[teamleaderleo]

This is slice S3 of Leo's sidebar-organization ask (item A, the PR badge). It builds on #21814 (the session tree's pull request badges). The owner of #21836 cleared it: the only overlap is two small spots next to each other, and whichever PR lands second takes the conflict.

What changes

Every workspace row now carries its pull request badge: a glyph and a count, in the most urgent state's color, with the hover card. Before this, the badge only drew with sidebar.sessionTree on.

• Unchanged: the tree's nesting, guide lines and Command-number keycaps still need sidebar.sessionTree, and manual order is untouched.
• How: SidebarSessionTreeFeed also runs for the badges and publishes them apart from the tree state (SidebarSessionTree.pullRequests). The layout draws them whether or not the tree is on.
• Redraws: a badge change redraws the list, and a group fold keeps its members' badges. The first review caught both.
• Cost: by default, gh now runs for anyone with agent chats, not only people who turned the tree on. It uses the reader the chat header's chip already shares: two calls at a time, never prompting, at most once every five minutes per folder or URL, with one batched GraphQL read for URLs. A failed read, including a missing or logged-out gh, keeps the last answers and backs off by doubling, up to 80 minutes.
• Toggle: Debug toggle sidebar.pullRequestBadges, on by default. Turning both it and the tree off stops the feed and its gh pr view checks, as before.

Tests

Tests/CmuxNextAppTests/SidebarPullRequestBadgeTests.swift (committed first) covers three things:

• the feed publishes badges but no tree while the tree is off;
• a row draws its badge with no nesting and no keycaps;
• a badge that arrives or leaves after the first draw is drawn;
• the layout carries the badge in manual order.

No capture: a capture slot has no GitHub credentials, so no row there gets pull request data. The badge view and its drawing are unchanged from #21814, and the view-level tests above cover a drawn badge with the tree off.

Observed / Not run

• Ran:
  • scripts/cmux-next/prepush-gate.py: no new violations.
  • Fleet compile-check --app: passed on 6d8ca2b.
  • Independent review: BLOCK at 6d8ca2b (redraw and fold), fixed in c510bb2.
• Not run locally: the Swift suites, which run in this PR's CI.

Changelog

• Changed: Sidebar rows show their pull request badge even with the session tree off.

Agent session: cosmic-gray

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

---

Summary by cubic

Every sidebar row now shows its pull request badge even with the session tree off; previously the badge only drew when sidebar.sessionTree was enabled.

• The session tree feed now runs just for the badges and publishes them separately from tree state (SidebarSessionTree.pullRequests).
• A new Debug toggle, sidebar.pullRequestBadges, is on by default; the tree's nesting, guide lines, and Command-number keycaps still require sidebar.sessionTree.
• A badge change redraws the list, and folding a group keeps its members' badges.
• gh now checks pull requests for anyone with agent chats, not only tree users: two concurrent calls, never prompting, at most once per five minutes per folder or URL, with backoff (doubling, up to 80 minutes) on failure.
• Tests cover the publish path, a row's badge with the tree off and in manual order, and badges that arrive or leave after first draw.

^{Written for commit 5bcbcd0. Summary will update on new commits.}

[teamleaderleo]

Owned by chief-sidebar on cosmic-gray · session

[github-actions]

cmux status _{(one comment, edited in place; each line is updated by the job that owns it)}

Review 5bcbcd0: HQ feed card · gallery before/after runs · checks

App of this head (app.zip, kept 1 day)

gh workflow run reload-build.yml -R manaflow-ai/cmux --ref feat-cmux-next -f tag=pr23495 -f ref=leo/sidebar-pr-badges -f platform=macos -f backend_mode=local

Not verified: macOS jobs did not run: compile and app tests were skipped on 5bcbcd09dc; ci-status is green without them.

CI passes on 5bcbcd09dc (run 39654562065 attempt 1).
macOS jobs did not run: compile and app tests were skipped: Fast static checks skipped (the macos job needs both to succeed). Re-run CI once they pass, before merging.

Written by scripts/ci/classify_failures.py (ci-failure-attribution.yml); signatures are its SIGNATURES table. A machine verdict is the runner's fault, not this PR's; yours means the failing file is one this PR changes, also red on main that main's latest full suite fails the same way, seen on other PRs that it failed on another pull request's run lately.

red: cmux-next checks (god files, concurrency, crash safety, l10n); base red: the same failure on the base 8c47a2e; does not block

[teamleaderleo]

Review (claude-review-subagent): BLOCK at 6d8ca2b

Read from a local worktree at the head; nothing built or run.

Blocking

1. When the tree is off, a badge change never redraws the list. In SidebarView+Observation.swift render(_:), RenderState.pullRequests is part of the equality, so a change gets past guard state != lastState. But rowsChanged checks lastState?.tree != state.tree and history, not pullRequests. So listChanged is false and nothing reloads. The new value goes into lastState and is never drawn.

