J1: the campaign set up and its first session — lane 3's acceptance, and two defects it found (#726, #729) #787

Merged

claude-bot merged 4 commits from build/lane3-j1 into main 2026-09-23 20:56:32 +00:00

Conversation 0 Commits 4 Files changed 15 +786 -54

claude-bot commented 2026-09-23 19:33:26 +00:00

Collaborator

Copy link

The last piece of lane 3. #729's acceptance requires J1 and J6 green on desktop, phone and keyboard-only with axe clean. J6 landed in #770. J1's lane-3 half was a test.skip with a TODO addressed to lane 3 by name: time zone, invite, Discord, and the first session. It's written now.

sh scripts/journeys.sh, whole suite: 27 passed, 3 skipped (J2's lane-4 skip in each project), 0 failed. a11y-baseline.js is still {}, with no tolerated entries. vitest 1928/1928; lint 0 errors; test_sessions, test_pagination and test_series 88/88; ruff clean.

Two defects the journey found, fixed here

1. Settings → Discord wiped what you'd typed when you tested the channel, and its verdict never appeared. The same cause made "Saved" vanish the moment it showed, on Discord and on Recording, retention & AI.

• After a test or a save, the section called useSettingsAccess().reload. That goes back through the loading state, so the campaign becomes null, SettingsSection draws Busy, and the form unmounts. It then remounts from what's stored: unsaved values gone, verdict gone.
• The fix is a quiet refresh that keeps the campaign on screen while it re-reads. It's only honoured over the load it was made against, so a stale re-read can't overwrite a newer page load. reload stays as the retry after a failure, where there's nothing on screen to keep.
• Why the unit tests missed it: a mock that resolves immediately is batched into one render, so the form never visibly unmounted. The new tests keep the re-read pending, and they fail on the old code.

2. The sessions list said "0 times" under every open vote. Sessions.jsx read time_slots, which SessionListItem has never carried; the unit fixture had invented it. The list response now has time_slot_count, counted in one grouped query and only for votes, so it's not an N+1. It's additive on the user API, not /api/bot/*, so there's no contract bump. "1 time" is no longer "1 times".

What J1 now walks

1. Dana creates a campaign under a run-unique name and opens it from the Home card, landing on This week with the four-item checklist.
2. Checklist → Settings → General. Set the zone (Asia/Tokyo) and add a reminder. The reminder offsets are read from the campaign and never hardcoded (repo CLAUDE.md); after a reload, the form must equal exactly what's stored.
3. Checklist → Invites & joining. Make a labelled link.
4. Contents → Discord.
   • The "Nothing is posted anywhere" banner (D67) shows, and "Send a test message" is aria-disabled with a reason. It's still focusable, and pressing it sends nothing (D36).
   • Fill in the channel and test: the fake bot's log grows by exactly one request with this run's guild and channel.
   • Queue a refusal on the fake. The test shows that exact message in role=alert, and This week shows the delivery banner (#369).
5. Checklist → New session, in vote mode with two candidate times. The fake bot receives session_proposed for this session with 2 slots.
6. D17, below.
7. The invitation. Sign out, open the invite, sign in as Wren. Wren lands on the campaign, not Home.
8. Wren leaves via Members & characters. The three projects share Wren, and J1's first test needs Wren in no campaign; a finally cleans up if a run dies between joining and leaving.

Axe is clean at WCAG 2.2 AA on every page reached, in all three projects. That covers This week in four states and as the invited player, the Settings sections, the session form, the Sessions list, the public invitation page, and Members & characters as a player.

Step 1 was written before D17 was corrected, and was reinterpreted

The TODO said to check that "every time shown afterwards is in [the campaign's zone]". That's the retired behaviour. D17 says one time, always the reader's own, with a line reconciling it to the table's day when the two differ.

• The table is in Asia/Tokyo, set in Settings. The reader is in America/Chicago, set with Playwright's timezoneId on this test only, so no other journey's dates move.
• A 10:00 AM Tokyo session is the previous evening in Chicago, with or without daylight saving, so the reconciling line is owed on every run whatever the date.
• Asserted: the stored instant reads exactly <day> 10:00 in Tokyo (set on the table's clock), and the page shows the browser's Chicago reading with a zone marker that's non-empty and not the table's (shown on the reader's clock). The table's own reading is absent from main, and "The table calls this 's session — you are N hours behind." appears.
• No abbreviation is spelled in the test. A zone's abbreviation is locale data, not a constant.

