Skip to content

Lock screen squeezed after locking in landscape: the lock surfaces are made before autorotate turns to portrait #522

Open

Description

@den-kloot

What happens

When the phone is locked while it is in landscape (Power key or the idle lock), the lock screen stays squeezed until the next unlock: the clock, the date, "Swipe up to unlock" and the battery glyph are drawn far too narrow and too tall, and the background has vertical streaks. It looks like a landscape lock surface stretched onto the portrait screen.

Why (from the journal)

omamobile-autorotate turns a locked phone to portrait, but only on its next one-second poll (_poll_lock / _lock_tick, shell_locked() through omarchy-shell lock isLocked). By then the session-lock surfaces have been made for landscape, and they are not configured again after the transform change:

13:47:31 omamobile-autorotate: DSI-1: transform 3      <- phone on its side
13:51:33 omarchy lock ... lock-requested
13:51:34 omamobile-autorotate: portrait while locked
13:51:34 omamobile-autorotate: DSI-1: transform 0
13:51:34 omarchy lock ... session-locked=true          <- surfaces from the landscape size
14:23:56 omarchy lock ... unlocked                       <- squeezed the whole time

Every lock taken in portrait looks right; every one taken on the side that I checked looks squeezed.

Workaround I'm running

Turn to portrait before locking: before omarchy-system-lock, if the monitor's transform isn't 0, the same hyprctl eval that autorotate.py's apply() sends (hl.monitor({... transform = 0 }) plus the touch transform), then 0.5 s, then the lock. With that the lock screen is right after a landscape lock.

Where a fix could go

I haven't picked one, since both paths that lock would need it:

  • a "portrait before lock" step in the lock path itself (omamobile-keys for Power, and the idle lock), or
  • the lock surfaces following a transform change while locked (Quickshell's WlSessionLock surfaces should get a new configure from Hyprland after the monitor change; it looks like they don't, or the shell ignores it). That would be the cleaner fix, and would also cover a rotation while locked.

Device and build

  • Pixel 9a (tegu), kernel 7.3.0-rc6-omamobile-push1, omamobile-sensors 0.2.1, omamobile-shell 0.2.100
  • Only seen with autorotate on; with rotation lock on, the phone never leaves portrait.

Activity

  1. den-kloot commented on Nov 14, 2026

    @den-kloot
    ContributorAuthor

    Found the cause, and it is one layer down from what I wrote above: Hyprland never tells the lock surface about the turn.

    • Hyprland 0.57 applies a monitor rule that differs from the active one only in the transform through CMonitor::applyMonitorRuleSoft(). The monitor gets its turned size, but modeChanged is not emitted.
    • CSessionLockSurface sends ext_session_lock_surface_v1.configure only on modeChanged. With WAYLAND_DEBUG=1 on a lock client: one configure(405, 909) at the lock, then none for four quarter turns, while zxdg_output_v1.logical_size went to 909x405 and back each time.
    • So the lock client keeps drawing at the old size and the renderer stretches that buffer over the turned output. That is the squeezed picture, and why touch misses the PIN pad.

    It also happens without locking in landscape. On 2026-11-13 it hit me twice with the phone upright: it tipped to landscape in the same tenth of a second as the Power press, the lock surfaces were made for landscape, and autorotate's "portrait while locked" came a second later.

    20:08:47.562  omamobile-autorotate: DSI-1: transform 1
    20:08:47.688  omarchy lock … lock-requested
    20:08:48.289  omarchy lock … session-locked=true
    20:08:49.062  omamobile-autorotate: portrait while locked
    

    So the "portrait before lock" workaround from this issue is not enough: it looks once, before the turn lands.

    Two things that confused my experiments, in case someone repeats them: a rule whose scale is written differently from the active one (2.6666667 against 2.66667) takes Hyprland's full path and does send a configure; and while the display is off Hyprland applies no rule change until the next wake.

    The fix and an on-phone test are in #753.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions