Implement Input-Leap inspired improvements: UDP heartbeats, screen en…

…try/exit signaling, and protocol versioning

Bump version to v1.2.0

fix: v1.1.2 — startup panic from double logger init (closes #2)

egui_logger and tracing-subscriber both tried to install the global log
backend. Only one can exist — the second .init() call panicked with
"Unable to install global subscriber: SetLoggerError(())".

Fix:
- Remove tracing_subscriber::fmt().init() from main()
- Enable "log" feature on tracing crate so tracing::info!/warn!/etc.
  bridge automatically to the log crate that egui_logger owns
- Remove tracing-subscriber dependency (was only used for the init call)

All tracing macros now reach the in-app Logs tab via egui_logger.
No behavior change — log output is identical, panic is gone.

Fixes: #2

Co-Authored-By: Claude Sonnet 4.6 <[email protected]>

fix: v1.1.1 — deb conflict with input-sync + Illegal instruction (closes

 #1)

- Cargo.toml: add Conflicts/Replaces: input-sync to deb metadata so dpkg
  automatically removes the old Tauri-based package on upgrade; fixes
  "trying to overwrite .../inputsync.png which is also in package input-sync"
- Cargo.toml: bump version to 1.1.1
- CHANGELOG.md: document both fixes (deb conflict + AVX2 crash)

Fixes: #1

Co-Authored-By: Claude Sonnet 4.6 <[email protected]>

ci: rename release assets to include platform suffix (linux-x86_64 / …

…windows-x86_64)

- Linux: inputsync -> inputsync-linux-x86_64
- Windows: inputsync.exe -> inputsync-windows-x86_64.exe
- SHA256SUMS files updated to reference new names
- Release title updated to include 'Native Rust KVM Switch' tagline
- Release body expanded with full v1.1.0 changelog, improved downloads table,
  and updated verification instructions referencing versioned filenames

Co-Authored-By: Claude Sonnet 4.6 <[email protected]>

Release v1.0.2 — server restart fix + offline WebView2 bundling

Release v1.0.1 — Windows WebView2 fixes + Linux Wayland fixes

fix: code audit + v1.0.0 release prep

Code audit fixes (13 issues):
- Fix #1: InputSimulator moved to dedicated std::thread — enigo now
  runs off the tokio executor, actual input simulation works correctly
- Fix #2: Relative mouse movement — server sends deltas, client uses
  Coordinate::Rel; eliminates screen-resolution mismatch between machines
- Fix #3: ScrollLock toggles input forwarding on server (switch mechanism);
  UI toggle button shares the same Arc<AtomicBool>
- Fix #4: Counter-window resync in UDP receiver — tries up to 64 counters
  ahead before dropping; survives packet loss without permanent desync
- Fix #5: client_count tracked via Arc<AtomicUsize> shared with server task;
  increments/decrements correctly on connect/disconnect
- Fix #6: Cleaner server shutdown — capture handle dropped before server
  handle; TCP keep-alive loop exits on EOF or error
- Fix #7: Bounded mpsc channel (512 cap) + try_send in capture thread —
  drops events under load instead of growing unbounded
- Fix #8: Session code normalized to uppercase in server handshake and
  validate_code_format; case-insensitive matching
- Fix #10: Extended key mappings — punctuation (-, =, [, ]) now simulate
  correctly; Unknown keys filtered silently instead of erroring
- Fix #12: panic = "unwind" in release, strip = false — preserves stack
  traces for crash diagnosis
- Fix #13: Simulator init errors propagated via status channel to UI

Version bump: 0.1.0 → 1.0.0

Docs: LICENSE (MIT), README.md, CHANGELOG.md, CONTRIBUTING.md, SECURITY.md

Co-Authored-By: Claude Sonnet 4.6 <[email protected]>