Repository navigation
fix(members): name the folded crewmate switcher as the roster - #18219
Conversation
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Check that the page title "Crewmates" still fits in the chip next to the identity pill in long locales. The screenshot covers English, dark theme only. [DESIGN-REVIEWED] b1c5d3a Description read[DESCRIPTION-READ] c4dbeb3e1c8e93625805296416e469d00a762a97ec3ef72c777f23ec38ba14ad That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Check that dropping "Needs your approval or answer" from the chip's tooltip is intended. The tooltip now always reads "Switch crewmate". What this change shipsInventory (4 items) — 2 justifiedIntent: let a first-time visitor who lands in a reopened thread see where the crew list is. This is a FIX (#18123, 5 sightings in the nightly GUI user test).
[FIRST-PRINCIPLES-REVIEWED] b1c5d3a Description read[DESCRIPTION-READ] e9d2392d50f30cc32f5930e08287d6cda1099dd43196ac293416bb2ea73e3974 That sha256 names the PR description this verdict read. Recompute it with |
UX Review (Opus 5.5) — 🟡 CONCERNSUX-level review of UX-Verdict: CONCERNS The blind reader identified the "Crewmates 4" chip as the crew switcher but rated it only "a guess", and confused it with the "radar Working" pill beside it. Watch
Suggestions
[UX-REVIEWED] b1c5d3a Description read[DESCRIPTION-READ] e0f54916688436692d951abcf783d39d7d984ac91fddb6ce67302daae786abcb That sha256 names everything this verdict read: the PR description, plus the 4 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
|
Dispositions for the advisory findings on
No code change for any of these, so no push. CI on this head keeps running. |
CrysisDeu
left a comment
There was a problem hiding this comment.
Approving at b1c5d3a. Read the full diff plus the owning spec; nothing blocking. Non-blocking notes, worth a follow-up or a quick fix before merge:
- [MINOR] website/src/pages/members/CrewmateSwitcher.tsx:170 -- the issue's stated expectation is "a list of member cards to choose from"; the PR keeps the auto-reopen of the last thread (citing #17808 / #16317) and fixes the DISCOVERABILITY of the folded roster, not the arrival behaviour the tester stalled on. That is a defensible product call and the body declares it, but it is a symptom-level fix relative to the issue text: the GUI-user lane will open a fresh issue if the next tester still reads "/members dropped me into a chat". Owner should confirm the label alone is the intended close for #18123 rather than, e.g., opening the roster popover once on first desktop arrival.
- [NIT] website/src/pages/members/CrewmateSwitcher.tsx:170 -- the label span has no
truncate/min-w-0and the button isshrink-0inside agrid-cols-[1fr_minmax(0,auto)_1fr]left cell (MembersPage.tsx:3703). With 3 stacked faces + label + count + chevron, a long locale at themdbreakpoint can push the chip into the centred identity pill. Real catalogs are short (maxfr"Coéquipiers", 11 chars;en-XApseudo is 26 and is the only one likely to collide), so this is a watch item, not a defect; evidence is English/dark only as the Design lane already noted. - [NIT] website/src/pages/members/CrewmateSwitcher.tsx:32-34 -- the re-wrapped doc comment leaves one line ("crewmate, the current one marked, and a "New crewmate" footer that opens the same create dialog the") well past the surrounding wrap width; re-flow it.
Problem / Motivation
Goal: A first-time visitor who opens Crewmates and lands in a reopened conversation can see where the crew list is.
On desktop, a bare
/membersreopens the crewmate you last chatted with, and the roster folds into the header switcher chip. That chip showed only stacked faces, a number and a chevron, with no visible words. The nightly GUI user test read the page as "Crewmates skipped the member list and dropped me straight into nova-sky's existing chat" (5 sightings).Why it matters
Every first look at the Crewmates page on desktop goes through this. A user who has chatted with a crewmate once cannot tell how to reach the others, because the only way back to the roster is a chip that does not say what it is.
Not a goal
md, where nothing auto-opens and the roster is the first screen.What changed (motivation → approach → change)
CrewmateSwitcher's closed chip carried its name only inaria-label/title. Sighted users saw faces + a count with no text.pages.membersPage.titlekey, so no new strings or locale edits. Following WCAG 2.5.3 (label in name), the accessible name now starts with that visible word ("Crewmates", or "Crewmates, Needs your approval or answer"). The action "Switch crewmate" moves to the tooltip. The tooltip now always shows only that action. The needs-you signal stays in the visiblewarndot on the faces and in the accessible name, but it is no longer repeated in the tooltip.docs/system-specs/modules/crew-mode.md(switcher section) updated to match.Backwards compatibility
Compatible: same chip, same test ids, same popover and switching path. Only the visible text and the accessible name change.
Removes nothing: the change adds a visible label and re-words the chip's accessible name. No path, key, record or capability is removed.
Tests
website/src/pages/members/CrewmateSwitcher.test.tsx: the closed-chip test now checks that the visiblecrewmate-switcher-labelreads "Crewmates", that the accessible name is "Crewmates", and that the tooltip is "Switch crewmate". The needs-you tests now expect "Crewmates, Needs your approval or answer". Mutation check: removing the label text makes the closed-chip test fail (1 failed / 15 passed). With the change in place, 16/16 pass (npx vitest run src/pages/members/CrewmateSwitcher.test.tsx).npx eslinton both changed files: clean.Manual verification
I ran
website/scripts/capture-members-remember-last.mjs(realMembersPage, gateway-free) against a local vite. Frame01-arrivalis the issue's exact state: desktop/membersreopens a thread, and the chip now reads "Crewmates 4 ▾" next to the identity pill, with no overlap. I captured dark theme only. The script's later back-navigation frame (04) timed out on a click and never reached a screenshot. It does not cover this chip. I did not run macOS or Windows.Screenshots / video
After: desktop, bare
/membersreopens the last thread, and the header chip names the roster.Before: see the screenshot in #18123 (faces + count only).
Related Issues
Closes #18123
Pattern harvest
Not generalizable: one chip had dropped its visible label when the roster column folded into it. Other icon-only header controls use
aria-label/titleon purpose and are not a lost-label case.Decision for the owner
Clicking Crewmates should keep reopening the last conversation, and the issue asks how a first-time user is supposed to find the other crewmates from there. This draft keeps the reopen and labels the folded roster chip "Crewmates", the smallest form of the triage's recommended option, instead of opening the roster with nothing selected, which would cost every returning user an extra click.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)Contribution License Agreement
Contributions are accepted under the project's Apache-2.0 license; see the Licensing section of CONTRIBUTING.md.