umputun / agterm / 29120682989

98%

push

github

fix: clear the active agent-status glyph on Ctrl-C, not just Escape (#185)

* fix: clear the active glyph on Ctrl-C, not just Escape

the active agent-status glyph cleared only on a physical Escape (keyCode 53).
Ctrl-C dismisses a pending prompt in Claude Code and most TUIs exactly like Esc,
yet arrived as .control+"c" and left the working glyph stranded after an interrupt.

keyDown now computes the flag via isInterruptKeystroke — Escape, or a bare Ctrl-C
(control held, base letter c, no command/option) — and the host-free clear flag is
renamed isEscape -> isInterrupt through AgentStatus.clearedByKeystroke, clearedBy,
the surface closure, and their tests, so the name no longer lies. blocked/completed
already cleared on any key, so only active changes; Ctrl-C to a working agent stops
it, so clearing there is correct.

the Ctrl-C wiring lives in the app target, which host-free tests can't reach, so a
focused e2e (testCtrlCClearsActiveStatus) covers it. rule docs updated.

* fix(review-loop): iteration 1 addressed 5 findings

- [major] [GhosttySurfaceView+Input.swift:39] Ctrl-C was matched only by the
  produced character ("c"), which fails on non-Latin layouts where the physical C
  key yields e.g. Cyrillic "с" — the stale active glyph never cleared there. Match
  the base letter OR the physical c key (keyCode 8), the same layout-independent
  fallback the super+key_c binds use; the char check still covers Dvorak.
- [minor] [GhosttySurfaceView+Input.swift:42] Ctrl-Shift-C cleared active (shift
  makes charactersIgnoringModifiers "C", lowercased to "c"); exclude .shift so only
  a bare Ctrl-C qualifies.
- [test-gap] the classifier's false branches (ordinary key, Cmd-C, Ctrl-Shift-C,
  Ctrl-D) were untested — the e2e would pass even if it returned true for all keys.
  Extract the pure classification into agtermCore (InterruptKeystroke over keyCode +
  a host-free KeyModifiers option-set + base character) and unit-test the full truth
  table; isInterruptKe... (continued)

19 of 19 new or added lines in 2 files covered. (100.0%)

4404 of 4469 relevant lines covered (98.55%)

6727556.37 hits per line