Fr0zkaandClaude Opus 5.5
9c6d261d01
Standalone only: fixtures, liminal_road_override (fillDais), --audit-liminal-road-walk, capture-only distance fog. Road 65 m x 6 m, 22 m up, capsule walk 127/127. Owner hashes unchanged (v227), fast tier, Clang/GCC green. Hills repeat every 80 m but the capture reaches 56 m, so only one rise shows. Also R48 (who owns the tunnel mouth): recommends a bounded joint plan of passage ports and layer landing candidates. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016dhxM43dnWvUu2B5Ph1iG1
37 lines
2.2 KiB
Markdown
37 lines
2.2 KiB
Markdown
# VoxelForge — research R48 (read-only): who owns the tunnel mouth?
|
||
|
||
You are a researcher. **Read only:** don't edit, create or delete any file, and don't build, run or test anything.
|
||
Read code, reports and design docs. Be thorough, verify by hand, and mark each claim [CODE] file:line, [MEASURED]
|
||
report, or [INFERENCE]. This checkout (`Saved/Linux/r48-tree`) is branch `wip/l134-cave-planner` at `4f93f3f`.
|
||
|
||
## The problem
|
||
|
||
Inter-strate passages (tunnels between layers) are planned first, and each layer's content is planned afterwards
|
||
around their fixed mouths and tubes.
|
||
- **Tunnel Network** (L134–L138) works, because caves are carved into rock and can bend to the mouths.
|
||
- **Vertical Shafts** (L139): 15/16 seeds pass; seed 2 fails the shell and seams at the arrival mouth. Three local
|
||
geometry fixes failed.
|
||
- **Floating Islands** (L133 twice, then L140): 0/16 seeds pass.
|
||
- In 14 of 16 seeds, no island arrangement clears the fixed tubes and reaches both mouth heights.
|
||
- One seed has a composed collision.
|
||
- Wall ripple is 6–7 cm against the 3.5 cm limit on the passages.
|
||
|
||
Read the L133, L139 and L140 reports and patches in `Standalone/Research/WorldAssemblerM3/`.
|
||
|
||
## Questions
|
||
|
||
1. Where and in what order are passages and layer contents planned (adapter, `ResolvedLayoutInternal.h`, the
|
||
passage planner, the island and cave planners)? What does each stage fix that later stages must obey?
|
||
2. Which designs would make mouths solvable by construction for every layer type? For example:
|
||
- **content first:** the layer plans its landing spots, and passages are routed to target them;
|
||
- **negotiated:** passages reserve a mouth landing region, and every layer planner must provide support there,
|
||
a "landing pad" contract;
|
||
- **a joint plan:** a corridor and the island anchors are solved together.
|
||
For each: what changes, the risks (determinism, the existing audits, the passage smoothness/ripple work), and the
|
||
size.
|
||
3. Why does ripple fail at 6–7 cm on the upper passages with native islands (and in L67)? Is it the passage, or the
|
||
host field under it?
|
||
4. Recommend one design, with steps and the evidence that would prove or kill it.
|
||
|
||
Write plain Markdown, ≤ ~1,200 words.
|