Root cause H2 confirmed: the gun-barrel / file-select vertical gradient is
built from 0.75-native-pixel-tall gSPTextureRectangle row strips (title2.c
(<<2)-1 bottom idiom). On the N64 RDP the scan converter floors the U10.2
edges to whole scanlines, so each strip rasterizes to exactly one scanline;
gfx_draw_rectangle passed the fractional edges straight to the GPU's
continuous rasterizer, aliasing the 0.75px pitch into a periodic 'comb'
(fixed ~3-native-row period; window period = 3*scale — confirmed by a
3-scale 640/1024/1280 GE_PCDUMP test).
Fix (port-layer, surgical): after the sub-pixel offset, floor the top edge to
its scanline and the bottom edge to the last covered scanline (the RDP
scanline range [top_scan, bot_scan+1]) before NDC conversion. No-op for
whole-pixel rects (edges already 4k/4k-1), so only sub-pixel strips change
and normal fills are untouched.
Also: remove the temporary GE_D397SH / D397SHM env-gated shader probes
(gfx_opengl.cpp, gfx_pc.cpp) now that the root cause is settled.
Docs: findings.md D397 -> RESOLVED (+3-scale confirmation + fix record),
findings-index.csv regenerated, porting-notes.md section D15 marked resolved
with the landed fix.
Verified by playtest: gun-barrel + file-select gradient render clean, comb
gone.
New env-gated D397SHM probe in gfx_generate_cc dumps which G_CCMUX
inputs a combine feeds as vInput1..N. For the gun-barrow row-strip
shader (id0=0000000010001218): vInput1=PRIMITIVE, vInput2=ENV_ALPHA,
so the front.c LERP renders out = PRIM + (TEXEL-PRIM)*(envA/255).
The PD ground-truth port feeds the identical env_color.a/255 factor for
the same ENV_ALPHA LERP input, so the LERP factor is not a port-side
scaling bug (H1's 1/255-vs-1/32 variant refuted). A uniform factor
cannot produce the *periodic* comb, which points to H2 (row-strip
geometry / 256-tri vertex-ring wrap). Next: Q3 gfx_flush() probe +
Q5 native FBO capture. Findings §D397 + handoff updated.
Per scratch/d397_fs/HANDOFF-D397.md: the stale-RENDERTILE hypothesis is
REFUTED (D161 already routes CI + G_TT_NONE to the I8 import; the GE_TEXI
logs show palfmt=0 on all 299 row strips), so the D397 block added to
import_texture had no effect (12px frame diff) and is removed.
The finding's true state: data path proven clean (byte-exact I8 uploads
vs the ground-truth PGM); MSAA/mips/noise/quant/cache ruled out.
Leading cause H1 = the file-select custom-CC LERP feeding ENV_ALPHA 0x14
as a 20/255 mix factor (black texels render ~0.92x prim = the measured
27/28/17 census); the N64 RDP 5-bit coefficient scaling (gmain.s / PD
CC engine) is the open root-cause question. H2 = possible separate
row-strip-geometry comb. GE_D397SH probe stays in-tree until the root
cause lands. findings table + tail section + CSV corrected.
Root-caused the D397 interlaced/comb symptom on the file-select
background and intro gun-barrel (first display, not round-trip):
titleRenderFolderMenuBackgroundLines loads 299 one-pixel 440-byte I8
rows with gDPLoadTextureBlock(I, 8b) and samples G_TX_RENDERTILE but
never emits gSPSetTile, so the tile's fmt/siz is whatever the previous
screen left (CI/4b spectrum-screen leftover, D116 probe). fast3d ran
the I8 bytes through import_texture_ci4 against a stale/all-zero
16-entry palette -> periodic black rows (GL texture-cache artifact,
not a ground-truth dither — no ROM RLE decode needed).
- gfx_pc.cpp import_texture: route the import through the source's
fmt/siz when a CI/RGBA tile genuinely disagrees with an I8 8b load
(mirrors the D229 case); real CI/RGBA textures untouched.
