Agent: night-review (inspiron-3670). Read: HANDBOOK, BRIEF, all four docs/night2/reviews.
Do:
Review every PR into night/2026-11-17 within about 15 minutes of status/needs-review. End with VERDICT: APPROVED or VERDICT: CHANGES plus a list.
Check:
tests pass on Node 24 (run them in node:24-alpine on a merge preview)
no consensus or protocol changes (see HANDBOOK)
no Wikimedia calls in tests
no information leaks in viewFor/logFor
ownership is enforced on the server
generated text follows BRIEF decision 5: templates only, real data only, no invented facts about real people (lean in otherwise)
QA from 23:15: play against bots in the browser with Playwright at 360×740 and 1280×800. From 00:30, run PvP end to end on the testnet nodes (http://192.168.1.32:3101, 3102 and 3103) with the alice and bob wallets that testnet-ops provides. File bugs as comments on the owning team's issue.
Final gate after the 01:15 freeze: full tests and build on the head SHA. Post the result with the SHA here.
**Agent:** night-review (inspiron-3670). **Read:** HANDBOOK, BRIEF, all four docs/night2/reviews.
**Do:**
- **Review every PR into `night/2026-11-17`** within about 15 minutes of `status/needs-review`. End with `VERDICT: APPROVED` or `VERDICT: CHANGES` plus a list.
- **Check:**
- tests pass on Node 24 (run them in `node:24-alpine` on a merge preview)
- no consensus or protocol changes (see HANDBOOK)
- no Wikimedia calls in tests
- no information leaks in `viewFor`/`logFor`
- ownership is enforced on the server
- generated text follows BRIEF decision 5: templates only, real data only, no invented facts about real people (lean in otherwise)
- **QA from 23:15:** play against bots in the browser with Playwright at 360×740 and 1280×800. From 00:30, run PvP end to end on the testnet nodes (`http://192.168.1.32:3101`, 3102 and 3103) with the alice and bob wallets that testnet-ops provides. File bugs as comments on the owning team's issue.
- **Final gate after the 01:15 freeze:** full tests and build on the head SHA. Post the result with the SHA here.
claude
added this to the Night 2: Magic-style duel, PvP, polite Wikimedia milestone 2026-11-17 18:07:19 +00:00
[night-review] starting (21:40). Clone ok, Gitea MCP ok, node:24-alpine image present (host node is 20, so all tests run in Docker).
Plan:
Baseline: npm ci && npm test && npm run build on night/2026-11-17 head in node:24-alpine (under flock /tmp/night-heavy.lock), so later PR failures can be attributed.
Poll open PRs into night/2026-11-17 every ≤15 min; for each status/needs-review PR: merge-preview locally, run tests+build on Node 24, check the HANDBOOK hard rules (no consensus/protocol diffs, no network in tests, tests/<agent>-*.test.ts, no new deps), plus viewFor/logFor leaks, server-side ownership, BRIEF decision 5 text rules. Post VERDICT: APPROVED / VERDICT: CHANGES + list.
From 23:15: Playwright bot matches at 360×740 and 1280×800 against a local build; bugs go as comments on the owning team's issue.
From 00:30: PvP end to end on testnet (192.168.1.32:3101-3103) with alice/bob.
After 01:15 freeze: full tests + build on head SHA, result + SHA posted here; final report before 02:00.
[night-review] starting (21:40). Clone ok, Gitea MCP ok, `node:24-alpine` image present (host node is 20, so all tests run in Docker).
Plan:
- Baseline: `npm ci && npm test && npm run build` on `night/2026-11-17` head in `node:24-alpine` (under `flock /tmp/night-heavy.lock`), so later PR failures can be attributed.
- Poll open PRs into `night/2026-11-17` every ≤15 min; for each `status/needs-review` PR: merge-preview locally, run tests+build on Node 24, check the HANDBOOK hard rules (no consensus/protocol diffs, no network in tests, `tests/<agent>-*.test.ts`, no new deps), plus viewFor/logFor leaks, server-side ownership, BRIEF decision 5 text rules. Post `VERDICT: APPROVED` / `VERDICT: CHANGES` + list.
- From 23:15: Playwright bot matches at 360×740 and 1280×800 against a local build; bugs go as comments on the owning team's issue.
- From 00:30: PvP end to end on testnet (192.168.1.32:3101-3103) with alice/bob.
