Sylpheed

Files

sylph-decoder 0d42a1d8d1 re: rest_plateau() picks the wrong plateau -- and it is the whole residual

rest_plateau() selects the LONGEST run of identical adjacent poses, which need
not be the run covering the screen's settle instant. rest_vs_settle left a 21.9 %
disagreement that I recorded as ambiguous by construction. It is not.

  CONTROL  exactly one plateau, covering the settle instant:
           3 072 / 3 072 agree (100.0 %)
  TEST     more than one plateau, at least one covering:
           1 622 elements, agree on 586 (36.1 %)
           of the 1 036 disagreements, rest() landed on a run NOT covering the
           settle instant: 1 036 -- all of them, no exceptions

Both poses are genuinely held in these cases -- they are plateau cases, not
transients -- so this is rest() returning a pose the screen has ALREADY LEFT by
the time it settles.

This corrects my own METHOD entry of two iterations ago, which said a candidate
cannot be adjudicated against the incumbent it replaces. Too strong. The bare
comparison cannot; the comparison plus a structural property that independently
says which side is wrong in each disagreement can. What I lacked was not an
oracle but a discriminator.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Wuu56cE8vJGTBtn1ppsk8v

2026-08-30 12:38:18 +00:00

_all2.rs

re: a nested record's +0x08 is its loop length -- the plate's period is 120, not 105

2026-08-29 19:40:23 +00:00

_keys.rs

formats: a keyless primitive that would hide the screen is forced to paint first

2026-08-29 21:02:17 +00:00

_pbafc.rs

re: primitive blend mode -- undecodable with reach, and no longer a risk

2026-08-29 21:58:25 +00:00

_rechdr.rs

re: a nested record's +0x08 is its loop length -- the plate's period is 120, not 105

2026-08-29 19:40:23 +00:00

achievements_map.rs

re(challenge): the gate's bit space is the game's 24 ACHIEVEMENTS

2026-08-13 19:32:43 +00:00

adv_region_extend.rs

re: resolve_movie_voice_region starts INSIDE the first stream, 8 of 10 multichannel regions

2026-08-30 08:41:22 +00:00

adv_voice_dump.rs

re: which ADV stream sits where -- settled by level, not by waveform

2026-08-30 08:32:14 +00:00

arsenal.rs

game_data: Arsenal loader — player weapon options per hardpoint

2026-07-23 20:30:13 +02:00

bank_streams.rs

re: the two unexplained XMA streams are BGM_102, and BGM_103 s sizes survive

2026-08-29 16:16:13 +00:00

battle.rs

game_data: UnitRoster loader — per-mission battle compositions

2026-07-23 20:11:32 +02:00

bench_ships.rs

perf(mesh): memoise the whole-container decode

2026-08-12 18:57:50 +00:00

better_home.rs

fix(xbg7): degenerate index runs 582 -> 1 (grouped path + prefer a clean candidate)

2026-08-13 08:52:41 +00:00

bgm_dump.rs

re: the menu BGM loops at 61.93 s, and the game never reaches the fade

2026-08-30 09:34:42 +00:00

bounds_in_descriptor.rs

fix(mesh): filtered decode must not depend on the requested subset

2026-08-12 18:47:41 +00:00

campaign.rs

localization: resolve the game's UTF-16 text keys to display strings

2026-07-23 19:49:42 +02:00

caption_coverage.rs

re: read all eight caption families — 15x more text, and the same lesson twice

2026-08-26 02:47:13 +00:00

capture_ib_truth.rs

re(xbg7): a capture names a never-decoding family, and the cause is mixed strides in one pool

2026-08-13 10:52:35 +00:00

capture_index_bytes.rs

fix(xbg7): the index run was one element late for 575 sub-meshes

2026-08-13 07:26:52 +00:00

capture_match.rs

ship_capture: bake e106 COMPLETE from the 2026-07-26 F10 capture — all 8 parts

2026-07-26 18:26:23 +02:00

capture_truth_scan.rs

re: the capture is Stage_S02 -- provenance corrected, oracle grows to 46 buffers

2026-08-12 02:59:40 +00:00

capture_verify.rs

ship: engine placement CONFIRMED by 43-instance capture; exhaust markers added

2026-07-26 19:44:12 +02:00

challenge_map.rs

re(challenge): the EX missions are a MODE, not a stage number

2026-08-13 18:59:52 +00:00

challenge_screen.rs

re(challenge): the unlock is a bit test, and the mission table is on disc

2026-08-13 19:14:48 +00:00

composite_geometry.rs

re: the consistency figure is mostly composite bounding boxes -- real disagreement is 1

