Bring dock monitors back when an undock leaves them black at 0x0 until reboot
The mechanism holds: aquamarine 386 (closed 2026-09-15) describes the guard and the orphaned CRTC, PR 410 merged 2026-09-15, v0.15.1 was published 2026-09-17 with #410 in its notes, and Arch extra's JSON shows 0.15.1-1 updated 2026-09-17. Re-checked the mirrors on 2026-10-05 by downloading extra.db: stable-mirror and rc-mirror still list aquamarine-0.15.0-2, and mirror.omarchy.org (the edge mirrorlist in /usr/share/omarchy/default/pacman/mirrorlist-edge) lists 0.15.1-1. This workstation runs 0.15.0-2. omarchy 11019 is open and supports both recovery methods, but it also shows the reports are broader than the cause says (amdgpu, xe and NVIDIA reproductions, not only i915), a 2026-10-02 MST case where a VT switch alone left the outputs at 0x0 until an explicit mode was re-applied, and that hyprctl reload did not retry the modeset. The 'disable before undocking' prevention is a commenter's untested suggestion ('should avoid'), so it is now labelled that way. The danger was overstated: two commenters (2026-09-18 and 2026-10-03) installed Arch's 0.15.1-1 with pacman -U and confirmed the fix, the package provides the same libaquamarine.so=14 as the installed 0.15.0-2, and its library dependencies (libhyprutils.so=13, libdisplay-info.so=3) match what is installed here. Read /usr/bin/omarchy-update-pacman-guard: it blocks only transactions with both -S and -u, so pacman -U passes. Nothing was installed or run against a dock.
Switching Omarchy to the edge channel moves every package to the rolling mirror and replaces omarchy with omarchy-dev, not only aquamarine. Installing Arch's aquamarine with pacman -U puts one package ahead of Omarchy's tested snapshot. Reporters confirmed it works with Hyprland 0.56.2-2, but it is outside what Omarchy tests, and omarchy update will report the local package as newer until stable catches up. Do not use pacman -U with any aquamarine build whose libaquamarine.so version differs from the one your installed hyprland was built against (pacman -Qi aquamarine | grep Provides).
Symptom
Booting docked works, and plugging the dock in after a clean boot works. But unplug the dock, plug it back in, and the external monitors never come back. hyprctl monitors all lists them at 0x0@60, and the Hyprland log repeats atomic drm request: failed to commit: Invalid argument, flags: ATOMIC_ALLOW_MODESET ATOMIC_TEST_ONLY, walking down to 720x400 and 640x480 and still failing. Only a reboot fixes it. Some people get the same thing after every lock-screen blank on DisplayPort monitors with no dock at all. Downgrading the kernel does not help.
Cause
aquamarine 0.15.0, Hyprland's DRM backend, added a guard that rejects commits to disconnected outputs. On unplug the connector is marked disconnected before Hyprland commits the disable, so the disable itself is refused and the kernel keeps a CRTC active on a dead connector. That orphaned pipe holds the display pipe and link resources (on Intel Type-C ports, the intel_tc link reference), so every later modeset on the new connector fails with EINVAL. The log line to look for is Cannot commit a disconnected output, followed by clearing stale crtc. Upstream tracked it as hyprwm/aquamarine#386, fixed by #410 (merged 2026-09-15) and released in aquamarine 0.15.1 on 2026-09-17. Arch extra carries 0.15.1-1. As of 2026-10-05 Omarchy's stable and rc mirrors still serve aquamarine 0.15.0-2 and only the edge mirror serves 0.15.1-1. omarchy 11019 has reproductions on i915, xe, amdgpu and NVIDIA, from undocking, from lock-screen and idle DPMS blanking with directly attached DisplayPort, and from switching cables. omarchy 11066 has the MST lock-screen case.
Fix
**1. Confirm it is this bug:**
pacman -Q aquamarine # 0.15.0-2 is affected
grep -c 'Cannot commit a disconnected output' "$XDG_RUNTIME_DIR/hypr/$HYPRLAND_INSTANCE_SIGNATURE/hyprland.log"
for c in /sys/class/drm/card*-DP-*; do echo "${c##*/} $(cat $c/status) $(cat $c/enabled)"; done
A connector that reads disconnected enabled is the orphaned pipe.
**2. Recover without rebooting.** Methods reported to work in omarchy 11019:
- Switch to another VT and back (Ctrl+Alt+F2, then the VT your session runs on). It can take two rounds.
- On an MST dock the VT switch can free the pipes but leave the monitors at
0x0on the remembered fallback mode. Then re-apply an explicit mode to each external output. MST connectors get new names on every re-dock, so take the names and modes fromhyprctl monitors all:
hyprctl monitors all -j | jq -r '.[] | "\(.name) \(.width)x\(.height) \(.availableModes[0])"'
hyprctl eval 'hl.monitor({ output = "DP-3", mode = "1680x1050@59.95", position = "auto", scale = 1 })' # repeat for each external output
- On i915, one reporter recovered by disabling every physical output, then re-applying the monitor config. This also works over ssh while the screens are black:
hyprctl eval 'hl.monitor({ output = "DP-1", disabled = true })' # repeat for each external output
hyprctl eval 'dofile(os.getenv("HOME") .. "/.config/hypr/monitors.lua")'
hyprctl reload on its own was reported not to retry the modeset.
**3. Possible prevention until you have 0.15.1 (suggested in omarchy 11019, not confirmed).** Disable the external outputs before you pull the dock, so the disable is committed while the connector still exists, then reload after redocking:
hyprctl eval 'hl.monitor({ output = "DP-5", disabled = true })' # each dock output, before unplugging
# after plugging back in:
hyprctl reload
**4. The fix is aquamarine 0.15.1.**
- **Plain Arch:**
sudo pacman -Syu. - **Omarchy 4:** run
omarchy update -yand checkpacman -Q aquamarine. As of 2026-10-05 the stable and rc mirrors still serve 0.15.0-2. The edge channel (omarchy-channel-set edge) has 0.15.1-1, but see the danger note. Two reporters in omarchy 11019 installed Arch's build directly and confirmed it fixes the bug. It provides the same library version (libaquamarine.so=14) as 0.15.0-2, so the installed Hyprland loads it:
sudo pacman -U https://archive.archlinux.org/packages/a/aquamarine/aquamarine-0.15.1-1-x86_64.pkg.tar.zst
Then log out and back in so Hyprland loads the new library.
Verify
pacman -Q aquamarine reports 0.15.1 or newer. Undock, redock, and hyprctl monitors shows the externals at their native modes with no new Cannot commit a disconnected output lines in the log.
Sources
- https://github.com/omacom/omarchy/issues/11019
- https://github.com/omacom/omarchy/issues/11066
- https://github.com/hyprwm/aquamarine/issues/386
- https://github.com/hyprwm/aquamarine/pull/410
- https://github.com/hyprwm/aquamarine/releases/tag/v0.15.1
- https://archlinux.org/packages/extra/x86_64/aquamarine/
- https://archive.archlinux.org/packages/a/aquamarine/aquamarine-0.15.1-1-x86_64.pkg.tar.zst
- https://stable-mirror.omarchy.org/extra/os/x86_64/extra.db
- https://mirror.omarchy.org/extra/os/x86_64/extra.db