- After 01:15 freeze: full tests + build on head SHA, result + SHA posted here; final report before 02:00.
Baselinenight/2026-11-17 @ 2af2e7f (Node 24, --network none): 241 pass, 1 fail, 1 skipped. The failure is testnet-ledger "peer sync pages at most 32 blocks…", a 5 s until() deadline under load avg 9. Re-run alone it passes 25/25, so it's a load flake and not a regression. Build OK.
Method: a merge preview per PR, then tsc --noEmit and the PR's own test files in node:24-alpine with --network none, so any Wikimedia call in a test would fail.
The heavy lock was held for 15+ min by builds, so the quick checks run at --cpus=1 nice 19 without it.
The full suite and build now run under the lock on a combined preview of all 10 PRs. The #73/#74 server/testnet/http.ts edits auto-merge cleanly, so ignore my conflict warning on those two.
Next: combined suite result here, then reviews as new PRs arrive, then Playwright bot-match QA once the engine alpha and ui-duel land.
For night-lead: nothing blocks. A suggested merge order is #72 → #70 → #69 → #66 → the rest, so the dependent PRs (fromDuel adapter, engine alpha) can branch.
[night-review] Status at 22:13.
**Baseline** `night/2026-11-17` @ 2af2e7f (Node 24, `--network none`): 241 pass, 1 fail, 1 skipped. The failure is `testnet-ledger` "peer sync pages at most 32 blocks…", a 5 s `until()` deadline under load avg 9. Re-run alone it passes 25/25, so it's a load flake and not a regression. Build OK.
**Reviewed, all `VERDICT: APPROVED`, each with notes:**
| PR | Team | Notes worth knowing |
|---|---|---|
| #66 | design | discards are public, Featured vs Garrison, the rivalry wording (since fixed) |
| #67 | autodeploy | |
| #68 | identity | |
| #69 | story | seed secrecy, life tracking via `lifeGain` |
| #70 | cards | `ConnectKeys` (duel-core has since added `qid`/`cites`) |
| #71 | sim | stand-in numbers must not reach the REPORT |
| #72 | contracts | seed secrecy (now in DUEL_API.md as `storySeed`) |
| #73 | wikimedia | |
| #74 | ownership | |
| #75 | compose | |
**Method:** a merge preview per PR, then `tsc --noEmit` and the PR's own test files in `node:24-alpine` with `--network none`, so any Wikimedia call in a test would fail.
- The heavy lock was held for 15+ min by builds, so the quick checks run at `--cpus=1 nice 19` without it.
- The full suite and build now run under the lock on a **combined preview of all 10 PRs**. The #73/#74 `server/testnet/http.ts` edits auto-merge cleanly, so ignore my conflict warning on those two.
**Next:** combined suite result here, then reviews as new PRs arrive, then Playwright bot-match QA once the engine alpha and ui-duel land.
**For night-lead:** nothing blocks. A suggested merge order is #72 → #70 → #69 → #66 → the rest, so the dependent PRs (fromDuel adapter, engine alpha) can branch.
The preview is night/2026-11-172af2e7f with all 10 open PRs merged on top: #66, #67, #68, #69, #70, #71, #72, #73, #74 and #75 (preview 19d7fc2). Run in node:24-alpine with --network none:
npm test: 329 tests, 328 pass, 0 fail, 1 skipped
npm run build: OK
There are no conflicts between any of them, and #73 and #74 auto-merge cleanly in server/testnet/http.ts. night-lead can merge them in any order.
QA prep, also posted on #63: the LAN testnet (http://192.168.1.32:310x) is not a secure context, so crypto.subtle is missing and the browser wallet fails. Chromium with --unsafely-treat-insecure-origin-as-secure=… (full chromium channel in Playwright) works: alice's backup restores on node-a.
[night-review] **Combined gate, 22:27: green.**
The preview is `night/2026-11-17` 2af2e7f with all 10 open PRs merged on top: #66, #67, #68, #69, #70, #71, #72, #73, #74 and #75 (preview 19d7fc2). Run in `node:24-alpine` with `--network none`:
- `npm test`: **329 tests, 328 pass, 0 fail, 1 skipped**
- `npm run build`: OK
There are no conflicts between any of them, and #73 and #74 auto-merge cleanly in `server/testnet/http.ts`. night-lead can merge them in any order.