Found, not fixed: three items for the owner

1. The Contents rail doesn't pick up a campaign made in the same tab. AppFrame's useCampaigns fetches once per app load, so creating, renaming, joining or leaving without a full reload leaves a stale list. That's lane 1's shell. Re-fetching on pathname change would be cheap, given the client's 15 s cache and invalidate-on-mutation. The spec enters through the Home card for now and says why.
2. Sessions list, Past tab and search. SessionListItem has no summary or description, so a played row always says "No recap yet" and search only matches titles. The unit fixtures invent both fields. Summaries aren't gated on read, so returning them isn't a leak; it's a size choice between a server-side first sentence for the row and full text for search.
3. Home's fresh-campaign card sends both "Schedule the first session" and "Invite players" to /campaigns/:id, beside a comment saying "until lane 3 moves it". Lane 3 has moved it. Repointing needs a matching spec change.

Not walked

• Confirming a vote (lane 4, J2).
• The session page and its scan (lane 4, already recorded against J6 in #775).
• Nothing needed /campaigns/:id/manage.

🤖 Generated with Claude Code

**The last piece of lane 3.** #729's acceptance requires J1 and J6 green on desktop, phone and keyboard-only with axe clean. J6 landed in #770. J1's lane-3 half was a `test.skip` with a TODO addressed to lane 3 by name: time zone, invite, Discord, and the first session. It's written now. `sh scripts/journeys.sh`, whole suite: **27 passed, 3 skipped** (J2's lane-4 skip in each project), 0 failed. **`a11y-baseline.js` is still `{}`**, with no tolerated entries. vitest 1928/1928; lint 0 errors; `test_sessions`, `test_pagination` and `test_series` 88/88; ruff clean. ## Two defects the journey found, fixed here **1. Settings → Discord wiped what you'd typed when you tested the channel, and its verdict never appeared.** The same cause made "Saved" vanish the moment it showed, on Discord and on Recording, retention & AI. - After a test or a save, the section called `useSettingsAccess().reload`. That goes back through the loading state, so the campaign becomes null, `SettingsSection` draws Busy, and the form unmounts. It then remounts from what's *stored*: unsaved values gone, verdict gone. - The fix is a quiet `refresh` that keeps the campaign on screen while it re-reads. It's only honoured over the load it was made against, so a stale re-read can't overwrite a newer page load. `reload` stays as the retry after a failure, where there's nothing on screen to keep. - **Why the unit tests missed it:** a mock that resolves immediately is batched into one render, so the form never visibly unmounted. The new tests keep the re-read pending, and **they fail on the old code**. **2. The sessions list said "0 times" under every open vote.** `Sessions.jsx` read `time_slots`, which `SessionListItem` has never carried; the unit fixture had invented it. The list response now has `time_slot_count`, counted in one grouped query and only for votes, so it's not an N+1. It's additive on the user API, not `/api/bot/*`, so there's no contract bump. "1 time" is no longer "1 times". ## What J1 now walks 1. Dana creates a campaign under a run-unique name and opens it from the Home card, landing on This week with the four-item checklist. 2. **Checklist → Settings → General.** Set the zone (Asia/Tokyo) and add a reminder. The reminder offsets are **read from the campaign and never hardcoded** (repo `CLAUDE.md`); after a reload, the form must equal exactly what's stored. 3. **Checklist → Invites & joining.** Make a labelled link. 4. **Contents → Discord.** - The "Nothing is posted anywhere" banner (D67) shows, and "Send a test message" is `aria-disabled` with a reason. It's still focusable, and pressing it sends nothing (D36). - Fill in the channel and test: the fake bot's log grows by exactly one request with this run's guild and channel. - Queue a refusal on the fake. The test shows **that exact message** in `role=alert`, and This week shows the delivery banner (#369). 5. **Checklist → New session**, in vote mode with two candidate times. The fake bot receives `session_proposed` for *this* session with 2 slots. 6. **D17**, below. 7. **The invitation.** Sign out, open the invite, sign in as Wren. Wren lands on the campaign, not Home. 8. **Wren leaves** via Members & characters. The three projects share Wren, and J1's first test needs Wren in no campaign; a `finally` cleans up if a run dies between joining and leaving. **Axe is clean at WCAG 2.2 AA on every page reached, in all three projects.** That covers This week in four states and as the invited player, the Settings sections, the session form, the Sessions list, the public invitation page, and Members & characters as a player. ## Step 1 was written before D17 was corrected, and was reinterpreted The TODO said to check that "every time shown afterwards is in [the campaign's zone]". That's the retired behaviour. D17 says **one time, always the reader's own**, with a line reconciling it to the table's day when the two differ. - The **table** is in Asia/Tokyo, set in Settings. The **reader** is in America/Chicago, set with Playwright's `timezoneId` on this test only, so no other journey's dates move. - A 10:00 AM Tokyo session is the previous evening in Chicago, with or without daylight saving, so the reconciling line is owed on every run whatever the date. - Asserted: the stored instant reads exactly `<day> 10:00` in Tokyo (**set on the table's clock**), and the page shows the browser's Chicago reading with a zone marker that's non-empty and not the table's (**shown on the reader's clock**). The table's own reading is absent from `main`, and "The table calls this *<weekday>*'s session — you are N hours behind." appears. - **No abbreviation is spelled in the test.** A zone's abbreviation is locale data, not a constant. ## Found, not fixed: three items for the owner 1. **The Contents rail doesn't pick up a campaign made in the same tab.** `AppFrame`'s `useCampaigns` fetches once per app load, so creating, renaming, joining or leaving without a full reload leaves a stale list. That's lane 1's shell. Re-fetching on pathname change would be cheap, given the client's 15 s cache and invalidate-on-mutation. The spec enters through the Home card for now and says why. 2. **Sessions list, Past tab and search.** `SessionListItem` has no `summary` or `description`, so a played row always says "No recap yet" and search only matches titles. The unit fixtures invent both fields. Summaries aren't gated on read, so returning them isn't a leak; it's a size choice between a server-side first sentence for the row and full text for search. 3. **Home's fresh-campaign card** sends both "Schedule the first session" and "Invite players" to `/campaigns/:id`, beside a comment saying "until lane 3 moves it". Lane 3 has moved it. Repointing needs a matching spec change. ## Not walked - Confirming a vote (lane 4, J2). - The session page and its scan (lane 4, already recorded against J6 in #775). - Nothing needed `/campaigns/:id/manage`. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

