stack/agents/filbert/work/queue-181/r1/body.md at dev-console

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

4.5 KiB

Raw Permalink Blame History

Filbert review: row 181 round 1 (#1679), approve

Request 29370. Candidate manifest 46a4433ca4108682af2027dc27ad72359b45709006f809fbf45c4b45dfacc283 on base a757f8fe6. The brief is blob aafa471ae, and the rulings are in sage-1163.

Verdict: approve. No findings block. N1 and N2 are notes. F1 from BUILD.md stays a follow-up.

What I ran

I built ~/filbert-scratch/r181-r1/wt from a757f8fe6, then ran git apply --index build.patch.
- sha256sum -c on the manifest passed 3/3, and the manifest file hashes to 46a4433c...dfacc283.
- The patch-id is 1fef78100554: 3 files, +129/-16.
Base drift: git log a757f8fe6..refactor (fcfb82da2) on the three paths is empty.
Hash pins and path readers: no test under packages//tests or scripts/test-.sh reads scripts/stack-release.sh or docs/TOOLS.md, or pins a 64-hex hash of them, except stack-release.test.mjs itself.
I took the slot after sage-1185, with the /proc busy check CLEAR before and after, and TMPDIR ~/fs177.
- stack-release.test.mjs: 116/116 pass, 0 fail, rc 0, from 04:29:04Z to 04:32:48Z. The counts are an awk sum over the spec log.
I ran two mutants of my own on scripts/stack-release.sh, each against T45 only. Each was restored by git checkout, and the tree went back to the 3 staged paths.
- M1: a table's rows go into the text but not the total (if(n){text+=...}). Killed: rc 1, 5 of 6 failed. Every subtest failed except (c).
- M2: a missing listed table throws instead of counting 0 (the sqlite_master check removed). Killed: rc 1, 5 of 6 failed, the same subtests.
Darkwing's run (116/116 and 14 mutants killed) is their receipt, not mine.

Item 1: the lists, checked at `a757f8fe6`

I grepped the INSERTs in packages/bus/src. They match BUILD.md's table:
- broker.mjs: :561 for events, plus :999, :1129, :1159, :1551, :1575, :1687, :1696, :1734, :1769, :1806, :1894, :2257, :2301, :2331, :2362, :2377, :2470, :2483, :2518, :2535, :2565, :2596 and :2645.
- usage.mjs:92.
- store.mjs, for meta.
No other code under packages/ or scripts/ inserts into these tables.
Every counted broker writer stamps at with this.#now(), the monotonic ISO clock. usage.mjs stamps it with new Date().toISOString() inside its own transaction (:88-98).
decision_events is rightly in COUNTED: the decision.resolve event at :1778 is written only if (s.human).
decisions (list 3): the insert at :1696 and the action.allowed event at :1716 both run inside request's this.#store.transaction(() => ...) (:1295). The claim holds.
The UPDATE list is empty, as BUILD.md says. schema.sql has 16 CREATE TABLE statements, 16 BEFORE UPDATE triggers and 16 BEFORE DELETE triggers. The only UPDATEs are the migration step at store.mjs:213-214.
I agree with deviations 1-3 (deliveries, projects at, 13 tables).

Items 2 and 3

Every caller of the count gets COUNTED:
- count at :912;
- heldcount at :907;
- the in-run recount at :845.
The two version reads (:736, :743, :831) are unchanged.
The refusal at :917 prints the full text, and --events must equal the total. The usage text (:39) and TOOLS.md (:375, step 7) say so.
T45:
- (a) covers both modes for each of the 11 tables from a literal list, and the brief's 1/1/1 case. A 3f store counts 0 for the missing table.
- (b) refuses with no --events and with --events 2, before any stop and with the fingerprint unchanged. --events 1 then restores to B/3g. Both tables are covered.
- (c) covers the unreadable table.
- (d) covers the in-run recount through the stop-writes-row fault (:139).
On comparing only the total in-run (Darkwing's point): I agree it's enough. Every listed table has an abort trigger on UPDATE and DELETE, so between the pre-stop count and the recount, counts can only rise.

Item 4

T46 (:1268) builds packages/bus/schema.sql in memory. It checks that every table is in exactly one of the four lists, that no listed table is missing from the schema, and that the at split is right.
It reads lists 1-2 from the script, so a COUNTED change shows up there.

Notes

N1 (code text, not doc-only, doesn't block): the heldcount recheck's refusal at stack-release.sh:909 still says "the bus stopped while the events after STAMP were counted". It now counts rows in 12 tables, so "the rows after" would match the new :914 text. It's still true as written, and no test pins it.
N2: F1 (task_comment_cursors, :1129) stays the follow-up BUILD.md names. I checked that the table has no at column, so the guard can't count it without a schema change or an event.

4.5 KiB Raw Permalink Blame History