**QA prep, also posted on #63:** the LAN testnet (`http://192.168.1.32:310x`) is not a secure context, so `crypto.subtle` is missing and the browser wallet fails. Chromium with `--unsafely-treat-insecure-origin-as-secure=…` (full `chromium` channel in Playwright) works: alice's backup restores on node-a.
[night-review] Status at 23:45. Still no merges into night/2026-11-17 (2af2e7f), and night-lead is unresponsive. The orchestrator has logged this on #53 for Eric.
Combined gate 2: green. The preview is 2af2e7f plus the current heads of #66, #67, #68, #69, #70, #71, #73, #74, #75, #78 and #79, plus duel-core's engine-wip (which supersedes #72). Preview 4d1005b, node:24-alpine, --network none:
npm test: 376 tests, 375 pass, 0 fail, 1 skipped
npm run build: OK
Two real conflicts, both with #74 (posted on the PRs):
QA ahead of the merges (from pushed branches, no stacked PRs):
Engine soak of engine-wip: 300 random matches, with 0 hidden-information leaks in viewFor/logFor, exact replay and termination. The first player won 62.7% under random play, against sim-balance's 57% with heuristic bots and a design target of 48–54%. duel-core added the secondPlayerCitations lever (#55, #64).
Board: Playwright preview scenes at 360×740 and 1280×800. No overflow, no errors and no small targets. 4 P3 layout items, which ui-duel has fixed (#58).
Testnet: the browser wallet needs a secure origin on the LAN, and there's a workaround (#63). PvP end to end can't start until #72, #74 and #77 are merged and deployed.
Ops: a ui-duel sim has held the inspiron heavy lock for 38+ min, which is blocking builds. I've asked them on #58 to check it.
[night-review] **Status at 23:45.** Still no merges into `night/2026-11-17` (2af2e7f), and night-lead is unresponsive. The orchestrator has logged this on #53 for Eric.
**Combined gate 2: green.** The preview is 2af2e7f plus the current heads of #66, #67, #68, #69, #70, #71, #73, #74, #75, #78 and #79, plus duel-core's `engine-wip` (which supersedes #72). Preview 4d1005b, `node:24-alpine`, `--network none`:
- `npm test`: **376 tests, 375 pass, 0 fail, 1 skipped**
- `npm run build`: OK
**Two real conflicts, both with #74** (posted on the PRs):
- #77 conflicts in `server/game/index.ts`.
- #76 conflicts in `src/App.tsx`.
- Merge #74 first, then rebase those two. Every other pair merges cleanly.
**Reviewed since the last status, all `VERDICT: APPROVED`:**
- #76 and #77, with follow-ups already addressed in 53736e1 and 3da3e5a
- #78 (mobile mining)
- #79 (ui-duel board)
**QA ahead of the merges** (from pushed branches, no stacked PRs):
- **Engine soak** of `engine-wip`: 300 random matches, with 0 hidden-information leaks in `viewFor`/`logFor`, exact replay and termination. The first player won 62.7% under random play, against sim-balance's 57% with heuristic bots and a design target of 48–54%. duel-core added the `secondPlayerCitations` lever (#55, #64).
- **Board:** Playwright preview scenes at 360×740 and 1280×800. No overflow, no errors and no small targets. 4 P3 layout items, which ui-duel has fixed (#58).
- **Testnet:** the browser wallet needs a secure origin on the LAN, and there's a workaround (#63). PvP end to end can't start until #72, #74 and #77 are merged and deployed.
**Ops:** a ui-duel sim has held the inspiron heavy lock for 38+ min, which is blocking builds. I've asked them on #58 to check it.
[night-review] Final gate at 01:03: green, on an integration preview, not on the branch head.night/2026-11-17 is still 2af2e7f: nothing has merged tonight. So the "full tests and build on the head SHA" gate ran on a local preview of everything that's ready.
Preview SHA: 72f9ba3. It's base 2af2e7f plus these heads, merged in this order:
#76 (ui-pvp): a real conflict with #74 in src/App.tsx. Merge #74 first, then rebase #76.
Merge hazards for whoever merges:
part2-wip rewrote server/testnet/http.ts with LF line endings, while the base is CRLF, so a plain merge with #73 conflicts on every line. I asked host-pvp on #56 to restore CRLF.
#77 and #74 conflict in server/game/index.ts. part2-wip already resolves this.
Logs are on inspiron at ~/worker/jobs/night2/night-review/logs/final.{test,build}.log.
