Implemented mouse config by NeoTheFox · Pull Request #347 · feschber/lan-mouse

NeoTheFox · 2025-10-31T16:49:46Z

Add two options:
mouse_mod - sets a linear acceleration for the emulated mouse
invert_scroll - inverts the scrolling direction for the emulated mouse

Hopefully it helps to make the pointer more consistent across multiple devices.

Previously discussed in #339

feschber · 2025-11-01T12:48:16Z

There is one more thing:

Since we are making changes to the configuration, we should think about the names a bit more.
I'd prefer these options to be under a different section and name them a bit differently like this:

# transformations applied to input events received from other devices
[event_post_processing]
mouse_sensitivity = 1.0
invert_scroll = true

This would later enable a similar section for event pre transformations (when being sent to other devices):

# transformations applied to input events sent to other devices
[event_pre_processing]
mouse_sensitivity = 1.0
# linear_acceleration_factor = 1.2
# scroll_sensitivity = 0.3
invert_scroll = false

Maybe you can come up with some better names, but I find mouse_mod a bit indescriptive and would prefer sensitivity here.

feschber · 2025-11-01T12:52:14Z

Maybe input_post_processing would be a better name.

NeoTheFox · 2025-11-01T13:23:52Z

Wouldn't a simple input suffice?

feschber · 2025-11-01T14:36:37Z

Not if we want to add pre processing later on

NeoTheFox · 2025-11-01T14:49:02Z

That's fair. Done and done.

NeoTheFox · 2025-11-01T15:45:58Z

A note on my formatter - my cargo version is
cargo 1.90.0 (840b83a10 2025-07-30) (Arch Linux rust 1:1.90.0-4.1)
If the linter is mandatory I implore you to consider adding a clear .rustfmt.toml to the repository, as my version sees no issues here and forces the problematic line into that format.

feschber · 2025-11-01T16:00:32Z

A note on my formatter - my cargo version is cargo 1.90.0 (840b83a10 2025-07-30) (Arch Linux rust 1:1.90.0-4.1) If the linter is mandatory I implore you to consider adding a clear .rustfmt.toml to the repository, as my version sees no issues here and forces the problematic line into that format.

I have the same version and it did complain about the formatting when running cargo fmt --check.

Would you mind opening a separate PR for the rust-fmt.toml (I'm not against that)

feschber · 2025-11-02T23:32:04Z

Ah one last thing: The input_post_processing section in the config should also be optional, not just the individual settings.

feschber · 2025-11-02T23:43:11Z

Oh (in case you are unaware): The dynamic updates to the sensitivity and scrolling inversion do not work. You are passing the values to the EmulationProxy but not from there to the actual input-emulation.

feschber · 2025-11-02T23:43:56Z

If you don't want to do it, I can fix it myself as well.

NeoTheFox · 2025-11-03T12:51:16Z

I'll probably only be able to get back to this on the weekends, so if it's a quick fix - up to you. I'll do it myself in a few days otherwise

10a5er · 2025-11-14T11:18:09Z

I’ve been testing this intensively: macOS (Apple Silicon) ↔ Hyprland/Arch (Apple Silicon), both directions. In this setup it finally makes lan-mouse fully usable for real work.

I’m seeing one "maybe-issue" and one point that’s still unclear:

1.	Scroll wheel stops sending events

A gaming mouse tethered to macOS originally forwarded scroll events to the Linux client, but at some point it stopped completely. The wheel still works locally on the macOS host. No combination of config changes or fresh builds (selfbuilt and artefact's from here) brought it back.
This might be on me — I experimented with libinput and Karabiner earlier (trying to invert the wheel direction with upstream), so it’s possible I broke something that im not able to backtrack. Mentioning it in case it’s relevant.

2.	Unclear configuration surface

Not an issue — I just haven’t fully mapped out which configuration options are included in this PR.
Specifically:
• Is scroll sensitivity part of this, or was it only an example?
• How does the layering work between host-side and client-side configs?
• Which namespace is actually used now: input-, event-, or a mix (depending on -pre-, -post-)?

i hope this has any use and is not a complete waste of space.

NeoTheFox · 2025-11-30T13:35:50Z

Sorry for the delay, had a lot of other work to do recently

Revoxandco · 2025-12-01T16:42:35Z

Hey, can you explain how it works ? I set the variables in the config file for the host and the client but it doesn't work.

NeoTheFox · 2025-12-01T16:44:55Z

It's the client config that matters for these settings. You don't put them into the client section on the host, you put them into the config file on the client, like so

[input_post_processing]
mouse_sensitivity = 1.5
invert_scroll = false

Revoxandco · 2025-12-01T17:00:35Z

It's the client config that matters for these settings. You don't put them into the client section on the host, you put them into the config file on the client, like so
[input_post_processing]
mouse_sensitivity = 1.5
invert_scroll = false

Just putting invert_scroll = false will do it ? Or disable natural in macos settings too is needed ?

