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

fix(client): Sprint 22 — fog rendering fix and seed config #78

Closed
jpmschweitzer wants to merge 0 commits from client into main
pull from: client
merge into: :main
Conversation 1 Commits - Files Changed -
jpmschweitzer commented 2026-02-28 16:28:46 +01:00
Owner
Copy Link
Copy Source

Summary

Sprint 22 client team deliverables:

  • #569 — Fix fog system: blocky edges and zero visibility in explored areas

    • Root cause: update_from_state() used visible_tiles (empty in live server mode) instead of visible_positions for bounds calculation
    • Shader fix: removed guard that cut off Gaussian gradient at unexplored boundaries, doubled blur step for D-066 compliant 6-8 tile gradient
    • Added debug_exploration mode for diagnostic rendering
    • 19 acceptance tests

  • #175 — Entanglement ratio configuration (world_seed IPC)

    • StartupMessage struct with world_seed: u64 added to bridge types
    • Full protocol flow: client SessionManager → GameState.world_seed → SimBridge sends after handshake → server seeds SimRng
    • Protocol changes across local.rs, tcp.rs, sim_bridge.gd, protocol.gd

  • #178 — Seed-based variation

    • EntanglementConfig with from_rng() / from_seed() in server/src/content/entanglement.rs
    • flat ∈ [25,35]%, intrigue ∈ [15,25]%, mundane derived as remainder (D-029)
    • 10 Rust tests: determinism, variation ≥90%, bounds, sum invariant

Test results

  • 1061 server tests pass (including 10 new EntanglementConfig + 3 StartupMessage tests)
  • 19 new GDScript fog tests + 9 entanglement stubs + 2 encode tests

Known items

  • Visual QA sign-off from Araminta needed for fog (#569) before merging — headless verification insufficient
  • Mundane ratio spec note: D-029 says [45,55]% but independent sampling gives [40,60]% — documented with TODO

Decisions referenced

D-059 (fog shader spec), D-015 (vision cone), D-066 (dual-scale grid), D-029 (entanglement ratios), D-010 (deterministic simulation), D-020 (dumb renderer)

## Summary Sprint 22 client team deliverables: - **#569 — Fix fog system: blocky edges and zero visibility in explored areas** - Root cause: `update_from_state()` used `visible_tiles` (empty in live server mode) instead of `visible_positions` for bounds calculation - Shader fix: removed guard that cut off Gaussian gradient at unexplored boundaries, doubled blur step for D-066 compliant 6-8 tile gradient - Added `debug_exploration` mode for diagnostic rendering - 19 acceptance tests - **#175 — Entanglement ratio configuration (world_seed IPC)** - `StartupMessage` struct with `world_seed: u64` added to bridge types - Full protocol flow: client SessionManager → GameState.world_seed → SimBridge sends after handshake → server seeds SimRng - Protocol changes across local.rs, tcp.rs, sim_bridge.gd, protocol.gd - **#178 — Seed-based variation** - `EntanglementConfig` with `from_rng()` / `from_seed()` in server/src/content/entanglement.rs - flat ∈ [25,35]%, intrigue ∈ [15,25]%, mundane derived as remainder (D-029) - 10 Rust tests: determinism, variation ≥90%, bounds, sum invariant ## Test results - 1061 server tests pass (including 10 new EntanglementConfig + 3 StartupMessage tests) - 19 new GDScript fog tests + 9 entanglement stubs + 2 encode tests ## Known items - Visual QA sign-off from Araminta needed for fog (#569) before merging — headless verification insufficient - Mundane ratio spec note: D-029 says [45,55]% but independent sampling gives [40,60]% — documented with TODO ## Decisions referenced D-059 (fog shader spec), D-015 (vision cone), D-066 (dual-scale grid), D-029 (entanglement ratios), D-010 (deterministic simulation), D-020 (dumb renderer)
jpmschweitzer added 3 commits 2026-02-28 16:28:46 +01:00
fix(client): fix fog system — blocky edges and zero explored visibility (#569) ebc87d8e74 
Root cause: update_from_state() used visible_tiles (always empty in live
server mode) instead of visible_positions for bounds calculation. Bounds
never grew beyond 64x64, so tiles outside that area rendered as solid
unexplored black.