- gfx_opengl.cpp: GE_D397SH env-gated shader-formula probe for the
file-select gunbarrel combine (label initially written as D398,
renamed to avoid the PC-Options-label finding collision).
- findings: D397 row -> FIX LANDED (by-eye + PCDUMP verification
owed); tail section notes the fix. CSV regenerated.
Cross-ref D182(2) (same symptom family), D159/D183 (disproven leads),
D229 (mirror-case precedent).
- D398: replace D348 procedural slider glyph with ROM-native IMG_DOT
texture (texSelect + display_image_at_position, Copy/Erase pattern);
superseded by D400.
- D399: file-select bottom row shifted 15px left as a block (user
sign-off); selectpos.f[0] behind a new #ifdef PORT in front.c.
- D400: icon removed on user sign-off; PC Options is now a text-only
label exactly like Copy/Erase (one textMeasure + one textRender,
game's packed 0xEBD879FF gold). Fixed two measurement bugs:
textMeasure(textheight, textwidth) args were passed swapped (label
35px above the bar, hit band transposed) and the bare label measured
h=0 vs 14 for newline-terminated Copy/Erase (7px low); height now
measured with kLabelNL. LABEL_GAP 20->28, LABEL_X 343->351.
PCDUMP frame 195: tops aligned 279.8 vs 279.3, 31px clearance from
Erase. By-eye check owed.
- docs: findings.md had been truncated to its section table (686
lines), losing all ## Dxx tail sections D376-D400. Recovered:
table D376-D400 rows + body/sections D001-D375 from HEAD;
D376-D387, D396-D399 sections recovered verbatim from pi session
transcripts; D398 section recovered from the 2026-09-28 session;
D388-D395 rebuilt as explicit reconstructed stubs from the
table/CSV one-liners (original detail unrecoverable); D400 section
written fresh. findings-index.csv regenerated and verified
byte-identical to the pre-truncation copy.
Build clean (ntsc-final). Verified: PCDUMP frame 195 (no overlaps,
tops aligned), CSV matches pre-truncation backup.
bondview2.c reads crouchDown = insightaimmode && (buttons & (D_JPAD|D_CBUTTONS))
(5318/5579): crouch is C-down WHILE IN AIM MODE, never the B button. M3's
IA_CROUCH -> GE_CONT_B (and the CrouchMode=1 latch) was inert, so crouch
never fired in the GEPD layout. Hold + latch now emit GE_CONT_D (0x0004).
N64-faithful side effect: S (IA_BACK, also C-down) crouches while aiming.
User re-test: aim (RMB/Left Shift) + Left Ctrl; hold or F10 Toggle.
M2 (the wave's only rule-2 item, sign-off recorded via user direction):
#ifdef PORT hook in gunfire.c gunDrawSight -- Video.CrosshairHide skips
the draw, Video.CrosshairColor tints the RGB (7 named tints, 0 = the
N64 white); non-PORT builds keep the original literals. F10 GRAPHICS
rows + kResetDefaults.
D374 (M3 playtest fix, user decision): GEPD is now the DEFAULT key
layout (Input.Layout 0 = GEPD, 1 = N64-era binds; a mouse FPS layout is
the natural default and the N64 never had a keyboard). The preset is
now the true effective default: only EXPLICIT per-key [Input.Bind]
overrides beat it; persisted values equal to the N64 defaults count as
unset. This fixes the playtest bug where a stale persisted
Fire=Left Ctrl masked the GEPD preset and made Ctrl fire+crouch at the
same time. Boot log reports the applied layout. One-off: pre-D374
Layout=1 (ex-GEPD) inis mean N64 now -- flip to 0 (release note).
Video.LowEndMode (default off) in F10 VIDEO: ON writes FpsCap 30 +
MSAA x1 (restart row); OFF restores the compiled defaults (60/2) rather
than per-user pre-preset values. The key records the last selected
preset; the effect flows through the two real knobs.
