Merge remote-tracking branch 'origin/main' into task-browser-up-down-arrow-keys-merge

# Conflicts:
#	cmuxUITests/BrowserPaneNavigationKeybindUITests.swift

Author: lawrencecchen

.github/workflows/build-ghosttykit.yml

@@ -6,6 +6,10 @@ on:
       - main
   pull_request:
 
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+
 jobs:
   build-ghosttykit:
     # Never run WarpBuild jobs for fork pull requests (avoid billing on external PRs).

.github/workflows/ci-macos-compat.yml

@@ -130,7 +130,9 @@ jobs:
             xcodebuild -project GhosttyTabs.xcodeproj -scheme cmux-unit -configuration Debug \
               -clonedSourcePackagesDirPath "$SOURCE_PACKAGES_DIR" \
               -disableAutomaticPackageResolution \
-              -destination "platform=macOS" test 2>&1
+              -destination "platform=macOS" \
+              -skip-testing:cmuxTests/AppDelegateShortcutRoutingTests/testCmdWClosesWindowWhenClosingLastSurfaceInLastWorkspace \
+              test 2>&1
           }
 
           set +e

.github/workflows/ci.yml

@@ -6,6 +6,10 @@ on:
       - main
   pull_request:
 
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+
 jobs:
   workflow-guard-tests:
     runs-on: ubuntu-latest
@@ -74,7 +78,7 @@ jobs:
     # Never run WarpBuild jobs for fork pull requests (avoid billing on external PRs).
     if: github.event_name != 'pull_request' || github.event.pull_request.head.repo.full_name == github.repository
     runs-on: warp-macos-15-arm64-6x
-    timeout-minutes: 20
+    timeout-minutes: 30
     steps:
       - name: Checkout
         uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4
@@ -171,14 +175,17 @@ jobs:
             xcodebuild -project GhosttyTabs.xcodeproj -scheme cmux-unit -configuration Debug \
               -clonedSourcePackagesDirPath "$SOURCE_PACKAGES_DIR" \
               -disableAutomaticPackageResolution \
-              -destination "platform=macOS" test 2>&1
+              -destination "platform=macOS" \
+              -skip-testing:cmuxTests/AppDelegateShortcutRoutingTests/testCmdWClosesWindowWhenClosingLastSurfaceInLastWorkspace \
+              test 2>&1
           }
 
-          # xcodebuild exits 65 even for expected failures (XCTExpectFailure).
-          # Capture output and fail only if there are unexpected failures.
+          # Stream output via tee so CI logs are visible in real time, while still
+          # capturing for post-run analysis of expected vs unexpected failures.
           set +e
-          OUTPUT=$(run_unit_tests)
-          EXIT_CODE=$?
+          run_unit_tests | tee /tmp/test-output.txt
+          EXIT_CODE=${PIPESTATUS[0]}
+          OUTPUT=$(cat /tmp/test-output.txt)
           set -e
 
           # SwiftPM binary artifact resolution can occasionally fail on ephemeral
@@ -190,12 +197,12 @@ jobs:
             mkdir -p ~/Library/Caches/org.swift.swiftpm
             rm -rf ~/Library/Developer/Xcode/DerivedData/GhosttyTabs-*
             set +e
-            OUTPUT=$(run_unit_tests)
-            EXIT_CODE=$?
+            run_unit_tests | tee /tmp/test-output.txt
+            EXIT_CODE=${PIPESTATUS[0]}
+            OUTPUT=$(cat /tmp/test-output.txt)
             set -e
           fi
 
-          echo "$OUTPUT"
           if [ "$EXIT_CODE" -ne 0 ]; then
             SUMMARY=$(echo "$OUTPUT" | grep "Executed.*tests.*with.*failures" | tail -1)
             if echo "$SUMMARY" | grep -q "(0 unexpected)"; then
@@ -231,9 +238,9 @@ jobs:
 
   tests-build-and-lag:
     # Build the full cmux scheme and run the lag regression on WarpBuild.
-    # XCUITests cannot run on WarpBuild (Virtualization.framework limitation:
-    # XCUIApplication stuck "Running Background", 62s activation timeout per
-    # test). Interactive UI tests run via test-e2e.yml on GitHub-hosted runners.
+    # Keep lag validation separate from UI regressions so functional UI failures
+    # and performance regressions stay isolated. Broader interactive UI suites
+    # still run via test-e2e.yml on GitHub-hosted runners.
     if: github.event_name != 'pull_request' || github.event.pull_request.head.repo.full_name == github.repository
     runs-on: warp-macos-15-arm64-6x
     timeout-minutes: 20
@@ -343,6 +350,7 @@ jobs:
           VDISPLAY_PID=$!
           echo "VDISPLAY_PID=$VDISPLAY_PID" >> "$GITHUB_ENV"
           sleep 3
+          kill -0 "$VDISPLAY_PID"
 