   • With the tree on this works, because the badges are inside tree. With it off (the new default path), a gh answer that arrives after the first draw does not show until something unrelated reloads the list. The same goes for a badge going away and for the tunable being turned off.
   • Fix: add || lastState?.pullRequests != state.pullRequests to rowsChanged, or patch the rows whose badge changed.
   • The test doesn't catch this. aRowDrawsItsBadgeWithTheTreeOffInManualOrder sets SidebarSessionTree.shared.pullRequests before it builds the view and then calls list.reload itself. A test that sets the map after the first renderNow(), calls renderNow() again and expects the badge, would be red at this head.

2. Folding a group drops its members' badges. SidebarListView+Fold.swift foldGroup builds options from model.listOptions(), which never sets pullRequests. SidebarLayout.fold rebuilds every member row with o.rowContent(...), so a group's members lose their badge when it is opened or closed, and get it back only at the next full reload.

   • With the tree on this was already true (no sessionTree or keycaps on that path either), but this PR turns the badge on for everyone, so it is now on the default path.
   • Fix: in foldGroup, add o.pullRequests = SidebarSessionTree.shared.pullRequests; the keycaps could be added there too.
   • The other places that build layout options are fine. Row patches (SidebarRowIndex uses options(includeGap: true)), in-place reorder, drag and tab drop all go through options(includeGap:). SidebarModel+VisibleRows and +ItemOrder only need order and parents, and the badge does not change row height (metrics.height(for:) doesn't read it).

Checked, fine

• RenderState.init has a single init, and it sets pullRequests.
• Reading the tunable in the feed is observed: Tunable.value registers an Observation dependency on its key, as in SidebarLaneStatus. With nests || ... short-circuiting, the tunable is not tracked while the tree is on, which is correct because the badges come from the tree then.
• publish keeps SidebarSessionTree.state nil while the tree is off, so keycaps (options(includeGap:)), ItemOrder and VisibleRows (via state?.parents) and nesting stay tree-only. SidebarChatHistoryFeed has its own sessionTree guard and does not depend on this feed. The hover card reads the row's content.
• Cost: the feed and SidebarPullRequestStore now run for everyone. The store runs at most two gh reads at a time, only for folders and URLs of workspaces with agent chats, with a 300 s recheck of open or draft PRs. That seems acceptable, but the PR body should say that the default now makes gh calls for every user with agent chats. It should also confirm that a missing or unauthenticated gh fails quietly and does not retry in a tight loop.

Tests

• child.frame.minY > parent.frame.minY || child.frame.maxY < parent.frame.maxY only checks that the two frames differ, which two separate rows always do. Either drop it or assert what was meant, for example that lane sits right after other at depth 0 (the layout test already covers that).
• The keycap isHidden checks hold: with the tree off, rowContent never sets keycap, and the test's DesignSettings() leaves sessionTree off.
• publish and the layout test are fine.

[teamleaderleo]

Review (claude-review-subagent): SHIP at c510bb2

I read the fixup commit (6d8ca2b..c510bb2) and re-checked the whole PR diff against merge-base ed547f0. Local git only; I did not build or run anything.

Both earlier blockers are fixed

1. SidebarView.render: rowsChanged now includes lastState?.pullRequests != state.pullRequests, and RenderState snapshots SidebarSessionTree.shared.pullRequests. With the tree off, a badge change now reaches showModelChange → reload(fromModel: true).
2. foldGroup sets o.pullRequests = SidebarSessionTree.shared.pullRequests, so a spliced fold builds member rows that keep their badges. With the tree on, the fold's options have no sessionTree, which was already true before this PR. The fallback pullRequests[ws.id] reads the same map, because publish always writes pass.state.pullRequests, so the badges still match.

No other place draws rows without the badges. All row drawing goes through options(includeGap:): displayLayout, drag, tab drop, in-place reorder and the SidebarRowIndex patch path. That function sets pullRequests, and so does the fold. The other SidebarLayoutOptions() builders are SidebarModel+VisibleRows and SidebarModel+ItemOrder. They only compute row order and keycap order and never draw row content, so they don't need the badges.

The new test aBadgeThatArrivesLaterIsDrawn should pass:

• On the first renderNow(), lastState is nil, so both rowsChanged and selectionChanged are true, and render takes the list.reload(... fromModel: true) branch. apply creates views for rows inside realizationRect(), the same way the explicit list.reload does in the first test, so rowViews is populated. The observe() Task does not run before then, because the test never suspends.
• On the second and third renderNow(), only rowsChanged is true (the selection did not change), so render calls showModelChange. holdsModelChangeNow is false because window == nil, so it reloads. apply runs configure(view, row:) on every existing row view. configureSessionTree then calls pullRequests.configure(shown.pullRequests), which sets badge and isHidden right away, with no animation in between.
• let row = { try #require(...) } followed by try row() is valid. The closure is inferred as () throws -> WorkspaceRowView and stays on the main actor, #require works inside closures, and #expect(try ...) accepts throwing expressions.
• The meaningless frame assertion is gone.

