https://bugs.kde.org/show_bug.cgi?id=522961
[email protected] changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |[email protected] --- Comment #8 from [email protected] --- I can reproduce this on the same platform (CachyOS) and on a newer version than the original report (6.7.5 vs 6.7.2), so it is not fixed in 6.7.5. I also narrowed down which layer is at fault. Environment - CachyOS, KDE Plasma 6.7.5 (plasma-desktop 6.7.5-1.1) - KDE Frameworks 6.30.0 (bluez-qt 6.30.0-1.1) - bluedevil 6.7.5-1.1 - BlueZ 5.87-2.1 - Headset: Sony WH-1000XM5, the only paired Bluetooth device - UI language: French Symptoms - For about 2-3 months, each time I come back to the computer after being away, the Bluetooth widget shows additional copies of my headset. I have seen up to 3 entries at once (see attached screenshot). - The copies have the same address as the real device and show Paired: Yes, Trusted: Yes, Adapter: CachyOS (hci0). - The real entry is named "WH-1000XM5". The copies are named "WH-1000XM5 - hci0". - In the screenshot, the real entry is under "Connected", one copy is under "Available" with a "Connect" button, and another is stuck on "Connecting" with a spinner. - A reboot clears them: one device again, name back to "WH-1000XM5". Diagnosis While the duplicates were visible in the widget (no restarts done beforehand): - `bluetoothctl devices` lists only ONE device. - `busctl tree org.bluez` shows only ONE device object under /org/bluez/hci0 (with its sep1/sep2/sep3 endpoints). - The bluetoothd journal shows nothing unusual (no errors, no repeated device add/remove messages). Running `systemctl --user restart plasma-plasmashell` immediately removes all the duplicates, no reboot needed. Conclusion: BlueZ, and therefore the kernel and the adapter, only know one device. D-Bus object paths are derived from the address, so BlueZ cannot expose two objects for the same device on hci0. The duplicates exist only inside plasmashell, either in the bluez-qt device model or in the Bluetooth widget. The " - hci0" suffix suggests the ghost entries are built through a different code path (for example a fallback name when the device is re-added or its properties are momentarily unavailable) than the real entry. Not yet determined - The exact trigger: suspend/resume, screen lock, or the headset auto-reconnecting after power off/on. I did not capture D-Bus signals (InterfacesAdded/InterfacesRemoved) at the moment the duplicate appears. Workaround - systemctl --user restart plasma-plasmashell -- You are receiving this mail because: You are watching all bug changes.