2026-08-12 18:02:06 +00:00

consensus_check.rs

re: all 89 remaining disagreements are 24-vertex boxes -- no real mesh disagrees

2026-08-12 18:11:20 +00:00

correlate_capture.rs

test: fix the stale default disc paths -- three tests could never run unaided

2026-08-12 01:18:10 +00:00

correlate_frames.rs

ship: index-less brg/eng/sld parts never matched their GN frame — 34 ships assembled without a bridge

2026-08-10 19:16:06 +00:00

coverage_audit.rs

fix(mesh): grouped pools no longer emit sub-meshes with out-of-range indices

2026-08-12 04:41:17 +00:00

cue_unit_check.rs

re: cue times ARE seconds — and "the rate does not converge" was my error

2026-08-26 00:14:31 +00:00

decl_word_probe.rs

re(ui): a RATC child's name is stated, not inferred -- and it was hiding every menu background

2026-08-29 07:26:56 +00:00

deep_map.rs

game_data: typed loaders for the combat / character / mission tables

2026-07-23 19:49:21 +02:00

default_owners.rs

re: runtime DATA SHEET reads the live weapon record (Route B, first capture)

2026-07-28 18:01:25 +00:00

defaulted_fields.rs

re: runtime DATA SHEET reads the live weapon record (Route B, first capture)

2026-07-28 18:01:25 +00:00

desc_dump.rs

feat(xbg7): read the per-sub-mesh vertex declarations (XBG7_SUBMESH_DECLS, off by default)

2026-08-13 11:45:20 +00:00

dialogue.rs

game_data: DemoMessage loader — the dialogue backbone

2026-07-23 19:59:14 +02:00

dm_extract.rs

game_data: typed loaders for the combat / character / mission tables

2026-07-23 19:49:21 +02:00

dm_roster.rs

game_data: typed loaders for the combat / character / mission tables

2026-07-23 19:49:21 +02:00

dm_rows.rs

game_data: typed loaders for the combat / character / mission tables

2026-07-23 19:49:21 +02:00

dossier.rs

localization: resolve the game's UTF-16 text keys to display strings

2026-07-23 19:49:42 +02:00

edge_cap_sweep.rs

re: 98.7% coverage, and the scale-free degeneracy fix is refuted

2026-08-12 06:19:38 +00:00

envelope_screen.rs

re: containment screen fails too -- the static assembler inflates a shared turret

2026-08-12 17:20:52 +00:00

filter_consistency.rs

re: CORRECTION -- the static assembler is not at fault, the pre-fix decoder was

2026-08-12 17:28:18 +00:00

find_ib.rs

re: eng_02_l's real block is adjacent -- the connectivity heuristic rejects it

2026-08-12 02:31:53 +00:00

find_mirror.rs

test: lock in the twin invariant -- 0 unrelated pairs disc-wide, 1 known collapse

2026-08-12 03:44:29 +00:00

find_stream_by_size.rs

re: the two unexplained XMA streams are BGM_102, and BGM_103 s sizes survive

2026-08-29 16:16:13 +00:00

flights.rs

re: rebuild game_data on the IDXD record table — 966 misses and 596 flattened reads

2026-08-26 01:42:53 +00:00

focus_alpha_census.rs

re: 210 focus records pulse, not 2 -- and 8 of them fail in a way that looks right

2026-08-29 19:45:17 +00:00

forced_backdrop_ink_thresholds.rs

re: reconcile the two ink counts -- we agree at >1 and not at >0

2026-08-30 07:07:21 +00:00

forced_backdrop_key_source.rs

re: not ONE of the 80 forced instances has a file-read key

2026-08-30 07:11:12 +00:00

forced_backdrop_necessity.rs

re: the forced-backdrop rule's pixel cost -- 38 screens go black without it

2026-08-30 07:04:05 +00:00

forced_backdrop_pixel_cost.rs

re: the forced-backdrop rule's pixel cost -- 38 screens go black without it

2026-08-30 07:04:05 +00:00

gate_histogram.rs

re: coverage is 96.4% (6069/6294) -- and the 225 misses get a gate breakdown

2026-08-12 04:22:43 +00:00

grouped_report.rs

re(xbg7): a capture names a never-decoding family, and the cause is mixed strides in one pool

