D207/D241: shot-sound handle probe; live captures refute drop/starvation

TEMP GE_D207S (gunfire.c, #ifdef PORT, read-only) logs every player
shot-sound decision (H1/H2/SKIP). Four user captures incl. a 7-weapon
stress test at 60 fps: 0 SKIP, 0 voice drops, audio production at rate.
The one captured drop-out ran under GE_AUDIOTRACE, whose audio-path file
I/O is a likely cause. Still open as intermittent.

docs/dev/GE-ENV-PROBES.md

@@ -68,6 +68,7 @@ Two classes:
 | `GE_D245_FIXEDSHIFT` | `src/game/sky.c:946` (`#ifdef PORT`) | D245 test override: pins the water tc shift `k` to the given integer (0..SKY_TC_MAX_SHIFT) instead of the computed per-vertex value, to A/B the allowShift=TRUE behavior. Off by default (unset = current behavior). | **live, test-only** (D245 OPEN) |
 | `GE_D252` | `port/fast3d/gfx_pc.cpp:733` (`import_texture`, cached `s_d252`) | D252 rainbow-particle diag: one-shot per (address,size) import log of the tile state (tmem/fmt/siz/addr/size/line) for small textures so a suspect particle tile can be checked against its source bytes by hand. Remove once D252 is resolved. | **live** (D252 OPEN — non-reproducible) |
 | `GE_D252POOL` | `src/game/image.c` (`texLoad`), `src/game/initmttex.c`, `port/src/gimgfixup.c` | **TEMP** (M-201): one line per new texture load (pool main/other, free bytes) + `POOL-FULL` events + stage pool range; `D252SYNC: N slots re-synced, M changed` per stage load (the D252 fix at work: M>0 from the 2nd stage on). Load-time only, no fps cost. | **TEMP** (D252 fixed; remove after release) |
+| `GE_D207S` | `src/game/gunfire.c` (player shot-sound block, `#ifdef PORT`) | **TEMP** (M-201): one line per player shot-sound decision: `D207S t=<g_GlobalTimer> hand item rate -> H1|H2|SKIP` (SKIP = both per-hand sound handles still held). Read-only, uncapped, cheap; use instead of `GE_AUDIOTRACE` for live captures (that probe does file I/O on the audio path). | **TEMP** (D207/D241) |
 | `GE_D306C="f0 f1 [wmax]"` | `port/fast3d/gfx_pc.cpp` (after `zf_note_emit`) | **TEMP** (M-201): depth-state census of every emitted triangle in frames f0..f1 (GL depth mode, geometry_mode, other_mode_l, tile-0 texture, NDC centroid, z/w range, min w); optional `wmax` keeps near geometry only. Cap 60000 lines. | **TEMP** (D306 closed; remove post-release) |
 | `GE_INTROCAM=<n>` | `src/game/bondview_r.c` (`#ifdef PORT`, intro camera pick) | **TEMP/test-only**: force the level intro's first camera index (Dam truck shot = 5). The PRNG draw is still consumed. Logs `D306: intro camera index` when set (or with `GE_D306C`). | **TEMP** (useful repro tool) |
 | `GE_D235` | `src/boss.c` (per-stage `while (!done)` block, after `memaReset`; `#ifdef PORT`) | D235 evidence probe: logs how many GL-texture-cache entries the per-stage `videoResetTextureCache()` call purges at each level load. A cold boot into a level should log 0; a death/restart (or any transition from a level that uploaded textures) logs >0 — direct evidence of the stale-entry mechanism. The clear itself is **not** gated on this var (it's the fix, see below); the var only controls the log line. | **live** (D235) |

docs/dev/findings.md

@@ -11377,6 +11377,18 @@ the Surface 1 spawn capture is unchanged across the strip.
 - **Harness limits:** `gunRequestHandWeaponChange` / the level-start equip call did not switch weapons mid-level from the harness, so the automatic-weapon case (AK47/Klobb) is untested headless.
 - **Next:** capture the user's real scenario with `GE_AUDIOTRACE=1 GE_D322=1 GE_D204=1` (`scratch/playtest_audio.bat`), noting level, weapon and time of each drop-out; any `[D207-DROP]` line names the layer. If none appear, the "drop-out" is the faithful stop-then-restart of the per-hand shot sound, not a lost voice.
 
