A new /dev/input node is root:root 0600 until udev applies GROUP=input. The
scan probed each new node once and marked it seen even when the open failed, so
a node caught in that gap was never opened. Behind a KVM, a switch brings back a
hub of devices at once: on the Frame, four nodes failed with EACCES in one switch,
the keyboard was never grabbed, and its keys went to gamescope instead of the
desktop screens. A node that isn't readable yet now waits for the next scan.
From jlneal's PR: a trigger press on a desktop screen stays put until the
laser moves more than 8 logical pixels, so controller jitter doesn't turn
a click into a drag. On top of it, only a hand controller's press starts
the filter: the 3D mouse's laser reaches the screens the same way, and the
PR as sent turned the mouse's short drags into clicks.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The 3D mouse drives SteamVR's laser through the ft_pointer virtual
controller, so its events reach the screens the same way a controller's
do. The filter held every press, which turned the mouse's short drags
(selecting a character or two, nudging a slider) into clicks. Mark
button events from hand controllers and start the filter only on those.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
container-up.sh starts the dev container in a scope of its own, so
distrobox enter finds it running and skips its wait for distrobox-init.
On a fresh install, init was still setting up passwordless sudo when
dev-container.sh ran sudo dnf install, and sudo asked for a password
with no terminal to read it from. container-up.sh now waits for
container_setup_done itself, and the container's sudo calls use -n,
so a password prompt fails at once with a clear message.
Fixes#9
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A user on a fresh install got "Calibration failed: only 0 of 21 dots" with
no reason. The installer never installed our tracker, so gaze used
SteamVR's, and the only way SteamVR's tracker rejects a dot is losing an
eye for most of the look. The panel just showed a red ring.
- install.sh: step 9/10 installs our tracker (gaze/tracker/install.sh)
after gaze mode, yes by default; it needs sudo, so --yes runs it only
when sudo won't prompt. If it fails, gaze keeps SteamVR's tracker.
Configs that still say GAZE_TRACKER=steam (the old template) are asked
whether to switch.
- GAZE_TRACKER=auto, the new default: ours when it's installed (the
frame grabber, its unit, and ft-eyes' Python), else SteamVR's.
ft-gazed rechecks every second, so installing it switches over. Input
Settings lists Own tracker first as recommended, and says how to
install it when it's missing (checking the host's /etc through
/run/host from the dev container).
- The calibration panel has a note line, orange over the instructions:
why a dot wasn't taken (an eye lost, a blink, the eyes disagreeing for
SteamVR's tracker, from steady_samples' new drop counts; ft-eyes' reply
for ours), what a click is still waiting for after 1.5 s, and a failed
calibration's most common reason, which the Gaze page shows too.
steady_samples keeps the same samples as before (checked on 2037
windows of recordings); a lost eye is named before a blink, since its
openness reads 0.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Turning gaze mode on without a calibration opened Calibrate only on the
off-to-on change, and only if it could open right then. With the headset
off, the eye tracker silent, or the panel not built, or with gaze mode
already on when the gaze service started, nothing opened and nothing said
why: the pointer just stayed a mouse.
- The gaze service now checks every second: gaze mode on, no calibration
for the tracker in use, eyes seen -> the full calibration opens. One
that closes unfinished opens again only after the headset comes off and
on, gaze mode off and on, or Calibrate, so it doesn't loop. A start that
fails retries every 10 s.
- Its status says why gaze mode can't work yet (checks.problem): not
calibrated and opening, open, closed unfinished, or can't open and why.
- Input Settings shows that under the Gaze pointer switch, along with the
gaze service not installed or not running and our tracker missing its
frame grabber.
- ft-gazectl on notes a missing calibration.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
hands/ft-cutouts on|off|status starts ft-camd and ft-hands as transient
user units with ft-hands' new --no-gestures: hands are published for
ft-screens' cutouts, but no pinch or grip is detected, so nothing clicks
or drags and a closing hand doesn't raise the tracking rate. It needs a
build and ft-camd's capabilities, not hands/run.sh install. Its units
conflict with ft-handsctl's, so each stops the other, and they stop with
SteamVR.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A Meta+key combination (Meta+J for gaze_left, say) hides Meta's release
from the desktop, and the relay skipped share_key for it too. frame-voice
saw Meta go down on @frametop_keys and never come up, so it held all
dictated text back, waiting for that release. Keys of grabbed keyboards
are now shared as pressed, before key_binding() decides what the desktop
gets.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Key combinations (and mouse and controller buttons) get two new actions:
- steam_menu: Open Steam menu / close dashboard (steam/ft-steam menu).