Fix: new _grow_bounds_from_positions() method reads visible_positions
(always populated from server snapshots). Shader fix: removed the
(explored < 0.01 && vis_raw < 0.01) guard that cut off the Gaussian
gradient at unexplored tile boundaries. Doubled blur step size for
D-066 compliant 6-8 tile soft gradient. Added debug_exploration mode
for diagnostic rendering of the exploration texture.

19 acceptance tests covering exploration persistence, bounds grow-only
invariant, gradient margin, Forward-only visibility writes, and
exploration data surviving texture resize.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
feat(simulation): add world_seed IPC and EntanglementConfig (#175, #178) d591f44b35 
Implements the StartupMessage protocol: client generates world_seed in
SessionManager.new_game(), sends it after handshake, server uses it to
seed SimRng and sample EntanglementConfig.

EntanglementConfig samples flat ∈ [25,35]%, intrigue ∈ [15,25]%, mundane
as remainder (D-029). Same seed produces identical config (D-010
determinism). Different seeds produce distinct configs in ≥90% of pairs.

Protocol flow: HandshakeMessage (server→client) → StartupMessage with
world_seed (client→server) → SimRng initialization → tick loop.

10 Rust tests (determinism, variation, bounds, sum invariant).
9 GDScript test stubs + 2 encode tests for client-side pipeline.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
chore(meta): update changelog b89c4d5c9d 
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
jpmschweitzer commented 2026-02-28 18:53:35 +01:00
Author
Owner
Copy Link
Copy Source

Review: client → main (type: code)

Hoshe (Code Quality): REQUEST_CHANGES

Solid implementation with good test coverage. Three warnings.

# File Severity Issue
1 session_manager.gd:38 warning randi() produces u32 only — upper 32 bits of u64 world_seed always zero. Document limitation or widen entropy.
2 entanglement.rs:183-199 warning mundane_ratio test asserts [40,60] but D-029 specifies [45,55]. Spec divergence shipped as TODO — must resolve or update D-029.
3 fog_state.gd:11 suggestion VIS_PERIPHERAL = 180 is dead code — no longer written anywhere. Remove it.
4 test_fog_sprint22.gd:16-20 suggestion Private field access via get() + push_warning silently passes if fields rename. Consider test-only accessors or fail().
5 fog_state.gd:161-187 suggestion _grow_bounds() is dead code — replaced by _grow_bounds_from_positions(). Remove or document.
6 layer3.rs:88 suggestion Only tests seed=42 — zero and max-u64 not exercised at integration level.

Tyre (Architecture): REQUEST_CHANGES

Clean architecture, seed IPC is well-designed. Two warnings block merge.

# File Severity Issue
1 entanglement.rs:187-202 warning mundane_ratio bounds violate D-029 spec [45,55] — test widens to [40,60] with a TODO. Must fix sampling or update D-029.
2 session_manager.gd + main.rs warning world_seed not persisted to save file — resume_game() sends seed=0 to server, breaking deterministic replay (D-010 principle 4).
3 fog.gdshader:9,93-96 suggestion exp_fade comment says "soft transition" but filter_nearest on exploration_tex makes it a hard step. Comment is misleading.
4 entanglement.rs + main.rs suggestion EntanglementConfig::from_rng() never called in production — seed plumbing works but config isn't consumed yet. Intentional scope boundary? Needs tracking ticket.
5 session_manager.gd:24-25 suggestion game_id hex and world_seed are consecutive outputs from same PRNG — correlated. Mild metagaming concern per D-029.

Verdict: CHANGES REQUESTED

Shared concerns across both reviewers:

  1. mundane_ratio spec divergence — both flagged the [40,60] vs D-029's [45,55]. Fix the sampling or update the spec.
  2. world_seed not in save file (Tyre) — loaded sessions always get seed=0, breaking deterministic replay.
  3. u32 seed entropy (Hoshe) — document the limitation or widen to full u64.

The fog fix (#569) is clean. The entanglement plumbing (#175/#178) is architecturally sound but has these two data integrity gaps that need resolution before merge.