[night-review] **Final gate at 01:03: green, on an integration preview, not on the branch head.** `night/2026-11-17` is still **2af2e7f**: nothing has merged tonight. So the "full tests and build on the head SHA" gate ran on a local preview of everything that's ready.
**Preview SHA: `72f9ba3`.** It's base 2af2e7f plus these heads, merged in this order:
| Branch | Head | Covers |
|---|---|---|
| `host-pvp/part2-wip` | ae28050 | #77, #74, #72 + `engine-wip` 0eb80d2, #70, #68, and host-pvp part 2 |
| `duel-design/design-v1` | f881e2c | #66 |
| `testnet-ops/autodeploy` | 2f494e3 | #67 |
| `duel-design/story` | 39ffe40 | #69 |
| `sim-balance/duel-harness` | 51bdbe0 | #71 |
| `wikimedia/client` | f544019 | #73, merged with `-X ignore-space-at-eol` (see below) |
| `testnet-ops/compose-wikimedia` | da37378 | #75 |
| `mobile-mining/android-miner` | 15470d5 | #78 |
| `ui-duel/board` | 3e6d597 | #79 |
**Result** (`node:24-alpine`, `--network none`):
- `npm test`: **396 tests, 395 pass, 0 fail, 1 skipped**
- `npm run build`: **OK**
**Not in the preview:**
- **#76 (ui-pvp)**: a real conflict with #74 in `src/App.tsx`. Merge #74 first, then rebase #76.
**Merge hazards for whoever merges:**
1. `part2-wip` rewrote `server/testnet/http.ts` with LF line endings, while the base is CRLF, so a plain merge with #73 conflicts on every line. I asked host-pvp on #56 to restore CRLF.
2. #77 and #74 conflict in `server/game/index.ts`. part2-wip already resolves this.
Logs are on inspiron at `~/worker/jobs/night2/night-review/logs/final.{test,build}.log`.
[night-review] Final report (01:15). I'm stopping here.
Summary
Every PR opened tonight is reviewed and carries VERDICT: APPROVED. The final gate on an integration preview of everything except #76 is green: 395/396 pass, build OK. But nothing merged into night/2026-11-17: it is still 2af2e7f. night-lead never responded after kickoff. The orchestrator has logged the stall on #53 for Eric, and nobody else may merge. So the testnet never got Night 2 code, and the PvP end-to-end on the testnet with alice and bob was not run.
Final gate, preview 72f9ba3 (details in comment 5885): 396 tests, 395 pass, 0 fail, 1 skipped; build OK.
QA (from pushed branches, without stacking PRs)
Engine soak of duel-core/engine-wip, 300 seeded random games each at 5cd0ffb and 0eb80d2:
0 hidden-information leaks in viewFor/logFor, checking hand/library uids and cardIds, the seed and rng.
Exact replay, a serialize round-trip, and termination.
The first player won 62.7%, which led to balance v1. At 0eb80d2 it's 51.0%.
turnCap 40 verified.
Board: Playwright at 360×740 and 1280×800, all preview scenes. No overflow, errors or tiny targets. 4 P3 layout items, all fixed by ui-duel.
P1 found and fixed: host-pvp's toDuelCard fallback typed every card as a Doctrine whenever there was no frozen identity, which is the situation on the testnet tonight. That made PvP unplayable. Fixed in part2-wip dfb8abe. Verified on all 93 real testnet blocks: Figure 25, Location 22, Institution 42, Event 2, Doctrine 2.
Testnet: the LAN nodes are plain HTTP, so they aren't a secure context, and the browser wallet can't work. The workaround is in TESTNET.md via #68. alice's backup restores with it.
Ops:
The heavy lock was held 15–40 min at a time by builds and sims, so per-PR checks ran at 1 CPU without the lock.
Part2-wip had a CRLF→LF rewrite of http.ts, which host-pvp has since fixed.
My incident: a broken fetch loop sent about 490 /api/canonical reads to node-a in about 2 min, around 00:48. They were rate-limited by the node, with no Wikimedia calls, and node-a stayed healthy. Details are on #56.
Pre-bake identities on the testnet (prebake-identity.ts --live) before opening PvP.
Pick two eom.dev accounts to link to alice and bob (#53). Then run PvP end to end on 3101–3103: invite, match, recap, Elo. I didn't get to do this tonight.
Eric: the RandomX verifier latch (server/testnet/randomx.ts:140), from testnet-ops.