- command:CMD: run CMD with sh -c, as the relay's service, with layout/,
float/ and steam/ on its PATH. Input Settings offers it for key
combinations as Run a command...
Both work without pointer mode.
A modifier on its own is now a key combination too: a tap, pressed and
released with no other key, mouse button, or scroll in between. A bound
tap sends the desktop F24 before the release, so Plasma's launcher stays
shut. The defaults gain a Meta tap for the Steam menu; this replaces
META_DASHBOARD, which only worked in pointer mode.
input/test/keys-test.py runs the relay against fake devices with every
outgoing socket renamed, so it's safe next to the live relay.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
steam/ft-steam menu opens the SteamVR dashboard on Steam's menu, or closes
the dashboard if it's up, without pointer mode: it asks Steam's UI over its
debugging port to show its dashboard overlay (ShowVROverlay, what Steam
calls itself) and focus the Steam frame's left menu (MenuStore.OpenMainMenu).
ft-steam check says whether those calls still exist, and update-check.py
runs it, since a Steam client update can rename them.
The CDP client moves from display-settings/steam_settings.py to
steam/steamui.py, so both use it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Typing on a pass-through keyboard tells the helper "typing". With the
helper not running (SteamVR off), that send raised ConnectionRefusedError
and the relay exited, dropping every grab until systemd restarted it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- desktops.sh uninstall also removes Launch as Standalone's app copies
and the title bar decoration, and Display Settings' uninstall the
profiles' launcher entries; its install writes them back from the
saved profiles (new: ft-layout launchers)
- the README says which settings an uninstall leaves
- the install command is now curl -fsSL https://deejanuz.github.io/
frametop/get.sh | bash (GitHub Pages, from main)
- ft-gaze logs "action manifest ...: ok" instead of "error 0"
Found in a clean install test on the Frame (2026-10-01): uninstall,
then get.sh from the headset into a fresh clone; everything installed
and doctor.sh passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- install.sh no longer offers hand tracking (it's heavy on the CPU and
needs more work); hands/ still builds and installs by hand
- the build container drops python3-opencv and python3-numpy, which
only hand tracking's tools used: Fedora's OpenCV pulls in over a GB
- README: a new opening and feature list, the Use section by topic
(screens, mouse, floating windows, profiles, gaze), and new known
limitations
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- get.sh: curl ... | bash asks for stable (main) or experimental,
clones or updates ~/frametop, and runs install.sh; run it again to
update or switch
- install.sh offers gaze mode (step 8, yes by default) and notes what
to do if an SSH connection drops
- Input Settings' Gaze page says when the gaze service isn't installed
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- frame_sudo (scripts/_env.sh) replaces three copies of sudo_run. On
the Frame it asks in the terminal even when stdin isn't one, which
install.sh's steps never had: saying yes to the Bluetooth fixes or
hand tracking stopped the install with "no terminal for sudo".
From a PC it asks through ssh -t when .env has no password.
- start_with_steamvr: the pointer, power, and gaze services restart
when SteamVR runs (a re-install runs the new code) and are left to
start with it when it doesn't, instead of failing the install.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- floating-windows.md and profiles.md describe what's built, with what
isn't listed as such; the plans, phases, and branch notes are gone
- hands-migration.md is gone: the move is done; its open items are in
hands/README.md's Known issues
- reference.md: Layout & profiles, every action, key combinations,
floating windows, the gaze pointer, and hand tracking as they are
- design.md gets the KWin findings from floating-windows.md
- README, gaze/README, AGENTS, hazards, gaze-controllers, and the
example config catch up with gaze, hands, and the stuck-key fix
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- gaze/tracker/lab/eyes_track.py: nothing used it
- Input Settings: the pointer role backend, whose page was removed
- ft-screens: no log line for every floating-window resize
- hands: ft-hands --help gives the palm-down default (1: off), the
uninstall removes ft-handsctl's link, .frame-job is ignored
- Display Settings: "Save as profile…", not "Save current arrangement"
- two stale comments (ft-pointer's grabprobe, ft-gaze)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With no head pose (the headset off), ft-layout use stopped before
ft-floatd opened the apps. The screens now stay where they are, the
profile's hidden screens still hide, and the apps open.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The gaze checks and calibration in a headset panel (quick, five, full, headset
fit; shared-buffer drawing, click-to-capture), keyboard and mouse gaze clicks
(Meta+J/K, the mouse's buttons alike, mouse moves only while a button is held,
double right/Meta+K tilts a drag), one 55 degree learning limit with a quick
check past it, eye presence for "the headset went on", the gaze probe as a
development tool, and the relay releasing keys the desktop has down that no
keyboard holds.