       - name: Run workspace churn typing-lag regression
         run: |
@@ -386,7 +394,16 @@ jobs:
           CMUX_LAG_KEY_EVENTS=180 \
           python3 tests/test_workspace_churn_up_arrow_lag.py
 
-  ui-display-resolution-regression:
+      - name: Cleanup virtual display
+        if: always()
+        run: |
+          set -euo pipefail
+          if [ -n "${VDISPLAY_PID:-}" ]; then
+            kill "$VDISPLAY_PID" >/dev/null 2>&1 || true
+          fi
+          rm -f /tmp/create-virtual-display
+
+  ui-regressions:
     if: github.event_name != 'pull_request' || github.event.pull_request.head.repo.full_name == github.repository
     runs-on: warp-macos-15-arm64-6x
     timeout-minutes: 25
@@ -438,8 +455,8 @@ jobs:
         uses: actions/cache@5a3ec84eff668545956fd18022155c47e93e2684 # v4
         with:
           path: .ci-source-packages
-          key: spm-ui-display-resolution-${{ hashFiles('GhosttyTabs.xcodeproj/project.xcworkspace/xcshareddata/swiftpm/Package.resolved') }}
-          restore-keys: spm-ui-display-resolution-
+          key: spm-ui-regressions-${{ hashFiles('GhosttyTabs.xcodeproj/project.xcworkspace/xcshareddata/swiftpm/Package.resolved') }}
+          restore-keys: spm-ui-regressions-
 
       - name: Resolve Swift packages
         run: |
@@ -461,26 +478,225 @@ jobs:
             sleep $((attempt * 5))
           done
 
-      - name: Run display resolution churn UI regression
+      - name: Build for testing (display resolution)
         run: |
           set -euo pipefail
           SOURCE_PACKAGES_DIR="$PWD/.ci-source-packages"
+          xcodebuild -project GhosttyTabs.xcodeproj -scheme cmux -configuration Debug \
+            -clonedSourcePackagesDirPath "$SOURCE_PACKAGES_DIR" \
+            -disableAutomaticPackageResolution \
+            -destination "platform=macOS" \
+            build-for-testing
+
+      - name: Create persistent virtual display
+        run: |
+          set -euo pipefail
           HELPER_PATH="/tmp/create-virtual-display"
-          MANIFEST_PATH="/tmp/cmux-ui-test-display-harness.json"
+          clang -framework Foundation -framework CoreGraphics \
+            -o "$HELPER_PATH" scripts/create-virtual-display.m
+
+          VDISPLAY_READY="/tmp/cmux-vdisplay-persistent.ready"
+          VDISPLAY_ID_PATH="/tmp/cmux-vdisplay-persistent.id"
+          rm -f "$VDISPLAY_READY" "$VDISPLAY_ID_PATH"
+
+          "$HELPER_PATH" \
+            --modes "1920x1080" \
+            --ready-path "$VDISPLAY_READY" \
+            --display-id-path "$VDISPLAY_ID_PATH" \
+            > /tmp/cmux-vdisplay-persistent.log 2>&1 &
+          echo "VDISPLAY_PERSISTENT_PID=$!" >> "$GITHUB_ENV"
+
+          echo "Waiting for persistent virtual display..."
+          for i in $(seq 1 24); do
+            if [ -f "$VDISPLAY_READY" ]; then break; fi
+            sleep 0.5
+          done
+          if [ ! -f "$VDISPLAY_READY" ]; then
+            echo "ERROR: Persistent virtual display not ready after 12s" >&2
+            cat /tmp/cmux-vdisplay-persistent.log 2>/dev/null || true
+            exit 1
+          fi
+          echo "Persistent virtual display ready: ID=$(cat "$VDISPLAY_ID_PATH")"
 
