D337: fold aim modes (N64 via PD integration default, Centred (PC) opt-in); Q/L + Toggle aim routing; per-device mouse integration; scripted-mouse aim harness

- Input.AimMode 0=N64 (D332 PD-integration, default) / 1=Centred (PC); legacy
  AimStyle/PdMouseAim migrated; hidden Input.AimLegacyGepd fallback.
- Mouse aim routing follows RMB/Aim, Q/L (the game's second aim button), and
  insightaimmode (Toggle aim, pad): removes the accidental Q 'FPS aim'.
- The PD game hook only feeds the mouse while the mouse is the active aim
  device, so pad/keyboard stick aim is not swallowed.
- Harness: GE_INPUTSCRIPT MDX/MDY/RHOLD/LHOLD/RREL + GE_AIMLOG; measured old
  vs new: same screen range and edge-scroll, jitter equal headlessly.
Golden 3/3; Win + Linux builds.

docs/dev/findings-index.csv

@@ -214,3 +214,4 @@ D333,OPEN,"#104 (v0.3.0, Linux):","IMPLEMENTED (port/src/input.c + optionsoverla
 D334,?,"Native widescreen, world side (WIDESCREEN-FOV-PLAN Option B Phase 2):",IMPLEMENTED (Rule-2 approved 2026-09-23); 4:3 identity (golden 3/3); native 16:9 centre matches the 4:3 frame (mean diff 4.3 vs 12.4 stretched); 21:9 clean on Bunker/Facility/Dam.
 D335,?,"Native widescreen, HUD + front end (WIDESCREEN-FOV-PLAN Phase 3):","IMPLEMENTED (Rule-2 approved 2026-09-23); 4:3 byte-identical (no emission at ≤4:3); verified at 16:9/21:9 on HUD, watch, title card, legal/file-select/gun-barrel."
 D336,FIXED,"Frigate (and any level) ""all polygons break except the gun"" after turning:","FIXED (src/game/bondhead.c #ifdef PORT, ABI/layout); headless repro (Runway → exit → Frigate → turn, D294 2× room pool) clean, D156 0 fires; golden 3/3; Win + Linux builds."
+D337,OPEN,Aim modes folded (user decision):,IMPLEMENTED (port/src/input.c + optionsoverlay.c); headless-measured with new scripted-mouse harness; golden 3/3; Win + Linux builds.

docs/dev/findings.md

@@ -622,6 +622,7 @@ covers D24–D69; the log continues in §H (D32 procedure, D70–D121).
 | D334 | **Native widescreen, world side (WIDESCREEN-FOV-PLAN Option B Phase 2): the world is projected at the real window aspect (Hor+) by generalising GE's own N64 16:9 mode; `Video.NativeWidescreen` (default on).** — full `## D334` entry at file tail | IMPLEMENTED (Rule-2 approved 2026-09-23); 4:3 identity (golden 3/3); native 16:9 centre matches the 4:3 frame (mean diff 4.3 vs 12.4 stretched); 21:9 clean on Bunker/Facility/Dam. HUD still stretched → Phase 3. |
 | D335 | **Native widescreen, HUD + front end (WIDESCREEN-FOV-PLAN Phase 3): HUD elements anchored left/right/centre and the front end pillarboxed via the fast3d EXT aspect modes (PD pattern), `port/include/hudaspect.h`.** — full `## D335` entry at file tail | IMPLEMENTED (Rule-2 approved 2026-09-23); 4:3 byte-identical (no emission at ≤4:3); verified at 16:9/21:9 on HUD, watch, title card, legal/file-select/gun-barrel. Menu pointer mapped into the pillarbox. One cosmetic leak → GRAPHICS-BACKLOG D335a. |
 | D336 | **Frigate (and any level) "all polygons break except the gun" after turning: Bond's head-bob anim read two never-written placeholder fields (`field_5C0`, `animFlipFlag`) that on N64 are raw-offset aliases into the inline `player.model` (D100 follow-up; A1 class). Likely the D311 upstream anim-poison source.** — full `## D336` entry at file tail | FIXED (src/game/bondhead.c `#ifdef PORT`, ABI/layout); headless repro (Runway → exit → Frigate → turn, D294 2× room pool) clean, D156 0 fires; golden 3/3; Win + Linux builds. **User-confirmed 2026-09-24.** |
+| D337 | **Aim modes folded (user decision): N64 (default) = the N64 aim model with the mouse fed through the game's integrator at PD's damp (D332 made default); CENTRED (PC) opt-in; Q/L and Toggle-aim now route mouse aim like RMB; pad/keyboard aim no longer swallowed by the mouse integration.** — full `## D337` entry at file tail | IMPLEMENTED (port/src/input.c + optionsoverlay.c); headless-measured with new scripted-mouse harness; golden 3/3; Win + Linux builds. Open: N64-range calibration decision. |
 
 Phase 2 replaced the Phase-1 demo loop with the real `mainproc()` on real OS
 threads, compiled GE's real `src/sched.c`, and brought in PD's fast3d software
@@ -13282,3 +13283,18 @@ Implemented as `Input.PdMouseAim` (**default off**, because it changes aim feel)
 
