pi-crew 0.11.14 → 0.11.16
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md
CHANGED
|
@@ -1,5 +1,47 @@
|
|
|
1
1
|
# Changelog
|
|
2
2
|
|
|
3
|
+
## [0.11.16] — fix(bug #4): steer racing executor start no longer wiped by on-spawn truncate (2026-10-11)
|
|
4
|
+
|
|
5
|
+
Third live-caught fix from the 2026-10-10 full battery (run team_20261010172900).
|
|
6
|
+
|
|
7
|
+
### Fixed
|
|
8
|
+
|
|
9
|
+
- **`steer_subagent` claimed "Steer delivered" while the steering file stayed
|
|
10
|
+
0 bytes** — root cause: the F7 on-spawn truncate's race guard compared the
|
|
11
|
+
file mtime against `taskDispatchStartedAtMs` (captured when the executor's
|
|
12
|
+
attempt loop STARTS), but `steer_subagent` accepts a steer as soon as the
|
|
13
|
+
manifest + subagent record exist — which can be EARLIER (queue/semaphore
|
|
14
|
+
gap; live timeline: manifest 17:29:00.100, steer appended 17:29:05.088,
|
|
15
|
+
executor baseline captured later, truncate wiped it at 17:29:05.753). The
|
|
16
|
+
baseline is now the run's creation time (`steerRaceBaselineMs(manifest, …)`)
|
|
17
|
+
— the earliest instant a steer could have been accepted. Fresh per-run
|
|
18
|
+
artifacts roots cannot hold pre-run residue, so the baseline only moves
|
|
19
|
+
earlier: the original F7 post-dispatch race case is unchanged, pre-run
|
|
20
|
+
residue is still wiped. Regression: `child-executor-steer-race.test.ts`
|
|
21
|
+
(live timeline replay + boundary + legacy-manifest fallback + structural
|
|
22
|
+
wiring pin).
|
|
23
|
+
|
|
24
|
+
## [0.11.15] — U8 sqlite async parity: coalesced mirror + runner env allowlist (2026-10-10)
|
|
25
|
+
|
|
26
|
+
Second live-caught pair from the real-test battery (P2/P2b/P2c probes).
|
|
27
|
+
|
|
28
|
+
### Fixed
|
|
29
|
+
|
|
30
|
+
- **`saveRunTasksCoalesced` skipped the sqlite backend** — the hot path from
|
|
31
|
+
`persistSingleTaskUpdate` wrote only the tasks.json mirror; the authoritative
|
|
32
|
+
`state.sqlite` kept its createRun snapshot (P2 probe: manifest completed,
|
|
33
|
+
sqlite tasks all `queued`).
|
|
34
|
+
- **Detached runner env stripped `PI_CREW_STATE_BACKEND`** — even with the
|
|
35
|
+
mirror fix, async runs route task-state writes through the background
|
|
36
|
+
runner whose allowlist dropped the key: host created sqlite, runner updated
|
|
37
|
+
JSON (P2b re-probe still stuck). `PI_CREW_STATE_BACKEND`,
|
|
38
|
+
`PI_CREW_CHILD_PI_MODE`, `PI_CREW_CHILD_PI_POOL` are now allowlisted.
|
|
39
|
+
P2c end-to-end: sqlite tasks reach terminal status, tasks.json mirror and
|
|
40
|
+
events.jsonl audit spine intact, rpc transport + deterministic verify_gate
|
|
41
|
+
engaged.
|
|
42
|
+
|
|
43
|
+
# Changelog
|
|
44
|
+
|
|
3
45
|
## [0.11.13] — Reliability: runner survives run.lock contention + 2 fixes from live wave runs (2026-10-10)
|
|
4
46
|
|
|
5
47
|
Three reliability fixes discovered while shipping Waves 1-3 of the upgrade-spec
|