NeoTheFox · 2025-12-01T17:02:39Z

In OSX case, the "invert_scroll" should be set to "true" to disable natural scrolling. No other adjustments are necessary

Revoxandco · 2025-12-01T17:05:53Z

My config is correct I think but it doesn't work.. I am using latest development build that contains your fix

NeoTheFox · 2025-12-01T18:16:01Z

@Revoxandco make sure you're starting the right version and that your config is in the right path, I've been driving it daily since I opened this PR and just changed the "invert_scroll" from true to false, and it worked as intended. You can also try using the cli interface to toggle it. (lan-mouse cli command)

feschber · 2025-12-01T18:46:07Z

My config is correct I think but it doesn't work.. I am using latest development build that contains your fix

@Revoxandco this fix is not in the latest development release yet, unless you're specifically compiling this branch from source.

10a5er · 2025-12-28T15:21:38Z

you think the teased scroll sensitivity is a doable thing for the future?

…test upstream Rebased onto current main which includes: - Fix stuck modifiers (feschber#385) - remote key-up on triggered release (feschber#371) - fix inconsistent mouse capture on macOS (feschber#346) - improve reliability of connections (feschber#349) - macos: fix modifier capture (feschber#342) - macos: fix scroll capture (feschber#350) - and more Conflicts resolved in lan-mouse-ipc/src/lib.rs and src/service.rs (same SaveConfiguration variant conflict as original build). Also re-added #[derive(Clone)] on InputConfig and #[serde(alias = "input")] for config compat.

…gerprints Replaces the bare `description` value in `authorized_fingerprints` with `IncomingPeerConfig { description, natural_scroll, mouse_sensitivity }`. Each row in the receive-side authorization registry now carries its own scroll-direction and motion-sensitivity preferences, keyed by TLS certificate fingerprint — the only stable cross-session identity for an inbound peer. Backwards-compat: legacy configs that store a bare string per fingerprint deserialize via a custom `Deserialize` impl on `IncomingPeerConfig` that accepts either shape (untagged enum) and fills missing fields with defaults (`natural_scroll: false`, `mouse_sensitivity: 1.0`). No migration tooling needed; existing config.toml files keep working. Wires the schema through: - lan-mouse-ipc: new `IncomingPeerConfig` type; `AuthorizedUpdated(HashMap<String, IncomingPeerConfig>)`; `SetIncomingPeerNaturalScroll(fingerprint, bool)` and `SetIncomingPeerSensitivity(fingerprint, f64)` requests. - src/config.rs + src/listen.rs + src/service.rs: type updates through the `Arc<RwLock<HashMap<...>>>` shared with the DTLS listener, plus per-fingerprint mutators in Service. - lan-mouse-gtk: `KeyObject` carries the new properties; window re-emits them on `AuthorizedUpdated`. Receive-side application of the new settings (looking up by fingerprint at handle creation, pushing to InputEmulation) and the expandable per-row GTK controls land in follow-up commits to keep the diff focused. The deprecated global `SetNaturalScroll` / `NaturalScroll` IPC variants stay in place until the global UI toggle is removed in the cleanup commit. `FrontendRequest` loses its `Eq`/`PartialEq` derives because `f64` isn't `Eq`. No callers were comparing requests for equality. Sensitivity-multiplier algorithm pattern in InputEmulation::consume (applied in a follow-up commit) is borrowed from feschber#347 by Raidon Chrome / NeoTheFox, where the global form was originally proposed. Co-Authored-By: Raidon Chrome <soniczerops@gmail.com>

