stack/agents/filbert/work/queue-114/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

4.6 KiB

Raw Permalink Blame History

Filbert review: row 114 round 1 (#1608): approve

Candidate manifest sha256 68c60a79914692f3f6d4923f7bcff41de109ccca33bf6fba687fb88efa24bcac (7 files), as named in request 28496. I rebuilt it in my own worktree at BASE 72c9ac36c with build.patch from agents/rocko/work/queue-114/r1/. All 7 hashes match, and git diff --check is clean. No manifest path has drifted from 72c9ac36c to HEAD. The added lines have no dashes, bidi or control characters. scripts/stack-copy.sh and packages/discord/systemd/[email protected] are unchanged.

Sage's five checks

run --offline exits 4 before config. parse throws the usage error (exit 4) right after the binding-name check, and main reads config only after parse returns. The node test deletes the config file first and still gets 4 with no network mark. The test-discord.sh wrapper case gets 4 through scripts/discord.sh with the fetch and WebSocket trap loaded.
Status reads the effective cwd. status_unit takes systemctl --user show <unit> -p WorkingDirectory --value, strips an optional WorkingDirectory= prefix, prints tree:, then copy: <id> only if the last segment is 40 hex and holds cli.mjs, otherwise checkout: <path>. No --copy is passed to status in either test, so the id can only come from the stub. I think the stub is enough for this row. It answers every show with the same prefixed line, so it doesn't prove the property name or the bare --value form. I checked the bare form with a read-only systemctl --user show [email protected] -p WorkingDirectory --value: it prints /mnt/storage/src/mosaic-stack with no prefix, which this code reports as checkout: /mnt/storage/src/mosaic-stack.
Render and refusals. Without --copy, the node test compares the output to the template with only @REPO@ and @PATH@ replaced. I also rendered from a base worktree and from the candidate, replaced each tree path with one token, and cmp found them identical (1635 bytes). With --copy, the test compares every line other than WorkingDirectory and ExecStart, and asserts neither of those names the checkout. install renders before mkdir, so an empty id, short, a missing 40-hex copy, a missing cli.mjs and a missing pi each exit 1, and no unit file appears in the target directory. Both the node test and the shell cases cover all of these.
check --offline is local only. The return sits after the STOP report and before createRest. prepare only computes the journal path, and stopRequested is an existsSync. I ran a fixture by hand. With no journal directory, check --offline exited 0 and created none. With a STOP file present, it exited 0, printed STOP PRESENT, and left the file byte-identical with no turns/ created. The test's trap marks any fetch or WebSocket use. Online check hits the trap, and --offline doesn't.
Untouched files. git diff --stat at 72c9ac36c shows neither scripts/stack-copy.sh nor the template.

Brief tests and mutants

Every item in brief section 5 has a test: copy render, the four refusals with no unit, byte-identical checkout render, status from a stubbed show, offline pass, 0644 refusal, missing context under --repo, and run --offline exit 4. I read mutants.py and out/mutants/. The three prescribed mutants each replace a string that occurs once and restore it in finally, and each one fails the test written for it. The ExecStart mutant runs $REPO/scripts/discord.sh, which is what the brief's "ExecStart that keeps @REPO@" renders to.

Doc-only, follow-up

docs/TOOLS.md: the last new sentence runs into the existing "An optional tools key" sentence on the same line, so the new text has no paragraph break. Decision 137 check: grep -rlF TOOLS.md packages/*/tests scripts/test-*.sh finds nothing.

Notes, no change needed

No test asserts that check --offline leaves the journal and STOP alone. The code can't do otherwise today, and I checked it by hand (item 4).
The status stub matches any show. See item 2.

What I ran

I ran the rebuild and the checks above. In my own worktree with DOCKER_HOST pointing at a dead socket, I ran node --test packages/discord/tests/copy-service.test.mjs packages/discord/tests/offline-check.test.mjs: 5 of 5 passed. I also ran the base-against-candidate render cmp, the journal and STOP fixture, and the one read-only systemctl show above. I didn't rerun the mutants, the Discord node suite or the shell suites. Those counts come from Rocko's receipts: Discord node 183, test-discord.sh 78, 3 of 3 mutants killed. Sage's landing gate runs the Docker-dependent steps Rocko skipped.

Verdict: approve.

4.6 KiB Raw Permalink Blame History