stack/agents/darkwing/work/queue-81-review/review-r2.md

Row 81, the Console shows the reply to a web send, round 2 review (Darkwing)

Issue #1575, request comment 27816, packet agents/dewey/work/queue-81/r2/ (commit e244f4f5). Round 2 is the rebase after row 61 (6c42344d) plus my two round 1 notes. Base 1e5f5958. Candidate manifest sha256 3ce7527f652472755662f8366c4908d8304031b4f890dbce91e478df030b1fd5, 13 files, matching the request. I'm the only reviewer (lead decision 97).

Verdict: approve. The rebase changes nothing of row 81's own and drops nothing of row 61's. The two follow-ups say what they should. No findings.

Method

• A detached worktree at 1e5f5958 with git apply --index of r2/build.patch. git diff --cached --binary 1e5f5958 there hashes the same as the packet's patch (054fb26c…), and sha256sum -c gives 13/13 OK. No commit from 1e5f5958 to 1166e0de touches the 13 paths.
• Round 1 against round 2: the added and removed lines of each patch, in order (r2/r1.lines, r2/r2.lines). The only differences (r2/lines.diff) are the cli README talk entry and the rewrapped bus.mjs comment. Every other + and - line is identical.
• Row 61's lines: each line 6c42344d added to serve.mjs, public/bus.js and the webui README is still in the candidate, 8, 15 and 11 lines, none missing. The candidate's removed lines in those three files are row 81's own targets from round 1 (the "doesn't show it yet" text, the old fetchRead, BUS_VERBS without replies); none is a row 61 line.
• Row 61's /api/task-fields route is its own branch in serve.mjs, and its section in bus.js reads through get.api. Row 81 changes busRead's verb list and fetchRead's query name, so the two don't meet. Dewey says the same in build.md, and I checked it in the files.

The two follow-ups

• packages/cli/README.md: "the reply may still come, and the next mosaic talk prints it". cli.mjs:212's timeout message says "it may still come, and the next mosaic talk shows it". Same claim, and true of talk, which prints unread replies after its send.
• packages/webui/src/bus.mjs: the comment is the same words, rewrapped. Its longest line is now 76 characters, the block's width.

Tests

Node suites on the candidate, sequential, TMPDIR in my scratch, after Filbert's row 77 run had finished (r2/*.txt):

Suite	Window (2026-10-12, UTC)	Result
webui	19:19:57 to 19:20:55	84 pass, 0 fail
bus	19:20:55 to 19:20:58	102 pass, 0 fail
cli	19:20:58 to 19:21:18	115 pass, 0 fail
control-board	19:21:18 to 19:21:19	124 pass, 0 fail

The counts match Dewey's gate. I didn't rerun tasks or the scripts suites: the candidate touches no path they load, Dewey's gate has them green, and Sage reruns the gate. I didn't rerun mutants. The code lines are byte for byte round 1's, where my nine mutants ran (eight killed, one survivor that note 3 covered), and Dewey's rerun of their 17 on this candidate killed all of them.