You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Fixed in main — added src/drivers/sway.py, driven through the swaymsg seat IPC (cursor move/set/press/release, button4/button5 for scroll). Detection now checks SWAYSOCK first, then XDG_CURRENT_DESKTOP, so the X11 fallback warning shouldn't fire on Sway anymore.
One caveat worth knowing: the sway IPC has no query for the cursor position, so Sway records relative motion (like the GNOME driver does). cursor set still warps absolute at replay for start_pos/pos events, but absolute recording isn't possible there. Details in the README and the issue link there.
Not tested by me — if you're on Sway, git pull && ./tools/install.sh and tell me if anything's off: clicks, scroll, or relative drift. Test suites cover the driver logic (mocked IPC) but not a live compositor.
Follow-up: I ran the driver against a live sway (1.12, headless on a private SWAYSOCK, so my own session was untouched) and it works — but the run caught two bugs the mocked tests could not:
swaymsg answers commands with a JSON array ([{success: true}]), not a bare object, so every cursor move, button and scroll was reporting failure.
Its getopt eats negative coordinates: cursor move -30 40 reached sway as cursor move 40.
A third one was in the driver all along: the command string was passed as a single argv entry, so -t get_seats arrived as a literal message type (Unknown message type get_seats) — which silently made seat detection fall back to seat0 and resolution to 1920×1080.
All three are fixed in main and covered by regression tests. Verified against the live compositor: seat seat0, resolution 1024×768 read from get_outputs, cursor move 10 -5 arriving intact, warp + clamp to the last pixel, button press/release, and button4/button5 scroll — every one answered success: true. README status table now lists Sway as author-tested instead of "implemented, not tested".
Activity
feat(drivers): add a Sway driver through the swaymsg seat IPC
JADRT22 commentedon Oct 10, 2026
Fixed in main — added
src/drivers/sway.py, driven through the swaymsg seat IPC (cursor move/set/press/release,button4/button5for scroll). Detection now checksSWAYSOCKfirst, thenXDG_CURRENT_DESKTOP, so the X11 fallback warning shouldn't fire on Sway anymore.One caveat worth knowing: the sway IPC has no query for the cursor position, so Sway records relative motion (like the GNOME driver does).
cursor setstill warps absolute at replay forstart_pos/posevents, but absolute recording isn't possible there. Details in the README and the issue link there.Not tested by me — if you're on Sway,
git pull && ./tools/install.shand tell me if anything's off: clicks, scroll, or relative drift. Test suites cover the driver logic (mocked IPC) but not a live compositor.JADRT22 commentedon Oct 10, 2026
Follow-up: I ran the driver against a live sway (1.12, headless on a private
SWAYSOCK, so my own session was untouched) and it works — but the run caught two bugs the mocked tests could not:[{success: true}]), not a bare object, so every cursor move, button and scroll was reporting failure.getopteats negative coordinates:cursor move -30 40reached sway ascursor move 40.A third one was in the driver all along: the command string was passed as a single argv entry, so
-t get_seatsarrived as a literal message type (Unknown message type get_seats) — which silently made seat detection fall back toseat0and resolution to 1920×1080.All three are fixed in main and covered by regression tests. Verified against the live compositor: seat
seat0, resolution 1024×768 read fromget_outputs,cursor move 10 -5arriving intact, warp + clamp to the last pixel, button press/release, andbutton4/button5scroll — every one answeredsuccess: true. README status table now lists Sway as author-tested instead of "implemented, not tested".