Conversation
Easy-Switch mouse-follow-keyboard support (HID++ 0x1814/0x1815) has been fully implemented in the agent since host_switch_targets was added to DeviceConfig, but nothing in the GUI ever let a user configure it — the only way to link a mouse to a keyboard's host-switch channel was hand- editing config.toml, undocumented anywhere. Add a "Follow host switches from" card to the Device tab, shown only for a persistent mouse/trackball with at least one other configured keyboard. Each keyboard gets a plain toggle; checking it adds the active pointing device's config key to that keyboard's host_switch_targets, unchecking removes it. The card reads and writes the other device's config entry rather than the current one, since that's where the field actually lives. No wire/schema change: host_switch_targets already round-trips through config.toml. New Config::host_switch_targets/set_host_switch_targets accessors mirror the existing dpi_presets pair. Reused existing locale keys where possible; the one new string (device.host_switch_targets) is added at the same position in every locale catalog, English text pending translation via Crowdin. Fixes AprilNEA#1145
 |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
host_switch_candidates() gated the active pointing device and its candidate keyboards on DeviceKind alone, so an unsupported pairing (a mouse without ChangeHost, or a keyboard with no armable host-switch control) could still be toggled on with no way for the agent to ever act on it. Adds two feature-derived capabilities: Capabilities::host_switch_target (HID++ 0x1814/0x1815 ChangeHost — required on the device being switched to) and Capabilities::host_switch_source (an armable host-switch-channel control in the 0x1b04 control table — required on the keyboard sending the switch), computed the same way as every other capability flag. Gates host_switch_candidates() on both instead of kind. Capabilities rides the IPC wire, so this bumps PROTOCOL_VERSION 31->32 and regenerates the wire_format golden. Addresses the Greptile P1 finding on AprilNEA#1449.
| @@ -155,6 +167,8 @@ impl Capabilities { | |||
| haptic_feedback: ids.contains(&0x19b0), | |||
| haptic_panel: false, | |||
| dpi_gestures: false, | |||
| host_switch_target: has(&CHANGE_HOST), | |||
There was a problem hiding this comment.
A device is marked as a host-switch target when it exposes either 0x1814 or 0x1815, but the runtime requires 0x1814 (ChangeHost) to perform the switch. A pointing device that exposes only 0x1815 (HostsInfo) will therefore appear in the GUI and can be saved as a target, but every attempted host change will fail as unsupported.
Knowledge Base Used: Device integration stack
| host_switch_source |= HOST_SWITCH_CIDS.contains(&info.cid.0) | ||
| && (info.flags.is_divertable() || info.flags.supports_analytics_key_events()); |
There was a problem hiding this comment.
This probe recognizes host-switch controls only by CID, while the runtime recognizes them by either CID or host-switch task ID. A keyboard that uses task IDs 0xAE–0xB0 with a different CID can be armed by the runtime but receives host_switch_source = false, so the GUI hides a supported keyboard and the user cannot configure the link.
Knowledge Base Used: Device integration stack
1 participant
Summary
crates/openlogi-agent-core/src/watchers/host_switch.rs,crates/openlogi-device/src/session/host_switch.rs), driven byDeviceConfig::host_switch_targets: Vec<String>— but there was no GUI to set that field. The only way to link a mouse to a keyboard's host-switch channel was hand-editingconfig.toml, which isn't documented anywhere. See [Bug]: Mouse does not follow keyboard on Easy-Switch host change #1145.Changes
crates/openlogi-core/src/config.rs:Config::host_switch_targets/set_host_switch_targetsaccessors, mirroring the existingdpi_presetspair.crates/openlogi-desktop/src/state/host_switch.rs(new):AppState::host_switch_candidates()lists other configured, persistent keyboards for the active pointing device, each with whether it's already following;AppState::set_host_switch_follow()adds/removes the active device from a given keyboard'shost_switch_targetsand persists.crates/openlogi-desktop/src/app/detail.rs: a "Follow host switches from" card in the Device tab, shown only when the active device is a persistent mouse/trackball with at least one other configured keyboard. One plainSwitchper keyboard — no new UI chrome, matching the existing settings-row style.crates/openlogi-ui/locales/*.toml: one new key (device.host_switch_targets), added at the same position in every catalog. Non-English catalogs currently carry the English string pending a real translation via Crowdin, per this repo's stated policy for that case.The card reads/writes the other device's (the keyboard's) config entry rather than the currently-selected device's, since that's where
host_switch_targetsactually lives — the relationship is keyboard-initiated.Testing
cargo fmt --all -- --checkcargo clippy --workspace --all-targets -- -D warningscargo test --workspace(all green, including two new tests instate::host_switch::testscovering candidate filtering and the follow/unfollow toggle)cargo test -p openlogi-ui locale(catalog parity) andcargo test -p openlogi-desktop i18n(key resolution)cargo doc --workspace --no-deps --document-private-items(excluding the GPUI crates)openlogi-desktopagainstopenlogi-agent-mockwith theopenlogi-canonical-synthetic-001fixture to confirm the binary starts correctly with these changes present.config.tomlas expected before merging.Fixes #1145