Repository navigation
Conversation
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks. The Unread pass is narrow, sits in the row model the list and board both read, and is backed by committed screenshots of this revision. Notes
[DESIGN-REVIEWED] 3878124 Description read[DESCRIPTION-READ] ca78af9fe3b02540d8753a4d0ae2a02befdf43cca126dffc6a9b0f6eecbb4d9b 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 |
UX Review (Opus 5.5) — ✅ PASSUX-level review of UX-Verdict: PASS Nothing blocks: the diff adds no controls, and the screenshots show both the kept row and the row dropping out later. The blind reader correctly read the highlighted row as "the chat I have open right now". Notes
[UX-REVIEWED] 3878124 Description read[DESCRIPTION-READ] 38d85720f0f3849577aa7f4669fffd0b259a378a3bc7ada7aeb73b3448d276cf That sha256 names everything this verdict read: the PR description, plus the 6 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: |
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Nothing blocks: the one new behaviour removes a reported harm (#36670) at the filter that causes it. A test pins it and fails without the change. Notes
What this change shipsInventory (4 items) — 2 justifiedIntent: when Unread is on, the chat being worked in stays in the list instead of vanishing once opening it marks it read. FIX framed as feat.
[FIRST-PRINCIPLES-REVIEWED] 3878124 Description read[DESCRIPTION-READ] 6aea648f3712c4bc6d58fdfd462a4be4e2de66905d7ef053a3279e612c4a9a6e That sha256 names the PR description this verdict read. 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: |
f6b5916 to
c4510ef
Compare
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Thanks for the narrow fix. The code looks correct, but one evidence gap blocks the merge.
-
No still screenshots of this head (website/src/pages/chat-sidebar/rows.ts:633-634, the visible change). The UX review on c4510ef says it could not evaluate the change because the PR has no still screenshots. A review that cannot run for missing evidence counts as a block, not a pass. The only attachment is a video, and the body says it was recorded from f6b5916. In that commit the open chat passed every status chip. The head passes it under Unread only, so the video cannot show how the head behaves under In progress or Recent. Smallest fix: attach stills taken from c4510ef:
- the sidebar with Unread on, showing the opened, now read chat next to a chat that is still unread;
- the same list after you open another chat, with the first chat gone;
- the board with Unread on, still showing the open, read chat.
Re-record the video from this head or remove it, then re-run the UX review. No code change is needed.
Optional:
Closes #36670closes the issue, but the In progress case (named in your own Problem section) stays unfixed. ConsiderPart of #36670plus a follow-up issue, or retitling to name the Unread chip.- The
leavestable in rows.test.ts has nolocal.openKeyrow. Without one, removingopenKeyfrom the dimensions memo deps at rows.ts:904 turns no test red, and ChatSidebar always builds incrementally. Adding['local.openKey', withLocal('openKey', 'c')]would pin it. - No test pins the new
hidesclause at rows.ts:691. A case with Unread on plus a search that drops the open chat, asserting that only the search dimension hides it, would cover it.
Code change approver deep review: A person should decideThe open decision comes from #36670: should the open chat stay listed under Unread by default, or only behind a setting? This PR makes it the default for every user. No maintainer has recorded that decision on this PR or on the issue yet. If the answer is "default", no code change is needed. The review found no blocking issue on 3878124. What it checked on this head:
Not verified: this review did not run the tests, and it could not open the three committed screenshots. Done well: the rule lives in one place, the row model that both the list and the board read. Each excluded case (search, a status hide, a peer row with the same key) has its own test. Optional:
This changes behaviour users already rely on, so it needs a design sign-off from a maintainer before it can merge. A maintainer will post the decision here. To propose a design, follow the request-for-change process in CONTRIBUTING.md (docs/request-for-change/). Tier: 2, held for a person: reviewer-finding: Design Review. A deep review by the code change approver, not by a person · commit |
c4510ef to
598629b
Compare
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Thanks for narrowing the change. The code looks correct on this head, but one evidence gap still blocks the merge.
- No design or UX review has seen this change rendered (website/src/pages/chat-sidebar/rows.ts:647, and the status
hideschange around rows.ts:700). With Unread on,passesChipkeeps the open chat in the list and on the board, so users see a change. The Design and UX reviews on f8f7124 both say they could not evaluate it: Design downloaded 0 images and UX saw each image as "[image removed]". The three images in the description return HTTP 404 when fetched without credentials, so a re-run cannot admit them either. A review that cannot run because evidence is missing counts as a block, even though its header reads PASS. Smallest fix: commit the three stills under.github/screenshots/, or re-upload them so the links are public. Then re-run the Design and UX reviews on this head. No code change is needed. The stills come from 598629b, but the rebase did not change your logic, so you do not need to retake them. One line in the description saying so is enough.
