You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I run the VNC backend on a Raspberry Pi 3B. During the experiment, I observed that certain areas of the window were missing on the screen. I am curious whether this issue might be caused by the tx->damage_region parameter.
#163 reworked the VNC backend: each scan-out now reports exactly the rectangle it wrote as damage (the region is initialized once instead of being re-created on every update), and pointer events are no longer dispatched from uninitialized state. Could you retest on the Raspberry Pi with current main and say whether windows still have missing areas? If they do, one remaining suspect is that the framebuffer handed to neatvnc is drawn into again while it may still be encoding, with no second buffer.
Activity
jserv commentedon Oct 19, 2026
#163 reworked the VNC backend: each scan-out now reports exactly the rectangle it wrote as damage (the region is initialized once instead of being re-created on every update), and pointer events are no longer dispatched from uninitialized state. Could you retest on the Raspberry Pi with current main and say whether windows still have missing areas? If they do, one remaining suspect is that the framebuffer handed to neatvnc is drawn into again while it may still be encoding, with no second buffer.
huaxinliao commentedon Oct 19, 2026
I would retest on the Raspberry Pi with current main on the weekend.