jpmschweitzer/settled-reach
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(ui): step-canvas map component — RTT terrain + stepped zoom (D-255, T-1182) #203

Closed
jpmschweitzer wants to merge 0 commits from t1182-client-map into main
pull from: t1182-client-map
merge into: :main
Conversation 1 Commits - Files Changed -
jpmschweitzer commented 2026-07-25 08:57:14 +02:00
Owner
Copy Link
Copy Source

Summary

The two-layer client rebuild per D-255(a)(b)(e), replacing the _canvas.scale continuous-zoom model with one viewer, one path, all six rungs. Net -9,400 lines.

  • step_canvas_protocol.gd: the T-1181 wire codec — incl. a discovered wire subtlety (rmp_serde without serde_bytes emits png_bytes as a msgpack int-array, not bin; verified against rmp_serde dispatch, repacked before Image.load_png_from_buffer) and the extent-echo rule (read the server-clamped extent, never assume the requested one).
  • step_canvas/ component: six-rung transport (cursor-anchored scroll steps, edge-scroll/WASD pan with edge-crossing re-request, hard reset-to-Global), RTT terrain layer (Image.set_pixel colorize per the c1 measured ruling, texture.update reuse, NEAREST coarse / LINEAR fine), unscaled screen-space annotation sibling (courses + markers at literal px), in-memory LRU (Tier 1; T-1183 layers disk beneath), request lifecycle with pending-hold (no blank frames).
  • Full _canvas.scale retirement in the same change: 10 source files + 14 dead test suites deleted (T-1157 dead-goldens rule; replacement visual-capture coverage owed under re-scoped T-1157).
  • Surviving surfaces kept per ticket — with one documented correction: the real LRU shape lives in atlas_window_cache.gd (kept, cited), not the retiring orchestrator atlas_window_tile_set.gd the ticket text named.

Verification

  • 7 new gdUnit suites (164 cases) — including a real extent-echo bug caught by its own test during implementation. Full client suite green through the gate (3,452 passed / 0 failed) with the SR_LIVE-gated suites RUNNING (worktree server built — a fresh-worktree pitfall diagnosed this session: no server binary -> live suites skip -> gdUnit orphan exit 101).
  • tooling/godot-cold-parse --run-menu: clean (autoload parse-order rule honored in the sim_bridge.gd edits).
  • Paired-set screenshot eyeball (cold server, main vs branch, incl. the lake window) queued next — merge waits on its verdict.

What you will SEE

Opening a body lands on the whole-body Global view; the wheel walks six discrete cursor-anchored steps down to 64 m Chunk; terrain is texel-exact, annotations constant-px; the previous step holds with a pending wash while the next loads. Not yet: disk cache (T-1183), step-cross tween (optional cosmetic), footprint-true settlement markers (server approximation inherited).

🤖 Generated with Claude Code

## Summary The two-layer client rebuild per D-255(a)(b)(e), replacing the _canvas.scale continuous-zoom model with one viewer, one path, all six rungs. Net -9,400 lines. - step_canvas_protocol.gd: the T-1181 wire codec — incl. a discovered wire subtlety (rmp_serde without serde_bytes emits png_bytes as a msgpack int-array, not bin; verified against rmp_serde dispatch, repacked before Image.load_png_from_buffer) and the extent-echo rule (read the server-clamped extent, never assume the requested one). - step_canvas/ component: six-rung transport (cursor-anchored scroll steps, edge-scroll/WASD pan with edge-crossing re-request, hard reset-to-Global), RTT terrain layer (Image.set_pixel colorize per the c1 measured ruling, texture.update reuse, NEAREST coarse / LINEAR fine), unscaled screen-space annotation sibling (courses + markers at literal px), in-memory LRU (Tier 1; T-1183 layers disk beneath), request lifecycle with pending-hold (no blank frames). - Full _canvas.scale retirement in the same change: 10 source files + 14 dead test suites deleted (T-1157 dead-goldens rule; replacement visual-capture coverage owed under re-scoped T-1157). - Surviving surfaces kept per ticket — with one documented correction: the real LRU shape lives in atlas_window_cache.gd (kept, cited), not the retiring orchestrator atlas_window_tile_set.gd the ticket text named. ## Verification - 7 new gdUnit suites (164 cases) — including a real extent-echo bug caught by its own test during implementation. Full client suite green through the gate (3,452 passed / 0 failed) with the SR_LIVE-gated suites RUNNING (worktree server built — a fresh-worktree pitfall diagnosed this session: no server binary -> live suites skip -> gdUnit orphan exit 101). - tooling/godot-cold-parse --run-menu: clean (autoload parse-order rule honored in the sim_bridge.gd edits). - Paired-set screenshot eyeball (cold server, main vs branch, incl. the lake window) queued next — merge waits on its verdict. ## What you will SEE Opening a body lands on the whole-body Global view; the wheel walks six discrete cursor-anchored steps down to 64 m Chunk; terrain is texel-exact, annotations constant-px; the previous step holds with a pending wash while the next loads. Not yet: disk cache (T-1183), step-cross tween (optional cosmetic), footprint-true settlement markers (server approximation inherited). 🤖 Generated with [Claude Code](https://claude.com/claude-code)
jpmschweitzer added 2 commits 2026-07-25 08:57:14 +02:00
feat(ui): step-canvas map component — RTT terrain + stepped zoom (D-255, T-1182) d28d24fd26 
The two-layer client rebuild per D-255(a)(b)(e), replacing the
_canvas.scale continuous-zoom model with one viewer, one path, all six
rungs:

