Fix macOS screenshot script

Author: kojix2

.github/actions/screenshot/macos-screenshot.sh

@@ -25,30 +25,106 @@ APP_PID=$!
 echo $APP_PID > app.pid
 echo "Application launched with PID: $APP_PID"
 
-# Wait for application startup
-sleep 3
+# CoreGraphics can expose the window id used by screencapture -l even when
+# Accessibility does not provide AXWindowNumber on GitHub-hosted runners.
+if [ ! -x get_cg_window_info ]; then
+  cat > get_cg_window_info.swift <<'SWIFT'
+import CoreGraphics
+import Foundation
 
-# Wait for the window to appear
-echo "Waiting for window to appear..."
-for i in {1..20}; do
-  EXISTS=$(osascript -e "tell application \"System Events\" to return exists process \"$APP_NAME\"" || true)
-  if [ "$EXISTS" = "true" ]; then
-    HAS_WIN=$(osascript -e "tell application \"System Events\" to tell process \"$APP_NAME\" to return (count of windows) > 0" || true)
-    if [ "$HAS_WIN" = "true" ]; then
-      break
-    fi
-  fi
-  sleep 0.5
-done
+let appName = CommandLine.arguments[1]
+let appPid = Int(CommandLine.arguments[2]) ?? 0
 
-if [ "$EXISTS" != "true" ] || [ "$HAS_WIN" != "true" ]; then
-  echo "ERROR: process or window not found"
-  osascript -e 'tell application "System Events" to get name of every process' | tr "," "\n" | head -50
-  # Fallback: full screen capture
-  screencapture -x "$OUTPUT_FILE"
-  exit 0
+struct Candidate {
+  let layer: Int
+  let area: Int
+  let id: Int
+  let x: Int
+  let y: Int
+  let width: Int
+  let height: Int
+}
+
+func intValue(_ value: Any?) -> Int {
+  if let number = value as? NSNumber {
+    return number.intValue
+  }
+  if let int = value as? Int {
+    return int
+  }
+  if let double = value as? Double {
+    return Int(double)
+  }
+  return 0
+}
+
+func doubleValue(_ value: Any?, default defaultValue: Double = 0) -> Double {
+  if let number = value as? NSNumber {
+    return number.doubleValue
+  }
+  if let double = value as? Double {
+    return double
+  }
+  if let int = value as? Int {
+    return Double(int)
+  }
+  return defaultValue
+}
+
+for _ in 0..<30 {
+  let info = CGWindowListCopyWindowInfo(.optionOnScreenOnly, kCGNullWindowID) as? [[String: Any]] ?? []
+  var candidates: [Candidate] = []
+
+  for window in info {
+    let ownerPid = intValue(window[kCGWindowOwnerPID as String])
+    let ownerName = window[kCGWindowOwnerName as String] as? String ?? ""
+    if ownerPid != appPid && ownerName != appName {
+      continue
+    }
+
+    let bounds = window[kCGWindowBounds as String] as? [String: Any] ?? [:]
+    let width = intValue(bounds["Width"])
+    let height = intValue(bounds["Height"])
+    let windowID = intValue(window[kCGWindowNumber as String])
+    let alpha = doubleValue(window[kCGWindowAlpha as String], default: 1)
+    if width <= 1 || height <= 1 || windowID <= 0 || alpha <= 0 {
+      continue
+    }
+
+    candidates.append(Candidate(
+      layer: intValue(window[kCGWindowLayer as String]),
+      area: width * height,
+      id: windowID,
+      x: intValue(bounds["X"]),
+      y: intValue(bounds["Y"]),
+      width: width,
+      height: height
+    ))
+  }
+
+  if let best = candidates.sorted(by: {
+    if $0.layer == $1.layer {
+      return $0.area > $1.area
+    }
+    return $0.layer < $1.layer
+  }).first {
+    print("\(best.id),\(best.x),\(best.y),\(best.width),\(best.height)")
+    exit(0)
+  }
+
+  Thread.sleep(forTimeInterval: 0.5)
+}
+
+print("ERROR: CoreGraphics window not found")
+exit(1)
+SWIFT
+
+  swiftc get_cg_window_info.swift -o get_cg_window_info
 fi
 
+# Wait for application startup
+sleep 3
+
 # Write AppleScript to a temporary file to avoid heredoc issues.
 # AXWindowNumber is the CoreGraphics window id used by screencapture -l, but
 # some GitHub-hosted macOS images do not expose it for every Accessibility
@@ -77,7 +153,32 @@ OSA
 sed -i '' "s/APP_NAME_PLACEHOLDER/$APP_NAME/g" get_window_info.applescript
 
 # Get window id and rect as id,x,y,w,h
-WINDOW_INFO=$(osascript get_window_info.applescript)
+echo "Waiting for window to appear..."
+WINDOW_INFO=$(./get_cg_window_info "$APP_NAME" "$APP_PID" || true)
+if [[ "$WINDOW_INFO" == ERROR:* ]] || [ -z "$WINDOW_INFO" ]; then
+  echo "$WINDOW_INFO"
+  echo "CoreGraphics lookup failed; trying Accessibility window lookup."
+  for i in {1..20}; do
+    EXISTS=$(osascript -e "tell application \"System Events\" to return exists process \"$APP_NAME\"" || true)
+    if [ "$EXISTS" = "true" ]; then
+      HAS_WIN=$(osascript -e "tell application \"System Events\" to tell process \"$APP_NAME\" to return (count of windows) > 0" || true)
+      if [ "$HAS_WIN" = "true" ]; then
+        break
+      fi
+    fi
+    sleep 0.5
+  done
+
+  if [ "${EXISTS:-false}" != "true" ] || [ "${HAS_WIN:-false}" != "true" ]; then
+    echo "ERROR: process or window not found"
+    osascript -e 'tell application "System Events" to get name of every process' | tr "," "\n" | head -50
+    # Fallback: full screen capture
+    screencapture -x "$OUTPUT_FILE"
+    exit 0
+  fi
+
+  WINDOW_INFO=$(osascript get_window_info.applescript)
+fi
 case "$WINDOW_INFO" in
   ERROR:*) echo "$WINDOW_INFO"; screencapture -x "$OUTPUT_FILE"; exit 0 ;;
 esac