+**2026-09-26 (M-201, cont.): live captures. Not reproducible with lightweight logging; the one captured drop-out ran with the heavy GE_AUDIOTRACE probe.**
+
+- **Run 1 (user, Bunker, AK47, alarm active, `GE_AUDIOTRACE`+D322+D204):** drop-out heard. `[D207-DROP]`=0, pools healthy (sfx <= 6/8, pv free >= 11/24). The per-sound `sndPlaySfx` trace is capped (~630 events) and stopped before the AK47 section, so it could not show the moment. `GE_AUDIOTRACE` does an uncached `getenv` plus a file write from the audio path on nearly every sound event (D250/D302 class), so this drop-out may be probe-induced.
+- **New TEMP `GE_D207S`** (`src/game/gunfire.c`, `#ifdef PORT`, read-only): logs every player shot-sound decision (`H1`/`H2`/`SKIP`, where SKIP = both per-hand handles still held because the stop is async) with `g_GlobalTimer`. Uncapped, cheap.
+- **Run 2 (FpsCap 30):** 315 decisions, 0 SKIP, no drop-outs heard. AK47 gaps 5–7 ticks.
+- **Run 3 (FpsCap 60):** 408 decisions, 0 SKIP, no drop-outs heard. AK47 exactly every 5 ticks (`SoundTriggerRate` 4 in the BUGFIX_R0 build).
+- **Run 4 (FpsCap 60, stress test, 7 weapons incl. AK47/ZMG/D5K/RC-P90/Klobb):** 2073 decisions, **0 SKIP**; SFX cap frequently 8/8 but `drop=0`; evtq max 21/64; D204 `rt` 0.999–1.000 throughout, audio queue never empty. User: "seemed okay".
+- **Refuted:** voice-cap drops, physical-pool exhaustion, per-hand handle starvation (the 60 fps shot-rate divergence is real, 5 vs 6 ticks for the AK47, but never starves the handles), output underrun during these runs.
+- **Open:** the drop-outs the user heard in the untraced 21:9 session remain unexplained and intermittent. Standing capture: `scratch/playtest_audio.bat` (`GE_D207S` + D322 + D204, no audio-thread file I/O); press F12 on a drop-out.
+
+**Status: OPEN (not reproducible) — drop paths and handle starvation refuted by measurement; awaiting a natural occurrence under lightweight logging.**
+
 ## D242 — Player-death sequence UX regression vs N64 original: PC requires an LMB click to proceed to the mission-failed/file-select screen; user recalls N64 auto-plays…
 
 **Player-death sequence UX regression vs N64 original: PC requires an LMB click to proceed to the mission-failed/file-select screen; user recalls N64 auto-plays 3 death-camera angles then transitions with no input needed (M-106, user QA report).** **The 3-angle auto-replay and its no-input auto-transition both exist, unmodified, in the decomp — confirmed, not a design difference removed for the port.** `bondviewSetCameraMode(CAMERAMODE_DEATH_CAM_SP)` (`src/game/bondview2.c:901-1010`) calls `pickDeathCameraAngles()` (`bondview2.c:641-782`, picks a fresh random valid camera angle each call) to stage each replay; when it returns 0 — no further valid angle found — the code calls `bossRunTitleStage()` directly, and the source comment at `bondview2.c:1006-1007` literally reads *"...or when the three death replays are finished."* The 3-cycle count is driven by `bondviewAdvanceCameraMode()` (`bondview2.c:1041-1078`): `CAMERAMODE_DEATH_CAM_SP` → `CAMERAMODE_DEATH_CAM_MP` → `camera_mode++`, looping back to `DEATH_CAM_SP` while `camera_mode < CAMERAMODE_SWIRL` (three iterations). Each individual replay's own advance is driven purely by animation/timer state (`bondview2.c:1467-1490`): the `g_CameraAfterCinema` flag that `bondviewAdvanceCameraMode()` acts on is set automatically once `modelGetAnimFrame(bodyModel) >= modelGetAnimEndFrame(bodyModel)`, or after a `camera_transition_timer >= 180.0f` tick-timeout fallback when there is no body model — **no button check in that branch; fully automatic.** **The button-press the user is hitting is a real, intentional N64 skip-ahead shortcut, not a port-added gate:** the same code block (`bondview2.c:1483-1489`) *also* fires the identical advance immediately on `(buttons & ~oldbuttons & (CONT_A | B_BUTTON | Z_TRIG | START_BUTTON))` whenever `bonddead && redbloodfinished && deathanimfinished` — i.e. on real N64 hardware, pressing A/B/Z/Start at any point during a death replay skips straight to the end, by design. **On the port, LMB (mouse fire) is mapped to `GE_CONT_G`, which is Z_TRIG** (`port/src/input.c:664-665`: `(mb & SDL_BUTTON(SDL_BUTTON_LEFT)) ... button \|= GE_CONT_G`; binding table `input.c:351`: `{"Z",GE_CONT_G}`) — exactly one of the four original skip buttons. So an LMB click during the death sequence is GE's own original cutscene-skip input, faithfully wired through the port's control remap, not a port-invented "click to continue" prompt. **What this investigation could not determine** (no build/runtime access in this worktree): whether the *automatic*, no-input path actually completes on PC in a reasonable time. `camera_transition_timer` accumulates via `g_GlobalTimerDelta` (`bondview2.c:1470,1493`), which is set from the same `g_ClockTimer`/`speedgraphframes`/`deltaFrames` chain flagged in D240 above and investigated under D155/D186/D193/D204/D209 — if any residual in that chain stalls or mis-scales specifically in this state, the auto-advance could take far longer on PC than on N64 (or effectively never complete before the player gives up and clicks), which would make an LMB click *feel* mandatory even though the ground-truth logic never requires one.