- step_canvas_protocol.gd: StepCanvasRequest/Response codec against
  the T-1181 wire contract — incl. the discovered png_bytes subtlety
  (rmp_serde without serde_bytes emits a msgpack int-array, not bin;
  decode repacks via PackedByteArray before load_png_from_buffer) and
  the extent-echo rule (read the server-clamped extent, never assume
  the requested one).
- step_canvas/ component: transport (six-rung ladder, cursor-anchored
  scroll steps, edge-scroll/WASD pan with re-request on edge crossing,
  hard reset-to-Global), RTT terrain layer (Image.set_pixel colorize
  per the c1 measured ruling, texture.update reuse on step-cross,
  NEAREST coarse / LINEAR fine per rung), unscaled screen-space
  annotation sibling (courses + settlement markers at literal px),
  in-memory LRU cache (Tier 1; T-1183 layers the disk tiers beneath),
  request lifecycle (pending retry, staleness gate, extent echo).
- Full _canvas.scale retirement in the same change: the zoom-scaled
  canvas model, the _zs compensation family, select_rung /
  MAX_COVERAGE_M / compute_tile_grid, the orbital-mosaic-vs-window
  two-path split, _view_zoom/_canonical_fit_zoom — 10 source files
  deleted; their 14 test suites deleted with them (T-1157 dead-goldens
  rule; replacement visual-capture coverage is re-scoped T-1157).
- Surviving surfaces kept per the ticket: atlas_window_cache.gd's LRU
  shape (the ticket's named file atlas_window_tile_set.gd was the
  retiring orchestrator; the real LRU shape lives in
  atlas_window_cache.gd — cited in step_canvas_cache.gd), overlay
  colors, legend/overlay-bar chrome, AtlasViewer descend geometry.

Determinism boundary per D-255(e): the client interpolates only within
the closed server-supplied input set. 7 new gdUnit suites (164 cases)
incl. a real extent-echo bug caught by its own test during
implementation. Full client suite green (exit 0) with the live-gated
suites running against a worktree server build.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
chore(meta): update changelog d7fc26bd2a 
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
jpmschweitzer added 1 commit 2026-07-25 09:35:06 +02:00
fix(ui): PR #203 review round — total retirement + bin decode + wired reset 76d82a6a51 
Tyre finding: the orphaned AtlasViewer cluster is now actually deleted
(atlas_viewer, atlas_marker_overlay, atlas_descend_geometry,
atlas_legend_panel — a sixth orphan found beyond the review list —
atlas_generation_proxy, atlas_generation_state; ~2,497 lines), with
reachability re-verified across preload/class_name/res:// strings,
every .tscn, and the standalone companion app. Test suites triaged,
not blanket-deleted: 5 pure AtlasOverlayColors tests relocated into
test_atlas_window_colors, the live type-identity regression guard
relocated into test_step_canvas_viewer, dead coverage deleted. A real
harness gap surfaced during diligence and RULED, not patched:
visual_scenarios/visual_capture golden shots call retired
continuous-zoom API — no shim (would resurrect what D-255 kills);
inventory recorded on re-scoped T-1157 (gate-invisible, manual
targets only).