claude-bot added 4 commits 2026-09-23 19:33:27 +00:00

fix(webapp): a Settings re-read keeps the form, and a vote in the list says how many times (#729) ... e5f2889530

Two defects in lane 3's own pages, both found by walking J1 in a real browser.

Settings: a Discord channel test, and a save on Discord or Recording, re-read
the campaign through useSettingsAccess's `reload` - the retry-after-failure
path, which goes back through "Loading...". That unmounted the section's form
and drew a fresh one from the stored campaign, so a test of an unsaved
channel wiped the channel and its verdict, and "Saved" vanished before it
could be read. useSettingsAccess now has a quiet `refresh` that keeps the
campaign on screen while it asks, honoured only over the load it was made
against; both pages use it. The unit tests missed it because a re-read that
resolves at once lands in the same render as its own start - the new ones
hold it on the wire, and fail on the old code.

Sessions list: a vote read `time_slots`, which the list endpoint never
carried, so every open vote said "0 times". SessionListItem gains an additive
`time_slot_count`, counted for the page's votes in one query.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

test(e2e): J1's lane-3 half - the campaign set up, and its first session (#729) ... 654cb0cb2e

Replaces the test.skip lane 1 left for lane 3. Dana makes a campaign and walks
its This week checklist: the table's time zone and one more reminder on
Settings > General (the rows are checked against what the campaign stores,
never a hardcoded timing), an invite link on Invites & joining, Discord -
D36's test-with-nothing-to-test, a real test the fake bot records, a queued
refusal the page reports in the bot's own words and This week's #369 banner,
and a passing test that clears it - then a vote on the session form, whose
session_proposed reaches the bot. Wren opens the invite signed out, signs in
through its own door, is admitted to the campaign, and leaves again through
Members so the next project still has a stranger.

Step 1 is written as D17 decides, not as the TODO worded it: the table plays
on Asia/Tokyo and the reader (Playwright's timezoneId) is in America/Chicago,
so a ten o'clock morning session is the evening before for the reader. It
asserts the instant was set on the table's clock, shown on the reader's with
a zone marker that is not the table's, and reconciled by "The table calls
this ...'s session". Markers are asked of the browser, never spelled. A vote
has no time to show, so that check is on a second session booked directly.

axe runs clean on every page reached. OIDC_ORIGIN is exported from fixtures
for the invitation's own sign-in.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Merge remote-tracking branch 'origin/main' into build/lane3-j1 4b2430aae8

docs(changelog): the two defects J1 found (#729) ... 783231ff03

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

claude-bot merged commit 71db38d943 into main 2026-09-23 20:56:32 +00:00

claude-bot referenced this pull request from a commit 2026-09-23 20:56:33 +00:00

Merge PR #787: J1, the campaign set up and its first session — lane 3's acceptance, and two defects it found (#726, #729)

claude-bot referenced this pull request 2026-09-23 20:56:52 +00:00

Build lane 3: the campaign core and Settings (CAMP-1 to CAMP-17, SH-7) #729

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels   

backend

Issue related to the backend

  

bot

Issue related to the bot

  

bug

Something is not working

  

duplicate

This issue or pull request already exists

  

enhancement

New feature

  

frontend

Issue related to the frontend

  

help wanted

Need some help

  

invalid

Something is wrong

  

question

More information is needed

  

recording

Issue related to the recording system

  

security

Security hardening / vulnerability work

  

wontfix

This won't be fixed

No labels

backend

bot

bug

duplicate

enhancement

frontend

help wanted

invalid

question

recording

security

wontfix

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

rbrooks/Quest-Board!787

No description provided.