Files
benilla-octo/crates
gromandClaude Opus 5 ca60826bec The creature tables, and three things in them worth more than a count
`CreatureDisplayInfo` is 13452 rows where 5875 had 10534 and `CreatureModelData` ships different
art under the same paths, so every census over them moved. Field counts are unchanged (12 and 16),
so the layouts are intact.

Three are findings, not bookkeeping:

- **One display row dangles.** Display 185's `modelId` is 0 — it points at no `CreatureModelData`
  row at all, which 1.12 had none of. The blood-tier walk skips it because there is no model to
  read a column from, and our client falls back the same way: no model, nothing to draw. Pinned by
  id so a *second* dangling row gets looked at rather than absorbed.
- **Five races' collision boxes no longer match their columns.** The realm ships its own character
  art under unchanged paths, and the boxes moved while the DBC columns stayed: OrcMale and
  TrollMale carry a 2.031 box (HumanMale's — a reused rig) under columns of 2.361 and 2.083,
  DwarfMale 1.607 under 1.667, TrollFemale 2.427 under 1.839, GnomeFemale 1.056 under exactly 1.0.
  The column stays authoritative, which is the part that matters: the reference reads it and not
  the model (`0x60b270` → `0x6174b0`), so the depth lines follow the column and an orc's swim line
  sits a little above where its art suggests. The exception list asserts they still *dis*agree, so
  a race coming back into line gets the entry deleted instead of a stale carve-out.
- **The prism floor is exercised now, not inert.** Five `CreatureModelData` rows scale below 1
  (2602, 2719-2722 at 0.4 down to 0.2) where 1.12 had none, which is exactly where
  `max(SCALE_X, display scale)` bites. The arithmetic it exercises already has its own test in
  `entities::collision_height`, so this one names the five rows and points at it; the test is
  renamed for what it now says.

Also: HumanMale's footprint is 10 inches long and 14 wide, where 1.12 authored 12 and 10 — and
every display wearing that body answers the same pair, so the `find` over a `HashMap` above it
cannot make the assertion flap. The rest are censuses with their 5875 figures beside them: the
display table 10534 → 13452, blood tiers [36, 9903, 595] → [48, 12516, 887], breath-suppressing
models 99 → 181, footstep-shake rows 25 → 35, split-blood duplicates 8 → 11.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn
2026-10-31 18:52:40 -06:00
..