test(M00): record Goldshire pose evidence

docs/CAMERA_POSE_SWEEP.md

@@ -14,10 +14,11 @@ This offline M00 diagnostic recovers reproducible checkpoint framing when an ori
   -Checkpoint elwynn_adt_boundary `
   -YawOffsets -15,-10,-5,0,5,10,15 `
   -PitchOffsets -10,-5,0,5,10 `
+  -ViewportWidth 1280 -ViewportHeight 960 `
   -WaitSeconds 2
 ```
 
-Each candidate receives its own output directory and comparison report. `ranking.json` sorts candidates by mean perceptual error and then changed-pixel ratio. Use `-PlanOnly` to validate the Cartesian grid and output paths without rendering. Run a coarse grid first, then a finer grid around the best candidate.
+Each candidate receives its own output directory and comparison report. The viewport must exactly match the reference image dimensions; the runner fails on `size_mismatch` instead of inventing a score. `ranking.json` sorts candidates by mean perceptual error and then changed-pixel ratio. Use `-PlanOnly` to validate the Cartesian grid and output paths without rendering. Run a coarse grid first, then a finer grid around the best candidate.
 
 ```mermaid
 flowchart LR
@@ -37,3 +38,11 @@ The smallest error is a candidate, not automatic fidelity approval. Geometry, ma
 ## Verification
 
 The headless capture dry-run must report requested yaw and pitch offsets. Comparator `--self-test` covers paired metrics. Sweep `-PlanOnly` covers deterministic grid expansion and output naming without requiring a display. A real ranked sweep requires a display and a private build 12340 reference image.
+
+## Goldshire Inn evidence — 2026-07-12
+
+The private `2560x1440` build 12340 `goldshire_inn_large_wmo` reference was evaluated at FOV 62 degrees. A coarse 5x5 grid over yaw/pitch `[-20, -10, 0, 10, 20]` reduced mean error from `0.099632` at `(0, 0)` to `0.077575` at `(-10, -20)`. Extending pitch produced `0.070901` at `(0, -40)` and `0.063574` at `(0, -60)` with changed-pixel ratio `0.549556`.
+
+This is not a camera-pose solution. Human inspection shows the original-client reference centered on the Goldshire Inn, while the zero-offset OpenWC candidate contains forest/roads and no inn; negative pitch increasingly fills the frame with grass. The monotonically improving full-frame metric rewards similar green color coverage rather than landmark alignment. Therefore the manifest pose is unchanged and this checkpoint remains spatial/placement-composition evidence. A future registration metric must use approved landmark or region masks before it can select camera pose automatically.
+
+During this run the sweep orchestration was hardened in two ways: expected comparator exit code `1` is now collected through an explicit child process, and viewport dimensions are explicit. A reference/candidate `size_mismatch` now fails before ranking instead of producing an empty metric.

docs/modules/world-renderer.md

@@ -230,6 +230,7 @@ Exact exported settings and cache versions remain documented in [`../../RENDER.m
 - Camera-occluder probe не нашёл camera containment в пяти точках; paired mismatch локализован прежде всего в manual look direction/target/FOV calibration, с явным ограничением по RID-only geometry.
 - Empirical FOV sweep выявил, что checkpoint camera должна явно вызывать `make_current()`; после исправления projection ranking остаётся inconclusive из-за неизвестного manual yaw/pitch/framing reference.
 - Camera-pose sweep can now rank bounded yaw/pitch grids without changing manifest defaults; perceptual ranking remains diagnostic and requires human framing approval.
+- Goldshire Inn pose evidence has no landmark-aligned optimum: full-frame error decreases as negative pitch replaces the missing inn composition with grass. The checkpoint remains a spatial/placement gap, not a validated camera offset.
 - D3D12 descriptor issues remain; the capture-path anonymous `RefCounted` shutdown leak is regression-covered by a clean verbose dry-run, while other RID/resource diagnostics still require independent evidence.
 - M2/WMO/material/particle/ribbon/portal parity incomplete.
 - Public API is mostly exported configuration rather than stable contracts.