Fix the Bluetooth panel stuck on 'Turned Off' while bluetoothctl says Powered: yes
Read omacom/omarchy#7573 in full with all six comments. It supports the record: the triage comment establishes the missing reconciliation from a Quickshell 0.3.1 source read, and two reporters later posted instrumented QT_LOGGING_RULES logs on different hardware (Intel AX210 8087:0032 and Qualcomm QCA9377 04ca:3015) showing PowerState Enabling with Powered false as the final write to Powered in the session, and zero GetAll calls for org.bluez against 165 and 187 for NetworkManager. Confirmed on this machine at omarchy 4.0.2-1 by reading /usr/share/omarchy/shell/plugins/panels/bluetooth/Panel.qml: line 24 is the Bluetooth.defaultAdapter binding, line 79 returns "Turned Off" on !adapter.enabled, line 512 gates the discovery retry timer on adapter.enabled, and lines 635 to 637 are the toggleBluetooth that sends a direction from the cached value. /usr/share/omarchy/bin/omarchy-bluetooth-power has the is-on subcommand the symptom cites, and its header comment confirms the rfkill soft block is the state that persists. The verify command ran here and printed b true with rfkill unblocked, so that command and its output format are confirmed. Two defects. First, the cause said BlueZ re-registers hci0 at boot and Quickshell drops the adapter object and builds a new one, which is wrong for the boot route: the first reporter's shell built its FIRST adapter object during a 1.44 second Intel firmware load, with no prior object to drop, so the two entry routes are different and the cause is rewritten to keep both. Second, the fix claimed an omarchy-bluetooth-power off then on cycle leaves the panel reading the cached value, and the thread contradicts that: an rfkill soft block removes the adapter object entirely and the panel renders "No adapter" (the #6956 state the report itself distinguishes), and one instrumented log shows an explicit unblock's Powered transition being caught normally 38 seconds after a subscription was installed. The off and on cycle is a coin flip against the same race, not a guaranteed no-op, and its real hazard is the persisted block, so the fix paragraph is rewritten. Verified today that no fix has landed: v4.0.2...v4.0.3 touches no file under shell/plugins/panels/bluetooth/, the last commit to that Panel.qml is 3af7675a on 2026-08-26 requiring textFormat on Text elements, no merged pull request covers it, and the newest upstream Quickshell tag is v0.3.1 which is exactly what Arch extra ships and what is installed here. NOT exercised: I did not suspend or resume this workstation, did not toggle the adapter, and did not run an instrumented shell, so I confirmed the code path and the contradiction rather than reproducing the latch.
Symptom
The Omarchy shell's Bluetooth panel shows Turned Off, its switch is unchecked, and clicking the switch does nothing. The AVAILABLE list never fills, so no new device can be paired from the panel. Underneath, Bluetooth is fully working: bluetoothctl show reports Powered: yes, rfkill list bluetooth shows no block, omarchy-bluetooth-power is-on exits 0, and bluetoothctl scan on finds devices. Already-paired devices can still connect and play audio. Each click on the dead switch logs another rfkill: unblock set for type bluetooth in the journal. Reported on Omarchy 4.0.0-1 and 4.0.1-1 with quickshell-git 0.3.0.r20.g28771c7 and packaged quickshell 0.3.1-1, on Intel 8087:0a2b and 8087:0032, Realtek RTL8852BU, Qualcomm QCA9377 and Broadcom BCM20702 controllers. It appears either at boot, when the controller's firmware finishes loading after the shell has started, or after a suspend and resume with Bluetooth left on.
Cause
Established in the thread by instrumented Quickshell logs from two reporters on different hardware and by a source read of Quickshell 0.3.1. Panel.qml:24 is readonly property var adapter: Bluetooth.defaultAdapter, and adapter.enabled is a live binding onto BlueZ's org.bluez.Adapter1.Powered through Quickshell's BluetoothAdapter. Quickshell fills that cache from two places only: the property snapshot carried in BlueZ's InterfacesAdded or GetManagedObjects payload, and later PropertiesChanged signals. The PropertiesChanged subscription is installed in the adapter object's constructor, which runs strictly later than the moment BlueZ built the payload, and BluetoothAdapter is one of only two of Quickshell's D-Bus property groups that never calls GetAll to reconcile afterwards. A Powered transition landing in that gap is in neither source, so the cache keeps false and nothing repairs it for as long as Powered then stays true.
There are two ways into that gap, and they are different events. At boot the shell can build its **first** adapter object while the controller is still loading firmware, which is the Intel 8087:0a2b case at 1.44 seconds of firmware load. On resume BlueZ re-registers hci0 after a firmware reload, Quickshell destroys the adapter object and builds a new one, and that snapshot arrives carrying PowerState: Enabling with Powered: false while BlueZ is still powering the controller up. Both instrumented logs record that exact pair as the last write to Powered in the whole session, and both logs count zero GetAll calls for org.bluez in a process that made well over a hundred for NetworkManager. The resume route does not need the USB device to be re-enumerated: one reporter's controller held the same bus address across every cycle and BlueZ still tore the adapter object down and rebuilt it after the rampatch reload.
Because toggleBluetooth() at Panel.qml:635-637 sends a direction derived from the cached value, the switch sends on to an adapter that is already on, which is a no-op. The defect is in Quickshell, not in Omarchy's panel: the v0.3.0 to v0.3.1 diff touches no file under src/bluetooth/ or src/dbus/, so quickshell-git 0.3.0.r20.g28771c7 and packaged quickshell 0.3.1 run identical code here. No fix had landed on either side as of 2026-09-11. Omarchy's bluetooth panel is unchanged through v4.0.3, and v0.3.1 is still the newest Quickshell release.
Fix
Restart the shell. A fresh process takes a fresh snapshot of the adapter:
omarchy restart shell
Every reporter in the thread confirmed this clears it for the session. It recurs on the next boot or resume that hits the same window. To pair a device without restarting, use the CLI, which talks to BlueZ directly:
bluetoothctl scan on
bluetoothctl pair <MAC>
bluetoothctl connect <MAC>
Do not reach for an off and on cycle instead. The panel switch cannot perform one, because in this state it believes the adapter is already off and sends on, so you would have to run the helper by hand:
omarchy-bluetooth-power off
omarchy-bluetooth-power on
That is a coin flip rather than a cure. off sets an rfkill soft block, which removes the adapter object and makes the panel read No adapter instead, and on builds a new object against the same snapshot race, so it can latch false again. The block is also the half that systemd-rfkill persists across reboots, so an on that does not complete leaves Bluetooth switched off at the next boot. omarchy restart shell has neither failure mode.
Verify
Before restarting, confirm the mismatch:
busctl --system get-property org.bluez /org/bluez/hci0 org.bluez.Adapter1 Powered
b true while the panel says Turned Off is this bug. After omarchy restart shell the panel shows the adapter on, the switch responds, and the AVAILABLE list fills when the panel is open.
Sources
- https://github.com/omacom/omarchy/issues/7573
- https://github.com/omacom/omarchy/blob/quattro/shell/plugins/panels/bluetooth/Panel.qml
- https://github.com/omacom/omarchy/blob/quattro/test/shell.d/bluetooth-test.sh
- https://github.com/omacom/omarchy/compare/v4.0.2...v4.0.3
- https://archlinux.org/packages/extra/x86_64/quickshell/
- https://git.outfoxxed.me/quickshell/quickshell