…gerprint Centralizes scroll-direction inversion and motion-sensitivity scaling into a single transform site in `InputEmulation::consume`, replacing the per-backend `set_natural_scroll(bool)` plumbing introduced earlier in this PR. The platform emulation backends (macos, wlroots, libei, x11, windows, xdg_desktop_portal) go back to platform- mechanics-only — no scroll-sign awareness, no shadow per-backend state, no duplicated transform code. The active settings are stored per-`EmulationHandle` on `InputEmulation` itself and applied by mutating the event before it reaches the backend's `consume()`. The match-arm pattern (multiplier on `Motion`, sign-flip on `Axis` / `AxisDiscrete120`) follows the shape used in feschber#347 by Raidon Chrome, ported to per-handle. Plumbing on the receive side: - `lan-mouse-ipc` exposes `IncomingPeerConfig` (commit 1). - `src/emulation.rs::ListenTask` maintains a small cache of `addr → fingerprint` (populated from `ListenEvent::Accept`) plus the latest `incoming_peers` snapshot pushed by Service. - On Accept, ListenTask resolves the right `ReceivePostProcessing` for the addr and forwards it via `ProxyRequest::SetPostProcessing` so EmulationTask can attach it to the handle the moment one is assigned. - On `EmulationRequest::SetIncomingPeers` (Service push triggered by user UI changes / authorize / remove / config reload), ListenTask re-resolves every known addr and re-publishes to keep currently-active sessions in sync. - `EmulationTask` caches `addr → ReceivePostProcessing` so a backend respawn (CGEventTap timeout, portal session restart) re-applies the right values to all known handles. `input-emulation` stays decoupled from `lan-mouse-ipc`; the conversion from `IncomingPeerConfig` → `ReceivePostProcessing` happens at the boundary in `src/emulation.rs` so the lower crate doesn't grow IPC dependencies. The deprecated global `FrontendRequest::SetNaturalScroll` / `FrontendEvent::NaturalScroll` IPC variants stay round-tripping through Service (persist + echo only, no emulation effect) so the existing GUI toggle continues to render its state until the cleanup commit removes the global UI surface. Co-Authored-By: Raidon Chrome <soniczerops@gmail.com>

…gerprints Replaces the bare `description` value in `authorized_fingerprints` with `IncomingPeerConfig { description, natural_scroll, mouse_sensitivity }`. Each row in the receive-side authorization registry now carries its own scroll-direction and motion-sensitivity preferences, keyed by TLS certificate fingerprint — the only stable cross-session identity for an inbound peer. Backwards-compat: legacy configs that store a bare string per fingerprint deserialize via a custom `Deserialize` impl on `IncomingPeerConfig` that accepts either shape (untagged enum) and fills missing fields with defaults (`natural_scroll: false`, `mouse_sensitivity: 1.0`). No migration tooling needed; existing config.toml files keep working. Wires the schema through: - lan-mouse-ipc: new `IncomingPeerConfig` type; `AuthorizedUpdated(HashMap<String, IncomingPeerConfig>)`; `SetIncomingPeerNaturalScroll(fingerprint, bool)` and `SetIncomingPeerSensitivity(fingerprint, f64)` requests. - src/config.rs + src/listen.rs + src/service.rs: type updates through the `Arc<RwLock<HashMap<...>>>` shared with the DTLS listener, plus per-fingerprint mutators in Service. - lan-mouse-gtk: `KeyObject` carries the new properties; window re-emits them on `AuthorizedUpdated`. Receive-side application of the new settings (looking up by fingerprint at handle creation, pushing to InputEmulation) and the expandable per-row GTK controls land in follow-up commits to keep the diff focused. The deprecated global `SetNaturalScroll` / `NaturalScroll` IPC variants stay in place until the global UI toggle is removed in the cleanup commit. `FrontendRequest` loses its `Eq`/`PartialEq` derives because `f64` isn't `Eq`. No callers were comparing requests for equality. Sensitivity-multiplier algorithm pattern in InputEmulation::consume (applied in a follow-up commit) is borrowed from feschber#347 by Raidon Chrome / NeoTheFox, where the global form was originally proposed. Co-Authored-By: Raidon Chrome <soniczerops@gmail.com>

…gerprint Centralizes scroll-direction inversion and motion-sensitivity scaling into a single transform site in `InputEmulation::consume`, replacing the per-backend `set_natural_scroll(bool)` plumbing introduced earlier in this PR. The platform emulation backends (macos, wlroots, libei, x11, windows, xdg_desktop_portal) go back to platform- mechanics-only — no scroll-sign awareness, no shadow per-backend state, no duplicated transform code. The active settings are stored per-`EmulationHandle` on `InputEmulation` itself and applied by mutating the event before it reaches the backend's `consume()`. The match-arm pattern (multiplier on `Motion`, sign-flip on `Axis` / `AxisDiscrete120`) follows the shape used in feschber#347 by Raidon Chrome, ported to per-handle. Plumbing on the receive side: - `lan-mouse-ipc` exposes `IncomingPeerConfig` (commit 1). - `src/emulation.rs::ListenTask` maintains a small cache of `addr → fingerprint` (populated from `ListenEvent::Accept`) plus the latest `incoming_peers` snapshot pushed by Service. - On Accept, ListenTask resolves the right `ReceivePostProcessing` for the addr and forwards it via `ProxyRequest::SetPostProcessing` so EmulationTask can attach it to the handle the moment one is assigned. - On `EmulationRequest::SetIncomingPeers` (Service push triggered by user UI changes / authorize / remove / config reload), ListenTask re-resolves every known addr and re-publishes to keep currently-active sessions in sync. - `EmulationTask` caches `addr → ReceivePostProcessing` so a backend respawn (CGEventTap timeout, portal session restart) re-applies the right values to all known handles. `input-emulation` stays decoupled from `lan-mouse-ipc`; the conversion from `IncomingPeerConfig` → `ReceivePostProcessing` happens at the boundary in `src/emulation.rs` so the lower crate doesn't grow IPC dependencies. The deprecated global `FrontendRequest::SetNaturalScroll` / `FrontendEvent::NaturalScroll` IPC variants stay round-tripping through Service (persist + echo only, no emulation effect) so the existing GUI toggle continues to render its state until the cleanup commit removes the global UI surface. Co-Authored-By: Raidon Chrome <soniczerops@gmail.com>

