From 76add0b9b2becf39fd99b910c9dc377e7c4784e5 Mon Sep 17 00:00:00 2001
From: Viktor Liu <viktor@netbird.io>
Date: Sun, 17 May 2026 16:47:53 +0200
Subject: [PATCH] Fix ExtendedClipboard auto-request by advertising all actions
 in Caps

---
 client/internal/engine_vnc_stub.go     |  1 +
 client/vnc/server/extclipboard.go      | 13 +++++++----
 client/vnc/server/extclipboard_test.go |  8 ++++++-
 client/vnc/server/session.go           | 32 ++++++++++++++++----------
 4 files changed, 37 insertions(+), 17 deletions(-)

diff --git a/client/internal/engine_vnc_stub.go b/client/internal/engine_vnc_stub.go
index b1d0ac262..688a588f7 100644
--- a/client/internal/engine_vnc_stub.go
+++ b/client/internal/engine_vnc_stub.go
@@ -10,6 +10,7 @@ type vncServer interface{}
 
 func (e *Engine) updateVNC(_ *mgmProto.SSHConfig) error { return nil }
 
+// updateVNCServerAuth is a no-op on platforms without a VNC server.
 func (e *Engine) updateVNCServerAuth(_ *mgmProto.VNCAuth) {}
 
 func (e *Engine) stopVNCServer() error { return nil }
diff --git a/client/vnc/server/extclipboard.go b/client/vnc/server/extclipboard.go
index 495f0dc63..171430e77 100644
--- a/client/vnc/server/extclipboard.go
+++ b/client/vnc/server/extclipboard.go
@@ -52,11 +52,16 @@ const (
 	extClipMaxPayload = extClipMaxText + 1024
 )
 
-// buildExtClipCaps emits the Caps payload advertising the formats we accept
-// and our maximum size per format. One uint32 size follows the flags word
-// for each format bit set, in ascending bit order.
+// buildExtClipCaps emits the Caps payload. The flags word advertises every
+// action we support in the high byte (Caps + Request + Peek + Notify +
+// Provide) and every format we accept in the low 16 bits. noVNC uses these
+// action bits to decide whether to auto-Request on Notify; without
+// Request in our Caps it silently drops our Notify messages. After the
+// flags word we emit one uint32 max size per format bit set, in ascending
+// bit order.
 func buildExtClipCaps() []byte {
-	flags := extClipActionCaps | extClipFormatText
+	flags := extClipActionCaps | extClipActionRequest | extClipActionPeek |
+		extClipActionNotify | extClipActionProvide | extClipFormatText
 	payload := make([]byte, 4+4)
 	binary.BigEndian.PutUint32(payload[0:4], flags)
 	binary.BigEndian.PutUint32(payload[4:8], uint32(extClipMaxText))
diff --git a/client/vnc/server/extclipboard_test.go b/client/vnc/server/extclipboard_test.go
index fd9601a79..43c278bc3 100644
--- a/client/vnc/server/extclipboard_test.go
+++ b/client/vnc/server/extclipboard_test.go
@@ -16,7 +16,13 @@ func TestBuildExtClipCaps(t *testing.T) {
 	require.Len(t, payload, 8, "Caps with one format should be 4 bytes flags + 4 bytes size")
 
 	flags := binary.BigEndian.Uint32(payload[0:4])
-	assert.Equal(t, extClipActionCaps, flags&extClipActionMask, "action should be Caps")
+	// noVNC checks individual action bits in our Caps to decide whether to
+	// auto-Request on Notify, so all supported actions must be advertised.
+	assert.NotZero(t, flags&extClipActionCaps, "Caps action bit must be set")
+	assert.NotZero(t, flags&extClipActionRequest, "Request action bit must be set")
+	assert.NotZero(t, flags&extClipActionPeek, "Peek action bit must be set")
+	assert.NotZero(t, flags&extClipActionNotify, "Notify action bit must be set")
+	assert.NotZero(t, flags&extClipActionProvide, "Provide action bit must be set")
 	assert.Equal(t, extClipFormatText, flags&extClipFormatMask, "should advertise text format")
 
 	maxSize := binary.BigEndian.Uint32(payload[4:8])
diff --git a/client/vnc/server/session.go b/client/vnc/server/session.go
index aa656c28c..e59fb2edd 100644
--- a/client/vnc/server/session.go
+++ b/client/vnc/server/session.go
@@ -877,24 +877,32 @@ func (s *session) handleExtCutText(payloadLen uint32) error {
 		}
 		return nil
 	case extClipActionProvide:
-		if len(rest) == 0 {
-			return nil
-		}
-		text, err := parseExtClipProvideText(flags, rest)
-		if err != nil {
-			s.log.Debugf("parse ext clipboard provide: %v", err)
-			return nil
-		}
-		if text != "" {
-			s.injector.SetClipboard(text)
-		}
-		return nil
+		return s.handleExtClipProvide(flags, rest)
 	default:
 		s.log.Debugf("unknown ext clipboard action 0x%x", action)
 		return nil
 	}
 }
 
+// handleExtClipProvide decodes a Provide payload and pushes the recovered
+// text into the host clipboard. Errors and other unsupported formats (RTF,
+// HTML, etc.) are swallowed so a malformed message doesn't tear down the
+// session.
+func (s *session) handleExtClipProvide(flags uint32, payload []byte) error {
+	if len(payload) == 0 {
+		return nil
+	}
+	text, err := parseExtClipProvideText(flags, payload)
+	if err != nil {
+		s.log.Debugf("parse ext clipboard provide: %v", err)
+		return nil
+	}
+	if text != "" {
+		s.injector.SetClipboard(text)
+	}
+	return nil
+}
+
 // sendExtClipProvideText answers an inbound Request(text) with the current
 // host clipboard contents, capped to extClipMaxText.
 func (s *session) sendExtClipProvideText() error {