notes + Finding 6 queue bullet)

tasks/pg-fix-dedupe-cleanup.md

@@ -1,7 +1,7 @@
 ---
 id: pg-fix-dedupe-cleanup
 name: Decode dedupe + small cleanups in queue/tx/notify (review 002 minor notes)
-status: pending
+status: completed
 depends_on: [pg-fix-tx-wake]
 scope: narrow
 risk: low
@@ -76,8 +76,41 @@ landing after avoids touching the same file regions blind.
 
 ## Notes
 
-> To be filled by implementation agent
+- The dedupe landed as the task pinned: `job_from_row` is now owned by
+  `queue.rs` (`pub(crate)`), `tx.rs` imports it; queue.rs's
+  `live_columns()`/`dead_columns()` went `pub(crate)` so `get_job_tx`
+  reuses them (the `format!`-bound `{dead}`/`{live}` select strings are
+  byte-identical to the inlined lists, so zero SQL change). Tx-side
+  imports ride the existing `crate::queue` module edge — fine
+  crate-internally (queue.rs already imports tx.rs's
+  `encode_payload_bytes`/`enqueue_row`; Rust module cycles are legal
+  within a crate).
+- Doc truth-telling choices: queue.rs's `job_from_row` doc now states
+  the one owner explicitly ("the tx reads (`crate::tx`) ... import it");
+  tx.rs's trailing decode-owner comment now names both decode owners
+  (stream decode in `stream.rs`, job decode in `queue.rs`) under
+  ADR-012 §2; the removed tx.rs duplicate carried `JobState`'s only
+  remaining use, so that import was dropped. queue.rs's `sweep_expired`
+  module-doc bullet now states the moved + retention-deleted sum and
+  the `in_tx` multi-statement frame (matching the SQLite twin's
+  semantics — the code already pinned the sum).
+- `bridge_capacity()` → `const BRIDGE_CAPACITY` with the rationale doc
+  comment carried; the closed-store error in `listen` now routes
+  through `database_error` (it was the only inline duplicate;
+  import already present).
+- No behavior change anywhere: the select SQL strings, error shapes,
+  and the bridge capacity value are identical.
 
 ## Summary
 
-> To be filled on completion
+Deduped the pg engine's job-row decode and cleaned the wave-4 minor
+notes (review 002): `job_from_row` has one owner (`queue.rs`), `get_job_tx`
+reuses `live_columns()`/`dead_columns()` instead of inlining the
+18-column lists, `sweep_expired`'s doc states the moved +
+retention-deleted sum in the `in_tx` frame (the sum the SQLite twin
+returns), `notify.rs`'s closed-store error routes through
+`database_error`, and `bridge_capacity` is a `const`. Verified: pg
+suite 140/140 green against the harness server (`pglo-poc` :15432)
+and 140/140 server-less, unchanged assertions; workspace
+`cargo build` / `cargo clippy --all-targets -- -D warnings` /
+`cargo fmt --check` / `cargo test --workspace` all clean.