-          rm -f "$MANIFEST_PATH"
-          trap 'rm -f "$MANIFEST_PATH"' EXIT
+      - name: Run display resolution churn UI regression
+        run: |
+          set -euo pipefail
+          SOURCE_PACKAGES_DIR="$PWD/.ci-source-packages"
+          HELPER_PATH="/tmp/create-virtual-display"
+          TOKEN="$(uuidgen)"
+          DIAG_PATH="/tmp/cmux-ui-test-display-churn-${TOKEN}.json"
+          DISPLAY_READY="/tmp/cmux-ui-test-display-${TOKEN}.ready"
+          DISPLAY_ID_PATH="/tmp/cmux-ui-test-display-${TOKEN}.id"
+          DISPLAY_START="/tmp/cmux-ui-test-display-${TOKEN}.start"
+          DISPLAY_DONE="/tmp/cmux-ui-test-display-${TOKEN}.done"
+          HELPER_LOG="/tmp/cmux-ui-test-display-${TOKEN}-helper.log"
+
+          cleanup() {
+            pkill -x "cmux DEV" 2>/dev/null || true
+            rm -f "$DIAG_PATH" "$DISPLAY_READY" "$DISPLAY_ID_PATH" "$DISPLAY_START" "$DISPLAY_DONE" "$HELPER_LOG"
+            rm -f /tmp/cmux-ui-test-prelaunch.json /tmp/cmux-ui-test-display-harness.json
+          }
+          trap cleanup EXIT
 
+          # Build display helper
           clang -framework Foundation -framework CoreGraphics \
             -o "$HELPER_PATH" scripts/create-virtual-display.m
 
-          cat >"$MANIFEST_PATH" <<EOF
-          {"helperBinaryPath":"$HELPER_PATH"}
-          EOF
+          # Find the app binary
+          APP_BINARY=$(find ~/Library/Developer/Xcode/DerivedData -path "*/Build/Products/Debug/cmux DEV.app/Contents/MacOS/cmux DEV" -print -quit 2>/dev/null || true)
+          if [ -z "$APP_BINARY" ]; then
+            echo "ERROR: App binary not found in DerivedData" >&2
+            exit 1
+          fi
+          echo "App binary: $APP_BINARY"
+
+          for attempt in 1 2; do
+            cleanup 2>/dev/null || true
+
+            # Launch display helper from shell (non-sandboxed).
+            # Use --start-delay-ms instead of --start-path because the XCTest
+            # runner is sandboxed and can't write to /tmp/ for the start signal.
+            # 10s delay gives the test time to capture baseline render stats.
+            "$HELPER_PATH" \
+              --modes "1920x1080,1728x1117,1600x900,1440x810" \
+              --ready-path "$DISPLAY_READY" \
+              --display-id-path "$DISPLAY_ID_PATH" \
+              --done-path "$DISPLAY_DONE" \
+              --iterations 40 \
+              --interval-ms 40 \
+              --start-delay-ms 10000 \
+              > "$HELPER_LOG" 2>&1 &
+            HELPER_PID=$!
+
+            # Wait for display ready
+            echo "Waiting for virtual display..."
+            for i in $(seq 1 24); do
+              if [ -f "$DISPLAY_READY" ]; then break; fi
+              sleep 0.5
+            done
+            if [ ! -f "$DISPLAY_READY" ]; then
+              echo "ERROR: Virtual display not ready after 12s" >&2
+              cat "$HELPER_LOG" 2>/dev/null || true
+              continue
+            fi
+            DISPLAY_ID=$(cat "$DISPLAY_ID_PATH")
+            echo "Virtual display ready: ID=$DISPLAY_ID"
+
+            # Launch app from shell (non-sandboxed, outside XCTest sandbox)
+            CMUX_UI_TEST_MODE=1 \
+            CMUX_UI_TEST_DIAGNOSTICS_PATH="$DIAG_PATH" \
+            CMUX_UI_TEST_DISPLAY_RENDER_STATS=1 \
+            CMUX_UI_TEST_TARGET_DISPLAY_ID="$DISPLAY_ID" \
+            CMUX_TAG="ui-tests-display-resolution" \
+            "$APP_BINARY" > /tmp/cmux-ui-test-app.log 2>&1 &
+            APP_PID=$!
+            echo "App launched: PID=$APP_PID"
+
+            # Wait for app diagnostics
+            echo "Waiting for app diagnostics..."
+            APP_READY=false
+            for i in $(seq 1 30); do
+              if [ -f "$DIAG_PATH" ]; then
+                if python3 -c "import json; d=json.load(open('$DIAG_PATH')); assert d.get('pid')" 2>/dev/null; then
+                  APP_READY=true
+                  break
+                fi
+              fi
+              if ! kill -0 "$APP_PID" 2>/dev/null; then
+                echo "ERROR: App crashed during startup"
+                cat /tmp/cmux-ui-test-app.log 2>/dev/null | tail -30 || true
+                break
+              fi
+              sleep 0.5
+            done
+
+            if [ "$APP_READY" != "true" ]; then
+              echo "Attempt $attempt: App not ready after 15s"
+              pkill -x "cmux DEV" 2>/dev/null || true
+              kill "$HELPER_PID" 2>/dev/null || true
+              if [ "$attempt" -eq 2 ]; then
+                echo "Display resolution UI regression failed after 2 attempts" >&2
+                echo "--- App log ---"
+                cat /tmp/cmux-ui-test-app.log 2>/dev/null | tail -50 || true
+                echo "--- Helper log ---"
+                cat "$HELPER_LOG" 2>/dev/null | tail -20 || true
+                echo "--- Diagnostics ---"
+                cat "$DIAG_PATH" 2>/dev/null || echo "(not found)"
+                exit 1
+              fi
+              sleep 3
+              continue
+            fi
+
+            echo "App started. Diagnostics:"
+            cat "$DIAG_PATH"
+
+            # Wait for render stats (terminal surface initialization)
+            echo "Waiting for render stats..."
+            RENDER_READY=false
+            for i in $(seq 1 40); do
+              if python3 -c "import json; d=json.load(open('$DIAG_PATH')); assert d.get('renderStatsAvailable') == '1'" 2>/dev/null; then
+                RENDER_READY=true
+                echo "Render stats available after $((i / 2))s"
+                break
+              fi
+              sleep 0.5
+            done
+            if [ "$RENDER_READY" != "true" ]; then
+              echo "WARNING: Render stats not available after 20s. Diagnostics:"
+              cat "$DIAG_PATH" 2>/dev/null || true
+              echo "--- App log ---"
+              cat /tmp/cmux-ui-test-app.log 2>/dev/null | tail -30 || true
+            fi
+
+            # Write manifests so test can find the pre-launched state
+            MANIFEST_PATH="/tmp/cmux-ui-test-display-harness.json"
+            cat >"$MANIFEST_PATH" <<MANIFEST_EOF
+          {"readyPath":"$DISPLAY_READY","displayIDPath":"$DISPLAY_ID_PATH","donePath":"$DISPLAY_DONE","logPath":"$HELPER_LOG"}
+          MANIFEST_EOF
 