The QA scripts are outside git on inspiron: ~/worker/jobs/night2/night-review/{bin,qa,wt/qa/qa-soak.ts}. I pushed no code tonight; reviewing and QA was the whole job.
[night-review] **Final report (01:15).** I'm stopping here.
## Summary
Every PR opened tonight is reviewed and carries `VERDICT: APPROVED`. The final gate on an integration preview of everything except #76 is green: 395/396 pass, build OK. **But nothing merged into `night/2026-11-17`: it is still 2af2e7f.** night-lead never responded after kickoff. The orchestrator has logged the stall on #53 for Eric, and nobody else may merge. So the testnet never got Night 2 code, and **the PvP end-to-end on the testnet with alice and bob was not run**.
## Code review: 14 PRs, all `VERDICT: APPROVED`
| PR | Team | Notable outcome of review |
|---|---|---|
| #66 | duel-design | Deleted is public; Garrison vs Featured; clock wording. All fixed in v1.1. |
| #67 | testnet-ops | |
| #68 | testnet-ops | Re-reviewed at 26ad4af with the live identity source (background lane only, freezes only complete profiles). |
| #69 | duel-design | Rivalry text that implied a real relationship between real people, fixed. Tokens never cited. `storySeed`. |
| #70 | duel-cards | ConnectKeys aligned with #72 (`qid`/`cites`). |
| #71 | sim-balance | Stand-in numbers must not reach the REPORT. |
| #72 | duel-core | **The seed is a hidden-information secret.** DUEL_API now forbids sending `seed`/`rng` before gameOver. |
| #73 | wikimedia | Checked against the full BRIEF politeness list. Mint path is transport-only; the mint bug is fixed. |
| #74 | ownership | Server-side ownership confirmed. Wallet address is a primary key (no shared wallets). |
| #75 | testnet-ops | |
| #76 | ui-pvp | No "decisive" claims in the fallback recap; `/recap/<id>` is canonical. **Conflicts with #74 in `src/App.tsx`.** |
| #77 | host-pvp | Quick match skips stale lobbies; draft pool shown to seats only. **Conflicts with #74 in `server/game/index.ts`.** |
| #78 | mobile-mining | Store ban documented, consent, mainnet disabled, idle-lock trade-off. |
| #79 | ui-duel | Browser bot matches must never feed Elo, confirmed. |
## Gates (all `node:24-alpine`, `--network none`, so any Wikimedia call in a test would fail)
- **Baseline** 2af2e7f: 241 pass, 1 flaky. `testnet-ledger` peer sync has a 5 s deadline under load; it passes alone.
- **Combined #66–#75:** 328/329 pass. **Combined 2** (heads plus engine): 375/376 pass.
- **Final gate**, preview `72f9ba3` (details in comment 5885): **396 tests, 395 pass, 0 fail, 1 skipped; build OK.**
## QA (from pushed branches, without stacking PRs)
- **Engine soak** of `duel-core/engine-wip`, 300 seeded random games each at 5cd0ffb and 0eb80d2:
- **0 hidden-information leaks** in `viewFor`/`logFor`, checking hand/library uids and cardIds, the seed and `rng`.
- Exact replay, a serialize round-trip, and termination.
- The first player won 62.7%, which led to balance v1. At 0eb80d2 it's **51.0%**.
- `turnCap` 40 verified.
- **Board:** Playwright at **360×740 and 1280×800**, all preview scenes. No overflow, errors or tiny targets. 4 P3 layout items, all fixed by ui-duel.
- **P1 found and fixed:** host-pvp's `toDuelCard` fallback typed **every card as a Doctrine** whenever there was no frozen identity, which is the situation on the testnet tonight. That made PvP unplayable. Fixed in part2-wip dfb8abe. Verified on **all 93 real testnet blocks**: Figure 25, Location 22, Institution 42, Event 2, Doctrine 2.
- **Testnet:** the LAN nodes are plain HTTP, so they aren't a secure context, and the browser wallet can't work. The workaround is in TESTNET.md via #68. alice's backup restores with it.
- **Ops:**
- The heavy lock was held 15–40 min at a time by builds and sims, so per-PR checks ran at 1 CPU without the lock.
- Part2-wip had a CRLF→LF rewrite of `http.ts`, which host-pvp has since fixed.
- **My incident:** a broken fetch loop sent about 490 `/api/canonical` reads to node-a in about 2 min, around 00:48. They were rate-limited by the node, with no Wikimedia calls, and node-a stayed healthy. Details are on #56.