Conflicts with the profiles: the relay keeps known_action/needs_pointer and
the gaze defaults (Meta+J/K and the float key); Input Settings keeps the
profile actions everywhere and the gaze actions in key combinations only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After a gaze calibration, Meta stayed down in KWin: typing opened the overview,
and clicks on the desktop did other things. A key can be left down there when
its device vanishes with it held (release_held only let go of it on the relay's
own devices; the Z3's keyboard dropped out twice right then) or a release goes
astray. The relay now remembers which keys it told ft-screens went down, and
once a second releases any no device holds (EVIOCGKEY), and the key
combinations forget a Meta or modifier no device holds. A relay that starts
releases the modifiers on the desktop, for keys an earlier one left down.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The panel drew each picture with SetOverlayRaw, a new texture upload every
time: the full calibration's 1024x768 picture flickered on every change, and
in a live test the headset kept showing the last calibration picture after the
panel had drawn the fit check (2026-10-01). Now, as screens/keyboard.cpp does,
it writes into three linear DMA-BUFs SteamVR imported once (the size of the
biggest panel; texture bounds show the part in use) and switches between them.
A show makes the overlay visible once its first picture is in.
The full calibration's and the five-dot check's dots wait for a left click or
Meta+J while you look at the dot, in place of capturing any steady gaze (which
can be a look somewhere else); no capture ring, no 8 s skip, and the check
closes after 2 minutes without a click. The quick check still captures itself.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Check headset fit now runs in the headset panel too ("fitcheck"): a card per
eye (tracked or lost, the tracker's signal, how much of the last 10 s it was
seen) and the hints, from the probe's fitcheck.py, updated at most twice a
second; left click or Meta+J runs its guided check, right click or Meta+K closes
it. So Quick check, Calibrate, and Check headset fit all happen in one place.
The gaze probe moves to the Gaze page's overflow menu as "Gaze probe
(development)", and its app menu entry says it's a development tool. Texts that
sent users to it ("use Calibrate… with Own tracker") point at Calibrate.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A drag begun by right during the left's held-back press (or Meta+K during
Meta+J's) now lasts while either button or key is held, so pressing the right
one again is free: it tilts, as a right press does during any mouse drag. Meta+K
during a keyboard drag (held still into one, or the second Meta+K) tilts while
held: the head turns the panel, and the mouse can too; let go and the head drags
again from there.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The right button's press is held back like the left's: hold it, the pointer
stops where you look, move onto the target, and the right click comes on the
release (held still, it's a real right press). Right during the left's held-back
press now starts a drag where the pointer is, as Meta+J then Meta+K does, in
place of a right click there; letting go of either drops it. So once you've
moved, the left alone only clicks, and the right starts a drag.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
POINTER_GAZE_MOUSE_MOVE=held, the default (the Gaze page's Mouse movement
switch, with why it's on): while the gaze has the pointer, moving the mouse
does nothing; it moves the pointer only while a button is held, as a
correction. A bumped or drifting mouse can't pull the pointer off what you're
looking at, and every mouse move is a correction, so lessons aren't polluted by
mouse moves to somewhere else. With the gaze stale for a second, in a game, or
with the headset off, the mouse moves the pointer as usual. "free" is the old
behaviour.
Pressing the right button while the left one's press is held back right-clicks
where the pointer is instead (correct with the left, then right-click); both
releases are then nothing. Meta+J then Meta+K stays a drag.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Drops the quick check's second step (3629681): the user didn't want it, and it
re-entered itself every tick, re-placing and redrawing the panel, which
flickered in the headset.
POINTER_GAZE_NUDGE_MAX is the one learning limit, now 55 degrees by default:
half of the 109 the Frame shows across. The helper measures the correction
itself (raw gaze to click), not the mouse's path; past the limit it isn't
learned and the helper sends "recheck" for the quick check, for the mouse and
keyboard alike. ft-gazed's own limits (8 and 25) follow the setting.
Test, our tracker's 409 clicks since its Sep 29 calibration: a one-dot check
set from any one of them puts the next 2 minutes' clicks within 15 degrees (99%
within 4.2) and the next 10 minutes' within 25, so it gets well under 55.