Input.Layout (0=FPS default, 1=GEPD) + Input.CrouchMode (hold/toggle) rows
in F10 INPUT (Turok-standard functional section; the planned MODERN bucket
is dropped per D356). The GEPD preset acts as the per-key effective
default under [Input.Bind] overrides; new IA_RELOAD/IA_CROUCH actions emit
GE_CONT_B (unbound by default, N64-original). Toggle crouch latches B with
a menu-boundary latch drop. optionsoverlay rowSetCommit hook re-derives
the binds via inputLayoutApply(). inputRebuildBinds no longer warns on
intentionally-empty defaults. Boot + ini round-trip + F10 page verified;
in-game feel-check owed (D371).
'Game file' (the manual's term) reads, in modern PC usage, as the
program's own files, not the user's save. 'Profile' is the clearest
end-user term for what these files actually are -- a per-player
bundle of progress + per-file settings -- and '(per profile)'
expresses the tag's scope where '(save)' read as a verb.
Strings (both UIs: front options + F10 overlay):
- front top row: 'Save file' / 'File N' -> 'Profile' / 'N'
- section annotation: '(File N)' / '(no file)' -> '(Profile N)' / '(none)'
- per-file row tag: '(game file)' -> '(per profile)'
- F10 unavailable value: 'Select game file' -> 'Select a profile'
Display layer only: identifiers and ini keys (Bond.* rows,
watchSettings*, ini) are untouched. D357 plan section 4 is the
terminology record; new D359 finding + index row; manual
'game file' -> 'profile' sync noted as a TODO (port-owned doc,
not in this repo). Build + headless smoke clean.
Full-string audit of the settings UI (front options + F10 overlay):
- front top-row label 'Save file' -> 'Game file' (the instruction
manual's term)
- per-file row tag '(profile)' -> '(game file)', both UIs
- F10 unavailable value 'Select file' -> 'Select game file'
- kept: 'File N' values, the '(File N)' section annotation and the
'(no file)' unavailable state (short forms; the annotation's host
word is the section title, which says the file number)
- 'Bond file' now appears only in identifiers/config keys
(Bond.* rows, watchSettings* names, ini keys), never user-visible
Docs: D357 plan section 4 is now the terminology record (final string
table); D358 finding notes the (profile) -> (game file) supersession.
Build + headless smoke clean.
- The front options screen numbers items 0..s_rowN (0 = the top
save-file row, k = content row k-1), but the level-1 draw loop painted
content row k at rowY(k) -- the save row's own line -- and mapped
s_hl (item-numbered) straight into s_rowIdx (content-numbered): the
'Save file' row overlapped the first content row, highlights/adjusts
hit the wrong row, the last content row was dead, and s_rowIdx[s_hl]
could read one past the built rows. Content rows now draw at rowY(k+1)
and all s_hl -> s_rowIdx uses take s_hl - 1 (A-press, L/R repeat,
reset-maintain); hit-test/cursor were already item-numbered and are
untouched.
- Per-file row tag '(save)' -> '(profile)' (maintainer request), both
UIs (front options + F10 overlay) plus the stale comments.
- docs/dev/D357-SETTINGS-VALUES-PLAN.md (PROPOSAL): row-by-row audit of
current defaults/ranges/steps/displays vs the PD port (local checkout)
and the Turok PC list (maintainer notes); proposals: mouse sensitivity
as a x multiplier, deadzone as a whole-% key (D333 legacy-alias
pattern), texture filter '3-Point' -> Bilinear, audio step 128 -> 328
(whole-%), MSAA +16x, FPS-cap preset grid, FOV degrees; kResetDefaults
lockstep gate + GE_WSPROBE_RESET re-verification.
Build clean (248 targets), headless menu boot smoke clean; visual
check of the front options screen + F10 owed to the maintainer.