Hoshe finding 1: decode_png_field now detects the [Error,
PackedByteArray] bin-shape from messagepack.gd explicitly — a genuine
msgpack bin payload decodes correctly instead of silently collapsing
to [0,0]; test built from a real round-tripped bin decode.

Hoshe finding 2: the hard zoom-out reset is wired — ascend at rung 0
with a drifted view triggers _reset_to_global (the restored HARD
condition), behavioral tests through the real input path.

Notes folded: refloat + edge-scroll test coverage, legend smoke suite,
Vector2i narrowing-safety comment with computed headroom.

gdlint clean on touched files; full client suite 3364/3364.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
jpmschweitzer commented 2026-07-25 09:35:37 +02:00
Author
Owner
Copy Link
Copy Source

Review: t1182-client-map -> main (type: code, client)

Tyre (Architecture): REQUEST_CHANGES (1 finding — fixed)

D-255(e) verified clean — the component is a genuinely pure map-art function (server-sourced RTT colorize, unscaled annotations, no per-pixel invention; LINEAR filter correctly scoped as sanctioned texture-to-viewport resize) and the five-item retirement executed correctly inside the reachable path, with load-bearing doc comments throughout. Finding: ~3,600 lines of the AtlasViewer/_canvas.scale/_view_zoom cluster survived fully ORPHANED (zero production entry points — the orbital view is system_screen's own drawing), contradicting the one-viewer-one-path claim, plus ~1,300 lines of tests exercising the orphaned surface (the T-1157 dead-goldens anti-pattern).

Hoshe (Code Quality): REQUEST_CHANGES (2 findings — fixed)

Verified solid: D-243 spacing table matches scale.rs exactly, demux ordering provably safe, extent-echo + cache-key discipline mirror the server, deletion of the original 10 files clean, sim_bridge autoload edit follows the parse-order rule.

# File Issue
1 step_canvas_protocol.gd (decode_png_field) Silent byte-loss on genuine msgpack bin payloads: messagepack.gd's bin path returns [Error, PackedByteArray], which passes the Array guard and collapses to [0,0] — latent until a future serde_bytes annotation, then corrupts every texture silently.
2 step_canvas_viewer.gd (_reset_to_global) The documented hard zoom-out reset is dead code — zero production callers; regresses the explicit HARD condition the retired viewer honored.

Notes: refloat/edge-scroll/legend coverage gaps; Vector2i narrowing comment.

Author response — all fixed in 76d82a6a5

Tyre: reachability independently re-verified (incl. .tscn/string paths and the standalone app — surfaces the original grep could not see), then the cluster deleted TOTALLY: six files (~2,497 lines — a sixth orphan, atlas_legend_panel.gd, found beyond the review list). Test suites triaged, not blanket-deleted: 5 pure AtlasOverlayColors tests relocated into test_atlas_window_colors, the live type-identity regression guard relocated into test_step_canvas_viewer, dead coverage deleted. One REAL gap surfaced during diligence and RULED rather than patched: visual_scenarios/visual_capture golden shots (6 of 12, gate-invisible manual targets) call retired continuous-zoom API — no shim (would resurrect exactly what D-255 kills); full inventory recorded on re-scoped T-1157, which owns the (body, rung)-keyed harness redesign.

Hoshe 1: the [Error, PackedByteArray] bin-shape is now detected explicitly; test built from a REAL round-tripped bin decode (would have caught the original bug).

Hoshe 2: wired — ascend at rung 0 with a drifted view triggers _reset_to_global (new _is_global_view_drifted guard); behavioral tests through the real input path, plus the no-op case.

Notes: 4 refloat tests, 5 edge-scroll tests, a 10-test legend smoke suite, and the narrowing-safety comment with computed headroom — all folded, no parking lot.

Cold-parse clean post-deletion; full client suite 3,364/3,364 through the gate on re-push.

Screenshot eyeball: PASS (all four gates)

Cold start clean (Global fully drawn before 2s); the six-rung ladder walkthrough lands coherently with constant-px annotations and visible pending-hold; world-content agreement with the old viewer confirmed at the Arbour lake; untouched orbital screens byte-identical (AE=0). Rung-ladder capture set: .cache/screenshots/t1182-eyeball/rung-ladder/. Two non-blocking findings ticketed: T-1188 (blocky lake shorelines — server-side tone/refinement question, not viewer plumbing), T-1189 (Region-rung longitude wrap on small bodies).