The panel's dot is still and its capture ring fills in quarters, so it's drawn
again only a few times per dot.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After the quick check's capture, in gaze mode, the dot stays where it is in the
room (the panel's new "lock") and the gaze has the pointer again. You put the
pointer on the dot as you'd correct a click, with the mouse or Meta+J and the
head. That click clicks nothing: the helper sends it as "calverify", learned
whatever its size. Over POINTER_GAZE_NUDGE_MAX, the capture runs again, up to
3 times. A right click or Meta+K skips it.
POINTER_GAZE_NUDGE_MAX is now the one limit: 15 degrees by default (was 8),
and ft-gazed's own limits (8 for SteamVR's tracker, 25 for ours) follow it. A
keyboard correction past it isn't learned but opens a quick check ("recheck"),
in place of the 30 degree keyboard limit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A Meta+J hold's correction is always meant, but it went through the mouse's
8 degree nudge limit. Live, our tracker was 12 degrees off and every correction
was dropped without a word. Debug output now says when one is.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Steam's eyetracking.txt writes "HMD on" every minute or so with nobody in the
headset, and SteamVR said it was worn for 12 hours straight, so the quick check
opened about 40 times an hour at an empty headset. It now opens when SteamVR's
tracker sees eyes for 3 s after none for 3 s, and closes when they're gone 2 s.
The log reader also drops repeated "HMD on" lines and sub-second offs, so
lessons stop counting as from an older wear every minute.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
remap_volume stopped at the first EVIOCSKEYCODE_V2 that failed, so the
volume entries after it were never tried. On a keyboard where swapping
volume up failed, volume down stayed a real key that gamescope reads,
and one press with nothing focused aborts gamescope and the VR session.
Restoring had the same hole: a failed entry left the later stand-ins in
place after the relay exited.
Now every entry is tried and the first failure is raised at the end.
take_volume already marks the device remapped on that error, so the
relay routes the stand-ins that did swap and restores them on exit.
Checked against a fake keymap with a stubbed fcntl.ioctl: with volume
up's swap failing, volume down now swaps (it stayed real before), the
same for restore, and full swaps, restores and the no-keymap case come
out as before. Found by 0x1f6 in PR #8.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
From 0x1f6's robustness PR: survive a malformed control datagram (adapted
to wrap experimental's textfield command too) and plan a layout when the
rows setting exceeds what the screens fill. The PR's third commit, the
volume-key swap rollback, is left out: rolling back a partial remap leaves
every real volume key exposed to gamescope instead of some, and on restore
it turns already-restored keys back into stand-ins.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ft-layout crashed - and so did arranging the screens - whenever the row
setting in the layout preset is above what the screens fill. plan()
builds a grid with cols = ceil(count / rows) columns; when that leaves
fewer full rows than the setting asked for, the extra rows are empty,
and the per-row height max() over an empty row raised
ValueError: max() arg is an empty sequence
at plan, line 300 (flat) and 312 (curved)
The smallest real case is 4 screens with 3 rows: cols = ceil(4/3) = 2,
the first two rows hold all 4 screens, the third row is empty. Both the
flat wall and the curved preset crashed, so apply, plan and the
auto-arrange at desktop start all failed until the row setting was
lowered. The Rows spinner in Frametop Display Settings (main.qml) allows
any row count up to the screen count, and clamping to the screen count
does not prevent this - 3 rows for 4 screens is within that range and
never fits a full grid - so the crash was reachable from the UI as
shipped.
Reproduced by calling plan() directly with 4 screens and 3 rows: both
kinds raised. Also verified the whole placement matrix (counts 1-10,
rows 1-4) places every screen after the fix.
The fix trims rows to the number of rows the screens actually fill,
rows = ceil(count / cols), after cols is computed. The row count is
used again for stacking (gap times rows - 1, the sum of row heights),
so the stacking matches the trimmed grid: no empty row is ever built,
and a rows setting that can't be honoured degrades to the tightest fit
instead of failing.
(cherry picked from commit f2bdbd97a0)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
One bad datagram on the control socket ended the whole relay. Its
dispatch in handle_control ran unguarded in the main loop, so any
exception propagated out of main() and the process exited. The simplest
trigger is "watch abc": float(words[1]) raises ValueError, and the relay
died with
ValueError: could not convert string to float: 'abc'
at handle_control, via the bare handle_control(now) call in the
select loop
The control socket is an abstract socket bound to @frametop_relay.
Abstract sockets carry no permissions, so every local process can send
to it; nothing authenticates the sender. Dying from a datagram was bad
in three ways:
- Every grab is lost. Physical devices go back to gamescope and SteamVR,
which read them themselves, so the mouse types into Steam and the
desktop at once, and Meta+Shift shortcuts fire on both sides.