- 5 sections (INPUT/GAMEPLAY/GRAPHICS/AUDIO/VIDEO, all below MAX_PROWS):
DISPLAY+GRAPHICS merged into GRAPHICS; AUDIO carries the watch music/FX
sliders out of the Bond-file page; VIDEO keeps the pure video rows;
__HdrMouse/__HdrPad merged into INPUT; __HdrBond/__BondFile retired for
the front-only top save-file row (frontoptions.c)
- save-file scoping: per-file sections (GAMEPLAY, AUDIO) annotated
'(File N)'/'(no file)' via the new Row.saveScoped flag +
watchSettingsActiveFolder(); per-file rows in mixed sections carry a dim
'(save)' tag; the D353 optionsRowIsBondChooser special-case is
generalized (both UIs)
- N64-watch rows Bond.Look / Bond.AimControl commented out of the menu
surface (D181/D216 style: keys + watch paths stay live)
- per-section 'Reset to defaults' replaces the Turok rows (plan Gate E):
edge-triggered two-step arm->confirm within 3 s, one commit, disarm on
timeout / navigate-away / close (optionsResetMaintain/Clear);
rowResetSection walks the section's declared range: watch rows ->
BLANKSAVEDATA through the D352 commit path (front direct, stage queue),
ini rows -> kResetDefaults[] (the port's C initializers); __Resolution
is a documented exclusion (action-backed)
- GE_WSPROBE_RESET probe: front tick-120 prepare/dispatch/verify, stage
tick-120 dispatch / tick-121 verify; a direct -level_XX boot starts its
timer from phase 0 (front prepare is front-only)
- docs: findings.md D356 + index + CSV, WATCH-SETTINGS-PLAN status,
OPTIONS-MENU-PLAN pointer, plan doc status -> LANDED
Verified headless: GE_WSPROBE_RESET front (all 5 sections failures=0,
__Resolution untouched, file-2 scope isolation OK) and -level_33 (6
commands queued, all sections verified post-drain, saved music/sfx = 0xFF);
GE_WSPROBE / GE_WSPROBE_FRONT regressions clean; 248/248 build + link.
Add a game-thread bridge and field-scoped per-file persistence for PC UI edits, with an explicit file chooser; keep watch-only edits session-only. Fix D351 row initializer offsets for units and auto-hide. Document verification and review follow-ups.
No settings icon exists in the game art (mainfolderimages holds only
copy/del/select-arrow/X/check/dot), so a miniature slider control is drawn
instead: three menu-box tracks with offset knobs in the 14px gap after Erase
(10px wide, 2px margins), INK_DIM tracks, knobs matching the label ink
(white / gold while hot). Tracks the dynamic label x (JP-safe), text position
unchanged; hot region extended to include the glyph. PCDUMP pixel-verified.
D-pad left/right and stick-X now adjust the selected row's value (A/X/B/Y
kept); 18/4-frame hold-to-repeat added to value adjust and up/down nav,
matching the options screen's D345(e)/(f) cadence. Port-only; user
controller test verified.
D345 (a-g): re-assert window on menu change; cursorToItem skips snap while
the 1:1 pointer is live; W/S emit stick in menus under NaturalPitch;
F10-style discrete row stepping + arrow left/right value adjust replace raw
drift (first pass shipped broken, fixed); screen-placed cursor parks in the
left gutter off the row text.
D346 (+b): labels/values aligned to Nightdive/Turok + PD-port conventions
(title-case On/Off, Frame rate cap, Anti-aliasing/None, Anisotropic
filtering, units in values not labels, deadzone as %, self-explanatory
auto-FOV toggles); VIEW section merged into VIDEO (MAX_PROWS 12->15,
ROW_DY 20->18). Display-only: config keys and stored values untouched.
Port-only; no src/game edits. User playtest verified.
