Fix #3708: redraw cmux window on focus regain even when cursor is over the sidebar resize handle - #3879

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
cmux	Canceled		May 12, 2026 1:20am
cmux-staging	Building	Preview, Comment	May 12, 2026 1:20am

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.

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

Greptile Summary

This PR fixes a focus-regain redraw regression where the pointer being over the NSSplitView sidebar divider could suppress the standard AppKit window-focus invalidation path. It introduces a MainWindowKeyRegainRefresh helper that unconditionally walks the full view tree marking every non-hidden view and its backing CALayer as needing display before restoring keyboard focus, and adds a regression test using only local view-frame coordinates.

• Production fix (CmuxLifecycleEventPublishing.swift): replaces the direct keyboardFocusCoordinator call in handleCmuxWindowBecameKey with a full-tree display invalidation walk followed by the coordinator call, ensuring the redraw invariant is upheld regardless of cursor tracking state.
• Regression test (WindowAndDragTests.swift): sets up an off-screen split-view window, fires resign+becomeKey notifications, and asserts contentView.needsDisplay is true; uses only local coordinates so the assertion is deterministic in headless CI.
• Cursor rect scope: window.invalidateCursorRects(for: contentView) resets only the root content view's cursor rects; NSSplitView subview cursor areas remain untouched by this call.

Confidence Score: 5/5

Safe to merge — the change is localized to the key-regain path and adds a regression test; the tree-walk invalidation is unconditional and isolated behind MainWindowKeyRegainRefresh.

The tree-walk display invalidation is correct and self-contained. The only gap is that invalidateCursorRects(for: contentView) does not recurse into subviews, so NSSplitView cursor areas are not reset — but this does not affect display correctness and does not regress existing behavior. The regression test uses only local view-frame coordinates and is reliable in headless CI.

No files require special attention; the production change is small and the test is straightforward.

Important Files Changed

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[NSWindow.didBecomeKeyNotification] --> B[handleCmuxWindowBecameKey]
    B --> C[contextForMainTerminalWindow]
    C --> D[setActiveMainWindow]
    D --> E[publishCmuxWindowLifecycle]
    E --> F{context != nil?}
    F -- No --> G[Done]
    F -- Yes --> H[MainWindowKeyRegainRefresh.refresh]
    H --> I[invalidateContentDisplayTree]
    I --> J[invalidateDisplayTree recursive walk]
    J --> K[view.needsDisplay = true / view.layer?.setNeedsDisplay]
    K --> L{more subviews?}
    L -- Yes, not hidden --> J
    L -- No --> M[window.invalidateCursorRects for contentView]
    M --> N[keyboardFocusCoordinator.restoreTargetAfterWindowBecameKey]
    N --> G

_{Reviews (2): Last reviewed commit: "Clarify key-regain refresh ownership" | Re-trigger Greptile}