Status of my earlier review on c4510ef:
- Stills of the head: still open. They are attached, but no review could load them (item 1 above).
Part of #36670and a title that names Unread: fixed.local.openKeyrow in the memoleavestable: fixed.- Test for the new
hidesclause: not added (optional, see below).
Optional:
- No test fails if you delete the new
hidesclause at rows.ts:704. In the "Unread plus search 'beta'" case, assert that only the search dimension reportshides(opened)as true. - The test "stays listed under Unread after opening it marked it read" never makes the open row unread first. It actually pins a wider rule: any open chat shows under Unread, even one that was never unread. If that is the intent, rename the test and add one sentence to What changed.
- The commit subject and the describe block still say "status filters" or "status chips", but only Unread gets the pass. Consider renaming both, and marking the In progress sentence in the Problem section as left for #36670.
f8f7124 to
9273c6e
Compare
|
Addressed at head
No production code changed in this push. |
|
This review found no blocking issue on 9273c6e. It cannot be approved yet because PR Readiness is red on this head. The failing checks are Backend Tests shard 8 (Linux and Windows) and the Coverage Gate that follows it. The failing test is What the review checked on this head:
Done well: the rule lives in one place, the row model that both the list and the board read. Status of my earlier requests:
The issue asked whether this should be the default or sit behind a setting. The PR makes it the default. The approving maintainer should confirm that choice. Optional:
|
Withdrawn: every item in this review is fixed or answered on the current head (see the follow-up comment). Other holds on this PR are listed there.
9273c6e to
3b5077d
Compare
3b5077d to
3878124
Compare
|
This changes behaviour users already rely on, so it needs a design sign-off from a maintainer before it can merge. A maintainer will post the decision here. To propose a design, follow the request-for-change process in CONTRIBUTING.md (docs/request-for-change/). The open decision comes from #36670: should the open chat stay listed under Unread by default, or only behind a setting? This PR makes it the default. The review found no blocking issue on 3878124. What it checked on this head:
Done well: the rule lives in one place, the row model that both the list and the board read, and each part has a focused test. Status of my earlier requests:
Optional:
|
Withdrawn: every item in this review is fixed or answered on the current head (see the follow-up comment). Other holds on this PR are listed there.
|
@CrysisDeu could you make the design call this PR is waiting on? The question, from #36670: with the Unread filter on, should the chat you have open stay in the list after opening it marks it read?
Every check and review lane passes at |
Problem / Motivation
Goal: The chat you have open stays in the session list and on the board while a status filter such as Unread is on.
With Unread on, clicking an unread chat opens it and marks it read. The row then fails the Unread chip and vanishes from the list and the board, while the person is still working in that chat. The same happens with In progress when the open chat's turn ends.
Why it matters
Working through an Unread inbox makes the list jump on every click, and the row you are in cannot be renamed, pinned or moved from the list. Mail clients avoid this by keeping the open message in an unread-only view until you leave it.
Not a goal
What changed (motivation → approach → change)
The Unread chip now passes the open chat: the row model takes the active slot as
LocalRowIndex.openKey, andChatSidebar.tsxpassesactiveSlotin (makeDimensionsinwebsite/src/pages/chat-sidebar/rows.ts). Only Unread gets the pass, because opening a chat is what changes it; the other chips never change on open, so they still decide as before. The pass checks the local key only, so a peer row with the same key gets nothing. Opening a different chat drops the old one under normal rules. The list and the board read the same filtered rows, so both keep it.This PR covers the Unread chip only. The In progress case named above is unchanged and stays open in #36670.
Backwards compatibility
Compatible: no stored setting, API or type outside the sidebar row model changes; a row that was shown before is still shown.
Removes nothing: the change only lets one more row through the status chips.
Tests
website/src/pages/chat-sidebar/rows.test.ts, new block "the open chat is exempt from the status chips":openKeyalone rebuilds the filtered rows (the memo leaves table);Removing the pass in
rows.tsmakes the first test fail (1 failed, 152 passed); with it, 153 pass. The sidebar, ChatSidebar and sidebar integration suites pass, includingChatSidebarNotInProgressFilter(185 files, 2336 tests).tsc -p tsconfig.app.jsonis clean.Manual verification
Rendered the real
ChatSidebarin an isolated Playwright capture with Unread on, in the list and on the board. The open, read chat stays listed next to the chats still unread; after another chat is opened, it leaves the list. Unread stays on because other chats are still unread.Screenshots / video
List, Unread on: the open chat ("On-call ticket triage", highlighted, no unread dot) stays next to two unread chats.
List, after opening "Design doc questions": the first chat is gone.
Board, Unread on: the open, read chat stays in its column.
The stills are committed under
.github/screenshots/37210/. They were taken from598629b60a(scripted Playwright over the realChatSidebar, dark theme); the rebase onto main since then did not change the logic.Related Issues
Part of #36670
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)