#7: broadcast_mirror_click called hypr::cursor_pos().await?.0 then
.1 -- two hyprctl roundtrips. Replaced with a single
hypr::cursor_pos().await? returning (cx, cy).
#8: mouse_press always spawned the per-button repeat loop, even
when neither Mode::Mirror nor mouse_broadcast was on (a no-op press).
Added an early bail: when both are off, return Ok(()) without
firing the initial click or spawning the repeat task.
#10: refresh_slots re-enumerated the matched vec every 400 ms with
enumerate().map(|(i, c)| ((i + 1) as u32, c)), which shuffled slot
IDs mid-flight whenever one slot briefly hid -- regressing Bug #11.
Reuse the previous slot ID for any client whose wl_address is still
present in the new matched set; only assign fresh IDs (1..=n) to
genuinely new clients; drop disappeared clients.
#14: mouse_repeat_ms_clamps_env_var set ENBOXER_MOUSE_REPEAT_MS
without a serial guard. cargo test runs unit tests in parallel; a
concurrent test touching the same env var would race our reads.
Added a std::sync::Mutex<()> static to serialise the test.
cargo test 99+/0; clippy clean.
0906e86594