Rule-2 layout edit (user sign-off 2026-09-27): Copy 247/225 -> 222/200,
Erase 357/335 -> 304/282 under #ifdef PORT; their click boxes follow.
"PC Options" is drawn by the port hook right after Erase's measured text
end (min x 352), so wider localised labels push it right. The top-right
label is gone.
The file-select Options label opens a panel drawn in the same hook:
sections as tabs, rows from the F10 overlay's table via an exported row
API, the game's crosshair as pointer (input.c passes the stick, hands the
buttons to the panel). Quit to desktop now goes through videoRequestQuit.
Also repairs two escaped-newline typos in the D343 docs.
One Rule-2 #ifdef PORT draw hook in constructor_menu05_fileselect (user
sign-off 2026-09-27). The label's hit test and open request live in
optionsoverlay.c; the open is applied on the input thread. While the
overlay is open on file select, g_MenuTimer is held at 0 so the screen no
longer drops to the legal screen after 30 s.
The crop tested loaded_texture.line_size_bytes, which is the whole
LoadBlock (1400 B), so the water still uploaded 32x43 while UVs were
normalised to 32 rows; the mip rows showed as coloured dashes. Test the
tile line size instead. User-verified on Frigate; golden 3/3.
The PC wallet model (0x1664C B, 0x17000 reservation) loads at the start of
ptr_logo_and_walletbond_DL, but the file-select scratch DL and the
briefing text still used the N64 offset +0xA000 -- inside the model.
After main menu -> back the model is not reloaded, so the overwritten
folder bodies/photos disappeared. Both now use +0x17000 (same 0x6e000
budget). User-verified; golden gate 3/3. D341 also user-verified.
Two fast3d defects. (1) The 1-cycle blood texture rectangle sampled a
stale tile-1 CI8 declaration: the title leaves G_TL_LOD set and the D236
detail-base rule redirected texunit 0 to tile 1. LOD tile selection is
2-cycle-only on the RDP, so texrects outside 2-cycle now sample their own
tile. (2) The drip image is regenerated in place in the dyn pool (reached
via the KSEG0 mirror); the address-only cache key froze it on its first
image. Dyn-pool textures now key on a content hash. Headless-verified,
golden 3/3. A/B: GE_DYNTEXHASH_OFF.
The IsWater 32x32 CI8 image (drawn through an RGBA16 tile, D229) was
imported with its mip chain as 32x43 and normalised as 16 wide, so the
pattern ran at 2x and wrapped every 43 rows instead of the N64's 32:
fold steps made it jump. Import the base level only and use CI8
geometry in the tri path. Sky/water fans now go through a port-only
G_FLOATVTX_EXT (float clip positions + S/T) instead of s16 Vtx, removing
the tc shift and wScale quantisation jitter. User-verified vs 1964;
golden 3/3; Win + Linux build. A/B: GE_D245_OLDVTX / GE_D245_OLDTEX.
The op-22 DisplayListPrimaryRecord count is the number of star-flash
arms; dorottex (and PD's modelRenderNodeStarGunfire) read count*4
vertices. d43_emit/d43_convert sized the array as count vertices, so
3/4 of every muzzle star came from neighbouring sidecar bytes and drew
as long spikes (M16 "shooting up"). Size it 16*4*count. Emitter
round-trip checks all pass; M16 renders a clean star headless.
Also TEMP GE_STARTWEAPON (bondview_r.c) for weapon repros. Re-run
tools_pc/d43_emit.py to regenerate data/pcmodels-*.
CCTVRecord's own s32 `pad` (the look-at pad) shadows the inherited
ObjectRecord `pad` (mount pad) under IDO. The PC struct renamed it
`lookpad` to satisfy GCC, which silently re-pointed setupCctv's reads at
the mount pad, so every camera's look vector was its lens offset and the
lens faced the wall. Read `lookpad` on PC (#ifdef PORT, A1 class);
matches PD's setupCreateCctv (lookatpadnum). User-verified vs 1964,
golden gate 3/3. Porting-notes gains the duplicate-member rename trap.
