[teamleaderleo]

Summary

Ports the still-wanted half of #22202 (owner gone, CONFLICTING) onto feat-cmux-next. In a translucent window (appearance.backgroundOpacity below 1), a dragged tab turned from glass into a solid theme block the moment it lifted:

• the lifted tab in its strip painted windowBackground at full alpha (TabCell),
• the tab's drag image was drawn on a solid windowBackground card, and the drag ghost's glass stayed at alpha 0 until the preview card unfolded (TabStripView.snapshot, TabDragGhostPanel).

WindowBackdrop.dragPreview(scope) (CmuxNextDesign) now decides both from the source scope's backdrop. Over glass the lifted tab's fill is a thicker piece of the window tint, and the tab image has no card and stands on the ghost's glass from the first frame. Opaque windows and Reduce Transparency keep the solid card.

Dropped from #22202: its DragLiftView per-window glass/tint. Base now lifts a sidebar row on the one overlay glass (OverlaySurfaceView, Lawrence 2026-10-20, u5506). What stays from that half: a snapshot does not draw the glass, so SidebarDragHandoff.imageOnGlass (from DragLiftView.isGlass, any overlay material but opaque) tells the ghost to stand the row's image on glass. Base also renamed Motion.* to Motion.tabs.* in TabDragSession, so that conflict was resolved to base's names.

Testing

DragGlassTests (CmuxNextAppTests, from #22202 minus the superseded sidebar lift cases):

• liftAndCardFillsAreGlassOnlyInATranslucentWindow
• aLiftedSidebarRowHandsItsGlassToTheGhost (new: glass for liquid glass and vibrancy, not for opaque)
• aLiftedTabLetsTheWindowsGlassReadThrough
• theDragGhostStandsOnGlassFromItsFirstFrame
• aWorkspacesOwnBackgroundDecidesTheGhostsGlassNotTheApps
• aNewSecondaryWindowHasTheMainWindowsBackdrop

SubtleElevationTests' ghost cases pin onGlass: false.

Observed / Not run

• prepush-gate.py: no new violations.
• Fleet package-test-lane.sh for DragGlassTests|SubtleElevationTests at this head: result recorded in a comment when it finishes.
• Not run: live drag in a tagged build with a translucent theme (no Mac builds on this machine).

Changelog

Fixed: Dragging a tab in a translucent window keeps it glass instead of turning it into a solid theme-colored block

Localization: no user-facing strings changed.

Supersedes #22202.

🤖 Generated with Claude Code

Agent session: cosmic-gray

---

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

---

Summary by cubic

Fixes dragged tabs turning from glass into solid theme-colored blocks in translucent windows.

Drag previews now follow the source scope's backdrop (WindowBackdrop.dragPreview), not the app's, so a workspace that pins its own background decides its own glass.

• The lifted tab's fill is a thicker piece of the window tint (60% coverage) instead of full alpha.
• The tab's drag image is drawn without a card, and the drag ghost stands on glass from the first frame until the preview card unfolds.
• Opaque windows and Reduce Transparency keep the solid card.
• A lifted sidebar row's handoff now tells the ghost to stand the row's image on glass, since a snapshot does not draw glass.
• Supersedes Keep dragged tabs and sidebar rows on glass in translucent windows #22202.

^{Written for commit 63625dd. 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 63625dd: 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=pr22807 -f ref=cmux/tab-drag-ghost-glass -f platform=macos -f backend_mode=local

Not verified: macOS jobs did not run: compile and app tests were skipped on 63625dd7bf; ci-status is green without them.

CI passes on 63625dd7bf (run 39534224443 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.

web and size checks: green

[teamleaderleo]

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

[teamleaderleo]

owner: 01a15ba7-3152-78b2-bfdf-bb6eb0d5a053

[teamleaderleo]

Merged after review PASS. Fleet: DragGlassTests and SubtleElevationTests passed (step 04c9fe0c4d32, on this change plus the #22821 base compile fix, since base did not compile). The one red, generated drift, died on a GitHub API rate limit (HTTP 403) before checking anything. Compile and swift test checks were still pending at merge.

[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)