 **Verified:** the same Runway → Frigate → turn repro at the 2× pool renders correctly throughout, D156 0 fires; bunker1 golden 3/3; Windows + Linux builds clean. **D311 link:** D311 (upstream anim-poison source, open) named `bheadAdjustAnimation` as the one path where a corrupted value reaches `model->speed`. This is very likely that source; re-check D311's Caverns signature in play before closing it. **D294 note:** its "linear 3.0× variant corrupted model anim state on the menu-driven path" is this same bug (a bigger pool moved the player struct onto dirtier memory), not a pool-size problem. The 2× cap can be revisited. **User-confirmed (2026-09-24, native Windows, shipping settings incl. DD/LOD 400):** "Looks good when doing Facility > Frigate and in same session Runway > Frigate." Session log: 14 level loads (Bunker 1, Facility, Silo, Dam, Runway, Frigate…), **0 D156 fires** (every prior session had ≥1), 0 texture pool-full.
 
+## D337 — Aim modes folded; Q/L + Toggle routing; per-device mouse integration; headless aim measurements (2026-09-24)
+
+**User decision:** one N64 aim mode (the PD-integration of D332, made default) plus the opt-in "Centred (PC)" mode from #104. The old GEPD overwrite survives only as the hidden `Input.AimLegacyGepd=1` for one release. **Setting:** `Input.AimMode` 0 = N64, 1 = Centred (PC). Legacy `Input.AimStyle` 2 migrates to 1; `AimStyle` 0/1 and `Input.PdMouseAim` map to N64. F10 row "Aim style". `GE_PDMOUSEAIM=0/1` still forces the path at launch.
+
+**#104 origin, corrected:** the reporter was describing the port's existing Q behaviour, not requesting a new design. In GE's 1.1/1.2 control schemes the aim buttons are `L_TRIG | R_TRIG` (bondview2.c `aimButtons`), so Q ("LeanLeft", which sends L; the bind name is a misnomer) is a second aim button. The port's mouse routing only recognised RMB/Shift, so Q put the game in aim mode while the mouse kept turning the camera: an accidental "FPS aim". **Fix:** mouse aim routing is now "RMB/Aim bind **or** Q/L **or** the game's `insightaimmode`". That also covers the Aim Control "Toggle" option and gamepad aim, which the held-button check never did. The R emit stays tied to the physical button, so Toggle still sees edges. Verified: `LHOLD` + scripted mouse moves the drawn crosshair N64-style (160 → 280 px) with the camera turning only via edge-scroll.
+
+**Device-aware integration (would have been a regression when made default):** the D332 game hook replaced the stick's turn whenever the mode was on and the mouse was grabbed, freezing a gamepad or keyboard user's crosshair. PD chooses its damp per input device; the port now tracks the last-used aim device (mouse motion sets it; pad stick deflection or keyboard turn keys clear it), and the hook feeds the mouse integration only while the mouse is active.
+
+**Headless measurement harness (new):** `GE_INPUTSCRIPT` tokens `MDX<n>`/`MDY<n>` (sustained scripted mouse px per poll, treated as a grabbed mouse), `RHOLD`/`LHOLD`/`RREL` (sustained aim hold), and `GE_AIMLOG=1` (per-poll: aim state, active device, the game's drawn crosshair x `crosshair_angle.f[0]`, `crosshair_x/y_pos`, camera angles). **Results (Bunker 1, 640×480, 3 px/poll sweep):**
+- *Screen range and edge-scroll, old GEPD vs new:* the drawn offset is `crosshair_x_pos × (1 − damp) × W/2` (gunfire.c), so GEPD's raw ±5.16 at damp 0.8 and PD's ±1.0 at damp 0.01 are the same screen edge. Sweep speed on screen 1.00 vs 1.21 px/poll; edge-scroll starts at 0.745 vs 0.719 of half-width; edge-scroll rate identical (1.478°/poll, sd 0). Equivalent.
+- *Jitter:* drawn-crosshair cv 0.18 (GEPD) vs 0.16 (new), one stall and one jump each in 75 polls. The GEPD step jitter needs irregular ticks per poll (real vsync pacing), which this headless run rarely produces, so the jitter benefit rests on D332's analysis plus hands-on feel, not this test.
+- **Accuracy finding (open, user decision):** N64 stick aim turns the crosshair by `stick × 0.65/80` per tick, so at full deflection the drawn crosshair settles at **65% of the half-width regardless of weapon damp**, and the camera starts turning past stick 60/80 (≈49% of half-width). Both mouse paths let the crosshair travel to about 100% of the half-width, with edge-scroll from about 72%. A faithful calibration would map the mouse range to ±65% and start the camera turn at ≈49%.
+
+**Verified:** golden 3/3; Windows + Linux builds clean. **Owed:** user feel pass (N64 default vs Centred; Q now aims); decision on N64-range calibration.
+