[jira] [Created] (GUACAMOLE-2321) guacd 1.6.0: GUAC_ASSERT failure in guac_display_remove_layer() (display-layer-list.c:327) terminates active RDP connections

["Diederick Stoffers (Jira)" <ji...@apache.org>]

Diederick Stoffers created GUACAMOLE-2321:
---------------------------------------------

             Summary: guacd 1.6.0: GUAC_ASSERT failure in guac_display_remove_layer() (display-layer-list.c:327) terminates active RDP connections
                 Key: GUACAMOLE-2321
                 URL: https://issues.apache.org/jira/browse/GUACAMOLE-2321
             Project: Guacamole
          Issue Type: Bug
         Environment: guacd 1.6.0 (official image guacamole/guacd:1.6.0, Alpine base)
guacamole-client 1.6.0 (guacamole/guacamole:1.6.0)
Protocol: RDP
Deployment: Docker, guacd and client in separate containers
            Reporter: Diederick Stoffers


Since upgrading to 1.6.0, guacd intermittently aborts the connection
handler for an active RDP session with:

    GUAC_ASSERT in guac_display_remove_layer() failed at display-layer-list.c:327.

The assertion is always immediately followed by the connection being torn
down, e.g.:

    GUAC_ASSERT in guac_display_remove_layer() failed at display-layer-list.c:327.
    guacd[1]: INFO:  Connection "$33598ecb-d41f-48ea-b8e8-24b0c98b9dff" removed.
    guacd[1]: INFO:  Creating new client for protocol "rdp"

From the user's point of view the RDP session simply disappears and must
be reconnected. Because guacd forks per connection, only the child process
dies -- the guacd container itself stays up with a passing healthcheck, so
the failure is invisible to ordinary process-level monitoring.

Observed 7 times across 5 distinct days on a single guacd instance:

    2026-08-16 22:52:57
    2026-08-18 09:26:55, 09:27:25, 09:29:17, 10:43:16
    2026-08-19 12:27:42
    2026-08-20 09:42:28

The cluster on 2026-08-18 (three failures within ~2.5 minutes) appears to
be a user reconnecting and hitting the same fault repeatedly.

Six of the seven occurrences terminated the connection immediately. In the
remaining case the same connection continued logging afterwards.

A possibly relevant correlation: the majority of occurrences are preceded
by repeated CLIPRDR warnings on the same connection:

    guacd[8]: WARNING:  Received clipboard data cannot be sent to the RDP
    server because the RDP server has requested a clipboard format which
    was not declared as available. This violates the specification for the
    CLIPRDR channel.

Not all occurrences have this preamble -- one follows normal RDPDR device
redirection activity ("RDPDR user logged on") instead -- so clipboard
activity may be coincidental rather than causal. Flagging it in case the
layer lifecycle and the clipboard channel interact.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)