Verdict: CHANGES REQUESTED -> resolved; merging (after PR #202, which holds the queue's outlet-stub check)

## Review: t1182-client-map -> main (type: code, client) ### Tyre (Architecture): REQUEST_CHANGES (1 finding — fixed) D-255(e) verified clean — the component is a genuinely pure map-art function (server-sourced RTT colorize, unscaled annotations, no per-pixel invention; LINEAR filter correctly scoped as sanctioned texture-to-viewport resize) and the five-item retirement executed correctly inside the reachable path, with load-bearing doc comments throughout. Finding: ~3,600 lines of the AtlasViewer/_canvas.scale/_view_zoom cluster survived fully ORPHANED (zero production entry points — the orbital view is system_screen's own drawing), contradicting the one-viewer-one-path claim, plus ~1,300 lines of tests exercising the orphaned surface (the T-1157 dead-goldens anti-pattern). ### Hoshe (Code Quality): REQUEST_CHANGES (2 findings — fixed) Verified solid: D-243 spacing table matches scale.rs exactly, demux ordering provably safe, extent-echo + cache-key discipline mirror the server, deletion of the original 10 files clean, sim_bridge autoload edit follows the parse-order rule. | # | File | Issue | |---|------|-------| | 1 | step_canvas_protocol.gd (decode_png_field) | Silent byte-loss on genuine msgpack bin payloads: messagepack.gd's bin path returns [Error, PackedByteArray], which passes the Array guard and collapses to [0,0] — latent until a future serde_bytes annotation, then corrupts every texture silently. | | 2 | step_canvas_viewer.gd (_reset_to_global) | The documented hard zoom-out reset is dead code — zero production callers; regresses the explicit HARD condition the retired viewer honored. | Notes: refloat/edge-scroll/legend coverage gaps; Vector2i narrowing comment. ## Author response — all fixed in 76d82a6a5 **Tyre**: reachability independently re-verified (incl. .tscn/string paths and the standalone app — surfaces the original grep could not see), then the cluster deleted TOTALLY: six files (~2,497 lines — a sixth orphan, atlas_legend_panel.gd, found beyond the review list). Test suites triaged, not blanket-deleted: 5 pure AtlasOverlayColors tests relocated into test_atlas_window_colors, the live type-identity regression guard relocated into test_step_canvas_viewer, dead coverage deleted. One REAL gap surfaced during diligence and RULED rather than patched: visual_scenarios/visual_capture golden shots (6 of 12, gate-invisible manual targets) call retired continuous-zoom API — no shim (would resurrect exactly what D-255 kills); full inventory recorded on re-scoped T-1157, which owns the (body, rung)-keyed harness redesign. **Hoshe 1**: the [Error, PackedByteArray] bin-shape is now detected explicitly; test built from a REAL round-tripped bin decode (would have caught the original bug). **Hoshe 2**: wired — ascend at rung 0 with a drifted view triggers _reset_to_global (new _is_global_view_drifted guard); behavioral tests through the real input path, plus the no-op case. **Notes**: 4 refloat tests, 5 edge-scroll tests, a 10-test legend smoke suite, and the narrowing-safety comment with computed headroom — all folded, no parking lot. Cold-parse clean post-deletion; full client suite 3,364/3,364 through the gate on re-push. ## Screenshot eyeball: PASS (all four gates) Cold start clean (Global fully drawn before 2s); the six-rung ladder walkthrough lands coherently with constant-px annotations and visible pending-hold; world-content agreement with the old viewer confirmed at the Arbour lake; untouched orbital screens byte-identical (AE=0). Rung-ladder capture set: .cache/screenshots/t1182-eyeball/rung-ladder/. Two non-blocking findings ticketed: T-1188 (blocky lake shorelines — server-side tone/refinement question, not viewer plumbing), T-1189 (Region-rung longitude wrap on small bodies). ### Verdict: CHANGES REQUESTED -> resolved; merging (after PR #202, which holds the queue's outlet-stub check)
jpmschweitzer closed this pull request 2026-07-25 10:39:14 +02:00

Pull request closed

This pull request cannot be reopened because the branch was deleted.
Sign in to join this conversation.
Reviewers
No Reviewers
Labels
No items
No labels
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
Clear assignees
jpmschweitzer (Jeroen Schweitzer)
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: jpmschweitzer/settled-reach#203