[lawrencecchen]

Summary

Lawrence, 2026-10-18, on NIGHTLY: after Cmd-N the new row first drew above the "Accounts & billing" group, then moved below "Scratch".

Root cause: on the sidebar document (sidebar-doc-v1), SidebarBridge.place (Windows/SidebarBridge+Windows.swift) sent the workspaces.newPlacement slot to the personal order (placePersonal, workspace_placement). The document path does not show that order. The new workspace listed in the document's unplaced tail, after every group (SidebarDocSections.sections), and the daemon's legacy-to-document sync then put it where its personal neighbour mapped. The press row (PendingWorkspaceRows) drew at the real slot (top: first in the machine section, which here starts with "Accounts & billing"), so the row jumped.

Fix:

• On the document, SidebarBridge.place goes to SidebarDocMachines.place: one sidebar.apply (adopt + move, no undo step) at the slot the row drew at. The store then rebuilds shown in the same turn (SidebarDocStore.reshow), so the sidebar never lists the workspace in the unplaced tail first. applyPendingPlacements runs synchronously from the workspace-list change, before any observer.
• bottom outside a group is now placed at the end of the loose rows, where its press row draws. Before, it was left to the daemon, whose place on the document is the end of the machine section, after every group. Now the press row and the placement use one slot under every setting value.
• A drop from another window (accept) also goes through place, so on the document it now lands where it was dropped.

Bead: cx-qbktz.

Testing

• Red commit: RapidWorkspaceCreateTests.cmdNRowStaysWhereItDrewAtThePress (same-tree daemon, app action newTab, samples the rows the sidebar model draws until 2 s after the row is the workspace's own). It expects one place.
• Results: see the comments below.

Changelog

• Fixed: A new workspace's sidebar row no longer jumps to another place after it appears.

---

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

---

Summary by cubic

Fixes the cmux-next sidebar so a Cmd-N workspace's row stays where it drew at the press instead of jumping to another slot after it appears, on the sidebar document.

Changes

• SidebarBridge.place now routes document placements through SidebarDocMachines.place: one sidebar.apply (adopt + move, no undo step), then rebuilds shown in the same turn, so the new row never draws in the unplaced tail first.
• bottom outside a group places at the end of the loose rows, matching the press row, and drops from other windows land where dropped on the document.

^{Written for commit f976c7b. Summary will update on new commits.}

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

Receipt for f976c7b (this PR's head; a new push replaces it)

• Visual review: HQ feed card · gallery before/after runs
• Checks: this head
• 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=pr21914 -f ref=nx-cmdn-row-jump -f platform=macos -f backend_mode=local

[github-actions]

red: cmux-next CmuxNextApp size cap; do not merge until green; see lane-rules

[github-actions]

Not verified: macOS jobs did not run: compile and app tests were skipped on f976c7ba32; ci-status is green without them.

CI passes on f976c7ba32 (run 39199963766 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.

[blacksmith-sh]

Found 3 test failures on Blacksmith runners:

Failures

Test	View Logs
/tmp/tmplmygr12v/summary-fail.json (live):/summary-fail.json (live):	View Logs
/tmp/tmps7ywmopp/summary-missing.json (live):/summary-missing.json (live):	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

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

[lawrencecchen]

Correctness review (subagent), findings and resolution

Verdict at the first head (f976c7b): REVISE. No blocker. The reviewer confirmed the root cause and the frame order: reconcileMembership -> applyPendingPlacements runs synchronously before both observation hops, so reshow sets shown before the sidebar first lists the workspace.

#	Finding	Resolution
1	SHOULD-FIX: reshow called show(live:connectedRemotes: lastConnectedRemotes), which reruns the remote-close check with an old machine set. That could send close for a remote machine that just disconnected.	Fixed: reshow now only rebuilds shown and asks for adopts. It does not touch lastLive or lastConnectedRemotes, and it sends no close.
2	SHOULD-FIX: .bottom(anchor: nil) resolves to index 0 when the section has no loose rows (every workspace in a group), so bottom would put the new workspace first.	Fixed: bottom outside a group goes to the end of the section when the section lists no loose row (NewWorkspaceDefaultSlot.bottom). The press row uses the same slot.
3	SHOULD-FIX: the test did not require the document path.	Fixed: the test waits for SidebarDocWiring.isActive and requires it. It also builds the reported layout: a group first in the list, and the shown workspace loose after it. In the first version the list had no group, and the code before the fix did not jump there.
4	SHOULD-FIX: no red or green run.	Done on the fleet. Results are in the next comment.
5	SHOULD-FIX: the absolute index can change when an unrelated row appears.	Fixed: the test counts only the rows that were listed before the press and sit above the new row.
6	NIT: on the document path place reads model and not the sections argument.	Kept. accept passes before only because moveWorkspaces does not rebuild the model in the same turn. The document target is computed from item anchors, so the model at call time gives the same neighbours.
7	NIT: a mirror refusal showed a refusal HUD for a placement the person did not ask for.	Fixed: place uses sendChecked. A refusal returns false, shows nothing, and the older path runs.
8	NIT: bottom now sends one placement for each create on the older paths.	Accepted. It is one command for each create, and it makes the press row and the final slot the same.
9	NIT: undoable: false	Accepted. This matches the older path.
10	NIT: more than one pending placement in one reconcile is applied in dictionary order.	This was already true before this change. Not changed here, because WorkspaceLanes sends creates one at a time.
11	NIT: size-cap advisory	CmuxNextApp is under its cap (check-app-size.sh: 498 lines of headroom before this change, which adds about 15).

[austinywang]

The same-turn reshow from this PR is carried into #23071, with you credited as co-author, because the new-workspace jump it fixes still reproduced on feat-cmux-next.

Since this branch was written, SidebarBridge.place already sends the document move (SidebarDocWiring.handle(.reorder), #22593), so 23071 keeps that path and adds only the reshow.

That reshow rebuilds only shown: no adopt, and no remote-close check over a stale machine set. This branch's head still had that remote-close bug (review finding 1). The remote-close check now keeps its own baseline (lastShownItems).

Your cmdNRowStaysWhereItDrewAtThePress test is merged into RapidWorkspaceCreateTests there. The bottom slot change from this PR is not included.

[lawrencecchen]

Closed by the perf lead (agent, cmux-ad order 2026-10-31): stale, with no owner. This is an agent PR (lawrencecchen account). It was opened on 2026-10-19, its last commit is from 2026-10-19, no person pushed or commented in the last 3 days (automated review and sweep notes do not count), no live session names it in the lead status of the last 24 h, and it conflicts with feat-cmux-next (DIRTY).

The branch nx-cmdn-row-jump stays. To reopen: click Reopen (or gh pr reopen 21914 -R manaflow-ai/cmux), merge origin/feat-cmux-next into the branch, get a review receipt on the new head, then add ready-to-land.