## Review: client → main (type: code) ### Hoshe (Code Quality): REQUEST_CHANGES Solid implementation with good test coverage. Three warnings. | # | File | Severity | Issue | |---|------|----------|-------| | 1 | `session_manager.gd:38` | warning | `randi()` produces u32 only — upper 32 bits of u64 `world_seed` always zero. Document limitation or widen entropy. | | 2 | `entanglement.rs:183-199` | warning | `mundane_ratio` test asserts [40,60] but D-029 specifies [45,55]. Spec divergence shipped as TODO — must resolve or update D-029. | | 3 | `fog_state.gd:11` | suggestion | `VIS_PERIPHERAL = 180` is dead code — no longer written anywhere. Remove it. | | 4 | `test_fog_sprint22.gd:16-20` | suggestion | Private field access via `get()` + `push_warning` silently passes if fields rename. Consider test-only accessors or `fail()`. | | 5 | `fog_state.gd:161-187` | suggestion | `_grow_bounds()` is dead code — replaced by `_grow_bounds_from_positions()`. Remove or document. | | 6 | `layer3.rs:88` | suggestion | Only tests seed=42 — zero and max-u64 not exercised at integration level. | ### Tyre (Architecture): REQUEST_CHANGES Clean architecture, seed IPC is well-designed. Two warnings block merge. | # | File | Severity | Issue | |---|------|----------|-------| | 1 | `entanglement.rs:187-202` | warning | `mundane_ratio` bounds violate D-029 spec [45,55] — test widens to [40,60] with a TODO. Must fix sampling or update D-029. | | 2 | `session_manager.gd` + `main.rs` | warning | `world_seed` not persisted to save file — `resume_game()` sends seed=0 to server, breaking deterministic replay (D-010 principle 4). | | 3 | `fog.gdshader:9,93-96` | suggestion | `exp_fade` comment says "soft transition" but `filter_nearest` on `exploration_tex` makes it a hard step. Comment is misleading. | | 4 | `entanglement.rs` + `main.rs` | suggestion | `EntanglementConfig::from_rng()` never called in production — seed plumbing works but config isn't consumed yet. Intentional scope boundary? Needs tracking ticket. | | 5 | `session_manager.gd:24-25` | suggestion | `game_id` hex and `world_seed` are consecutive outputs from same PRNG — correlated. Mild metagaming concern per D-029. | ### Verdict: CHANGES REQUESTED **Shared concerns across both reviewers:** 1. **`mundane_ratio` spec divergence** — both flagged the [40,60] vs D-029's [45,55]. Fix the sampling or update the spec. 2. **`world_seed` not in save file** (Tyre) — loaded sessions always get seed=0, breaking deterministic replay. 3. **u32 seed entropy** (Hoshe) — document the limitation or widen to full u64. The fog fix (#569) is clean. The entanglement plumbing (#175/#178) is architecturally sound but has these two data integrity gaps that need resolution before merge.
jpmschweitzer added 1 commit 2026-02-28 22:11:26 +01:00
fix(client): address PR #78 review comments 50aba3adf6 
- Widen world_seed entropy from u32 to full u64 by combining two randi()
  calls (Hoshe warning #1)
- Persist world_seed to save directory and restore on resume_game() so
  loaded sessions maintain D-010 deterministic replay (Tyre warning #2)
- Constrain EntanglementConfig intrigue range based on flat value so
  mundane_ratio stays within D-029 spec [45,55]% (both reviewers)
- Remove dead VIS_PERIPHERAL constant and _grow_bounds() method
- Update test_client_p1 peripheral test for forward-only simplification
- Fix misleading exp_fade shader comment (filter_nearest = hard step)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
jpmschweitzer added 1 commit 2026-02-28 23:24:51 +01:00
fix(client): prevent signed overflow in world_seed generation and load 6fac224d55 
GDScript int is i64 — when randi() returns a value with bit 31 set,
left-shifting by 32 sets bit 63, producing a negative i64. MessagePack
encodes this as a negative integer, which Rust rmp_serde rejects when
deserializing as u64, causing ~50% startup failure rate.

Fix: mask bit 31 before shifting in new_game() to cap entropy at 63
bits. Also mask the sign bit in _read_seed_file() to handle save files
written before this fix.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
jpmschweitzer closed this pull request 2026-02-28 23:34:41 +01:00

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#78