szn

Bug 380 — Raw history-length subtraction at six mouse-path sites

380. The #293 pattern reintroduced in the scroll/copy-mode handlers

File: src/server/server.zig:1602, 1616, 1643, 1742, 1746, 1791 Severity: MEDIUM Status: ✅ FIXED

Fix (applied)

All six raw history.items.len - history_start subtractions in the mouse/copy-mode handlers replaced with the underflow-guarded grid.historyLen() (grep now reports zero raw sites in server.zig).

Test: mouse scroll uses guarded historyLen even with corrupt history_start — bug #380 (corrupts the ring-buffer invariant, then drives a wheel-up event through processInput — panicked with the old code).

Problem

const hist_len: u32 = @intCast(press_pane.screen.grid.history.items.len - press_pane.screen.grid.history_start);

Grid.historyLen() (grid.zig:117-120) guards history_start > items.len (added by bug #293 after a real underflow panic). These six sites — added around the copy-mode mouse handling — repeat the unguarded subtraction. Safe only while scrollUp’s compaction invariant holds; latent panic otherwise.

Fix (proposed)

Replace all six with @intCast(grid.historyLen()) and add a debug assert in historyLen so violations surface immediately.