Hi Phil,
I followed the SSCE program steps in
https://bugs.openjdk.org/browse/JDK-8365960 and I also confirmed this
with our own software. There is a program there that continously draws
to screen.
Let's call the machine being connected to the "host" and the machine
connecting the "guest". The host machine has to have a physical video
card for this to work. In my case I connected two machines with Windows
11 installed (one laptop guest to a stationary host).
There are two ways to reproduce this:
A.
1. On the host, start a Java GUI
2. On the guest, connect to the host with an RDP connection -- with the
same user account
(this overtakes the session, kicking the host user out)
B. You may also trigger this with the opposite:
1. Connect with the guest machine to the host, then
2. Start a Java GUI program
3. Login to the same account again on the host (kicking out the guest)
Result:
1. The JVM crashes, with the violation exception.
2. The RDP session remains connected and everything looks fine (in A).
Let me know if I can provide any more information to help with this!
--
Edward
On 30/03/2026 22.05, Philip Race wrote:
Thanks for the info.
I'm not sure your scenario is quite the one that was originally reported.
RDP and two physical machines, I see you say that, but what is the
exact sequence ?
Are you using RDP to start a Java app on the remote system ? Does that
remote system have a video card + display ?
At that time I presume the app is displaying windows (ie headful) ?
And does the RDP connection get dropped ?
If it does, when does the application crash ?
I don't think we've ever been able to reproduce the problem in any
form here. Maybe getting this info will help.
Also making it a bit of a tricky one in that "headless" isn't designed
to be something that changes during the life of the application.
-phil
On 3/27/26 12:35 AM, Edward wrote:
Small update: I got in contact with jdk-updates-dev because I thought
it was fixed in older versions but it was just reverted.
The original issue which seems to have caused the bug was
https://bugs.openjdk.org/browse/JDK-8185862
Then, this was reverted specifically for 17 and 21 in
https://bugs.openjdk.org/browse/JDK-8348625
But the same issue still remains in 23+
The target issue that I was pointed to:
https://bugs.openjdk.org/browse/JDK-8336862. This is medium prio
(P3), but was unassigned recently. Some of the linked issues mention
RDP, but some are about VM or headless. But they all seem to be
related to each other.
My personal findings regarding RDP specifically:
1. This bug only occurs when two physical Windows machines are
involved. Connecting with a virtual machine or from Linux to Windows
do not seem to have this issue. Windows seems to choose software
rendering in these cases.
2. No JVM flags that disable hardware acceleration or similar seem to
work, and neither does disabling hardware acceleration in Windows
Settings or the group policy editor for RDP connections.
3. Just minimizing the java application window does **not** work as a
workaround either.
4. The only actual workaround I've found is installing a virtual
display driver and then forcing RDP to use that display
(https://github.com/VirtualDrivers/Virtual-Display-Driver
<https://urldefense.com/v3/__https://github.com/VirtualDrivers/Virtual-Display-Driver__;!!ACWV5N9M2RV99hQ!O4jMVhbDZHM1cbGpKNHhtFIvGx0WfkSoFLCL45a4tr9abzFRDOQvsJONP9lf_CoE8PWp-KmPUZ9UKJx3LXh7mg$>)
But that workaround is too much of a hassle to propose for our users.
--
Edward