2026-08-13 10:52:35 +00:00

hokyu_demo_resolve.rs

Fix hokyu voice: resolve by subtitle demo-id, not ship category

2026-07-23 19:48:56 +02:00

hokyu_final.rs

Cutscene voice: resolve movies via continuous-stream cue regions

2026-07-23 06:53:55 +02:00

idxd_tokens.rs

re: generalize the two capture tools beyond weapons

2026-07-29 06:05:13 +00:00

index_hash_dump.rs

fix(xbg7): the index run was one element late for 575 sub-meshes

2026-08-13 07:26:52 +00:00

index_pad_check.rs

fix(xbg7): the index run was one element late for 575 sub-meshes

2026-08-13 07:26:52 +00:00

invert_capture.rs

re: the Stage-02 capture drew no capital ship at all — invert the match, then control range

2026-08-10 18:04:58 +00:00

iso_map.rs

game_data: typed loaders for the combat / character / mission tables

2026-07-23 19:49:21 +02:00

kf_flip_test.rs

re: the keyframe block s +4 and +8 are angles used almost always as a 180 flip

2026-08-29 18:30:29 +00:00

kf_timeline.rs

formats: a settled screen is one instant, not one hold per element

2026-08-29 19:28:27 +00:00

kf_unknown_census.rs

re: the keyframe block s +4 and +8 are angles used almost always as a 180 flip

2026-08-29 18:30:29 +00:00

leaf_alpha_compose.rs

re: a nested .rat leaf animates on its own timeline; parent alpha does not multiply

2026-08-29 18:03:19 +00:00

leaf_dump.rs

re: 125 percent is not the only odd scale, and ptlogo_eff2 is a pop not a steady state

2026-08-29 18:16:20 +00:00

leafdbg.rs

formats: teach the renderer to rotate (Option A) -- and report that it does not close the title

2026-08-29 18:59:26 +00:00

live_offsets.rs

re: a deeper window does not help -- the 0x820af844 object is not the definition record

2026-08-12 20:49:19 +00:00

locate_draw.rs

re: the break-composite alarm is false -- the container stores 3-6 copies of each buffer

2026-08-12 03:32:58 +00:00

loo_band.rs

formats: a settled screen is one instant, not one hold per element

2026-08-29 19:28:27 +00:00

manifest_map.rs

feat(formats): decode the movie manifest's full mission/phase structure

2026-07-21 22:49:19 +02:00

miss_targets.rs

re(xbg7): a capture names a never-decoding family, and the cause is mixed strides in one pool

2026-08-13 10:52:35 +00:00

mission_map.rs

game_data: typed loaders for the combat / character / mission tables

2026-07-23 19:49:21 +02:00

movie_map_csv.rs

re: rebuild the movie manifest on the record table — the old counts measured pool strings

2026-08-25 23:13:41 +00:00

name_from_hash.rs

re: the two unexplained XMA streams are BGM_102, and BGM_103 s sizes survive

2026-08-29 16:16:13 +00:00

name_resolution.rs

re(ui): a RATC child's name is stated, not inferred -- and it was hiding every menu background

2026-08-29 07:26:56 +00:00

node_dump.rs

ship: STATIC assembly is now EXACT — node-TRS off-by-one cracked via the capture

2026-07-26 19:00:49 +02:00

node_scale.rs

re: CORRECTION -- the static assembler is not at fault, the pre-fix decoder was

2026-08-12 17:28:18 +00:00

pad_shift_audit.rs

fix(xbg7): degenerate index runs 582 -> 1 (grouped path + prefer a clean candidate)

2026-08-13 08:52:41 +00:00

paint_order_audit.rs

re: the port was still being told SE audio is undecodable -- it is not

2026-08-29 02:31:37 +00:00

palogo_eff_check.rs

re: the port's two extra elements are PLATEAU cases -- refutation succeeds, and its point gets bigger

2026-08-30 11:54:43 +00:00

params.rs

feat(xbg7): read the per-sub-mesh vertex declarations (XBG7_SUBMESH_DECLS, off by default)

2026-08-13 11:45:20 +00:00

part_pos.rs

ship_capture: bake e106 COMPLETE from the 2026-07-26 F10 capture — all 8 parts

2026-07-26 18:26:23 +02:00

plateau_choice.rs

re: rest_plateau() picks the wrong plateau -- and it is the whole residual

2026-08-30 12:38:18 +00:00

plateauless_suppression.rs

re(ui): a static composite is only meaningful for a screen that settles

2026-08-29 04:37:59 +00:00

pool_window.rs

re: Size_Y is genuinely defaulted, and the loader leaves it at 0.0

2026-08-12 22:47:43 +00:00

prm_alpha_census.rs

formats: a keyless primitive that would hide the screen is forced to paint first

2026-08-29 21:02:17 +00:00

prm_colour_census.rs

re: the primitive colour census -- and it refutes 38 of my own 80 forced verdicts

