fix: add Xwayland startup and dynamic DISPLAY detection for GNOME screenshots

- Start Xwayland explicitly on :99 after gnome-shell's wayland-0 is ready
- Add findXwaylandDisplay() to detect X11 sockets dynamically
- Check displays :0, :1, :2, :99 in both XDG_RUNTIME_DIR and /tmp
- Add x11-utils package for xdpyinfo
- Better logging for screenshot fallback debugging

Note: Xwayland+scrot approach has limitations - only captures X11 windows,
not native Wayland apps. May need to switch to PipeWire-based capture
via remotedesktop-session.py for proper GNOME screenshots.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <[email protected]>

Author: lukemarsden

Dockerfile.ubuntu-helix

@@ -289,7 +289,7 @@ RUN apt-get update && apt-get install -y \
     fonts-noto-color-emoji fonts-dejavu-core \
     imagemagick \
     kitty \
-    scrot xclip \
+    scrot xclip x11-utils \
     grim slurp \
     && rm -rf /var/lib/apt/lists/*
 
@@ -519,7 +519,24 @@ fi
 
 if [ -x /usr/local/bin/screenshot-server ]; then
   gow_log "[start] Starting screenshot server with GNOME environment..."
-  # Uses org.gnome.Shell.Screenshot D-Bus API
+  # Start Xwayland explicitly on DISPLAY=:99 for scrot fallback
+  # GNOME headless may not auto-start Xwayland without X11 app demand
+  (
+    sleep 3  # Wait for gnome-shell's wayland-0 socket to be ready
+    if [ -S "\$XDG_RUNTIME_DIR/wayland-0" ]; then
+      gow_log "[start] Starting Xwayland on :99 for screenshot support..."
+      WAYLAND_DISPLAY=wayland-0 Xwayland :99 -auth /tmp/.Xauthority &
+      sleep 1
+      if [ -S "/tmp/.X11-unix/X99" ]; then
+        gow_log "[start] Xwayland started successfully on DISPLAY=:99"
+      else
+        gow_log "[start] WARNING: Xwayland may not have started properly"
+      fi
+    else
+      gow_log "[start] WARNING: wayland-0 not found, cannot start Xwayland"
+    fi
+  ) &
+  # Uses org.gnome.Shell.Screenshot D-Bus API, falls back to scrot via Xwayland
   WAYLAND_DISPLAY=wayland-0 XDG_CURRENT_DESKTOP=GNOME /usr/local/bin/screenshot-server >> /tmp/screenshot-server.log 2>&1 &
 fi
 

api/cmd/screenshot-server/main.go

@@ -476,19 +476,57 @@ func captureScreenshotGNOME(format string, quality int) ([]byte, string, error)
 	return pngData, "png", nil
 }
 
+// findXwaylandDisplay tries to detect the Xwayland display from socket files
+// GNOME/Mutter starts Xwayland on-demand, so we check for X socket files
+func findXwaylandDisplay() string {
+	// Check environment first
+	if display := os.Getenv("DISPLAY"); display != "" {
+		return display
+	}
+
+	// Check for X socket files in XDG_RUNTIME_DIR (Xwayland creates these)
+	xdgRuntimeDir := os.Getenv("XDG_RUNTIME_DIR")
+	if xdgRuntimeDir == "" {
+		xdgRuntimeDir = "/run/user/1000"
+	}
+
+	// Display numbers to check - includes :99 which we start explicitly for GNOME headless
+	displayNumbers := []int{0, 1, 2, 99}
+
+	// Look for X socket files (format: X{N} where N is display number)
+	for _, i := range displayNumbers {
+		socketPath := filepath.Join(xdgRuntimeDir, fmt.Sprintf(".X11-unix/X%d", i))
+		if _, err := os.Stat(socketPath); err == nil {
+			display := fmt.Sprintf(":%d", i)
+			log.Printf("[GNOME/X11] Found Xwayland socket at %s, using DISPLAY=%s", socketPath, display)
+			return display
+		}
+	}
+
+	// Also check /tmp/.X11-unix (traditional location)
+	for _, i := range displayNumbers {
+		socketPath := fmt.Sprintf("/tmp/.X11-unix/X%d", i)
+		if _, err := os.Stat(socketPath); err == nil {
+			display := fmt.Sprintf(":%d", i)
+			log.Printf("[GNOME/X11] Found X socket at %s, using DISPLAY=%s", socketPath, display)
+			return display
+		}
+	}
+
+	log.Printf("[GNOME/X11] No X11 display socket found, defaulting to :99")
+	return ":99"
+}
+
 // captureScreenshotGNOMEScreenCast captures a screenshot using scrot via X11.
 // DISABLED: The PipeWire-based gnome-screenshot.py conflicts with Wolf's video capture.
 // Reading from the same PipeWire node as Wolf causes stream interference and crashes.
 // Fall back to scrot which uses X11/Xwayland instead.
 func captureScreenshotGNOMEScreenCast(format string, quality int) ([]byte, string, error) {
 	log.Printf("[GNOME/X11] Screenshot D-Bus blocked, falling back to scrot via Xwayland...")
 
-	// Use scrot via X11 - this doesn't interfere with Wolf's PipeWire capture
-	// GNOME runs on Xwayland at DISPLAY=:0 (or :9 in some configs)
-	display := os.Getenv("DISPLAY")
-	if display == "" {
-		display = ":0" // Default for GNOME's Xwayland
-	}
+	// Detect Xwayland display from socket files
+	display := findXwaylandDisplay()
+	log.Printf("[GNOME/X11] Using DISPLAY=%s for scrot", display)
 
 	// Create temporary file for screenshot
 	tmpDir := os.TempDir()
@@ -510,6 +548,7 @@ func captureScreenshotGNOMEScreenCast(format string, quality int) ([]byte, strin
 
 	output, err := cmd.CombinedOutput()
 	if err != nil {
+		log.Printf("[GNOME/X11] scrot failed with DISPLAY=%s: %v, output: %s", display, err, string(output))
 		return nil, "", fmt.Errorf("scrot failed: %v, output: %s", err, string(output))
 	}