+            PRELAUNCH_PATH="/tmp/cmux-ui-test-prelaunch.json"
+            cat >"$PRELAUNCH_PATH" <<PRELAUNCH_EOF
+          {"diagnosticsPath":"$DIAG_PATH"}
+          PRELAUNCH_EOF
+
+            # Run test — app is already launched from shell
+            if xcodebuild -project GhosttyTabs.xcodeproj -scheme cmux -configuration Debug \
+              -clonedSourcePackagesDirPath "$SOURCE_PACKAGES_DIR" \
+              -disableAutomaticPackageResolution \
+              -destination "platform=macOS" \
+              -only-testing:cmuxUITests/DisplayResolutionRegressionUITests \
+              test-without-building; then
+              exit 0
+            fi
+
+            pkill -x "cmux DEV" 2>/dev/null || true
+            kill "$HELPER_PID" 2>/dev/null || true
+
+            if [ "$attempt" -eq 2 ]; then
+              echo "Display resolution UI regression failed after 2 attempts" >&2
+              exit 1
+            fi
+            echo "Attempt $attempt failed, retrying..."
+            sleep 3
+          done
+
+      - name: Run browser find focus UI regression
+        run: |
+          set -euo pipefail
+          SOURCE_PACKAGES_DIR="$PWD/.ci-source-packages"
           xcodebuild -project GhosttyTabs.xcodeproj -scheme cmux -configuration Debug \
             -clonedSourcePackagesDirPath "$SOURCE_PACKAGES_DIR" \
             -disableAutomaticPackageResolution \
             -destination "platform=macOS" \
-            -only-testing:cmuxUITests/DisplayResolutionRegressionUITests \
-            test
+            -maximum-test-execution-time-allowance 180 \
+            -only-testing:cmuxUITests/BrowserPaneNavigationKeybindUITests/testCmdFFocusesBrowserFindFieldAfterCmdDCmdLNavigation \
+            test-without-building
+
+      - name: Cleanup persistent virtual display
+        if: always()
+        run: |
+          if [ -n "${VDISPLAY_PERSISTENT_PID:-}" ]; then
+            kill "$VDISPLAY_PERSISTENT_PID" >/dev/null 2>&1 || true
+          fi
+          rm -f /tmp/cmux-vdisplay-persistent.ready /tmp/cmux-vdisplay-persistent.id /tmp/cmux-vdisplay-persistent.log

.github/workflows/nightly.yml

@@ -13,9 +13,8 @@ on:
 
 concurrency:
   group: nightly-build-${{ github.ref_name }}
-  # Only the newest nightly matters. Cancel older runs so a fresh main push
-  # does not sit behind an outdated build that would be discarded anyway.
-  cancel-in-progress: true
+  # Queue concurrent runs instead of canceling them so no build is lost.
+  cancel-in-progress: false
 
 permissions:
   contents: write