…gerprints Replaces the bare `description` value in `authorized_fingerprints` with `IncomingPeerConfig { description, natural_scroll, mouse_sensitivity }`. Each row in the receive-side authorization registry now carries its own scroll-direction and motion-sensitivity preferences, keyed by TLS certificate fingerprint — the only stable cross-session identity for an inbound peer. Backwards-compat: legacy configs that store a bare string per fingerprint deserialize via a custom `Deserialize` impl on `IncomingPeerConfig` that accepts either shape (untagged enum) and fills missing fields with defaults (`natural_scroll: false`, `mouse_sensitivity: 1.0`). No migration tooling needed; existing config.toml files keep working. Wires the schema through: - lan-mouse-ipc: new `IncomingPeerConfig` type; `AuthorizedUpdated(HashMap<String, IncomingPeerConfig>)`; `SetIncomingPeerNaturalScroll(fingerprint, bool)` and `SetIncomingPeerSensitivity(fingerprint, f64)` requests. - src/config.rs + src/listen.rs + src/service.rs: type updates through the `Arc<RwLock<HashMap<...>>>` shared with the DTLS listener, plus per-fingerprint mutators in Service. - lan-mouse-gtk: `KeyObject` carries the new properties; window re-emits them on `AuthorizedUpdated`. Receive-side application of the new settings (looking up by fingerprint at handle creation, pushing to InputEmulation) and the expandable per-row GTK controls land in follow-up commits to keep the diff focused. The deprecated global `SetNaturalScroll` / `NaturalScroll` IPC variants stay in place until the global UI toggle is removed in the cleanup commit. `FrontendRequest` loses its `Eq`/`PartialEq` derives because `f64` isn't `Eq`. No callers were comparing requests for equality. Sensitivity-multiplier algorithm pattern in InputEmulation::consume (applied in a follow-up commit) is borrowed from feschber#347 by Raidon Chrome / NeoTheFox, where the global form was originally proposed. Co-Authored-By: Raidon Chrome <soniczerops@gmail.com>

…gerprint Centralizes scroll-direction inversion and motion-sensitivity scaling into a single transform site in `InputEmulation::consume`, replacing the per-backend `set_natural_scroll(bool)` plumbing introduced earlier in this PR. The platform emulation backends (macos, wlroots, libei, x11, windows, xdg_desktop_portal) go back to platform- mechanics-only — no scroll-sign awareness, no shadow per-backend state, no duplicated transform code. The active settings are stored per-`EmulationHandle` on `InputEmulation` itself and applied by mutating the event before it reaches the backend's `consume()`. The match-arm pattern (multiplier on `Motion`, sign-flip on `Axis` / `AxisDiscrete120`) follows the shape used in feschber#347 by Raidon Chrome, ported to per-handle. Plumbing on the receive side: - `lan-mouse-ipc` exposes `IncomingPeerConfig` (commit 1). - `src/emulation.rs::ListenTask` maintains a small cache of `addr → fingerprint` (populated from `ListenEvent::Accept`) plus the latest `incoming_peers` snapshot pushed by Service. - On Accept, ListenTask resolves the right `ReceivePostProcessing` for the addr and forwards it via `ProxyRequest::SetPostProcessing` so EmulationTask can attach it to the handle the moment one is assigned. - On `EmulationRequest::SetIncomingPeers` (Service push triggered by user UI changes / authorize / remove / config reload), ListenTask re-resolves every known addr and re-publishes to keep currently-active sessions in sync. - `EmulationTask` caches `addr → ReceivePostProcessing` so a backend respawn (CGEventTap timeout, portal session restart) re-applies the right values to all known handles. `input-emulation` stays decoupled from `lan-mouse-ipc`; the conversion from `IncomingPeerConfig` → `ReceivePostProcessing` happens at the boundary in `src/emulation.rs` so the lower crate doesn't grow IPC dependencies. The deprecated global `FrontendRequest::SetNaturalScroll` / `FrontendEvent::NaturalScroll` IPC variants stay round-tripping through Service (persist + echo only, no emulation effect) so the existing GUI toggle continues to render its state until the cleanup commit removes the global UI surface. Co-Authored-By: Raidon Chrome <soniczerops@gmail.com>