[teamleaderleo]

Takes back #22294: Leo 2026-10-20 asked that the hover follow the pointer at once, one row at a time; #22294's rule kept the old row hovered on the next row's fill, which felt trapped. The corner and gap rules from rest stay.

Leo, 2026-10-20: moving the pointer between sidebar workspace rows felt trapped.

Was

• The old row's hover stayed lit while the pointer was already on the next row's fill, especially at its right end where the × sits.
• The × and ⋯ buttons faded out.
• Row pills ran to the sidebar's right edge.
• A row's hover card opened against the row's top-right corner, so a pointer moving down to the next row ran into the card and kept it.

Now

• Hover follows the pointer. Once a row is hovered, the pointer anywhere on another row's fill hands it the hover at once. Only the gap between rows keeps the previous row lit, so sweeps don't flicker (CmuxNext: sidebar rows tile so hover moves row to row with no flicker #20315). From rest, a row's rounded corners still don't light it. This replaces the 10-19 rule that held the old row until the pointer reached the next row's inner pill.
• Hover-out is instant. The × and ⋯ buttons clear with the fill. The fill already used hoverOut, which is 0 by default.
• Rows stop short of the edge. Every list row ends SidebarStyle.rowTrailingTrim (Metrics.space4) short of the right inset, Chrome-style. Headers get the same trim, so the trailing column stays aligned.
• The hover card stands off its row. Sidebar cards (.besideRow, .beside) sit HoverCardEscapeGeometry.rowGap (10 pt) off their row. The strip between row and card keeps the card only while the pointer is heading into it (safe triangle), so moving down past the row lets it go. Tab cards still touch their tab (Lawrence 2026-10-19).

Tests

• Commit 1 adds failing tests:
  • SidebarRowHoverFollowTests (hover follows the pointer; rows end short of the edge)
  • SidebarHoverCardEscapeTests.theCardStandsOffItsRow
  • SidebarHoverCardEscapeTests.movingDownPastTheRowsRightEndDismissesTheCard
• Commit 2 is the fix. It also updates the older tests that encoded the previous behavior: SidebarRowHoverInsetTests, WorkspaceRowHoverMotionTests, HoverCardEscapeGeometryTests, HoverCardFlushTests.

Observed / Not run

Not built or run locally: Mac builds run on CI only. Local git on the build Mac was stuck in disk wait, so both commits were made through the GitHub API, and the local pre-push gate didn't run. CI is the first compile.

Changelog

Fixed: sidebar hover follows the pointer between rows, row highlights stop short of the edge, and a row's hover card no longer catches the pointer.

Agent session: cosmic-gray

🤖 Generated with Claude Code

---

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

---

Summary by cubic

Fixes sidebar hover in cmux-next so a pointer moving between workspace rows hands the hover to the row under it, instead of leaving the previous row lit and trapping the pointer.

• Hover follows the pointer: once a row is hovered, any other row whose fill the pointer is on takes the hover at once; only the gap between rows keeps the previous row lit, so sweeps down or up the list don't flicker.
• The × and ⋯ hover buttons clear at once instead of fading out, so they never trail behind the pointer.
• Row pills end Metrics.space4 short of the sidebar's right edge, Chrome-style; headers get the same trim.
• Sidebar hover cards now stand 10 pt off their row. The gap keeps the card only while the pointer is heading into it, so a pointer moving down past the row dismisses the card instead of running into it.

^{Written for commit 52986bb. Summary will update on new commits.}

[github-actions]

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

Review 52986bb: 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=pr22733 -f ref=cmux/sidebar-hover-follow -f platform=macos -f backend_mode=local

Not verified: macOS jobs did not run: compile and app tests were skipped on 52986bb13a; ci-status is green without them.

CI passes on 52986bb13a (run 39528051667 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.

Tour of 52986bb13a (run): no media. The tour job failed before the tour ran (build or setup); the run log has the output. Both apps run cmux-tui from 856bc1b (tree 04453e44e8c6): the merge base's own tree 7d0ed68a3527 was not published yet.

[teamleaderleo]

Bead cx-8ctt4 created for this PR by the cmux coordinator.

[teamleaderleo]

cmux-next guards / defaults-snapshot is red on 21 of 21 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.3955987471201 (03960f5a64)

[github-actions]

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.3956132659001 (233cc81754)