fabi/Syplheed-Reborn
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

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

Browse Source 
examples/why_missed.rs reports the furthest gate per undecoded resource. Disc-wide
that is 47 rows / 43 distinct names, and they are mostly not ship geometry:

  30  e_rou_/_rou_ pose & proxy composites (24-vertex marker boxes, extent 0.010)
   6  .DAT particle composites in ptc_pack
   8  damage/LOD variants (e101_bdy_02_d, e901_wing_05_*_m, ...)
   3  props/other (g005 extent 0.196, _rou_f001_wep_05, e_rou_e005)

By gate: extent 32, coverage 8, winding 7. The extent bucket is almost entirely the
pose-proxy boxes, and lowering that floor was measured and refuted earlier. The one
genuinely interesting residual is e901_wing_05_L/R at winding 0.587/0.570 against the
0.70 floor — the signature of a thin double-sided sheet, unproven without a capture
with the boss on screen (flying stage 16 puts the container in memory but the unit
never appeared).

So: 6 247 / 6 294 resources decode (99.25 %), and every real mesh drawn in three
captured missions decodes at the GPU's own offsets with index runs matching byte for
byte (93/93 and 128/128).

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NptfmpjdpNCKEez6d2xvA9
This commit is contained in:
Claude (auto-RE)
2026-08-13 12:50:19 +00:00
parent 3b35c48fce
commit d1b50dc1e6
3 changed files with 47 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
docs/re/structures/xbg7-mesh.md
View File

@@ -1700,3 +1700,30 @@ The last row is the one that matters most: the four buffers that capture could n
name are the `n201` family, and now they decode **and** their index runs match the
GPU byte for byte. Suite green (`twin_pairs_do_not_share_a_buffer` included) apart
from the pre-existing known-failing cross-container consistency test.
### Where the anchor work stands: 6 247 / 6 294 (99.25 %), and what the residual 47 is
`examples/why_missed.rs` reports the furthest gate for every resource a container
fails to decode. Disc-wide that is **47 rows / 43 distinct names**, and they are
not a threshold away — they are mostly **not ship geometry**:
| class | rows | what they are |
|---|---|---|
| `e_rou_` / `_rou_` pose & proxy composites | **30** | scene-composite records like `e_rou_e901_attackA_start`, `…_guard_end`, `…_stand` — 24-vertex marker boxes whose extent is **0.010** |
| `.DAT` particle composites (`ptc_pack`) | 6 | `h309.DAT`, `m001_n.DAT`, … — the marker lists that provably do not map onto the stored blocks |
| damage / LOD variants | 8 | `e101_bdy_02_d` (indices reach 13 171 of 15 430 vertices), `e901_wing_05_*_m`, … |
| props / other | 3 | `g005` (extent 0.196), `_rou_f001_wep_05`, `e_rou_e005` |
By gate: **extent 32 · coverage 8 · winding 7**. The extent bucket is almost
entirely the pose-proxy boxes — a 0.010-unit cube is a *marker*, not a mesh, and
lowering the floor to admit them was measured before and refuted (it bought 2
resources and cost cross-container consistency). The `e901_wing_05_L/R` pair is the
one genuinely interesting residual: winding **0.587 / 0.570** against the 0.70 floor,
which is what a **thin double-sided sheet** looks like when face normals are compared
to averaged vertex normals — plausible but unproven, and the only way to settle it is
a capture with the boss on screen (flying stage 16 was not enough: the container is
resident but the unit never appeared).
So the anchor line of work is **done for practical purposes**: every real mesh the
engine draws in three captured missions decodes, at the offsets the GPU used, with
index runs matching byte for byte (93/93 and 128/128).