2026-08-29 21:51:06 +00:00

prm_forced_first.rs

formats: a keyless primitive that would hide the screen is forced to paint first

2026-08-29 21:02:17 +00:00

prm_occlusion_check.rs

formats: a keyless primitive that would hide the screen is forced to paint first

2026-08-29 21:02:17 +00:00

prm_span_sensitivity.rs

re: the forced-backdrop span -- 256 vs 211 is a bundle mismatch, and the hold

2026-08-29 21:14:58 +00:00

probe_anchor.rs

fix(xbg7): degenerate index runs 582 -> 1 (grouped path + prefer a clean candidate)

2026-08-13 08:52:41 +00:00

ptloop_leaf_sweep_at.rs

re: ask #2 -- t=357.7 came from a GPU draw capture, and a sweep cannot date a frame

2026-08-30 07:15:45 +00:00

rat_inspect.rs

formats: ui_layout — reassemble UI screens from .rat records

2026-07-29 20:27:35 +02:00

rat_leaf_placement.rs

re(ui): the focus ring SPINS, the game draws it, and the leaf owns the f record

2026-08-29 08:46:53 +00:00

ratc_child_names.rs

re(ui): a RATC child's name is stated, not inferred -- and it was hiding every menu background

2026-08-29 07:26:56 +00:00

ratc_optless_children.rs

re(ui): the 60 nameless RATC children are frames, not children -- .tan decoded

2026-08-29 07:39:09 +00:00

record_loop_length.rs

re: a nested record's +0x08 is its loop length -- the plate's period is 120, not 105

2026-08-29 19:40:23 +00:00

resolve_all.rs

Cutscene voice: resolve movies via continuous-stream cue regions

2026-07-23 06:53:55 +02:00

rest_fallback_audit.rs

re: my own 1697 audited -- and the first correction failed its control

2026-08-30 12:13:46 +00:00

rest_fallback_census.rs

re: the rest() fallback -- its example dissolved, the question got bigger

2026-08-30 11:50:07 +00:00

rest_fallback_title.rs

re: the rest() fallback -- its example dissolved, the question got bigger

2026-08-30 11:50:07 +00:00

rest_vs_settle.rs

re: propose settle-instant posing -- and my control cannot validate it

2026-08-30 12:22:44 +00:00

roster_target.rs

re(units): target missions by roster, harvest S09 — 36 units / 3 439 defaulted values

2026-08-13 13:48:12 +00:00

scale_census.rs

re: 125 percent is not the only odd scale, and ptlogo_eff2 is a pop not a steady state

2026-08-29 18:16:20 +00:00

screen_configs.rs

re(challenge): poking unlocks the whole story campaign; MISSION SELECT is story-only

2026-08-13 20:54:57 +00:00

screen_layout.rs

re: placement X/Y are SIGNED, the tail word is time, and the resting pose is max-dwell

2026-08-11 22:14:36 +00:00

se_wave_dump.rs

media: expose se_wave_riff -- the menu's SE cues, assembled where the format lives

2026-08-29 08:41:42 +00:00

settle_window.rs

formats: a settled screen is one instant, not one hold per element

2026-08-29 19:28:27 +00:00

shared_bank_takes.rs

re: RETRACTED "audio is missing" — a subtitle cue is a START time

2026-08-26 00:20:57 +00:00

shared_vbase_check.rs

re: the proven offsets ARE candidates -- these mis-anchors are selection failures

2026-08-12 02:11:41 +00:00

ship_audit.rs

viewer,ship: external parts ON by default + Ships catalog = composite ships only

2026-07-26 19:15:34 +02:00

ship_dump.rs

ship_capture: ingest the draw-logger format + quantify the static gaps

2026-07-25 20:57:08 +02:00

ship_render.rs

ship: engine placement CONFIRMED by 43-instance capture; exhaust markers added

2026-07-26 19:44:12 +02:00

sibling_rule_check.rs

re: the "sibling default" rules are a dedup artefact — WITHDRAWN

2026-08-25 23:38:55 +00:00

slab_screen.rs

re: the other four ships are clean; automating the slab check failed

2026-08-12 07:22:43 +00:00

slb_fmt_probe.rs

re: the .slb leading region is XMA1 MONO — 113x more audio than stereo

2026-08-25 23:53:10 +00:00

slb_hybrid_scan.rs

re: the .slb leading region is 1392+n*2048 — and my fix for it is withdrawn

2026-08-25 23:30:42 +00:00

spawn_survey.rs

game_data: typed loaders for the combat / character / mission tables

2026-07-23 19:49:21 +02:00

sprite_dims.rs

re: the title's build-in measured in the guest's draw stream -- the flashes are real

