stack/agents/filbert/work/queue-49/r1/body.md at sp2

0 Fork 1

Files

T

jason.woltje b9ea9304fd lead: commit filbert's untracked evidence (273 files: queue-101 queue-102 queue-105 queue-107 queue-108 queue-109 queue-110 queue-112 queue-113 queue-114 queue-115 queue-116 queue-119 queue-120 queue-122 queue-123 queue-125 queue-126 queue-128 queue-129 queue-130 queue-131 queue-132 queue-133 queue-134 queue-135 queue-136 queue-137 queue-139 queue-140 queue-141 queue-142 queue-145 queue-146 queue-147 queue-148 queue-149 queue-150 queue-151 queue-152 queue-153 queue-154 queue-155 queue-156 queue-162 queue-163 queue-165 queue-166 queue-167 queue-169 queue-170 queue-175 queue-177 queue-181 queue-182 queue-184 queue-185 queue-186 queue-187 queue-188 queue-49 queue-64 queue-66 queue-70 queue-71 queue-80 queue-84 queue-85 queue-87 queue-88)

Unclaimed files left in the canonical checkout. Inventory: canonical-cleanup-inventory.md (commit-as-evidence groups). detect-secrets scan clean on 2026-11-02: hits are sha256 pins, git object ids, test sentinels already in tracked tests and unit-file lines.

2026-11-02 15:05:14 -06:00

4.1 KiB

Raw Permalink Blame History

Row 49 round 1 review (Filbert): approve

Candidate: manifest sha256 87ef1321a1a08be64a20499889fc24b5549b47d6d33abfd86db95f62fb3cb6d0 (17 paths), packet agents/rocko/work/queue-49/r1/ in ea5e70fa8, base.txt 7b166a0f1.

Rebuild: a fresh worktree at 5f457b2ff (current refactor) with build.patch applied. Every manifest hash matched and git diff --check was clean. The added lines have no dash, curly-quote or control characters. The packet manifest checks. No commit from 7b166a0f1 to 5f457b2ff touches a manifest path, so the candidate is the same on the current base.

What I checked and found correct

Brief test. comment-restart.test.mjs test 1 stops the bus, posts a comment with the clock moved three hours before and after it, restarts, and sees exactly one event. A tick and a reconcile after that add none.
Durable state. Cursor rows, snapshots and events commit in one recordTask transaction. lastComment, commentsAt and commentCount advance only after record returns. Tests 3 and 4 cover a crash after commit and a failed event insert.
Row growth. A row is appended only when the observed maximum id rises. The forward trigger rejects anything else. Bot comments advance the cursor but don't count toward events.
Startup. The adapter restores cursors from taskView(name, 'comments'). The first reconcile reads the comment list for every task, because no count is cached yet.
Authority. recordTask refuses commentCursors from a capability caller (agent-required), so only the sync path writes them.
Migration. Normal startup refuses a 3b store. migrate-schema.mjs checks each step's source digest, metadata, WAL mode, integrity and times, and applies the DDL, version, digest and epoch in one transaction. It refuses a missing store. The meta_no_update guard is dropped and recreated inside that transaction, and the final digest check covers it. store.mjs replays the chain at import and refuses if it doesn't reach the schema digest.
land-work. read.mjs imports SCHEMA_VERSION instead of its own '3b', so the version has one source.

Sage's note (28559)

I'm approving as is, so the migration test's version pins stay and row 123's eight-line edit lands with row 123, per the ruling.

Notes (non-blocking)

A task read twice in one pass aborts the pass. foreignComments no longer updates lastComment during the loop. If one task id appears in two project reads in the same tick or reconcile and has a new comment, both reads push the same cursor. The forward trigger rejects the second insert, and the whole recordTask rolls back. My probe on the candidate store: two equal inserts for one task in one transaction give "comment cursor must advance" and zero rows. walk refuses duplicates inside one project, and ctx.tree has one entry per project, so this needs a move between two project reads. The next pass recovers and nothing is lost or duplicated. A follow-up could keep the highest cursor per task in out instead.
land-work and the live bus between landing and migration. Once this lands, read.mjs from any tree at or after it refuses a 3b bus with bus-schema until Mos migrates the live store. This fails closed. MIGRATION.md doesn't mention it, so Sage may want it in the window plan if a land-work run could fall in that gap.

What I ran and what I took from receipts

I ran, as single files with TMPDIR set and Docker disabled:

Test file	Result
bus comment-cursors	2/2
bus comment-migration	5/5
bus schema	1/1
bus store	6/6
bus tasks	9/9
bus task-read	3/3
tasks comment-restart	5/5
tasks sync	11/11
tasks tree	19/19
tasks adapter	8/8
tasks startup	9/9

I also ran the duplicate-cursor probe above.

From Rocko's receipts, not rerun by me:

the targeted gate in gate/results.json: bus 178, tasks 97, cli 195, webui 112, harness 84, seat 132, test-land-work 84/0;
22 of 22 mutants killed in MUTANTS.md.

I didn't run test-land-work.sh or any full suite, and I took no slot. The brief's full shell gate is Sage's landing gate.

4.1 KiB Raw Permalink Blame History