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.
optionsoverlay.c called atoi and lround through implicit int declarations
(GE's stub stdlib.h/math.h shadow the CRT, D324 class); atoi joins the
other K&R decls in port/shim/stdlib.h, lround gets a local prototype.
Windows + Linux builds clean of both warnings.
GE_INPUTSCRIPT gains ZHOLD/ZREL (sustained fire) for automatic-weapon
audio repros. D207/D241: headless burst-fire at 60 vs 30 fps with the
alarm forced logged zero voice drops at both caps; owed a real-play
capture. Golden gate 3/3.
D291 (user-captured) is D306: Dam intro camera 5, truck over the walkway
camera; reproduces at FovScale 100. All PC-specific leads refuted with
evidence: FOV, depth precision/DrawDistance, depth state (every truck
triangle is opaque z-tested), near-plane handling, wheel transforms. The
image is a depth-correct view of the authored model from under the truck.
GE_ZF's earlier "real fighting" was a +-255 z quantisation artefact.
TEMP tools: GE_INTROCAM=<n> pins the level intro's first camera
(bondview_r.c, #ifdef PORT, PRNG draw still consumed); GE_D306C depth-state
census in gfx_pc.cpp. Golden gate 3/3.
gimgSyncCompiledGlobalDLs (D68 shim) only copied texLoad's resolved
texture pointers into the compiled globalDL shadows while a slot still
held the 0xABCD IMAGESEG marker, i.e. only for the first stage (title).
Every later stage kept pointers into the title's freed texture pool, so
explosion smoke/fire/debris rendered as rainbow garbage once the stage
overwrote that memory (Facility: past the lockers). Learn the slots once,
re-sync them every stage. User-verified; golden gate 3/3.
Also: TEMP GE_D252POOL texpool/sync probe; GE_D157 F12 burst and caps
(uncapped colour match made live play 4-6 fps). D314 F10 flicker
user-confirmed fixed. Findings M-201, porting-notes C2.
- D252 (M-200): the 'rainbow confetti' is the explosion debris
(explosionInitFlyingParticles). Statically clean for colour; 4 headless
repros including the user's grenade-launcher trigger are all clean. Still OPEN.
- d320repro (TEMP): new explode= and cheat= actions for headless explosion
captures; documented in GE-ENV-PROBES.
- D314: optionsOverlayScroll (host thread) rebuilt s_visIdx/s_visN and wrote
s_sel while the scheduler thread draws the overlay. Wheel notches are now
queued atomically and applied in HandleInput (D287 pattern). Emit and
HandleInput share the scheduler thread. Headless file/mission select is
pixel-stable; likely fixed together with D287, needs a user check.
- D291: truck AI mapped (ai_9 / 0x040a, path 7); headless walk did not reach it.
input.c called SDL_SetRelativeMouseMode / SDL_ShowCursor from the scheduler
thread (every controller poll) and from the host thread's focus/click
handling at the same time. SDL's mouse API is not thread-safe; on Windows
relative mode starts/stops the raw-input thread, and clicking back into the
window after being unfocused is exactly when both threads flip it - the D287
freeze trigger. Callers now record the wanted state; inputApplyMouseRequests()
applies it on the host thread from videoPumpEvents(). Queue-deadlock leads
ruled out statically. Win+Linux build clean, golden 3/3; freeze re-test owed.
Four simultaneous drain monitors on the same USB interface: WASAPI (512 and
2048 samples) fell to 81-88% drain for ~30-50 s every ~317 s and lost 1.4% of
playback, while DirectSound stayed at exact real time. Each WASAPI shortfall
overflowed the audio queue and dropped blocks (the periodic garbled bursts).
audioInit() now hints directsound,wasapi; SDL_AUDIODRIVER still overrides.
Golden gate 3/3; 60 s GE_D204 run clean after load.
James Dansereau