- The volume-key takeover is lost. gamescope aborts on a volume key when
no window has keyboard focus (wlr_seat_keyboard_notify_enter asserts on
a null focus surface), which ends the whole VR session - the exact
failure the relay exists to prevent.
- systemd restarts the service, but Type=notify with READY only after
the virtual devices exist makes the restart a visible hiccup, and a
crash loop from repeated bad datagrams would flap SteamVR's input.
Reproduced by running the relay with stubbed uinput devices on a
non-Linux host and sending "watch abc" to the control socket: it exited
on the first datagram after answering "devices" correctly. After the
fix, the same exchange gets a log line ("bad control datagram ...") and
the relay keeps answering.
The fix wraps each datagram's dispatch in try/except inside
handle_control's loop, so one malformed message is logged and skipped
while the rest of the queue is still processed. Parsing of the common
helper commands (vrbtn, vrhello, gazeawake, keyboard) keeps its own
guards; the catch-all is only a backstop for anything the guards miss,
including float() on a non-numeric argument to "watch" and "vrcapture".
Adapted for experimental (from PR #8): the guard also wraps the textfield
command, which experimental added to this dispatch after the PR's base.
(cherry picked from commit ca7e58069b)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On the Frame, a SteamOS update replaces SteamVR, KWin, and gamescope with the
rest of the OS image. scripts/update-check.py, run by doctor.sh and report.sh,
checks what Frametop uses from it: the OpenVR interface versions the installed
programs were built against, the vrcmd --overlays format, the eye tracker's
shared memory layout, host files, services, sockets, and the driver
registration. doctor.sh --mark-good records the package versions once things
work, and later runs say what changed and what to try by hand.
The session no longer stops when mesavars.sh or flatpak.sh is missing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
KWin's PlacementTracker keeps each window's geometry, full screen and
maximized state per layout of the outputs, and puts windows back when a
layout it has seen comes back. A spare output resizes after its window,
so resizing a floating window back to an earlier size (or changing its
scale, or full screen) made the window and its output flip forever, and
floating or docking one window could move others onto or off a spare.
The KWin script now keeps where each window belongs, reports nothing
while KWin changes the outputs, and on screensChanged puts floating
windows back (and the screens' windows when only spares changed),
cancelling KWin's requests before the app sees them. A size asked for is
held for a second against late answers.
With that: a launched app and a profile get their remembered scale back,
scale steps keep the size in pixels, ft-floatd waits for the end of an
edge resize before resizing the output (KWin cancels the resize on any
output change), a window taken over after a restart keeps its app and
panel density, and `ft-float float ID` asks the script for the window's
current place.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- KWin 6.2's scripts have no maximize mode: a window counts as maximized
when it fills its output's maximize area.
- A late event for a window that just closed brought it back into
ft-floatd's table, so a profile "found" it open, floated a window that no
longer existed, and held a slot. The script doesn't report deleted
windows, and ft-floatd ignores events for ids it has seen close.
- Docking a window that came from a screen that's hidden now puts it on the
first screen that shows instead.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A profile is a named layout plus the screens it hides and its apps' windows
(docs/profiles.md). ft-layout save captures them (ft-floatd's "windows",
after the KWin script reports every window as it is now); use opens them:
the screens move, open windows of each app go to their places (on a screen,
maximized or not, or floating), and missing apps start, once and then again
for each window still missing 3 s after the first. Nothing closes.
The desktop starts in FT_PROFILE or default_profile (ft-layout start, from
the session script). Each profile gets a launcher entry (Frametop: NAME, in
SteamVR's Launch a program list) that switches to it or starts the desktop
in it. The relay's profile:NAME action and Input Settings' "Open profile"
entries put one on a key, mouse button, or controller button. Display
Settings' Layout page becomes Layout & profiles: Save as profile, Open
profile, the profile's apps, and Start in profile. Plasma's own session
restore is off in the Frametop session.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ft-screens: "conceal <screen|all>" hides a screen on its own, whatever the
visibility mode or the hotkey says, until "reveal"; "concealed" lists them.
(Not "hide N": older builds read anything starting with "hide" as the hotkey.)
ft-layout keeps it per screen ("hidden" in the layout), applies it when it
arranges the screens, and has hide/show N|all and hidden. Display Settings
gets a Shown switch per screen on the Visibility tab. ft-floatd floats a new
window that opens on a hidden screen, and puts a stray window on a screen
that shows. Profiles (next) use it to show only some screens.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CuriousJ