## Left, for night-lead or Eric
1. **Merge** in this order: #72, #66 → #70 → #69 → #73 → #74 → #77 (rebase) → #76 (rebase) → #67, #68, #71, #75, #78, #79. Then duel-core's engine PR and host-pvp part 2.
2. **Pre-bake identities** on the testnet (`prebake-identity.ts --live`) **before** opening PvP.
3. Pick **two eom.dev accounts** to link to alice and bob (#53). Then run PvP end to end on 3101–3103: invite, match, recap, Elo. I didn't get to do this tonight.
4. Eric: the RandomX verifier latch (`server/testnet/randomx.ts:140`), from testnet-ops.
The QA scripts are outside git on inspiron: `~/worker/jobs/night2/night-review/{bin,qa,wt/qa/qa-soak.ts}`. I pushed no code tonight; reviewing and QA was the whole job.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Agent: night-review (inspiron-3670). Read: HANDBOOK, BRIEF, all four docs/night2/reviews.
Do:
night/2026-11-17within about 15 minutes ofstatus/needs-review. End withVERDICT: APPROVEDorVERDICT: CHANGESplus a list.node:24-alpineon a merge preview)viewFor/logForhttp://192.168.1.32:3101, 3102 and 3103) with the alice and bob wallets that testnet-ops provides. File bugs as comments on the owning team's issue.[night-review] starting (21:40). Clone ok, Gitea MCP ok,
node:24-alpineimage present (host node is 20, so all tests run in Docker).Plan:
npm ci && npm test && npm run buildonnight/2026-11-17head innode:24-alpine(underflock /tmp/night-heavy.lock), so later PR failures can be attributed.night/2026-11-17every ≤15 min; for eachstatus/needs-reviewPR: merge-preview locally, run tests+build on Node 24, check the HANDBOOK hard rules (no consensus/protocol diffs, no network in tests,tests/<agent>-*.test.ts, no new deps), plus viewFor/logFor leaks, server-side ownership, BRIEF decision 5 text rules. PostVERDICT: APPROVED/VERDICT: CHANGES+ list.[night-review] Status at 22:13.
Baseline
night/2026-11-17@2af2e7f(Node 24,--network none): 241 pass, 1 fail, 1 skipped. The failure istestnet-ledger"peer sync pages at most 32 blocks…", a 5 suntil()deadline under load avg 9. Re-run alone it passes 25/25, so it's a load flake and not a regression. Build OK.Reviewed, all
VERDICT: APPROVED, each with notes:lifeGainConnectKeys(duel-core has since addedqid/cites)storySeed)Method: a merge preview per PR, then
tsc --noEmitand the PR's own test files innode:24-alpinewith--network none, so any Wikimedia call in a test would fail.--cpus=1 nice 19without it.server/testnet/http.tsedits auto-merge cleanly, so ignore my conflict warning on those two.Next: combined suite result here, then reviews as new PRs arrive, then Playwright bot-match QA once the engine alpha and ui-duel land.
For night-lead: nothing blocks. A suggested merge order is #72 → #70 → #69 → #66 → the rest, so the dependent PRs (fromDuel adapter, engine alpha) can branch.
[night-review] Combined gate, 22:27: green.
The preview is
night/2026-11-172af2e7fwith all 10 open PRs merged on top: #66, #67, #68, #69, #70, #71, #72, #73, #74 and #75 (preview 19d7fc2). Run innode:24-alpinewith--network none:npm test: 329 tests, 328 pass, 0 fail, 1 skippednpm run build: OKThere are no conflicts between any of them, and #73 and #74 auto-merge cleanly in
server/testnet/http.ts. night-lead can merge them in any order.QA prep, also posted on #63: the LAN testnet (
http://192.168.1.32:310x) is not a secure context, socrypto.subtleis missing and the browser wallet fails. Chromium with--unsafely-treat-insecure-origin-as-secure=…(fullchromiumchannel in Playwright) works: alice's backup restores on node-a.[night-review] Status at 23:45. Still no merges into
night/2026-11-17(2af2e7f), and night-lead is unresponsive. The orchestrator has logged this on #53 for Eric.Combined gate 2: green. The preview is
2af2e7fplus the current heads of #66, #67, #68, #69, #70, #71, #73, #74, #75, #78 and #79, plus duel-core'sengine-wip(which supersedes #72). Preview 4d1005b,node:24-alpine,--network none:npm test: 376 tests, 375 pass, 0 fail, 1 skippednpm run build: OKTwo real conflicts, both with #74 (posted on the PRs):
server/game/index.ts.src/App.tsx.Reviewed since the last status, all
VERDICT: APPROVED:53736e1and3da3e5aQA ahead of the merges (from pushed branches, no stacked PRs):
engine-wip: 300 random matches, with 0 hidden-information leaks inviewFor/logFor, exact replay and termination. The first player won 62.7% under random play, against sim-balance's 57% with heuristic bots and a design target of 48–54%. duel-core added thesecondPlayerCitationslever (#55, #64).Ops: a ui-duel sim has held the inspiron heavy lock for 38+ min, which is blocking builds. I've asked them on #58 to check it.
[night-review] Final gate at 01:03: green, on an integration preview, not on the branch head.
night/2026-11-17is still2af2e7f: nothing has merged tonight. So the "full tests and build on the head SHA" gate ran on a local preview of everything that's ready.Preview SHA:
72f9ba3. It's base2af2e7fplus these heads, merged in this order:host-pvp/part2-wipae28050engine-wip0eb80d2, #70, #68, and host-pvp part 2duel-design/design-v1f881e2ctestnet-ops/autodeploy2f494e3duel-design/story39ffe40sim-balance/duel-harness51bdbe0wikimedia/clientf544019-X ignore-space-at-eol(see below)testnet-ops/compose-wikimediada37378mobile-mining/android-miner15470d5ui-duel/board3e6d597Result (
node:24-alpine,--network none):npm test: 396 tests, 395 pass, 0 fail, 1 skippednpm run build: OKNot in the preview:
src/App.tsx. Merge #74 first, then rebase #76.Merge hazards for whoever merges:
part2-wiprewroteserver/testnet/http.tswith LF line endings, while the base is CRLF, so a plain merge with #73 conflicts on every line. I asked host-pvp on #56 to restore CRLF.server/game/index.ts. part2-wip already resolves this.Logs are on inspiron at
~/worker/jobs/night2/night-review/logs/final.{test,build}.log.[night-review] Final report (01:15). I'm stopping here.
Summary
Every PR opened tonight is reviewed and carries
VERDICT: APPROVED. The final gate on an integration preview of everything except #76 is green: 395/396 pass, build OK. But nothing merged intonight/2026-11-17: it is still2af2e7f. night-lead never responded after kickoff. The orchestrator has logged the stall on #53 for Eric, and nobody else may merge. So the testnet never got Night 2 code, and the PvP end-to-end on the testnet with alice and bob was not run.Code review: 14 PRs, all
VERDICT: APPROVED26ad4afwith the live identity source (background lane only, freezes only complete profiles).storySeed.qid/cites).seed/rngbefore gameOver./recap/<id>is canonical. Conflicts with #74 insrc/App.tsx.server/game/index.ts.Gates (all
node:24-alpine,--network none, so any Wikimedia call in a test would fail)2af2e7f: 241 pass, 1 flaky.testnet-ledgerpeer sync has a 5 s deadline under load; it passes alone.72f9ba3(details in comment 5885): 396 tests, 395 pass, 0 fail, 1 skipped; build OK.QA (from pushed branches, without stacking PRs)
duel-core/engine-wip, 300 seeded random games each at5cd0ffband0eb80d2:viewFor/logFor, checking hand/library uids and cardIds, the seed andrng.0eb80d2it's 51.0%.turnCap40 verified.toDuelCardfallback typed every card as a Doctrine whenever there was no frozen identity, which is the situation on the testnet tonight. That made PvP unplayable. Fixed in part2-wipdfb8abe. Verified on all 93 real testnet blocks: Figure 25, Location 22, Institution 42, Event 2, Doctrine 2.http.ts, which host-pvp has since fixed./api/canonicalreads to node-a in about 2 min, around 00:48. They were rate-limited by the node, with no Wikimedia calls, and node-a stayed healthy. Details are on #56.Left, for night-lead or Eric
prebake-identity.ts --live) before opening PvP.server/testnet/randomx.ts:140), from testnet-ops.The QA scripts are outside git on inspiron:
~/worker/jobs/night2/night-review/{bin,qa,wt/qa/qa-soak.ts}. I pushed no code tonight; reviewing and QA was the whole job.