Nits (not blocking)

• The badge tests read and write the process-wide SidebarSessionTree.shared. .serialized only orders tests inside this suite. If another suite running in parallel leaves shared.state non-nil, rowContent would take sessionTree.pullRequests before options.pullRequests. Resetting SidebarSessionTree.shared.state = nil in the setup would close that gap.

[teamleaderleo]

Bead cx-3by72 created for this PR by the cmux coordinator.

[cursor]

Bugbot needs on-demand usage enabled

Bugbot uses usage-based billing for this team and requires on-demand usage to be enabled.

A team admin can enable on-demand usage in the Cursor dashboard.

[cursor]

Bugbot needs on-demand usage enabled

Bugbot uses usage-based billing for this team and requires on-demand usage to be enabled.

A team admin can enable on-demand usage in the Cursor dashboard.

[teamleaderleo]

Review (claude-review-subagent): SHIP at 5bcbcd0

Scope: only the conflict resolution in the second base merge (5bcbcd0) against #23531, read with local git (git diff 5bcbcd09^2 5bcbcd09, git diff c510bb24 5bcbcd09). Nothing built or run.

• SidebarSessionTreeFeed.swift: the resolution is a faithful union. Base's readsPullRequests(sections) / pass.showsTree = sections.sessionTree are kept; the PR adds || SidebarTunables.pullRequestBadges.value to the guard and replaces the inline tree write with publish(_:to:), which writes tree.state only while showsTree (same as base) and always writes tree.pullRequests. With the tree off, tree.state stays nil exactly as on base.
• Red CI (SidebarSessionTreeFeed+Checks.swift, unchanged by the merge): checksFailing still guards on readsPullRequests(sections) (tree || Priority view), not on the badges toggle. With badges on and both tree and Priority off, the reads now run, but checksFailing still returns false, so the Priority view's Red CI and SidebarLaneStatus's failing-checks input report nothing base wouldn't. SidebarPriorityRedCITests expectations (!checksFailing / !readsPullRequests with both off) still hold.
• No leftover nests reference anywhere in Packages/macOS code (git grep '\bnests\b' finds only unrelated comments). The only Pass( constructions are the feed's Pass(state:asks:) and the tests' Pass(state:asks:showsTree:), which matches the memberwise init order state, asks, showsTree.
• SidebarTunables is a public enum in CmuxNextSidebar, which the feed already imports publicly.

Nits, not blocking: the checksFailing doc comment ("which run while the tree or the Priority view is on") no longer says that the badges also keep the reads running. Separately, with only badges on, a row's badge can show failing while the lane-status failing-checks chip stays off. That is base's gating, kept on purpose; it's worth a follow-up only if the two should agree.

[blacksmith-sh]

Found 9 test failures on Blacksmith runners:

Failures

Test	View Logs
/tmp/tmpaa4t19ir/summary-fail.json (live):/summary-fail.json (live):	View Logs
/tmp/tmpd6zao4f0/summary-missing.json (live):/summary-missing.json (live):	View Logs
[new_workspaces] debug.new_workspace timed 0 of 50 new workspace rows/ [new_workspaces] debug.new_workspace timed 0 of 50 new workspace rows	View Logs
[new_workspaces] the window draws no frames/[new_workspaces] the window draws no frames	View Logs
[tabs] debug.tab_switch timed 0 of 40 switches/ [tabs] debug.tab_switch timed 0 of 40 switches	View Logs
[tabs] the window draws no frames/[tabs] the window draws no frames	View Logs
convention violations found (ERROR lines above)./ convention violations found (ERROR lines above).	View Logs
test_a_file_a_test_reads_selects_that_test (main.PullRequestTiers.test_a_file_a_tes t_reads_selects_that_test)/ test_a_file_a_test_reads_selects_that_test (main.PullRequestTiers.test_a_file_a_tes t_reads_selects_that_test)	View Logs
test_a_shared_module_selects_its_dependents (main.PullRequestTiers.test_a_shared_mo dule_selects_its_dependents)/ test_a_shared_module_selects_its_dependents (main.PullRequestTiers.test_a_shared_mo dule_selects_its_dependents)	View Logs

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[teamleaderleo]

cmux-next crash-class rules / cmux-next crash-class rules is red on 9 of 9 fresh feat-cmux-next PR heads in the last 2:00:00, so it is most likely a base red, not this PR's. The base-reds lane has it; a base merge picks up the fix once it lands.

[teamleaderleo]

cmux-next guards / cmux-next localization debt ratchet is red on 17 of 18 fresh feat-cmux-next PR heads in the last 2:00:00, so it is most likely a base red, not this PR's. The base-reds lane has it; a base merge picks up the fix once it lands.

[github-actions]

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.3967575499801 (2a0a42ab31)
• cmux NEXT nightly 1.0.0-nightly.3967669311101 (2a3491e235)
• cmux NEXT nightly 1.0.0-nightly.3967869181901 (cd9b30e302)
• cmux NEXT nightly 1.0.0-nightly.3968001295301 (8cba1da27d)