2026-08-29 19:56:34 +00:00

sprite_of.rs

formats: teach the renderer to rotate (Option A) -- and report that it does not close the title

2026-08-29 18:59:26 +00:00

squadrons.rs

re: rebuild game_data on the IDXD record table — 966 misses and 596 flattened reads

2026-08-26 01:42:53 +00:00

starts_probe.rs

feat(xbg7): read the per-sub-mesh vertex declarations (XBG7_SUBMESH_DECLS, off by default)

2026-08-13 11:45:20 +00:00

starts.rs

re: n206 was fixed after all -- the audit was classifying by geometry, not buffer

2026-08-12 04:10:05 +00:00

submesh_dump.rs

fix(mesh): require exact pool coverage -- found by rendering, not by a metric

2026-08-12 06:56:49 +00:00

tie_break_pixel_cost.rs

re: the paint-order tie-break costs one pixel, on one screen we do not ship

2026-08-29 19:35:29 +00:00

tie_cost_over_time.rs

re: the paint-order tie-break costs one pixel, on one screen we do not ship

2026-08-29 19:35:29 +00:00

try_anchor.rs

re: the twins' correct block validates -- distinctness is a real fix; eng_02_l is not

2026-08-12 02:17:55 +00:00

twin_mirror_audit.rs

re: n206 was fixed after all -- the audit was classifying by geometry, not buffer

2026-08-12 04:10:05 +00:00

ui_screen.rs

re: crack the UI layout record — the retail UI rebuilds from the disc

2026-07-28 19:05:21 +00:00

undecoded.rs

re: coverage is 96.4% (6069/6294) -- and the 225 misses get a gate breakdown

2026-08-12 04:22:43 +00:00

unit_fields.rs

re: partial field map inside the live definition object, and why the automation failed

2026-08-12 20:36:57 +00:00

unit_signatures.rs

re(route-b): 13 defaulted Size_Y values read out of the running game

2026-08-12 20:28:55 +00:00

unit_values.rs

re: three probes for the parsed record all miss -- hypothesis: there is not one

2026-08-12 21:00:48 +00:00

validate_cues.rs

Cutscene voice: resolve movies via continuous-stream cue regions

2026-07-23 06:53:55 +02:00

vcount_index.rs

re: the Stage-02 capture drew no capital ship at all — invert the match, then control range

2026-08-10 18:04:58 +00:00

verify_fieldmap.rs

re(route-b): full field map verified 154/154, and 101 defaulted fields read out

2026-08-12 23:44:57 +00:00

voice_bank_dump.rs

re: the .slb leading region is 1392+n*2048 — and my fix for it is withdrawn

2026-08-25 23:30:42 +00:00

voice_bank_shape.rs

re: the .slb leading region is 1392+n*2048 — and my fix for it is withdrawn

2026-08-25 23:30:42 +00:00

voice_len_vs_subs.rs

re: the resupply banks really are missing audio — the subtitles prove it

2026-08-25 23:45:26 +00:00

voice_region_cap_sweep.rs

formats: drop the 1.5 MB cap that truncated 17 voice regions' first stream

2026-08-30 08:55:47 +00:00

voice_region_chunk_census.rs

re: count the voice-region population properly -- 25 three-chunk, and 17 of them were broken

2026-08-30 09:46:08 +00:00

voice_region_chunks.rs

re: the voice-region third chunk is a different structure from the BGM one

2026-08-29 14:56:17 +00:00

voice_region_fix_test.rs

formats: drop the 1.5 MB cap that truncated 17 voice regions' first stream

2026-08-30 08:55:47 +00:00

voice_region_start_audit.rs

re: resolve_movie_voice_region starts INSIDE the first stream, 8 of 10 multichannel regions

2026-08-30 08:41:22 +00:00

voice_region_start_why.rs

formats: drop the 1.5 MB cap that truncated 17 voice regions' first stream

2026-08-30 08:55:47 +00:00

voice_stream_cue_map.rs

re: a long voice cue is three presentations of one take, and our guard clips the first

2026-08-29 15:09:40 +00:00

voice_three_stream_sizes.rs

re: the dual-mono explanation for the extra stream bytes does not generalise

2026-08-29 15:37:46 +00:00

why_missed.rs

re(xbg7): characterise the residual 47 misses — 99.25% of resources decode

2026-08-13 12:50:19 +00:00

why_rejected.rs

re: n206 was fixed after all -- the audit was classifying by geometry, not buffer

2026-08-12 04:10:05 +00:00