From 871d72f5ef79ddc155b2be9cffd6277d4175edc6 Mon Sep 17 00:00:00 2001 From: Jason Woltje Date: Thu, 22 Oct 2026 10:09:30 -0500 Subject: [PATCH] duckworth: mos-1292 rulings recorded for row 253 M21 and row 269 refusedOnce --- .../work/row253/RULING-c4-integration-aca1.md | 11 +++++++++++ agents/duckworth/work/row269/REVIEW-design-early.md | 11 +++++++++++ docs/SESSIONS.md | 1 + 3 files changed, 23 insertions(+) diff --git a/agents/duckworth/work/row253/RULING-c4-integration-aca1.md b/agents/duckworth/work/row253/RULING-c4-integration-aca1.md index 5012ba920..863a717df 100644 --- a/agents/duckworth/work/row253/RULING-c4-integration-aca1.md +++ b/agents/duckworth/work/row253/RULING-c4-integration-aca1.md @@ -65,3 +65,14 @@ - The killing input is still in the test, at case1-accept.test.mjs:256 (`200 997 997 extra`, status 0). With M21 it matches :656 with valid pid, uid and gid. So in run3 the case most likely fails before the regex, in the supervisor path that cca936716 added (:116-247). - The loop at :262-266 asserts only `boundary=failed`, not `reason=probe`, so an earlier refusal passes the base and masks M21. That's the same shape as 264a's M2. - What decides it is the base P10 `reason=` for peer-malformed. If that reason isn't `probe`, this is a test gap that came in with C4, and the fix changes an old assertion, which goes to Mos (mos-1272). + +## M21 ruling (mos-1292 via sage-2747, 15:09Z) + +- First, Gander runs one diagnostic from a scratch copy on base 0604d91fc. It prints every P10 fault line with its `reason=`. +- Then every fault in the P10 loop asserts the `reason=` the design gives it, taken from the brief and BUILD.md, not from the diagnostic's output. peer-malformed asserts `reason=probe`. +- A fixture fix is allowed only where the stub hides the fault. A product change, or a reason the design doesn't name, goes to Mos. +- My review, on the new candidate and manifest: + - Each fault's expected reason traces to the design text, not to the diagnostic's output. + - The fixture diff is limited to what hides a fault. + - There's no product diff. +- The whole gate then reruns from the start, with nothing carried over. M21 has to be killed by the named P10 test, by assertion. diff --git a/agents/duckworth/work/row269/REVIEW-design-early.md b/agents/duckworth/work/row269/REVIEW-design-early.md index 6cbb044b8..00c70f349 100644 --- a/agents/duckworth/work/row269/REVIEW-design-early.md +++ b/agents/duckworth/work/row269/REVIEW-design-early.md @@ -119,3 +119,14 @@ Any other pre-existing test change goes to Mos first. - Input: gizmoduck-0243 about local commit d595570a6, which isn't pushed or read yet. Points 1-4 look right as described. I check them on the blob. - Point 5: `legacy-run-live` (BUILD.md:177-178 at e2e51bae1, Q2) has no test. It carries the brief's "a restart waits for proven prior-run exit" (t1-sessions.md:109), so a missing test is a blocker at the code round. - Suggested inputs that need no bwrap: a v1 leftover whose runner is a test-owned `sleep` child, or a registry entry with no session. Assert a refusal with 0 spawns, then a start after the child exits. + +## refusedOnce ruling (mos-1292 via sage-2747, 15:09Z) + +The refusedOnce actor_run pin (264a N1) is allowed on mos-1285's terms. At the round I check: +- The SELECT reads actor_run. +- The expected row has the exact run-name literal. +- The three callers pass the name. +- Nothing else changes. +- The commit names mos-1292 and quotes the old `['action.refused', {code:'not-holder'}]` row. + +DEFERRED.md:474-480 gets its closing line when 269 lands. diff --git a/docs/SESSIONS.md b/docs/SESSIONS.md index 1c9cb41e0..c3f14203a 100644 --- a/docs/SESSIONS.md +++ b/docs/SESSIONS.md @@ -1491,3 +1491,4 @@ are never rewritten or removed; corrections are new entries. 2026-10-22T14:53Z | Duckworth (Opus 5.5) | row 269 (264b) thread-launch.test:182 addition e2e51bae1 BUILD.md a4b8a1629 (duckworth-0153) | reason holds, Mos rules, no blockers | 2026-10-22T14:56Z | Duckworth (Opus 5.5) | row 269 (264b) mos-1289 :182 ruling recorded as manifest-round checks (sage-2738) | recorded, no verdict | 2026-10-22T15:07Z | Duckworth (Opus 5.5) | row 253 M21 static lead (0155) and row 269 legacy-run-live early warning (0156) | lead sent to Sage for Mos; no-test is a code-round blocker | +2026-10-22T15:09Z | Duckworth (Opus 5.5) | mos-1292 rulings recorded: row 253 M21 path, row 269 refusedOnce terms (sage-2747) | recorded, no verdict |